Tricarboxylic Acid Cycle and Glyoxylate Bypass

JOHN E. CRONAN, JR.,^1* AND DAVID LAPORTE²

[SECTION EDITOR: AUGUST BöCK]

Posted November 9, 2006

Departments of Microbiology and Biochemistry, University of Illinois, Urbana, IL 61801,¹ and Department of Biochemistry, Molecular Biology & Biophysics, University of Minnesota, Minneapolis, MN 55455²

*Corresponding author. Phone: (217) 333-0425, Fax: (217) 244-6697, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

The tricarboxylic acid (TCA) cycle plays two essential roles in metabolism (2). First, the cycle is responsible for the total oxidation of acetyl-CoA that is derived mainly from the pyruvate produced by glycolysis (Fig. 1). Second, TCA cycle intermediates are required in the biosynthesis of several amino acids. Although the TCA cycle has long been considered a "housekeeping" pathway in Escherichia coli and Salmonella enterica, the pathway is highly regulated at the transcriptional level (52, 69, 70). Regulation was first demonstrated by enzymatic assays for some gene products. However, other TCA cycle enzymes appeared to be expressed constitutively (46). Subsequent work demonstrated that this latter group of enzymes is also inducible and the appearance of constitutivity had resulted from differential regulation of two or more genes (50, 52). For example, the anaerobic repression of succinate dehydrogenase was masked by the anaerobic induction of fumarate reductase (50). Similarly, regulation of the aerobic fumarase, FumA, was masked by the anaerobic FumB enzyme and the unregulated FumC (154).

Fig. 1TCA cycle and glyoxylate bypass. The glyoxylate bypass is composed of isocitrate lyase and malate synthase. IDH and IDH-P indicate the dephosphorylated and phosphorylated forms of isocitrate dehydrogenase, respectively.

The expression levels of the TCA cycle enzymes respond primarily to the presence of oxygen (2) and to the carbon source (2, 46) and seem determined solely by mRNA levels. In E. coli, the fully expressed TCA cycle is seen only during aerobic growth on acetate or fatty acids (2). These culture conditions give the highest levels of TCA cycle enzymes, and the cycle provides all the energy and reducing potential needed to support growth. Growth on acetate or fatty acids, however, requires induction and function of an anaplerotic pathway, the glyoxylate bypass (77, 78), to replenish the dicarboxylic acid intermediates consumed in amino acid biosynthesis; this is also discussed. Note that the great bulk of recent work on the TCA cycle has been done by Guest and coworkers; these investigators have written several very valuable reviews (50, 52, 99). Due to space limitations we do not repeat the references to the original literature that can readily be found in these reviews, but instead we refer to the most relevant review. In contrast, most of the voluminous literature on the regulation, enzymology, and structures of the TCA cycle enzymes has not been reviewed recently, and if no review is available, we provide recent references as an entree into the literature. Note that recent data indicate that the TCA cycle functions as a full cycle during pathogenesis of mice by Salmonella enterica serovar Typhimurium and that the anaerobic metabolic enzyme fumarate reductase has no role in the processes (140).

It has long been known that, during growth on glucose under anaerobic conditions, the TCA cycle can no longer provide energy and instead functions as two biosynthetic pathways, a reductive pathway that produces succinyl-CoA acid and an oxidative pathway producing 2-ketoglutarate (Fig. 2). These pathways fail to form a cycle because 2-ketoglutarate dehydrogenase and succinate dehydrogenase are virtually absent under these growth conditions (2, 70, 128) (synthesis of these enzymes is the most severely repressed of all the TCA cycle enzymes). Under these fermentative conditions, the levels of the other enzyme activities (and enzyme proteins) are lower (as much as 10- to 20-fold) than those found during aerobic growth. This reduction in expression is appropriate since a much lower flux would be required when only biosynthetic functions are served (2, 46, 52). However, recent data indicate that excess glucose must be the carbon source to elicit the fully bifurcated pathway in wild-type E. coli strains (1). The pathway remains almost totally bifurcated even when an electron acceptor such as nitrate is provided in place of oxygen. Under anoxic conditions, however, when the carbon source is glycerol or lactate and nitrate is the electron acceptor, a functionally complete TCA cycle is formed but as in the case of glucose the poorer electron acceptors, fumarate and dimethyl sulfoxide, cannot replace nitrate. Hence, as glucose-grown cultures become anaerobic much of the switch from the cyclic pathway to the bifurcated pathway is due to oxygen availability, with the switch occurring in the microaerobic range (20 to 40% of totally aerobic conditions defined as giving complete oxidation of glucose) (1, 116, 119, 127). Hence, to understand the regulation of expression of the TCA cycle enzymes at least four growth conditions must be considered: (i) fully aerobic growth on glucose where full aerobiosis is defined as a complete lack of acetate excretion, (ii) microaerobic growth on glucose, (iii) anaerobic growth on glucose, and (iv) anaerobic growth on nonfermentable carbon sources plus nitrate. Note that the data supporting these conclusions are in vivo measurements of TCA cycle function based on determination of the metabolites that result from incomplete oxidation of the carbon source or by metabolic flux determinations (which are more sensitive). Although results were obtained for both batch and carbon source-limited chemostat cultures, the latter results seem more reliable since growth conditions can be more stringently controlled.

Fig. 2Function of the TCA cycle in anaerobic growth. The oxidative branch is to the right, and the reductive branch is to the left.

During anaerobic growth, several other enzymes are induced to augment the biosynthetic pathways formed by the TCA cycle enzymes. For example, fumarate reductase replaces succinate dehydrogenase to allow reductive production of succinyl-CoA (as well as providing an anaerobic respiratory pathway). Aspartase is induced to assist (together with the constitutive aspartate aminotransferase) in the conversion of oxaloacetate to fumarate. Furthermore, some carbon does flow between the branches by the action of the glyoxylate cycle enzyme, isocitrate lyase, which converts isocitrate to succinate and glyoxylate. The action of isocitrate lyase under these conditions is demonstrated by the finding that anaerobically grown cells display a succinate requirement only when both isocitrate lyase and fumarate reductase are inactivated by mutation (24).

The TCA cycle enzymes may also be subject to end product repression by amino acids (46, 52). Repression by amino acids would seem advantageous in the branched mode of the cycle and data suggesting a small (two- to fourfold) decrease in TCA cycle enzyme levels upon addition of amino acid mixtures to aerobic glucose-grown cultures have been reported (46). However, other data indicate that addition of glutamate (the major TCA cycle-derived amino acid) to such glucose cultures gave increased enzyme activity (2). These conflicting reports seem likely to result from differences in growth conditions. For example, cultures that were thought to be aerobic were actually microaerobic. When the TCA cycle must function as the primary source of energy, amino acid repression (if it occurs) would probably be a disadvantage. Consistent with this prediction, similar levels of 2-ketoglutarate dehydrogenase activity are seen when cultures are grown on either acetate or glutamate as sole carbon source (2).

GENE-ENZYME RELATIONSHIPS IN THE TCA CYCLE

All of the E. coli genes that encode the TCA cycle enzymes and the major auxiliary enzymes have been cloned, sequenced, and placed on the genetic and physical maps (52, 99) (Table 1). One 13-kbp chromosomal segment contains nine TCA cycle genes that encode the subunits of four of the cycle enzymes, whereas the genes encoding the remaining enzymes are scattered around the chromosome. In general, the genes encoding the subunits of multisubunit complexes are adjacent and cotranscribed, the exception being the gene (lpdA) encoding the E3 subunit of 2-ketoglutarate dehydrogenase which can be either cotranscribed with the genes (aceEF) encoding the subunits of pyruvate dehydrogenase (the E3 subunit is common to both 2-ketoacid dehydrogenases) or transcribed separately (61). Transcriptional mapping of the TCA cycle gene cluster has provided no strong rationale for clustering except for genes encoding subunits of a heteromeric enzyme (99). Indeed the gltA (citrate synthase) gene is transcribed in the direction opposite to the other TCA cycle genes. Most genes within this cluster are transcribed from two different promoters and those transcripts that encode subunits of two TCA cycle enzymes (e.g., sucAB and sucCD) provide the only rationale for clustering (30).

Table 1Genes and enzymes

Many of the TCA cycle genes are named for the enzyme encoded; others are named for the TCA cycle intermediate needed to bypass the enzyme deficiency. Some of these gene names illustrate the dual functions of the TCA cycle. For example, the citrate synthase gene is called gltA because mutants lacking the enzyme require 2-ketoglutarate, which is readily supplied as glutamate or proline. Similarly, mutations in sucAB, which encode two of the three 2-ketoglutarate dehydrogenase subunits, result in a requirement for succinate for aerobic growth on glucose (13, 14). However, some of the gene designations seem both confusing and misleading (although at this stage we are probably stuck with them). The sucCD genes encode the two subunits of succinyl-CoA synthetase and mutants defective in this enzyme are unable to utilize succinate as a source of succinyl-CoA when grown anerobically (59). Therefore, the suc designation has opposite meanings in the sucAB and sucCD cases. Moreover, the use of consecutive letters implies that the gene products are closely related, perhaps being subunits of a single enzyme complex (as for the adjacent operon, sdhABCD) rather than two successive enzymes of the pathway. We suggest that stcAB (for succinate thiokinase, a more proper name for succinyl-CoA synthase) would be a more fitting designation for sucCD.

In several cases, the deduced amino acid sequences of the E. coli TCA cycle enzymes provided the first complete primary structures of the corresponding enzymes. Subsequent isolations of the TCA cycle genes from other organisms have generally given deduced amino acid sequences that strongly resemble those of the E. coli proteins. Note that discovery of the methylcitrate cycle of propionate utilization has eliminated some of the confusing older data concerned with the TCA cycle enzymes (Chapter Two-Carbon Compounds and Fatty Acids as Carbon Sources). For example, the PrpD enzyme was originally misidentified as a third aconitase (AcnC). However, although PrpD can dehydrate both citrate and methylcitrate, it is unable to catalyze the second stage of the aconitase reaction, the rehydration reaction (8, 13, 14). Others of the enzymes of one pathway have some activity with substrates of the other pathway which thereby confused analyses of both pathways. The "second" citrate synthase of Pattan and coworkers (111), isolated as a suppressor of a gltA mutant, is identified as PrpC by alignment of the N-terminal sequence with the proteins of E. coli (J. E. Cronan, unpublished data). The mutation that gave rise to this activity seems likely to be a mutation in PrpR that results in constitutively activating transcription of the prpBCDE operon (107, 108) and thereby suppressing the gltA mutation.

MUTANTS DEFECTIVE IN THE ENZYMES OF THE TCA CYCLE

Mutants lacking any of the TCA cycle enzymes have a common phenotype; all fail to grow with acetate as sole carbon and energy source. Some TCA cycle mutants (sucAB mdh, sdhCDAB, and fumA) grow normally on glucose minimal medium under anaerobiosis, and this growth is attributed either to a lack of function of the enzyme in the branched mode or to other enzymes that supply the missing biosynthetic intermediates. A caveat to some of these findings is that much of this work was done before null mutants (e.g., transposon insertions or clean deletions) in bacterial genes could be readily isolated, and some of the mutants tested may have retained activity undetectable by enzymatic assay. The enzymes involved in the oxidative side of the branched pathway, citrate synthase, aconitase, and isocitrate dehydrogenase, are responsible for glutamate synthesis and thus should be essential for both aerobic and anaerobic growth on glucose minimal media. Mutants lacking citrate synthase or isocitrate dehydrogenase are known to be glutamate auxotrophs under aerobic conditions, but the question of auxotrophy under anaerobic conditions does not appear to have been addressed. Therefore, one of us (J.E.C.) tested two different gltA null mutants and found that under anaerobiosis neither mutant strain grew on glucose unless supplemented with glutamate.

All of the genes needed to encode the enzymes of the TCA cycle have been cloned and sequenced. Most of these genes were isolated by complementation of mutants having lesions within the target genes (99). Other TCA cycle genes were isolated because they were closely adjacent to previously isolated TCA cycle genes and identified by DNA sequencing (99). In some cases, reverse genetic approaches were required. These approaches included screening with oligonucleotides based on amino acid sequence data or by screening with antibodies (118, 136). Other genes were isolated when residual enzyme activity was found to remain in cell extracts of strains in which the gene encoding the primary enzyme had been inactivated or when an expected growth phenotype was not manifested (48, 154). Interpretation of enzyme assay data is also problematic for those enzymes that contain Fe-S centers, since these activities can be labile in conventional cell extracts.

THE ENZYMES OF THE TCA CYCLE

Citrate Synthase

E. coli citrate synthase is a hexamer of the gltA-encoded protein (151). The allosteric behavior of this enzyme is typical of the citrate synthases of gram-negative bacteria and distinct from the enzymes of other bacteria and mammalian mitochondria (151). E. coli citrate synthase is inhibited by 2-ketoglutarate and activated by NADH; properties consistent with the function of the TCA cycle in the branched and energy-producing modes, respectively (151). The allosteric properties of this enzyme are complex and remain under active investigation (36, 96, 103, 131). The crystal structures of wild-type E. coli citrate synthase as well as several mutant proteins (36, 96, 103) have appeared. Activation by NADH is intimately involved with structure of the monomer interfaces of the hexamer. Given these data it seems likely that the insensitive citrate synthase of Danson and coworkers (31) has altered interfacial interactions.

Use of controlled overproduction of citrate synthase shows that this enzyme catalyzes the rate-limiting step of the TCA cycle when E. coli is grown on acetate but is normally made in functional excess in aerobic cultures grown on glucose (146). However, altered production of citrate synthase could also affect the phosphorylation of isocitrate dehydrogenase (see below); thus, it would be of interest to repeat the glucose growth experiments in an aceK null mutant.

Aconitase

E. coli contains two genetically distinct aconitases, AcnA and AcnB (72, 143). The crystal structure of AcnB has been solved (152). The corresponding genes have been cloned and sequenced. AcnA is about 17% identical with AcnB. The active-site residues and the cysteine residues that serve as ligands for the iron-sulfur clusters are conserved in both AcnA and AcnB. The two genes have different patterns of expression which have led to the conclusion that AcnB is the major TCA cycle enzyme, whereas AcnA is an aerobic-stationary-phase enzyme that is specifically induced by iron and oxidative stress. As expected from its regulation, AcnA is much more stable than AcnB and dominates the activity found in conventionally prepared cell extracts. AcnA is both an enzyme in the presence of iron and an mRNA-binding posttranscriptional regulator in the absence of iron. Both AcnA and AcnB play roles in sensing intracellular iron concentrations. In the case of AcnB the switch between catalytic and regulatory modes involves iron-dependent dimer formation (137). Loss of AcnB results in increases in the levels of 2-ketoglutarate dehydrogenase and succinyl-CoA synthetase (139). AcnB also plays a role in posttranslational regulation of flagellin synthesis (and hence in bacterial motility) that involves binding of the transcript encoding the FtsH protease (138).

Isocitrate Dehydrogenase

Isocitrate dehydrogenase (IDH) is a dimeric protein encoded by the icdA gene. Phosphorylation of this enzyme plays a key role in the partition of carbon between the TCA cycle and the glyoxylate bypass. We discuss this enzyme mainly in the context of the bypass. Note that the reported "icdB" gene (85) has been shown to be an allele of gltA (57). Recent data indicate that the IDH of E. coli has been tuned to act in acetate utilization in that the reaction depends on NADP although most of the known IDHs use NAD. A form of E. coli IDH was engineered to strongly prefer NAD over NADP and the wild-type gene was replaced with the gene encoding the engineered protein (159). Relative to the wild-type strain, the resulting strain grew more slowly on limiting acetate, but more rapidly on limiting glucose. The rationale of this experiment was that bacteria growing on glucose produce high levels of NADPH by the pentose phosphate pathway, whereas bacteria growing on highly oxidized carbon sources such as acetate lack this facile avenue of NADPH production and must rely on IDH, transhydrogenase-catalyzed conversion of NADH to NADPH, and the maeB-encoded malic enzyme for NADPH synthesis. Hence, loss of NADPH production by IDH results in slower growth of the mutant strain (159). This scenario predicts that the growth defect of acetate should be intensified if these other sources of NADPH were eliminated. Indeed, introduction of deletions of the pntAB (which encode the membrane-bound transhydrogenase) or maeB genes into the icdA mutant strain resulted in slower growth on acetate relative to strain that carried the deletion but had a wild-type icdA gene. A strain carrying the icdA mutation plus deletions of maeB, pntAB, and sthA (sthA is also known as udhA and encodes the soluble transhydrogenase) grows on acetate at only 16% of the rate of a strain that carries the three deletions but retains the wild-type icdA allele (growth rate on limiting glucose gave only very modest differences among the various strains carrying either the wild-type or mutant icdA alleles) (159). The essentiality of an NADPH-dependent ICD for growth on acetate is also clear from prokaryotic genomics. The presence of a gene encoding an NADPH-dependent IDH is invariably accompanied by an ICL gene, whereas no ICL gene is found in any genome that encodes an NAD-dependent IDH (159).

2-Ketoglutarate Dehydrogenase

An appreciable amount of our information on E. coli 2-ketoglutarate dehydrogenase comes by analogy with the closely related enzyme, pyruvate dehydrogenase, with which it shares the lipoamide dehydrogenase subunit (the lpdA gene product) (49, 52, 99, 114, 115). Lipoic acid is attached to a specific lysine residue close to the N terminus of the E2 (sucB) subunit by lipoate ligase and this modification is essential for enzyme activity in that this coenzyme acts as a classical swinging arm in moving enzyme intermediates between two active sites. The protein-bound lipoate accepts the decarboxylation product from thiamine pyrophosphate cofactor of the E1 (sucA) subunit and delivers this product to the second active site in the catalytic portion of SucB to form succinyl-CoA. The E3 (lpdA) subunit then reoxidizes the E2-bound dihydrolipoate acid moiety. The three subunits (encoded by the sucAB and lpdA genes) that make up the active enzyme are found in huge complexes of molecular weight 5 to 10 million that form particles of 30 to 40 nm (significantly larger than ribosomes). The enzyme is too large for direct structural analysis by current methods other than electron cryotomography and related techniques which show that E1 and E3 subunits are flexibly tethered ~11 nm away from the E2 core (100). However, parts of the complex have been studied by crystallography and nuclear magnetic resonance with considerable success (114, 115). The enzyme consists of a core of 24 copies of the E2 (sucB) subunit together with 12 copies of each of the other two subunits. However, it seems that this enzyme may not have a strictly defined structure, and complexes of various stoichiometry probably exist in vivo (100). The close analogy between 2-ketoglutarate dehydrogenase and pyruvate dehydrogenase is evident not only from the reaction chemistries and deduced amino acid sequences but also from the common quaternary structure (100, 114, 115). Note that the lpdA gene product is not confined to carbon metabolism. LpdA also functions as an essential component of the glycine cleavage system (130).

Upon dual overexpression, phosphoenolpyruvate (PEP) carboxykinase and PEP carboxylase are capable of catalyzing a futile cycle that would result in the wasteful hydrolysis of ATP (18). Surprisingly, this futile cycle is enhanced in mutants lacking 2-ketoglutarate dehydrogenase (18), suggesting that the TCA cycle may provide a protection against excess ATP consumption (18). It would be interesting to reexamine the effects of overproduction of these enzymes under conditions in which the PEP-glyoxylate cycle (see below) is active.

Succinyl-CoA Synthetase

The enzymology of E. coli succinyl-CoA synthetase (more properly called succinate thiokinase) is well studied (105). The enzyme is an α₂β₂-tetramer, and the reaction proceeds through a phosphorylated enzyme intermediate in which a specific His residue of the a subunit is transiently modified. High-resolution X-ray structures of the wild-type enzyme in both phosphorylated and dephosphorylated forms as well as a mutant enzyme have been reported (5, 40, 41, 73).

During aerobic growth succinyl-CoA synthase generates ATP (GTP is also produced in vitro). This substrate-level phosphorylation serves to use the energetic thioester bond of the succinyl-CoA produced by 2-ketoglutarate dehydrogenase to produce a nucleotide triphosphate. During anaerobic growth the enzyme serves to convert the succinate produced by fumarate reductase to the succinyl-CoA needed for biosynthesis. This point was demonstrated by isolation of phage Mu insertions into sucCD (95). These strains fail to grow anaerobically on glucose unless supplemented with lysine, methionine, and δ-aminolevulenic acid. The sucCD genes were originally identified by the similarity of the open reading frames (ORFs) encoded downstream of the sucAB genes to those of the mammalian succinate thiokinase subunits (99), and thus the Mu insertions document the physiological function of these genes.

Succinate Dehydrogenase

The succinate dehydrogenase of E. coli is encoded by four genes in an operon located within the cluster of TCA cycle genes (for review, see reference 52). The sdhA gene encodes a flavoprotein subunit containing a covalently bound FAD moiety, whereas the sdhB gene encodes an iron-sulfur protein (52). The sdhC and sdh D genes encode two very hydrophobic membrane proteins that serve to anchor the hydrophilic flavoprotein and iron-sulfur protein subunits to the cytoplasmic membrane and also participate in electron transport (the SdhC protein is cytochrome b₅₅₆) (52).

The SdhA and SdhB proteins have remarkable sequence similarities to the proteins encoded by the frdA and frdB genes of the anaerobic fumarate reductase, suggesting a common evolutionary origin (52). (The flavoprotein and iron-sulfur protein subunits of fumarate reductase are also anchored to the membrane by a pair of very hydrophobic membrane proteins unrelated to Sdh C and SdhD.) Succinate dehydrogenase and fumarate reductase catalyze the same reaction, the interconversion of succinate and fumarate. Indeed, the original cloning of the fumarate reductase operon resulted from attempts to clone the succinate dehydrogenase genes (multiple copies of frdABCD permit sdh mutants to grow aerobically) (52). As discussed above, these two enzyme complexes are differentially expressed. Succinate dehydrogenase is present only in aerobically grown cells, where it functions to donate electrons to the respiratory chain. In contrast, fumarate reductase is present only in anaerobically grown cells, where the enzyme synthesizes succinate in the branched mode of the TCA cycle and also provides a terminal oxidase to aid mixed acid fermentation (and to provide energy for anaerobic growth on nonfermentable carbon sources when fumarate is present in the medium).

Fumarase

E. coli contains three distinct fumarases, encoded by the genes fumA, fumB, and fumC (50, 52, 154). The first two enzymes are 90% identical and belong to a novel class of fumarases that are dimeric, oxygen-labile, iron-sulfur proteins (50, 154). FumC is a typical oxygen-stable tetrameric fumarase that strongly resembles the mammalian and yeast mitochondrial enzymes. Several crystal structures of FumC are available (147, 148, 149). FumA is clearly the enzyme designed for TCA cycle function, whereas FumB is derepressed under anaerobic conditions and functions as a malate dehydratase in the branched TCA cycle and also converts malate to fumarate to be utilized as an anaerobic electron acceptor (154). FumC is encoded by a gene adjacent to fumA, but the two genes are regulated differently. fumA is regulated in the same manner as the other TCA cycle genes, whereas fumC expression does not respond to O₂ or glucose levels. However, fumC has been reported to be regulated by the SoxRS regulon, which functions to protect redox balance and metabolism from oxidative stress (87). Replacement of the oxidation-labile FumA with higher levels of the stable FumC would serve such a protective function. However, several other TCA cycle enzymes are labile iron-sulfur proteins, and thus increased FumC activity alone seems unlikely to spare the TCA cycle from oxidative stress.

Malate Dehydrogenase

The malate dehydrogenase (MDH) of E. coli is a dimer of the protein encoded by the mdh gene and the deduced amino acid sequence is closely similar to that of mammalian mitochondrial enzymes and distinct from the mammalian cytosolic enzymes (136, 145; for review, see reference 97). Crystal structures of the enzyme in the presence and absence of citrate have recently been reported (54) as well as the crystal structure of a mutant enzyme (6). A monomeric form of the MDH enzyme has also been engineered (11). Mutations in a locus distinct from mdh were reported to result in a specific loss of malate dehydrogenase activity (23). This second locus has been assigned a regulatory role (136, 145) but remains ill defined. These mutants should be further explored. The putative anaerobic role for the mdh gene product has not been demonstrated and it seems possible that the anaerobic conversion of oxaloacetate to malate could be catalyzed by consecutive action of aspartate transaminase and aspartase (52). The latter enzyme is known to be derepressed under anaerobic conditions (154). Another E. coli malate dehydrogenase, actually a malate:ubiquinone oxidoreductase, encoded by the mqo gene has been reported. Although deletion of mqo had no effect on growth, the presence of the enzyme was shown to relieve some of the deleterious growth defects of mdh deletion strains (141). Expression of the mqo gene is under ArcA control (see below) and is low in stationary-phase cells, suggesting that it cannot function anaerobically (141).

Lipoic Acid Synthesis and Attachment

In E. coli and other organisms lipoic acid is essential for function of several key enzymes involved in oxidative and single-carbon metabolism including pyruvate dehydrogenase (PDH), 2-oxoglutarate dehydrogenase (2-OGDH), and the glycine cleavage system (124). In each enzyme, a specific subunit is modified by attachment of lipoic acid to specific lysine residues within conserved domains of these subunits. In each of these domains an amide linkage is formed between the carboxyl group of lipoic acid and the ε-amino group of the specific lysine residue (76). During catalysis, the protein-bound lipoamide moieties serve as carriers of reaction intermediates among the multiple active sites of these multienzyme complexes (124). Strains lacking the ability to synthesize lipoic acid or to attach the cofactor to the cognate proteins lose TCA cycle function because of the lack of acetyl-CoA and succinyl-CoA (60, 144). If such cultures are supplemented with acetate, the physiology is that of a sucAB mutant strain. Our knowledge of the pathways of lipoic acid synthesis, attachment, and function has progressed rapidly in the past 10 years largely due to complementary genetic and biochemical analyses in E. coli. Since lipoic acid synthesis has been reviewed recently (28) (and in this website), it will suffice to point out that there are two pathways whereby lipoic acid becomes attached to its cognate enzymes. The lipoate protein ligase encoded by the lplA gene scavenges lipoic acid from the medium, whereas the lipA lipB pathway is responsible for de novo biosynthesis. The pathway is unusual in that LipB transfers octanoate moieties of the octanoyl-acyl carrier protein intermediate of fatty acid synthesis to the apo forms of the 2-oxoacid dehydrogenases and LipA then inserts two sulfur atoms into the protein-bound octanoate to give the active dehydrogenases (158). That is, this essential enzyme cofactor is assembled "on site."

TCA CYCLE MYSTERIES

Several unexpected aspects of the TCA cycle remain unexplained. First, icdA mutants are readily isolated by selection for strains resistant to very low levels (about twice the minimum inhibitory concentration) of the DNA gyrase inhibitor, nalidixic acid (58). At the time of its discovery this phenomenon was inexplicable. However, it is now possible to propose an explanation. Many E. coli K-12 strains carry a defective lambdoid prophage called e14 that integrates into the icdA gene (47). The integration event fails to inactivate icdA because the prophage contains a 216-bp DNA segment that is essentially homologous to the C-terminal end of the icdA gene. During the integration event this segment becomes fused to the upstream icdA sequence such that the protein acquires a new, fully functional C terminus that differs from the native protein by only two amino acid residues (61). Excision of the e14 prophage is known to be induced by DNA damage through the host SOS response (15). Nalidixic acid is a potent inducer of the SOS response (102) and thus of e14 excision. Some of these excision events will be imprecise and thereby generate inactivating mutations within icdA analogous to the mutations in the biotin and galactose operons engendered by imprecise excision of phage λ. Although this scenario can explain the icdA mutations, explanation of the nalidixic acid resistance is less straightforward. However, the level of nalidixic acid resistance in icdA mutants is much >10-fold lower than those given by alterations in DNA gyrase (gyrA mutations) or by mutations that activate efflux pumps. Hence, resistance could be due to indirect physiological effects. One possible scenario is that by slowing the rate of DNA replication the slow growth of icdA mutant strains decreases the level of DNA gyrase activity required for growth. Hence, the icdA cell might survive partial inactivation of gyrase by low levels of nalidixic acid.

Another mystery is finding that the succinate requirement for aerobic growth of sucA or lpdA mutants on glucose is suppressed by mutants within the sdh operon (13, 14). It therefore seems that E. coli has another source of succinate and that utilization of this supply for biosynthesis competes with consumption by succinate dehydrogenase. This suppression phenomenon was observed with lpdA deletion mutants (13), and thus the caveat of residual 2-ketoglutarate dehydrogenase activity does not seem germane. Therefore, it seems that this succinate is generated not by the TCA cycle but by an alternative route. The recently described phosphoenolpyruvate-glyoxylate cycle (39), which bypasses 2-ketoglutarate dehydrogenase, could be the succinate source. It seems likely, however, that the succinate dehydrogenase mutants tested would have had to be leaky to avoid blocking this new cycle. Another possibility is that the low levels of fumarate reductase present in aerobic cultures (51) could partially bypass the need for succinate dehydrogenase in the phosphoenolpyruvate-glyoxylate cycle. Note that there is no need for a large supply of succinate. Although succinyl-CoA is required for the methionine and lysine biosynthetic pathways and in a pathway of arginine degradation, the carbon atoms of the succinyl moiety are not incorporated and reappear as succinate. Hence, there is no drain of carbon from the TCA cycle, although the succinate must be converted to succinyl-CoA to have a role in amino acid metabolism.

It also seems, however, that E. coli may have a means of succinyl-CoA synthesis other than succinyl-CoA synthetase since sucCD mutants grow anaerobically on fucose (but not on glucose) and suppressors that bypass sucCD mutants that restore aerobic growth on acetate have been isolated (95). It seems possible that these mutations activate the anaerobic pathway used on fucose. Another possibility is transfer of the CoA moiety of propionyl-CoA to succinate (55).

Recent experiments aimed at obtaining an E. coli strain with a genome of minimal size showed that growth of E. coli on a rich medium (Luria-Bertani broth) can tolerate loss of 2-ketoglutarate dehydrogenase or succinyl-CoA synthetase but not loss of both enzymes (156). Since this medium is rich in both lysine and methionine but lacks the diaminopimelate needed for peptidoglycan synthesis, the obvious essential metabolite requiring succinyl-CoA for its synthesis is diaminopimelate. It is puzzling that this supplement was not tested despite prior work showing that diaminopimelate restores growth of sucA mutants (24, 25).

REGULATION OF THE EXPRESSION OF THE GENES OF THE TCA CYCLE

The levels of the TCA cycle enzymes have long been known to be regulated at the transcriptional level; this regulation is exerted by interaction of two global regulatory systems, the catabolite (glucose) repression and the ArcAB two-component system (50, 52, 70, 99). Catabolite repression of the TCA cycle enzymes seems to follow the typical pattern in that transcription and enzyme levels are decreased in cells grown with excess glucose, in crp mutants lacking the Crp transcriptional activator, or in cya mutants lacking the enzyme responsible for the synthesis of the coactivator cAMP (50, 129). However, note that a few of the known TCA cycle promoters lack a readily apparent Crp DNA-binding sequence (50, 129). The ArcAB system was discovered by the release of the anaerobic repression of the sdhCDAB operon (86). It was subsequently demonstrated that this system regulates the effects of electron acceptor availability on the levels of the TCA cycle enzymes. For details of the mechanisms of transcriptional regulation we defer to EcoSal chapters that deal with control of gene expression by Crp-cAMP and ArcAB. The derepression of fumarate reductase, fumarase B, and aspartase is due to regulation by the fumarate and nitrate reduction regulatory protein FNR (see elsewhere in EcoSal). An exception to simple ArcAB/Crp regulation is the transcription of the lpdA gene that encodes the E3 (dihydrolipoamide dehydrogenase) subunit of both the 2-ketoglutarate and pyruvate dehydrogenases. The lpdA gene has recently been shown to be transcribed from either of two promoters, the pdh promoter located upstream of the pdhR reguatory gene and the lpdA promoter (120, 121). Transcription initiating at the pdh promoter proceeds through pdhR and aceEF before lpdA is transcribed. PdhR is a negative regulator of both the pdh and lpdA promoters, and thus the synthesis of E3 is increased by the presence of pyruvate. The lpdA promoter is also regulated by the ArcAB and Crp systems, which have little effect on the pdh promoter (121).

Recent reports indicate that the regulation of expression of the TCA cycle enzymes is considerably more complex than given above. Many of these data are the products or outgrowths of microarray and proteomic studies. In most such studies expression of the TCA cycle enzymes was assayed due to the abundances of the proteins and their encoding transcripts. Metabolic profile studies have also given valuable information, but many studies cannot be interpreted because of unspecified growth and/or medium conditions. Historically, the ArcA-ArcB and CRP-cAMP global regulatory systems were thought to be the main players in determining the transcription of the TCA cycle genes and hence the levels of the cycle enzymes. This remains true as mentioned above, but recent data indicate extensive interplay both between these two systems and with a third global regulatory system, the RpoS stationary-phase sigma factor. Interplay is reported not only at the level of the promoters of the TCA cycle genes but also at a secondary level, that of one global regulatory system affecting another. At present we have many interesting leads, but the overall picture remains vague. Because the global regulatory systems are covered elsewhere in EcoSal, we restrict our coverage to those aspects specific to TCA cycle gene expression although recent data indicate that the TCA cycle genes are only a minor fraction of the genes controlled by these systems.

ArcA-ArcB

The ArcA-ArcB system (where Arc represents anoxic respiratory control) is a classical two-component system consisting of ArcB, a membrane-bound sensor kinase, and ArcA, a protein that binds DNA when phosphorylated (44, 86, 113). Under anaerobic growth conditions, DNA binding by phosphorylated ArcA represses transcription of the TCA cycle genes and activates transcription of genes involved in anaerobic or fermentative metabolism (86). During aerobic growth, the kinase activity of ArcB is inhibited by oxidized quinine, which acts to oxidize two active-site cysteine residues resulting in disulfide-linked ArcB dimers (44, 91). The classical picture of the ArcA-ArcB system has only two states, aerobic and anaerobic, but recent data argue that ArcA plays a major role under microaerobic growth conditions (1, 127). E. coli strains carrying arcA deletions show the greatest deviation from wild-type strains during microaerobic growth, whereas fully aerobic or anaerobic conditions give only minor differences between wild-type and ▵arcA strains (1, 127). The switch-over point seems to be at about 50% of the oxygen content of fully aerobic cultures (i.e., about 10% oxygen; air contains 21% oxygen) with fully aerobic conditions defined as those giving no detectable acetate excretion. Below 50% aerobiosis the levels of the TCA cycle gene transcripts and the metabolic flux through the cycle are about sixfold higher in ▵arcA strains (1, 127). Hence, ArcA has a sixfold range in regulating expression of the TCA cycle genes. It seems likely that ArcA-ArcB plays a role in the branched, biosynthetic form of the TCA cycle present during growth of aerobic batch cultures on excess glucose, in particular, if at least transiently microaerobic conditions occur within these cultures. Unfortunately no extant data seem to clearly address this possibility. Note that deletion of arcA has been reported to increase the flux through the TCA cycle in aerobic cultures with excess glucose (116). This would seem to argue that not all of the ArcA is ArcA-phosphate under these conditions. However, these shake-flask cultures produced significant amounts of acetate and thus may fall into the microaerobic range as defined by other workers using chemostat cultures.

As mentioned above, contrary to the classical picture, it is now clear that the TCA cycle can function cyclically under anaerobic conditions, but only when nitrate is required as an electron acceptor, and the carbon source must be present at limiting concentrations. It has long been known that E. coli grows anaerobically on glycerol as the sole carbon source when nitrate is present, but the efficiency of carbon utilization has been unclear. Prohl and coworkers (119) report that in wild-type strains both glycerol and lactate were oxidized with about 70% efficiency, whereas an arcA mutant strain fully oxidized both substrates. In contrast only a low, but significant, fraction of the glucose carbon was oxidized in the presence of nitrate that could be detected only by sophisticated labeling techniques (116). Since flux through the TCA cycle increased in an arcA deletion strain, ArcA regulates flux through the TCA cycle under anaerobic conditions (116). Note that, of the alternate electron acceptors, only nitrate supports growth on glycerol; the less favorable electron acceptors dimethyl sulfoxide and fumarate failed to allow growth (116, 119). All of the mapped promoters of the TCA cycle genes contain an ArcA-phosphate binding site, and ArcA-phosphate binding represses transcription from the promoter. The only exception to date is the fumB promoter, which lacks an ArcA-phosphate binding site but is bound by regulatory proteins (e.g., NarL-phosphate) that are not known to bind other TCA cycle gene promoters.

An interesting complication of the metabolic flux results was that only the ▵arcA strain showed an increase in metabolic flux through the TCA cycle; the ▵arcB strain analyzed in parallel showed a flux indistinguishable from that of the parental wild-type strain (116). These results raise the possibility that another kinase or a small-molecule phosphoryl donor can activate ArcA for DNA binding. Unfortunately, most other studies have assumed that deletions of arcB and arcA would be equivalent and thus tested only the latter mutant strain. Hence, the generality of the ▵arcB results cannot be assessed at present. Further work, perhaps using mutant strains defective in both the ArcA-ArcB and Crp-cAMP pathways, is needed.

Crp-cAMP

The levels of the TCA cycle enzymes have long been known to be repressed during aerobic growth with excess glucose (153). In the cases where this has been examined strains carrying deletions of either the cya (adenylate cyclase) or crp gene have markedly lower TCA cycle enzyme levels than their parental strains, consistent with positive regulation by the Crp-cAMP complex. Moreover, the enzyme levels and transcript levels correlate well, and each of the TCA cycle gene promoters (except fumB and fumC) contains a Crp-cAMP binding site.

RpoS

A new entry on the list of global regulators that regulate the TCA cycle is RpoS, the sigma subunit of RNA polymerase responsible for expression of genes in the stationary phase of growth (110, 150). Preliminary indications for such a role of RpoS in regulation of sdhA and acnB were made in the mid-1990s (29, 155), but only recently has the generality of this effect been established. Microarray experiments indicate that all of the TCA cycle transcripts are elevated by 2- to 3.5-fold in strains lacking RpoS (110, 150). Although these values seem reliable because they agree well with the prior sdhA and acnB data, note that another microarray experiment that compared wild-type and ▵rpoS strains failed to list the TCA cycle genes and a third listed only a subset of the genes. This may be because the values obtained are close to often chosen cutoff values for significance or to the well-known variation among microarray results. There is a recent physiological indication of a role for RpoS in TCA cycle function. Chen and coworkers (19) noticed that some wild-type strains of E. coli grew very poorly on succinate as the sole carbon source, whereas derivatives of these strains that carried null mutations in rpoS grew well. Upon plating of these wild-type strains on succinate, faster growing colonies were found at the frequency expected for single mutations and each of the independently isolated fast-growing colonies examined was found to have loss-of-function mutations in rpoS. The phenotype of better growth on poor carbon sources could explain the bewildering variety of rpoS alleles seen in laboratory and clinical strains of E. coli (4).

At present the mechanism of RpoS repression of TCA cycle gene transcription appears to lie in the interactions of the RpoS, ArcA-ArcB, and Crp-cAMP regulatory systems that are discussed below, although it is still possible that competition for core RNA polymerase between RpoS and the housekeeping σ70 for the promoters of the TCA cycle genes plays a role (38).

Other Regulators

The Cra (formerly FruR) protein has been reported to regulate expression of a few TCA cycle genes. However, the TCA cycle flux was found not to differ between a ▵cra strain and its wild-type parent (116). A similar picture was seen for Fnr (116). The Fur (ferric uptake regulator) has also been implicated in regulation of the levels of three Fe-S enzymes of the TCA cycle: AcnA, FumA, and SdhCDAB (93, 94). A small noncoding RNA called RhyB that is transcribed under negative regulation by Fur binds to the acnA, acnB, fumA, and sdhDAB mRNAs and results in degradation of the mRNAs (and of RhyB) (92, 93). Under conditions of iron limitation Fur is released from the rhyB promoter, resulting in accumulation of RhyB, which then blocks translation (via mRNA degradation) of the acnA, fumA, and sdhDAB messages (92). The function of this regulatory system is presumably to divert the available iron to more critical enzymes. This work explains the long-puzzling observation that fur mutants are unable to grow with succinate as the sole carbon source (56). A small protein called Hfq is required for stability of RhyB (and several other small noncoding RNAs). Hence, introduction of a ▵hfq mutation should allow a ▵fur mutant to grow on succinate (or fumarate), and this has been demonstrated (93).

Communication among the Regulators of TCA Cycle Gene Expression

Recent work has provided intriguing clues to an interlocking network of regulators known to have roles in expression of the TCA cycle genes. These are as follows.

• ArcA regulates PtsG (71), which in turn is involved in cAMP synthesis by its involvement in adenylate cyclase function (109).

• ArcB regulates RpoS by phosphorylation of RssB, a protein that targets RpoS for proteolysis. Small increases in ArcA levels protect RpoS (by competing with RssB for ArcB) and result in increased RpoS levels. ArcA directly represses RpoS transcription (98).

• RpoS expression is under Crp-cAMP regulation (79, 98).

• Fur expression is under Crp-cAMP regulation (32).

• ArcA (113) and RssB (10) are phosphorylated by acetyl-phosphate, a precursor of acetyl-CoA, the key substrate of the TCA cycle.

Given the number of regulators, a myriad of models can be put forth. Further experiments using strains and assays that test interactions among these regulators are needed, but combinations of regulatory elements are clearly at play. For example, strains carrying deletion mutations in both cya and fur are reported to have lower levels of sdhCDAB, sucAB, sucCD, and fumA transcripts than either of the single-mutant strains (157). Since both Crp (directly) and Fur (acting through RhyB) are positive activators of expression of these genes, this is the expected result. However, in the absence of the recent RhyB data, these results would have been unexplained due to the lack of Fur binding sites in the relevant promoters.

THE GLYOXYLATE BYPASS

The glyoxylate bypass has long been known to be essential for growth on carbon sources such as acetate or fatty acids because this pathway allows the net conversion of acetyl-CoA to metabolic intermediates. Strains lacking this pathway fail to grow on these carbon sources, since acetate carbon entering the TCA cycle is quantitatively lost as CO₂ (44), and thus there is no means to replenish the dicarboxylic acids consumed in amino acid biosynthesis. The pathway involves the synthesis of three enzymes. Two of these, isocitrate lyase and malate synthase A, act to convert some of the carbon of the TCA cycle from isocitrate to malate. Isocitrate lyase competes with the TCA cycle enzyme isocitrate dehydrogenase (IDH) for isocitrate, and the third enzyme, the bifunctional IDH kinase/phosphatase, is needed to decrease the activity of IDH to allow isocitrate lyase to effectively compete for isocitrate.

An important new finding is the discovery of a role for the glyoxylate enzymes in glucose utilization. Recent work employing ¹³C labeling to monitor flux rates has shown that a role of the glyoxylate bypass in direct glucose utilization that can operate only when the glucose supply is growth limiting (39, 117). Under these "hungry" conditions, the catabolite repression that precludes use of the glyoxylate bypass during growth on excess glucose is abolished and the bypass enzymes are highly expressed. This cycle, which is called the PEP-glyoxylate cycle, includes pyruvate carboxykinase (Fig. 3). This enzyme (encoded by the pckA gene) is generally ascribed a role in gluconeogenesis and catalyzes the following reaction:

Fig. 3The PEP-glyoxylate cycle (39).

Oxaloacetate + ATP ↔ CO₂ + Phosphoenolpyruvate + ADP

The PEP-glyoxylate cycle gives complete oxidation of glucose (as does the TCA cycle) and has a stoichiometry similar to that of the TCA cycle, although 1 fewer ATP is produced per glucose molecule (39). The PEP-glyoxylate cycle is proposed to play a role in redox-cofactor balancing between NADH and NADPH, since it is active under conditions of glucose excess in strains that overproduce NADPH-generating enzymes (39). Note that neither the glyoxylate bypass nor the PEP-glyoxylate cycle apparently is required for effective pathogenesis of mice by S. enterica serovar Typhimurium (140).

THE ENZYMES OF THE GLYOXYLATE BYPASS

The two enzymes that interconvert the intermediates of the glyoxylate bypass are isocitrate lyase and malate synthase A and are encoded by aceA and aceB, respectively. The AceB protein is called malate synthase A ("A" for induction by acetate) to distinguish it from a second malate synthase, malate synthase G ("G" for induction by glycolate), encoded by glcB, which functions in growth with glycolate or glyoxylate as the sole carbon source (Chapter Two-Carbon Compounds and Fatty Acids as Carbon Sources). The two malate synthases are readily distinguished by different stabilities and inhibitor patterns (106). Expression of glcB and the other glycolate operon genes is induced by acetate and by glycolate (112). The enzymology of malate synthase A seems little explored, although a crystal structure of malate synthase G is available. The enzyme does not appear to have been purified to homogeneity, and the subunit structure has not been reported. In contrast, isocitrate lyase, a tetrameric protein, has been purified and the active site and mechanism have been studied in some detail (18, 19). The enzyme has been crystallized, and structures of the unliganded (12) and an abortive ternary complex (3) are available. Active isocitrate lyase is phosphorylated on a histidine residue; this modification seems essential for activity (126). However, the identity of the histidine residue and the role of the modification are unclear. The crystal structure (12), together with the results of site-directed mutant studies (35), suggests that His-184 is a strong candidate for the site of phosphorylation. The third protein, the IDH kinase/phosphatase, provides the most novel aspect of the glyoxylate bypass and is discussed in detail below.

REGULATION OF ISOCITRATE DEHYDROGENASE AND THE GLYOXYLATE BYPASS

In E. coli and S. enterica, IDH is regulated by phosphorylation. The function of this phosphorylation cycle is to control the flow of isocitrate through the glyoxylate bypass (Fig. 1). During growth on acetate, about 75% of the IDH is converted to the inactive phosphorylated form. Inhibition of IDH slows the TCA cycle and thus forces isocitrate through the bypass (9, 84).

The first insight into the regulation of IDH came in the early 1970s when Bennett and Holms (7, 62) reported that the activity of this enzyme was increased when E. coli made the transition between growth on acetate and growth on preferred carbon sources such as glucose or pyruvate. Increased activity did not simply result from induction of IDH expression, since it occurred even in the presence of protein synthesis inhibitors (7, 63). Lowry and coworkers (88) came to similar conclusions but by quite a different route. They found that addition of glucose to a culture growing on acetate produced a metabolic crossover at IDH: the cellular level of isocitrate decreased while that of 2-ketoglutarate increased. This metabolic crossover provided indirect evidence that IDH had been activated during this transition (88).However, the mechanism responsible for regulation of IDH was not reported until 1979, when it was demonstrated that IDH activity was controlled by phosphorylation (42, 43).

Regulation of E. coli IDH by phosphorylation was a surprising result. Although protein phosphorylation had been known for many years in eukaryotes, conventional wisdom held that bacteria did not use this regulatory strategy. The lone exception appeared to be bacteriophage T7, a bacterial virus that encodes a protein kinase activity capable of phosphorylating RNA polymerase during the infection cycle (122). The discovery that IDH is regulated by phosphorylation and the simultaneous identification of a variety of other protein kinase activities in extracts of E. coli and S. enterica demonstrated that protein phosphorylation was, in fact, a universal regulatory strategy (74).

Structural Basis of the Inhibition of IDH

Phosphorylation of IDH results in total inactivation of the enzyme (9, 82). The extent of this inhibition is a striking contrast to most phosphorylated enzymes, which undergo more subtle changes in their properties. IDH is inactivated because the phosphorylated serine lies within the active site. In the active dephosphorylated enzyme, this serine forms a hydrogen bond with isocitrate, one of the substrates. Phosphorylation of IDH blocks isocitrate binding by disrupting this hydrogen bond and by introducing electrostatic repulsion between this phosphate and isocitrate (33, 34, 64, 65, 66, 67).

Enzymology of the IDH Phosphorylation Cycle

The enzymes that phosphorylate and dephosphorylate IDH exhibit several features that first seemed peculiar but are now commonplace. IDH kinase and IDH phosphatase are encoded by the same gene, aceK, and reside on the same polypeptide (75, 80, 81). This phosphatase activity has an absolute requirement for ATP or ADP but is not supported by nonhydrolyzable ATP analogues (82) (unpublished observations of D.C. LaPorte [1981]). IDH kinase/phosphatase also catalyzes a third activity: an ATPase that is more active then either IDH kinase or IDH phosphatase (133). IDH kinase/phosphatase appears to catalyze all three activities at the same active site. This model was first suggested by the observation that a variety of mutations have parallel effects on these activities. For example, one class of mutant proteins possesses kinase, phosphatase, and ATPase activities, but all three activities exhibit drastic reductions in affinity of phospho-IDH. Furthermore, all three activities of one of these proteins have reduced affinity for ATP (unpublished observations of T. P. Ikeda and D.C. LaPorte [1997]). Finally, a mutation in a conserved ATP binding site motif eliminates both IDH kinase and IDH phosphatase activities (134). Affinity labeling of IDH kinase/phosphatase with an ATP analogue also supports a single active site on IDH kinase/phosphatase (142). This analogue yielded parallel inhibition of IDH kinase and IDH phosphatase, yet, in a preliminary experiment, appeared to label a single peptide.

A likely model for the mechanism of this protein proposes that the IDH kinase and IDH phosphatase reactions occur in the same active site and that the phosphatase reaction results from the back reaction of IDH kinase tightly coupled to ATP hydrolysis. According to this model, the phosphatase reaction requires the formation of a ternary complex between IDH kinase/phosphatase, ADP and phospho-IDH. The phosphate group of phospho-IDH is then transferred to ADP (the kinase back reaction) and then to water (an ATPase reaction). This sequence of steps results in the net dephosphorylation of phospho-IDH (the phosphatase reaction) (132, 135).

Regulation of the Glyoxylate Bypass by IDH Phosphorylation

One role of the IDH phosphorylation cycle is to regulate the branch point between the glyoxylate bypass and the TCA cycle during steady-state growth on acetate or fatty acids (see above). Phosphorylation of IDH diverts some of the flux from the TCA cycle to the glyoxylate bypass. The immediate effect of phosphorylation is to inhibit IDH activity. The resulting increase in the level of isocitrate increases the velocity of isocitrate lyase, the first enzyme of the bypass. Mathematical analyses have demonstrated that saturation of IDH with isocitrate (which occurs during growth on acetate) makes it impossible to directly regulate isocitrate lyase activity. Regulation can only be achieved indirectly through control of IDH activity (84). Mutant strains that are deficient in IDH kinase failed to grow on acetate, suggesting that the phosphorylation of IDH is required for use of the glyoxylate bypass (83).

IDH kinase/phosphatase also controls the glyoxylate bypass during transitions between carbon sources. For example, addition of a preferred carbon source (e.g., glucose or pyruvate) to a culture growing on acetate renders the glyoxylate bypass unnecessary. Under these conditions, the cell shuts this pathway down by dephosphorylating IDH. Inhibition of the bypass results because the activation of IDH draws isocitrate through the TCA cycle. The resulting decrease in isocitrate concentration yields a proportional decrease in the velocity of isocitrate lyase (84).

What Controls the IDH Phosphorylation Cycle?

A variety of metabolites that affect IDH kinase/phosphatase in vitro have been identified. These metabolites activate IDH phosphatase and inhibit IDH kinase (81, 104). During growth on acetate, it appears that isocitrate and 3-phosphoglycerate participate in the control of the IDH phosphorylation cycle. These effectors probably act as general indicators of the levels of metabolic intermediates and thus of the need for isocitrate to be directed to the glyoxylate bypass. For example, depletion of these metabolites would result in increased phosphorylation of IDH, forcing more isocitrate through the bypass. However, isocitrate and 3-phosphoglycerate are not responsible for the dephosphorylation of IDH that results from the addition of preferred carbon sources such as glucose, since their levels fall under these conditions. Dephosphorylation of IDH during these metabolic transitions is probably promoted, at least in part, by pyruvate, since the level of this metabolite rises dramatically upon addition of glucose (21, 56). For these reasons it would be very interesting to determine the IDH phosphorylation state during growth under the hungry-for-glucose conditions that induce the PEP-glyoxylate cycle (Fig. 3).

Although the glyoxylate bypass can provide metabolic intermediates, the TCA cycle is more efficient in providing energy. The cell must, therefore, precisely balance the flux of isocitrate between these competing pathways during growth on acetate. The energy requirements of the cell appear to be monitored through AMP levels. Like the other metabolites discussed above, AMP activates IDH phosphatase and inhibits IDH kinase. AMP is a particularly attractive choice for monitoring the energy needs of the cell because, if the adenylate kinase reaction is at equilibrium, the level of AMP will vary as the square of ADP concentration. An increase in AMP, signaling a depletion of cellular energy, would yield a net dephosphorylation of IDH, diverting more isocitrate through the TCA cycle, whereas a surplus of energy, indicated by a low level of AMP, would have the opposite effect. An open question is the regulation of IDH activity during anaerobic growth when the TCA cycle functions in the branched mode.

Sensitivity Amplification and the Glyoxylate Bypass

The regulation of the glyoxylate bypass appears to be exquisitely sensitive to the metabolic state of the cell. This conclusion is supported by a variety of theoretical analyses and by experiments performed in vitro. A high degree of sensitivity amplification is achieved by simultaneous activation of IDH phosphatase and inhibition of IDH kinase by its effectors. This type of sensitivity amplification is termed a multistep effect. The IDH phosphorylation cycle also appears to be subject to zero-order ultrasensitivity, a form of sensitivity amplification that occurs in covalent regulatory systems when the concentration of the interconvertible protein (in this case IDH) exceeds the Michaelis constants of the converter enzymes (45, 81). A third mechanism of sensitivity enhancement, termed a branch point effect, results from the profound difference in the affinities of IDH and isocitrate lyase for isocitrate (Michaelis constants of 8 and 600 μM, respectively) (84). As a result of the branch point effect, the flux of the glyoxylate bypass is strikingly sensitive to the phosphorylation state of IDH. These three mechanisms for sensitivity amplification combine to produce a system in which very subtle changes in metabolic signals have the potential for producing profound changes in the flux through the glyoxylate bypass.

In addition to responding to changes in the external environment, the IDH phosphorylation cycle must be capable of adapting to the conditions that prevail inside the cell. For example, the cellular level of IDH can vary by at least a factor of 2 between different strains of E. coli. The effect of this difference in IDH levels would be amplified by the branch point effect (see above), potentially preventing growth on acetate. However, the IDH phosphorylation cycle responds to these differences by altering the fractional phosphorylation of IDH so that a constant level of IDH activity is maintained under these growth conditions. This cycle could even compensate for a 15-fold overproduction of IDH, a condition that clearly had the potential for pathological consequences (83). It seems likely that this response to the cellular level of IDH resulted because the control of IDH kinase/phosphatase by a variety of metabolites provides a particularly effective feedback control mechanism.

Is precise control of IDH phosphorylation necessary? The tolerance of E. coli to variation in IDH activity was tested using mutant strains having defects in IDH kinase/phosphatase. The effect of excess IDH activity was determined by comparison of strains with null mutations in aceK. A twofold increase in IDH activity reduced the growth rate on acetate, while a fourfold increase yielded nearly complete inhibition of growth (83). The minimum level of IDH required for growth on acetate was determined by coexpression of wild-type aceK with an allele of aceK, whose product selectively lacked IDH phosphatase activity. The cells tolerated a 50% reduction in IDH activity without apparent effect on growth rate. However, further reductions in IDH activity inhibited growth, with arrest occurring when this level dropped to 15% of the wild type (68). Thus, the striking precision exhibited by the IDH phosphorylation cycle in vivo is not absolutely essential. However, in nature, even small differences in growth rate can yield dramatic differences during competition between organisms for limiting resources.

Balancing Environmental Sensitivity and Futile Cycling

A regulatory system that employs covalent control must balance the benefits of a rapid response to a signal with the need to minimize the loss of cellular energy that results from the cyclic modification of the target protein (i.e., futile cycling). Insight into the minimum level of IDH kinase/phosphatase required for growth on acetate came from an unexpected source. Although mutation of the consensus ATP binding site reduced the IDH kinase and IDH phosphatase activities by factors of at least 100 both in vivo and in vitro, the altered protein retained sufficient kinase activity to support growth on acetate. It thus appears that wild-type IDH kinase/phosphatase is maintained in massive excess over the level required for steady-state phosphorylation of IDH (132). A likely function for this excess IDH kinase/phosphatase would be to allow rapid responses to changes in the available carbon source. For example, when pyruvate is added to cultures growing on acetate, the cells dephosphorylate IDH, increasing its activity, thus inhibiting the flux of isocitrate through the glyoxylate bypass (see above). The ability to control the expression of wild-type IDH kinase/phosphatase using a clone of a wild-type aceK gene provided a method for determining the effect of the level of this protein on the response rate. The rate of the pyruvate-induced dephosphorylation of IDH was proportional to the level of IDH kinase/phosphatase, demonstrating that IDH kinase/phosphatase represented the primary rate-limiting step in this metabolic transition (133).

GENETIC CONTROL OF THE GLYOXYLATE BYPASS

The genes that encode the metabolic and regulatory enzymes of the glyoxylate bypass reside in the same operon, aceBAK (21) (Fig. 4). The metabolic enzymes, malate synthase and isocitrate lyase, are encoded by aceB and aceA, whereas aceK encodes IDH kinase/phosphatase. The operon is expressed from a single promoter during growth on acetate (21). Expression is induced during growth on acetate or fatty acids but induction can be prevented by the presence of an excess preferred carbon source (e.g., glucose, glycerol, or pyruvate) (78). In the presence of severely growth-limiting concentrations of glucose, however, aceBAK expression is also induced (39, 117).

Fig. 4The glyoxylate bypass operon. Malate synthase, isocitrate lyase, and IDH kinase/phosphatase are encoded by aceB, aceA, and aceK. These genes are transcribed from a single promoter upstream of aceB.

Expression of the glyoxylate bypass operon also responds to the lack of aerobic respiration, with repression by the ArcA-ArcB system occurring under anaerobic conditions (69, 70).

The expression of aceBAK is affected by the integration host factor (IHF) protein, a histone-like protein that is abundant in E. coli. IHF binds to two sites that lie upstream of the promoter. Binding of IHF to the promoter-proximal site yields a fivefold increase in transcriptional activity (125). The aceBAK operon is regulated, at least in part, by a repressor protein encoded by iclR (77, 89, 90, 101, 135). The IclR protein binds to a site that overlaps the minus 35 region of the aceBAK promoter (21, 101). Release of this repression upon adaptation to growth on acetate, fatty acids, or limiting glucose is presumably responsible for induction of operon expression. By analogy with other systems, this release presumably involves binding of some metabolic intermediate. Acetate can probably be ruled out as the inducer, since fatty acids also induce expression of aceBAK without producing free acetate (101). However, the metabolic intermediate cannot yet be confidently identified, although high concentrations of isocitrate and free CoA have been suggested (37). In E. coli the iclR gene lies closely downstream of aceK but is transcribed from the other DNA strand. Between iclR and aceK is a gene called arpA that is transcribed in the same direction as iclR. This protein has been annotated as a regulator of acetyl-CoA synthetase, but we can find no data supporting the annotation in the literature. This gene is restricted to E. coli. The aceK-iclR intragenic region of S. enterica lacks arpA and is consequently much smaller than that of E. coli.

The response of aceBAK to fatty acids is mediated, in part, by FadR (89, 90). FadR was initially identified because it represses the genes encoding the enzymes of fatty acid degradation (27). FadR was subsequently shown to activate the transcription of fabA and fabB, genes whose products participate in unsaturated fatty acid biosynthesis (16, 59, 106). DNA binding by FadR is specifically antagonized by CoA esters of long-chain fatty acids (26, 59). Mutations in fadR result in increased expression of aceBAK on repressing media, although the effects of these mutations are much smaller than those observed for iclR (90). The mechanism is now known to be activation of the iclR promoter by FadR (53). Hence, a lack of FadR binding to the iclR promoter results in decreased levels of IclR and increased aceBAK expression.

Expression of the glyoxylate bypass operon, like TCA cycle gene expression, responds to aerobic respiration with repression occurring under anaerobic conditions (17, 86), and this regulation is mediated by the ArcAB global regulatory system. The expression of the glyoxylate bypass operon is also controlled by the catabolite repressor/activator protein encoded by cra, a global regulatory protein first identified by its effects on fructose metabolism and called FruR (20). In vitro, the Cra protein binds to a site upstream of the aceBAK promoter (123). However, the physiological significance of this site remains uncertain since deletion of the Cra binding region has only a slight effect on aceBAK expression (125). Expression of the glycolate (glcDEFGB) operon responsible for glycolate utilization including its malate synthase (GlcB) is also under control of the ArcA-ArcB and Crp-cAMP systems (112). Hence, together with induction of the glycolate operon during growth on acetate, these systems provide an appreciable level of coordinate expression of the two E. coli malate synthases (112).

Polarity in aceBAK

Although the genes of the aceBAK operon are expressed from the same promoter (Fig. 4), the relative cellular levels of malate synthase, isocitrate lyase, and IDH kinase/phosphatase are approximately 0.3:1:0.003 (22). The upshift in expression between aceB and aceA results from differences in translational efficiency. In contrast, inefficient translation and premature transcriptional termination contributed to the downshift in expression between aceA and aceK. Premature transcriptional termination occurs within aceK and appears to result from inefficient translation (22). The sequences responsible for inefficient expression of aceK lie within the ribosome binding site of this gene, a surprising result since this site contains a good match to the Shine-Dalgarno ribosome binding sequence (22).

FUTURE DIRECTIONS

Although significant gaps remain in our knowledge of the carbon flow of the TCA cycle and glyoxylate bypass, future work in this area seems likely to be targeted to the detailed enzymology at the three-dimensional structural level. Quantitative data on the fluxes through these pathways obtained under rigorous experimental conditions should be used to develop computational models. Such models may give further insights into the mechanisms that regulate expression of the genes of these pathways and the newly described PEP-glyoxylate cycle.