Biosynthesis of Pantothenic Acid and Coenzyme A

ROBERTA LEONARDI AND SUZANNE JACKOWSKI*

[SECTION EDITOR: JOHN CRONAN ]

August 13, 2007

Department of Infectious Diseases, St. Jude Children's Research Hospital, 332 North Lauderdale, Memphis, TN 38105-2794

*Corresponding author. Mailing address: Protein Science Division, Department of Infectious Diseases, St. Jude Children's Research Hospital, 332 North Lauderdale, Memphis, TN 38105-2794. Phone: (901) 495-3494, fax: (901) 495-3099, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

Pantothenate is vitamin B₅ and is the key precursor for the biosynthesis of coenzyme A (CoA) and proteins that have a phosphopantetheine prosthetic group. The phosphopantetheine moiety is donated to these proteins by CoA and is used to shuttle intermediates between the active sites of enzymes involved in fatty acid, nonribosomal peptide, and polyketide synthesis. CoA is an essential cofactor for cell growth and is involved in many metabolic reactions, including the synthesis of phospholipids, synthesis and degradation of fatty acids, and the operation of the tricarboxylic acid cycle. The prebiotic formation and stability of pantothenate precursors suggests that CoA function was important in the earliest metabolic pathways (90). According to genetic footprinting experiments in Escherichia coli, all five biosynthetic genes involved in CoA biosynthesis from pantothenate are essential (47), but the pantothenate biosynthetic genes can be dispensable in some organisms due to the expression of a pantothenate permease, which transports the vitamin into the cell. The supply of pantothenate and the level of expression of pantothenate kinase (PanK) together determine the cellular CoA levels (21, 121, 141). The CoA biosynthetic pathway is a popular target for development of novel antibacterial agents because of the distinctive differences between the bacterial and the mammalian proteins that catalyze the PanK reaction and the phosphopantetheine adenylyltransferase (PPAT) reaction. The CoA biosynthetic pathway networks with other vitamin-associated pathways, for example, by altering the metabolic requirements for nutrients such as thiamine (35) or by restricting pantothenate production by the availability of methylene tetrahydrofolate, which is derived from folic acid (12).

BIOSYNTHESIS OF PANTOTHENATE AND ITS PRECURSORS

Most bacteria, such as E. coli, Salmonella enterica serovar Typhimurium LT2, and Corynebacterium glutamicum (8, 57, 135), synthesize pantothenate from the amino acid aspartate and an intermediate in valine biosynthesis. In Salmonella, the acetohydroxy acid synthase isozyme I, followed by the dihydroxyacid dehydratase, provides most of the flux to valine and pantothenate (36), both of which are made from α-ketoisovalerate. Valine can be converted back to α-ketoisovalerate by the products of either the ilvE or avtA genes. The amino group of valine is deaminated to a keto moiety to yield α-ketoisovalerate, which, in turn, is converted to α-ketopantoate by transfer of a hydroxymethyl group. α-Ketopantoate is reduced to pantoate (Fig. 1). Aspartate is decarboxylated to yield β-alanine. β-Alanine and pantoate are then condensed to yield pantothenate (also called pantothenic acid). Pantothenate is either used for CoA biosynthesis or released from the cell. Comparative analysis of the sequences and structures of the individual enzymes involved in pantothenate biosynthesis indicate that the pathway likely evolved via a patchwork mechanism, whereby the activities were recruited separately from diverse protein families (78). The level of pantothenate is low in all the archaebacteria, including methanogens (75), compared with that in eubacteria (98). Metabolic engineering of C. glutamicum to increase pantothenate production reveals several constraints on the biosynthetic pathway (52, 110), most notably at the α-ketoisovalerate branch point (12).

Fig. 1Pathway for the biosynthesis of pantothenic acid (called pantothenate). Pantoate formation begins with the transfer of a methyl group to α-ketoisovalerate by ketopantoate hydroxymethyltransferase, the panB gene product, followed by reduction by ketopantoate reductase, the panE gene product. The α-ketoisovalerate is an intermediate in valine biosynthesis, and the relative demand for the amino acid versus pantothenate determines the interconversion between the two metabolites. The branched-chain-amino-acid transferase, the product of the ilvE gene, converts valine back to α-ketoisovalerate. β-Alanine is formed from aspartate by aspartate-1-decarboxylase, the product of the panD gene. Pantothenate is formed by the ATP-dependent condensation of β-alanine and pantoate by pantothenate synthetase, the panC gene product. Pantothenate is used for CoA biosynthesis or exits from the bacterium.

Ketopantoate Biosynthesis

The first step in the biosynthesis of d-pantoic acid is the formation of α-ketopantoate by the α-ketopantoate hydroxymethyltransferase (EC 2.1.2.11) (Fig. 1). The enzyme uses N₅,N₁₀-methylene tetrahydrofolate and catalyzes the conversion of α-ketoisovalerate to ketopantoate by transferring the carbon of the methylene moiety from the cofactor to the substrate (102). The idea that α-ketoisovalerate, an intermediate in the biosynthesis of valine, is also required for pantoate synthesis was first suggested by Maas and Vogel (84) and was later demonstrated in mutant strains with defects in ilvD, ilvE, or avtA (156). The hydroxymethyltransferase is the product of the panB gene of E. coli that encodes a 28-kDa polypeptide estimated to be a large multimer by several techniques (63, 102, 103, 132). The crystal structures of the native proteins of E. coli (152) and Mycobacterium tuberculosis (13) show that the enzymes form a decameric complex, with subunits in opposing pentameric rings held together by a swapping of their C-terminal α-helices. A magnesium ion is coordinated in the active site and is required for catalysis (102). The affinity for ketoisovalerate is about 1 mM for the E. coli enzyme and 240 μM for the M. tuberculosis enzyme (155). The ketopantoate hydroxymethyltransferase exhibits end-product inhibition by ketopantoate, thus helping to regulate pantothenate production (102, 137). The enzyme is also inhibited by pantothenate (>500 μM) and CoA (>1 mM) but only by concentrations that are much higher than those found in vivo (145), which argues against feedback regulation by these downstream components of the pathway.

The panB gene is in an operon together with the panC gene (63, 109), and in serovar Typhimurium a mutation of the panB promoter leads to increased expression of both the hydroxymethyltransferase and the synthetase that in turn increases pantothenate production (109). Overexpxression of the panB gene alone, but not the panC gene (109), or supplementation with exogenous pantoate (21) increases CoA by 40 to 50%, suggesting that the supply of endogenous pantoate is limiting for pantothenate biosynthesis, and pantothenate utilization, in turn, is regulated at the PanK step.

Pantoate Biosynthesis

d-Pantoate is synthesized from α-ketopantoate by α-ketopantoate reductase (EC 1.1.1.169), the product of the panE gene (Fig. 1), which transfers a hydrogen from NADPH to pantoate. The enzyme first purified from Stenotrophomonas (formerly Pseudomonas) maltophilia has a subunit molecular mass of 30.5 kDa and exists as a multimer in its native state (117). The monomer sizes of the reductase enzymes from E. coli and serovar Typhimurium are both about 33 kDa (43, 87, 170). The pH dependence of the reaction for the E. coli enzyme suggests the involvement of a general acid/base in the catalytic mechanism, where the conserved Lys176 and Glu256 residues are essential (169). The Lys moiety is located in a hinge region, or cleft, bending between the N- and C-terminal domains (17) and undergoes a conformational switch from a resting state to an active state that binds ketopantoate (17, 77). The enzyme discriminates between ketopantoate and pantoate only with NADPH bound, which promotes cooperative interaction with the substrate.

The acetohydroxy acid isomeroreductase (EC 1.1.1.86), which is encoded by the ilvC gene and is involved in the biosynthesis of isoleucine and valine, is also capable of catalyzing the reduction of α-ketopantoate both in vitro and in vivo. In C. glutamicum, the ilvC gene product is the only enzyme capable of producing pantoate (88). Primerano and Burns isolated α-ketopantoate reductase mutants and clarified the relationship between pantoate and branched-chain-amino-acid biosynthesis in serovar Typhimurium (104, 105). panE mutants do not require pantoate when the ilvC gene is abundantly expressed, but when ilvC expression is low, panE mutants require pantoate for growth. The panE gene maps at 10 min and is allelic to apbA in serovar Typhimurium (43, 44). A role for the apbA gene had been proposed in an alternative pathway for thiamine biosynthesis, but in vivo labeling showed that pantoic acid, the product of the ApbA-catalyzed reaction, is not a direct precursor to thiamine. The conditional requirement of panE mutants for either thiamine or pantothenate is manifest only when flux through the purine biosynthetic pathway is reduced due to the lack of thiamine pyrophosphate, not when CoA levels are decreased (42). Enhanced expression of the panE gene by threefold results in about a threefold increase in pantothenate excretion from E. coli and is further increased when the medium is supplemented with ketopantoate (33).

Alanine Biosynthesis

Aspartate was first suggested as the precursor to β-alanine on the basis of the conversion of aspartate to β-alanine by intact cells (23, 151). Williamson and Brown (160) and Cronan (20) independently characterized an l-aspartate-1-decarboxylase activity from E. coli that converts aspartate to CO₂ and β-alanine (EC 4.1.1.11) (Fig. 1). The decarboxylase has a molecular size of 58 kDa and has pyruvate moieties covalently bound to multiple identical subunits that associate as a tetramer. The pyruvate is involved in catalysis (159) and forms a Schiff base between enzyme and substrate (74) at the active sites that are located between adjacent subunits (3). The affinities of the enzymes from E. coli and M. tuberculosis for aspartate are similar, with K_m values estimated at 151 μM (106) and 219 μM (15), respectively. The decarboxylase is translated as an inactive pro-protein (π-protein) of 13.8 kDa, which subsequently undergoes an intramolecular rearrangement and is self-cleaved at the Gly24-Ser25 bond (106) to a mature form that has a β-chain (2.8 kDa) with a hydroxyl group at its C terminus and an α-chain (11 kDa) that becomes activated by formation of the pyruvoyl group at its N terminus. The α- and β-chains associate together and constitute a subunit of the tetramer (3, 74). A crystal structure of the tetramer shows three cleaved subunits containing pyruvoyl moieties and one subunit with the ester intermediate. The molecular mechanism of self-processing relies on the conformational freedom of a loop preceding the cleavage site along with stabilization of an oxyoxazolidine intermediate by a Thr residue in the loop to form the ester intermediate (113).

Aspartate decarboxylase is encoded by the panD gene located at 3.1 min on the E. coli chromosome; mutants require β-alanine or pantothenate for growth (20, 160). The panD locus in serovar Typhimurium is located at 4.5 min, and in both organisms the panD gene is adjacent to panB and panC. Insertional mutation of the panD gene in serovar Typhimurium confers sensitivity to suphometuron methyl, suggesting that CoA availability modulates the response to this herbicide (73). Inhibition of pantothenate production in intact cells by d-serine, β-hydroxyaspartic acid, or l-cysteic acid is consistent with inhibition of the decarboxylase enzyme (19, 30, 31 ,49, 83, 118). A screen for inhibitors of the aspartate decarboxylase revealed that the active site is highly stringent for the size of compounds and only the structural analogs of aspartate, glutamate, and 2-aminomethylsuccinate will bind (155). A second pathway for the synthesis of β-alanine from dihydrouracil was proposed (119, 120) but does not occur in E. coli, since Cronan showed that β-alanine auxotrophs (panD mutants) will not grow on dihydrouracil (20). The aspartate decarboxylase from C. glutamicum has more activity than the homolog from E. coli, and overexpression of either enzyme increases de novo pantothenate production in E. coli (32). The C. glutamicum enzyme is expressed at a higher level due to more efficient translation of the protein. The amino acid carrier CycA mediates the transport of β-alanine into E. coli with a low affinity (K_m 2.4 mM) ( 114), and supplementation of wild-type E. coli with high concentrations of β-alanine also increases pantothenate production (21, 32).

Pantothenate Biosynthesis

Pantothenate synthetase (EC 6.3.2.1) catalyzes the ATP-dependent condensation of d-pantoate with β-alanine to form pantothenate (Fig. 1). Pantothenate synthetase is encoded by the panC gene located at min 3.1 on the chromosome in E. coli and at min 4.5 in serovar Typhimurium (21, 82). Pantothenate synthetase from E. coli (92) and from M. tuberculosis (168) has been purified and characterized. The reaction proceeds via formation of a pantoyl adenylate intermediate (158) following the binding of ATP first, then d-pantoate and the release of pyrophosphate. The binding of β-alanine, the second substrate, can occur only after formation of the pantoyl adenylate (153) and triggers the release of the products, pantothenate and AMP. The three-dimensional structure of the M. tuberculosis enzyme in complex with the pantoyl adenylate was determined (154) as part of the TB Structural Genomics consortium (48), and the active site residues required for the formation and stabilization of the intermediate were identified by mutagenesis (171).

The detailed reaction mechanism provided by this information led to the idea that nonreactive analogs of pantoyl adenylate would be highly specific inhibitors of the pantothenate synthetase. Inhibitors of pantothenate synthetase that have potential as antimicrobial agents have, in fact, been identified (140, 157). Ten analogs of the reaction intermediate pantoyl adenylate, in which the phosphodiester is replaced by either an ester or sulfamoyl group, are all modest inhibitors, with the sulfamoyls being more potent (140). The clockwise gene order as determined by direct DNA sequencing is panB-panC-panD (21), where panB and panC lie adjacent to one another but are separated from panD by an open reading frame of unknown function, yadD, which is oriented in the opposite direction (89). A yadD knockout in the Keio mutant collection (5) grows well on minimal medium, whereas the pan mutants grow poorly, suggesting that the genes are not related. Pantothenate synthetase activity is not tightly regulated in vivo since E. coli secretes into the medium most of the pantothenate that is synthesized (24, 59, 83), thus providing the vitamin to the mammalian host. Pantothenate synthesis does not draw on the valine pathway substantially, and thus stringent regulation of this step may be metabolically inefficient (21).

Pantothenate Transport

Pantothenate is taken up by virtually all bacteria and is essential for growth of those organisms lacking de novo pantothenate biosynthesis, such as Streptococcus pneumoniae, Lactobacillus lactis, and Haemophilus influenzae (47). Bacillus subtilis is an exception, however, as efficient pantothenate uptake cannot be demonstrated (6). A pantothenate transport activity was first described in E. coli (95) and later identified as mediated by the pantothenate permease, also termed the PanF protein, encoded by the panF gene located at min 72 of the chromosome (Fig. 2). The panF gene is cotranscribed with the prmA gene, which encodes a protein that methylates ribosomal proteins (149). The PanF protein is predicted to contain 12 transmembrane hydrophobic domains connected by short hydrophilic chains, which is a topological motif characteristic of other cation-dependent permeases of the major facilitator superfamily of proteins (58, 64, 107). PanF uses a sodium-cotransport mechanism to concentrate pantothenate from the medium (58, 148), which is highly specific for pantothenate, with a K_t of 0.4 μM and a maximum velocity of 1.6 pmol/min/10⁸ cells (95, 148). A similar transport system is present in serovar Typhimurium (29). Overexpression of the PanF permease in E. coli produces a 10-fold increase in the rate of pantothenate uptake and concomitant elevation of the steady-state intracellular concentration of pantothenate (58). The CoA levels remain unaffected by overexpression, however, indicating that PanF activity does not regulate CoA biosynthesis. Conversely, CoA levels do not change the rate of pantothenate transport (58). Inactivation of the panF gene blocks uptake but does not reduce the exit of pantothenate synthesized in the bacteria (147), indicating the existence of a distinct efflux system. The active transport of pantothenate into bacteria would suggest that the biosynthetic pathway leading to pantothenate is of limited interest for the development of novel antibacterial compounds. However, the development of an attenuated live vaccine against M. tuberculosis has been enabled by engineering the bacterium to rely solely on pantothenate transport for CoA biosynthesis (111). The pantothenate auxotroph of M. tuberculosis is severely limited in its growth (112), and its reduced replication in the host animal is sufficient to elicit an effective immune response but does not give rise to pathogenic infection. Thus, inhibition of pantothenate production in this bacterium would be an effective antibacterial strategy.

Fig. 2Pathway for CoA biosynthesis and 4'-phosphopantetheine prosthetic group transfer and recycling. Pantothenate is either synthesized de novo (Fig. 1) or actively taken up into cells by a unidirectional sodium-dependent permease, the panF gene product. The exit of pantothenate is independent of permease activity. Pantothenate kinase isoforms type I, II, and III are the products of the coaA gene(s) and phosphorylate pantothenate. Cysteine and CTP are then added to 4'-phosphopantothenate in a two-step reaction to yield 4'-phosphopantetheine by the bifunctional product of the coaBC gene. The phosphopantetheine adenylyltransferase, the coaD gene product, adds the adenine moiety to 4'-phosphopantetheine to form dephospho-CoA, which in turn is phosphorylated by dephospho-CoA kinase, the product of the coaE gene, to form CoA. The CoA either becomes acylated by a fatty acid or organic acid or donates the 4'-phosphopantetheine moiety to a protein, such as acyl carrier protein (ACP), to activate it. The transfer of the 4'-phosphopantetheine prosthetic group is catalyzed by the product of the acpS gene. The prosthetic group is cleaved from the protein by an ACP-specific phosphodiesterase encoded by the acpH gene. The 4'-phosphopantetheine metabolic intermediate either is utilized for CoA synthesis or exits from the cell. Intact 4'-phosphopantetheine is not taken up by the bacterium.

CoA BIOSYNTHESIS

Pantothenate Phosphorylation

Pantothenate kinase (PanK, also called CoaA) (EC 2.7.1.33) catalyzes the ATP-dependent phosphorylation of pantothenate to 4'-phosphopantothenate, the first step in the biosynthesis of CoA (Fig. 2). Originally it was thought that the first phosphorylated intermediate in the pathway was 4'-phosphopantetheine, but this point was reinvestigated in 1958 by Brown (11), who convincingly showed that the formation of 4'-phosphopantothenate was required for the subsequent reactions in the CoA biosynthetic pathway. PanK is encoded by the coaA gene, and temperature-sensitive coaA mutants were isolated by Dunn and Snell in serovar Typhimurium (29) and subsequently by Vallari and Rock in E. coli (146). Historically, the first temperature-sensitive coaA mutants were isolated in 1966. However, the biochemical defect caused by the mutation was unknown and the mutated gene was called rts (39). The rts and coaA mutations are alleles of the same gene (123), as indicated by their nucleotide sequences (40, 123). The coaA mutants cannot be supplemented since E. coli does not incorporate extracellular phosphorylated intermediates such as 4'-phosphopantothenate, 4'-phosphopantetheine, or dephospho-CoA. The coaA gene is located at min 90 of the E. coli chromosome between the birA and thrU genes (121); it possesses its own promoter and produces a 1.1-kb transcript. The utilization of either of two translation initiation sites produces two PanK proteins that differ in eight amino acids at the amino terminus (121). The poor homology of the coaA promoter region to the consensus E. coli promoter sequences and the low frequency of optimal codon usage are consistent with a low level of PanK expression (123). A potential regulator of PanK has been identified as the YbjN protein, which is capable of suppressing the temperature-sensitive phenotype of several coaA mutants when it is overexpressed (14).

Jackowski and Rock (59) proposed PanK as a major rate-controlling step in CoA biosynthesis on the basis of the copious excretion of pantothenate from E. coli (24, 59, 83). Nonesterified CoA was five times more potent than CoA thioesters in inhibiting the kinase in extracts from E. coli (145). CoA is a competitive inhibitor with respect to ATP, thus providing a mechanism to coordinate CoA production with the energy state of the cell. The enzyme was purified to homogeneity from overexpressing strains and shown to be a homodimer of 36-kDa subunits (122). The purified enzyme exhibits cooperative binding of ATP and is competitively inhibited by CoA. Cooperative ATP binding made determination of the kinetic mechanism complicated, and to overcome the problematic interpretation of the kinetic data, intragenic complementation was used to produce a chimeric heterodimer between a wild-type and an inactive subunit. The inactive subunit was expressed by a coaA plasmid construct in which the codon of the critical Lys101 in the predicted ATP binding site was mutated to a Met residue. Kinetic analysis of these chimeric molecules indicated the absence of cooperative ATP binding and revealed that the kinase reaction proceeds by a sequential mechanism, with ATP as the first substrate interacting with the protein. The K_m constants for ATP and pantothenate are 136 and 36 μM, respectively.

The crystal structures of E. coli PanK bound to a nonhydrolyzable ATP analog and to free CoA show that the phosphate groups of the two ligands occupy the same space at Lys101 (165). As a consequence, the binding of CoA prevents the binding of ATP, explaining the competitive nature of the CoA inhibition. The structure also reveals that the free thiol group of CoA fits tightly in the inhibitor binding site, and thus there is little space to accommodate acyl chains. This seems likely to provide the structural basis for the more potent inhibition of PanK by nonesterified CoA than by CoA thioesters. In contrast to the CoA binding site, the pantothenate binding site is remarkably flexible, allowing for the binding and phosphorylation of a bulky pantothenate analog such as N-pentylpantothenamide (53). This molecule and similar N-alkylpantothenamides were first synthesized and found to inhibit E. coli growth in 1970 (18). However, the precise mechanism of action against the bacterium was elucidated more than 30 years later, when it was shown that N-pentylpantothenamide is a substrate for E. coli PanK and the downstream enzymes in the CoA biosynthetic pathway (128). N-Pentylpantothenamide and, in general, the N-alkylpantothenamides block fatty acid biosynthesis in E. coli by being converted to CoA analogs that are incorporated into and inactivate acyl carrier protein (150, 166).

The availability of genomic sequence information for many bacterial pathogens has uncovered new and different proteins that catalyze the first step in CoA biosynthesis. Three types of PanKs have been identified so far, and the well-characterized E. coli enzyme is considered a type I, or prokaryotic, PanK. The PanK from Staphylococcus aureus is a type II, or eukaryote-like, enzyme. The S. aureus coaA gene was identified by comparative genomics, and the encoded protein was found to be moderately related to eukaryotic PanKs and unrelated to the E. coli enzyme (47). S. aureus PanK has been expressed and purified to homogeneity (16, 76). The enzyme is a dimer of 29-kDa subunits with K_m constants for ATP and pantothenate of 34 and 23 μM, respectively. The structure of the enzyme reveals two solvent-exposed openings to the active site and suggests a nonordered mechanism, with ATP entering from one opening and pantothenate binding through the other opening in random sequence (51). The type II PanK accepts the N-alkylpantothenamides as substrates, and these compounds manifest more potent antimicrobial action against S. aureus than E. coli. (76, 150). Unlike the type I PanK, the type II PanK from S. aureus is not subject to feedback inhibition by free CoA or CoA thioesters, and the organism accumulates high concentrations of intracellular CoA limited only by the pantothenate supply in the medium (76). S. aureus lacks glutathione (96), and instead this bacterium utilizes nonesterified CoA as the major intracellular thiol and a unique CoA disulfide reductase to maintain the redox potential of the cell (26, 27). Therefore, the ability to accumulate CoA through lack of feedback inhibition of PanK represents a clear advantage in the physiological context of S. aureus.

Bacterial type II PanK sequences are restricted to very few organisms, including Bacillus anthracis (47); however, no 4'-phosphopantothenate was detected in reaction mixtures containing the predicted B. anthracis type II PanK and pantothenate (97). The existence of a third type of PanK was inferred by the fact that several organisms, including pathogens such as Pseudomonas aeruginosa and Helicobacter pylori, lack any recognizable PanK, although they possess all the remaining enzymes of the CoA biosynthetic pathway (46, 47, 101). The first gene encoding a new type of PanK was discovered in B. subtilis and initially was named coaXto distinguish it from the coaA gene encoding a type I PanK in the same organism (164). The observation that disruption of the coaA gene in B. subtilis gave a normal growth phenotype led to the conclusion that there could be another enzyme able to perform the same reaction. The coaX gene was identified by the screening of a B. subtilis gene library for the ability to complement a temperature-sensitive E. coli strain. B. subtilis coaX homologs were subsequently found in the vast majority of organisms classified as missing either a type I or a type II PanK. Type III PanKs from a variety of organisms have been expressed, purified, and crystallized (10, 51, 97). The most remarkable difference between these enzymes and the types I and II PanKs is the low affinity for ATP. The K_m values for ATP are in the millimolar range, and the weak binding of the phosphoryl donor absolutely requires a monovalent cation such as potassium or ammonium. The type III PanK from P. aeruginosa is a dimer of 29-kDa subunits. The individual subunits of the S. aureus and P. aeruginosa PanKs fold in a very similar manner but assemble distinct dimers with unique enzymatic properties; in particular, the poor affinity and selectivity of the type III PanK for ATP can be explained by the wide and solvent-exposed binding site revealed by the crystal structure of the dimer (51). Conversely, pantothenate binds in a buried pocket at the dimer interface that would not accommodate bulkier analogs like the N-alkylpantothenamides. These compounds are not toxic for P. aeruginosa because they are not substrates for the type III PanK and hence cannot enter the CoA biosynthetic pathway. Hong and coworkers also showed that P. aeruginosa resistance to the N-alkylpantothenamides results from the lack of uptake of these compounds (51). Similar to the type II S. aureus PanK, the type III enzymes are refractory to feedback inhibition by free CoA or CoA thioesters. Among the organisms that possess a type III PanK are those that rely on the CoA/CoA disulfide reductase system to maintain redox status of the cell, such as B. anthracis, but also organisms such as P. aeruginosa that utilize the more widespread glutathione/glutathione reductase system.

Formation of 4'-Phosphopantetheine

4'-Phosphopantetheine is formed in two enzymatic steps (Fig. 2) (11). First, 4'-phosphopantothenate is condensed with cysteine by 4'-phosphopantothenoylcysteine synthetase (EC 6.3.2.5) to form 4'-phosphopantothenoylcysteine, which is then decarboxylated by 4'-phosphopantothenoylcysteine decarboxylase (EC 4.1.1.36) to generate 4'-phosphopantetheine. The 4'-phosphopantothenoylcysteine decarboxylase activity in E. coli was initially identified in fractions enriched in a 35-kDa protein containing a covalently bound pyruvoyl group (163); however, the enzyme was not purified to homogeneity and the conclusion turned out to be incorrect. More than a decade later the dfp gene (124, 125), recently renamed coaBC, was found to encode a flavin mononucleotide-containing bifunctional enzyme responsible for both the 4'-phosphopantothenoylcysteine synthetase and the 4'-phosphopantothenoylcysteine decarboxylase activities in E. coli (69, 130). Years before the discovery of the function of the dfp gene, a conditional-lethal dfp mutant, dfp-707, was isolated by Spitzer and coworkers, who noticed that the mutation caused a slow cessation of DNA synthesis at 42°C and that the dfp-707 strain required either β-alanine or pantothenate to grow at 30°C in minimal medium (124, 125). Another mutation in the same gene, named dfp-1, conferred auxotrophy but not the conditional lethality of dfp-707, allowing the dfp-1 mutant to grow in rich medium at 42°C. The nutritional requirements of both mutants were reminiscent of the panD mutants, but the aspartate-1-decarboxylase activity was similar to the wild-type strain and the authors did not investigate the connection to CoA biosynthesis any further. The 4'-phosphopantothenoylcysteine synthetase (CoaB) and the 4'-phosphopantothenoylcysteine decarboxylase (CoaC) activities encoded by the E. coli dfp gene were discovered individually and independently (69, 130). These activities are fused together in almost all bacteria with the exception of enterococci and streptococci that possess separate genes (46, 47). B. anthracis contains a gene predicted to encode a monofunctional 4'-phosphopantothenoylcysteine synthetase in addition to the bifunctional CoaBC enzyme. E. coli CoaBC is a homododecamer of 43-kDa subunits (69). Amino-terminal sequencing of the endogenous protein identified ATG²⁵ as the start codon of the coaBC gene and revealed the cleavage of the initiator methionine from the mature protein (130).

The individual 4'-phosphopantothenoylcysteine synthetase and 4'-phosphopantothenoylcysteine decarboxylase domains of the E. coli CoaBC have been expressed and purified (67, 68). The flavin mononucleotide-binding amino-terminal domain of the enzyme spans residues 2 to 190, possesses 4'-phosphopantothenoylcysteine decarboxylase activity, and is responsible for the formation of dodecamers by the full-length bifunctional protein. Amplification and sequencing of the coaBC gene from the conditional-lethal dfp-707 mutant identified a single point mutation in codon 11 that inserts a glycine in place of aspartate. The mutant protein was expressed and purified from cells grown at 30 and 37°C (67). The 4'-phosphopantothenoylcysteine carboxylase activity of the mutant was not affected by the expression temperature, but only very small amounts of protein were obtained at 37°C, suggesting that the conditional lethality of E. coli dfp-707 results from decreased protein solubility or stability. From a mechanistic point of view, the decarboxylation of 4'-phosphopantothenoylcysteine proceeds via a thioaldehyde intermediate formed by the flavin mononucleotide-dependent oxidation of the cysteine moiety of 4'-phosphopantothenoylcysteine. Decarboxylation of this intermediate followed by reduction completes the catalytic cycle (127, 129, 131). The carboxy-terminal domain (residues 191 to 406) of E. coli CoaBC mediates dimer formation and also catalyzes the CTP-dependent condensation of 4'-phosphopantothenate and cysteine (68). The preference of CTP over ATP distinguishes the bacterial 4'-phosphopantothenoylcysteine synthetase from the human monofunctional homolog (11). Similar to the dfp-707 mutant, the sequence analysis of the dfp-1 mutant revealed a single point mutation that inserts Thr for Ala275 (70). This Ala is part of the CTP binding site (126), but the molecular reason for the temperature sensitivity of the dfp-1 mutant needs to be elucidated in more detail.

The formation of 4'-phosphopantothenoylcysteine occurs via two half-reactions: first, an activated 4'-phosphopantothenoyl-cytidylate intermediate is formed, which is then attacked by the amino group of cysteine to yield the final product (70). The 4'-phosphopantothenoyl-cytidylate intermediate copurifies with mutant forms of the E. coli enzyme containing different residues in place of Asn210 and has been unequivocally identified by mass spectrometry. The intermediate is also visible in the crystal structure of the Asn210Asp mutant obtained in the presence of CTP and 4'-phosphopantothenate (126). E. coli 4'-phosphopantothenoylcysteine synthetase discriminates between cysteine and serine or homocysteine, but it can couple cysteamine and cysteine methyl ester to 4'-phosphopantothenate. The available crystal structures of the Asn210Asp enzyme reveal a potential cysteine binding site where the thiol group would be tightly accommodated in a hydrophobic cavity that would have poor affinity for the more polar serine side chain and insufficient space to accommodate larger side chains. Conversely, the carboxylate group of cysteine is predicted to be exposed to the solvent, explaining why cysteamine and cysteine methyl esters can function as substrates. The crystal structure of the human 4'-phosphopantothenoylcysteine synthase is available (85); the major structural features of the human and the E. coli enzymes, which likely share the same enzymatic mechanism, are very similar. However, the two enzymes differ significantly in the nucleobase-binding parts of the respective nucleotide binding sites, explaining the preference of the human enzyme for ATP and of the bacterial enzyme for CTP.

Conversion of 4'-Phosphopantetheine to Dephospho-CoA

ATP:phosphopantetheine adenylyltransferase (EC 2.7.7.3), also known as CoaD or PPAT, catalyzes the Mg²⁺-dependent reversible transfer of the AMP moiety of ATP to 4'-phosphopantetheine to form dephospho-CoA (1, 50). The enzyme was first isolated from Corynebacterium ammoniagenes (formerly Brevibacterium ammoniagenes) as a trimer of 35-kDa subunits (86), but fractionation of E. coli cell extracts identified a smaller enzyme of 18 kDa with PPAT activity (45). N-terminal sequencing of the purified endogenous PPAT allowed the identification of kdtB as the coding gene, now renamed coaD. Recombinant E. coli PPAT forms homohexamers arranged as dimers of trimers and has K_m values of 220 and 7 μM for PP_i and dephospho-CoA, respectively, in the reverse reaction.

Metabolic labeling experiments in E. coli detect accumulation of both intracellular and extracellular 4'-phosphopantetheine, suggesting that, in addition to the primary control exerted at the PanK level, PPAT is a secondary regulatory point in CoA biosynthesis (59, 62) (see below). Similar to PanK, regulation of the PPAT activity probably occurs through feedback inhibition by unesterified CoA (59, 108, 144), and consistent with this theory, the E. coli enzyme is purified with 0.5 mol of CoA per mol of enzyme (45). Crystal structures of the E. coli PPAT complexed with the substrates, product, and the putative inhibitor CoA have been solved (54, 55, 56). These structures show that PPAT is an allosteric enzyme characterized by half-of-the-sites reactivity. In fact, only one of the trimers within the hexamer is in the substrate- or dephospho-CoA-bound conformation, whereas the other trimer is in its unbound, conformationally distinct state. This asymmetry in the binding to the two trimers is further confirmed by the CoA-bound structure of the enzyme. Specifically, the phosphopantetheine moiety of CoA binds to one trimer in a dephospho-CoA- or phosphopantetheine-like manner, with the adenylyl moiety disordered. This mode of binding, characterized by CoA occupying the substrate binding site, suggests a mechanism for the enzyme inhibition and produces a conformational change in the other trimer that forces CoA to bind in an ordered and unique conformation imposed by the presence of a phosphate group at the 3' position of the ribose (55).

Kinetic analysis of the reverse reaction catalyzed by the E. coli PPAT reveals the formation of a ternary complex between the enzyme and both pyrophosphate and dephospho-CoA before catalysis, and the structural analysis predicts a random ordered binding mechanism (45, 56). Superposition of the ATP- and 4'-phosphopantetheine-bound PPAT structures supports a model in which the α-phosphate of ATP undergoes nucleophilic attack by the phosphate group of 4'-phosphopantetheine in an in-line displacement mechanism (54). No direct participation of active site residues in acid-base or covalent catalysis is necessary, and the sole role of the enzyme seems to be to properly orient the substrates and to stabilize the transition state to lower the reaction energy barrier. Histidine-18 is particularly important in the stabilization of the transition state. This residue corresponds to the last histidine of the highly conserved H/TxGH motif that characterizes the nucleotidyltransferase α/β-phosphodiesterase superfamily (9) to which PPAT belongs (45). Crystal structures of the enzyme are also available from other organisms, and they all reveal the same protein fold as the E. coli PPAT (94, 136). In mammals, the PPAT is fused with dephospho-CoA kinase in a bifunctional enzyme designated CoA synthase (2, 22, 161, 172). Similar to PanK, the PPAT domain of the human CoA synthase does not share any significant sequence similarity with its prokaryotic counterparts (22, 46, 47). This fact and the availability of crystal structures make the enzyme an attractive target for the design of selective antibacterial drugs (167).

Phosphorylation of Dephospho-CoA to CoA

The last step in CoA biosynthesis is catalyzed by dephospho-CoA kinase (DPCK, also called CoaE) (EC 2.7.1.24) that adds an ATP-derived phosphate group to the 3'-hydroxyl of dephospho-CoA. The enzyme responsible for the dephospho-CoA activity in C. ammoniagenes was first isolated from cell extracts and used to determine the protein N-terminal sequence. A BLAST search using this information then allowed the identification of the yacEgene as encoding a highly homologous protein in E. coli. The yacE gene has been renamed coaE (91). E. coli dephospho-CoA kinase has been overexpressed and purified. The enzyme exhibits K_m values of 0.74 and 0.14 mM for dephospho-CoA and ATP, respectively. DPCK is purified as a 22-kDa monomer, but in the presence of sulfate ions it forms trimers both in solution and in the crystal structure (99). The monomer fold identifies the enzyme as a member of the P-loop-containing nucleotide triphosphate hydrolase superfamily, which includes several nucleotide and nucleoside kinases. The P-loop motif is involved in the binding of the ATP triphosphate group, as confirmed by the crystal structures of ATP-bound DPCK from H. influenzae and Thermus thermophilus HB8 (100, 116). The three structures are very similar, but while E. coli DPCK crystallizes as a trimer, the structures of the H. influenzae and T. thermophilus HB8 enzymes contain one and three independent monomers in the asymmetric unit, respectively. The structures also reveal the three-domain organization characteristic of nucleotide and nucleoside kinases: the nucleotide binding domain, the substrate binding domain, and the lid domain. The latter two domains are expected to be very mobile and, on binding of dephospho-CoA, to close over the catalytic site to direct the phosphate group transfer.

CoA METABOLISM

In proliferating bacteria with unlimited pantothenate availability, the CoA pool is large and more than sufficient to support rapid growth. Removal of the pantothenate supply does not result in a cessation of growth until the CoA pool is reduced by dilution, which requires several generations of bacterial growth (59). Regulation of the size of the CoA pool is largely accomplished by limitation of its synthesis through feedback inhibition rather than by degradation. Mechanisms to reduce CoA by degradation are in place, however, to respond to metabolic challenge, such as an acute change in carbon flux (61) or adaptation to pantothenate starvation (60).

Composition of the Intracellular CoA Pool

The composition of the E. coli intracellular CoA pool has been examined by using panD mutants to specifically label CoA and by estimating individual CoA thioesters by high-pressure liquid chromatography (145). The size and composition of the CoA pool vary depending on the carbon source (Table 1). The CoA pool is highest in cells growing on glucose, and acetyl-CoA is the predominant species. In contrast, the amount of total CoA is much lower in cells growing on casein hydrolysate as the carbon source, indicating that amino acid biosynthesis requires an elevated CoA pool, in particular, in the form of acetyl-CoA. Consistent with this interpretation, when pantothenate auxotrophs growing on glucose minimal medium are deprived of pantothenate, the succinyl-CoA pool becomes limiting for amino acid and protein synthesis, leading to growth arrest (61). Alternatively, growth on glucose may lead to acetate accumulation, which, in turn, may determine the size of the acetyl-CoA pool. The latter explanation is supported by the observation that the acetyl-CoA pool is rapidly reduced >30% upon withdrawal of supplement from an acetate auxotroph (144). The total amount of unesterified CoA drops correspondingly, with the excess CoA being hydrolyzed and 4'-phosphopantetheine effluxed from the cell. On the other hand, elevation of intracellular CoA by overexpression of PanK in E. coli causes an increase in carbon flux to acetate production, provided that sufficient pantothenate precursor is available (121, 141). Increased acetate production by the cell is reflected not only by increased acetyl-CoA levels but also by an increase in excreted acetate (142).

Table 1Intracellular CoA pool composition in E. coli

The concentration of nonesterified CoA in S. aureus reaches millimolar levels (26). In the absence of glutathione, this organism uses free CoA not only in metabolic cycles but also as the major low-molecular-weight thiol in the cell. The ratio of reduced to oxidized (disulfide) CoA is >100 and maintained by a specialized CoA disulfide reductase. Accumulation of CoA to levels significantly higher than in E. coli is possible because S. aureus possesses a PanK refractory to feedback inhibition by CoA (76). Glutathione is also missing in many bacilli (37, 38) that, instead, utilize a pantetheine 4',4″-diphosphate- or CoA-disulfide reductase system (133, 134). In the spore-forming bacterium Bacillus megaterium, CoA is also used to form CoA-protein disulfides during spore formation. Accumulation of these mixed disulfides is proposed to maintain metabolic dormancy in the spores and/or to contribute to heat and radiation resistance by protecting labile protein thiol groups (115). Although CoA-glutathione mixed disulfides have been reported in E. coli (79, 80), these compounds are likely due to oxidation prior to analysis, and there is no evidence that they occur in vivo.

Regulation of CoA Levels by Feedback Inhibition

The phosphorylation of pantothenate catalyzed by PanK is the primary rate-limiting step in CoA biosynthesis in E. coli. This reaction is controlled through feedback inhibition of the enzyme by CoA and CoA thioesters, the end products of the pathway. As previously mentioned, considerable differences exist in the size and composition of the CoA pool in E. coli cells grown on different carbon sources. A shift from glucose to acetate as the carbon source results in an increase in the nonesterified CoA/acetyl-CoA ratio from 0.7 to 4.3 (145) and in the reduction of ATP levels (81). This remodeling of the CoA pool composition is associated with the selective inhibition of pantothenate phosphorylation, consistent with nonesterified CoA being the most potent inhibitor of PanK in vivo. E. coli mutants that possess a PanK activity in crude extracts which is refractory to feedback inhibition by CoA have been isolated (144). Strains harboring this mutation [coaA16(Fr)] have CoA levels that are significantly (>twofold) higher than in strains containing the wild-type kinase. A similar result is obtained when a single mutation of Arg106 to Ala is introduced in E. coli PanK and the mutant is expressed at single-copy levels in the coaA15(Ts) strain background at elevated temperature (108). Furthermore, corroboration of the conclusion that modulation of PanK activity by feedback regulation is the critical factor controlling the intracellular CoA concentration comes from studies of the effect of PanK overexpression on the size of the CoA pool (123). Strains expressing 76-fold more wild-type kinase exhibited only a 2.7-fold increase in the steady-state CoA level. CoA regulates the PanK activity by competing for the ATP binding site (122, 145, 165); thus, the activity of the enzyme can also be coordinated with the energy state of the cell, where an increase in ATP levels would displace the competitive inhibitor and resume the biosynthetic activity. Therefore, changes in the composition of the CoA pool and ATP levels function in concert to modulate the rate of CoA biosynthesis.

The secondary regulatory step in CoA biosynthesis is catalyzed by PPAT. Regulation at this step in E. coli is proven by the secretion of 4'-phosphopantetheine to the medium (59, 62) and becomes more important when the primary regulatory step is disrupted (108) or when PanK is overexpressed (121). An increase in the amount of intracellular and extracellular 4'-phosphopantetheine under these circumstances reflects the restriction of the CoA precursor flux through PPAT. Excretion of 4'-phosphopantetheine is an irreversible event since E. coli is unable to transport the phosphorylated intermediates of CoA biosynthesis (62). The time- and concentration-dependent correlation between accumulation of intracellular CoA and exit of 4'-phosphopantetheine from the cells suggests that PPAT is regulated by free CoA (59, 144). Consistent with this hypothesis, the enzyme is isolated and crystallized with bound unesterified CoA (45, 55). S. aureus lacks both the PanK and the PPAT regulatory checkpoints. This conclusion is suggested by the fact that neither pantothenate nor 4'-phosphopantetheine accumulates inside or outside the cells in metabolic labeling experiments (76).

CoA Metabolism and Prosthetic Group Transfer

CoA is the source of the 4'-phosphopantetheinyl prosthetic group present in several proteins that function as acyl/aminoacyl/peptidyl group carriers. Examples are the carrier proteins of fatty acid synthases, nonribosomal peptide synthetases, and polyketide synthases (66). The transfer of the 4'-phosphopantetheinyl moiety of CoA to a conserved serine residue of these carrier proteins releases 3',5'-ADP and is catalyzed by a class of enzymes called phosphopantetheinyltransferases (PPTases) (71, 93). In E. coli, the acyl carrier protein (ACP) of fatty acid biosynthesis is specifically converted to holo-ACP by ACP synthase (EC 2.7.8.7) (34, 72), also termed AcpS1 (41), which is a PPTase encoded by the acpS gene. Sequencing of the acpS gene from the E. coli strain MP4 reveals a Gly4Asp mutation that reduces the catalytic efficiency of the enzyme about fivefold. The EntD PPTase of E. coli activates the enterobactin synthetase (71). A third PPTase identified in E. coli K-12 and other E. coli strains by homology searches is encoded by the gene acpT (25). The AcpT protein modifies two carrier proteins encoded in O-island 128, a cluster of fatty acid biosynthesis-like genes located adjacent to acpT in the genome of the pathogenic E. coli strain O157:H7, but it cannot substitute fully for AcpS in its activity. Overexpression of the acpT gene (formerly called yhhU) can suppress the phenotype of acpS mutants. Whereas in most cases the bacterial PPTases exhibit acceptor protein substrate specificity, in P. aeruginosa the single PPTase modifies all the 4'-phosphopantetheinyl-modified proteins (7).

The 4'-phosphopantetheine prosthetic group can be removed from E. coli ACP by AcpH, the ACP phosphodiesterase, also called ACP hydrolase (EC 3.1.4.14) (138, 139, 143). While AcpS is essential (47), AcpH is not and its distribution is limited to gram-negative organisms (138). AcpH, a noncanonical member of the HD phosphatase phosphodiesterase family, cleaves both unacylated and acylated ACP species with chain lengths of 6 to 16 carbons (138); its activity is dependent on Mn²⁺ ions (139). Metabolic radiolabeling following the starvation of pantothenate auxotrophs shows that the level of holo-ACP is maintained at the expense of CoA (4, 59, 62). The turnover of the ACP prosthetic group is four times faster than the rate of new ACP protein synthesis during recovery from CoA deprivation and drops an order of magnitude during exponential growth when the CoA levels are high (62). The 4'-phosphopantetheine released by the AcpH can either re-enter the CoA biosynthetic pathway (Fig. 2) or irreversibly exit from the cell and thereby regulate the size of the CoA pool.

CoA can also be hydrolyzed directly to yield 4'-phosphopantetheine and 3',5'-ADP, and this process does not involve ACP prosthetic group turnover (144). CoA degradation occurs when the level of acetyl-CoA falls, leading to a concomitant increase in nonesterified CoA. These results are corroborated by the finding that large amounts of 4'-phosphopantetheine are excreted also in an E. coli strain lacking AcpH (138). Recent data suggest that the pyrophosphatase responsible for CoA degradation might be a member of the nudix family (65, 162).