Chapter 34

Biosynthesis of Purine Nucleotides

HOWARD ZALKIN and PER NYGAARD

INTRODUCTION

This chapter reviews our current understanding of the biosynthesis of purine nucleotides by de novo and salvage pathways and the inverconversions of the purine bases, nucleosides, and nucleotides in Escherichia coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium). Readers may consult two recent articles that review de novo purine nucleotide synthesis in gram-positive bacteria (177) and in vertebrates (178). For earlier reviews and references to older literature on metabolism of purine bases and nucleotides, see references 50, 106, 111, and 116). It is only in E. coli, however, that there is now an understanding at the molecular level of the regulatory circuits that control the biosynthesis of purine nucleotides. This chapter provides recent information that explains how gene expression and enzyme activity are regulated by the pools of purine bases and nucleotides, respectively. Purine nucleotides may be derived from exogenous purines by the so-called salvage pathways, or they may be synthesized de novo from simpler precursors. In either case, the ribose 5-phosphate moiety is derived from 5-phosphoribosyl-α-1-pyrophosphate (PRPP).

PRPP

The synthesis of PRPP is catalyzed by phosphoribosylpyrophosphate synthetase as follows:

ATP + ribose 5-phosphate

PRPP is required for the biosynthesis, de novo and salvage, of purine, pyrimidine, and nicotinamide nucleotides and for the biosynthesis of histidine and tryptophan. In cells growing in minimal medium, 30 to 40% of the PRPP is used for the synthesis of purine nucleotides; an equal amount is used for synthesis of pyrimidine nucleotides. Histidine and tryptophan each consume about 10 to 15%; 1 to 2% is used for the synthesis of nicotinamide nucleotides (70). PRPP is of importance only for these metabolic pathways and is dispensable in a special genetic background (63).

PRPP Synthetase

Enzymology and Regulation.

PRPP synthetase has been purified to homogeneity from both S. typhimurium (154) and E. coli (64). The molecular mass of the subunit from both sources is 31 kDa. The native enzyme exists in several states of aggregation. There are only two amino acid differences between the enzymes from E. coli and S. typhimurium (10, 64). P_i is essential for enzyme stability and activity. Magnesium ions and ATP also stabilize the enzyme, but they cannot replace phosphate (154). Magnesium ions are needed both as chelators of ATP and as free cations.

Many nucleotides inhibit enzyme activity competitively with ATP. However, the only potent nucleotide inhibitor is ADP; it competes with ATP and is an allosteric inhibitor that binds to a site different from the active site (64). The pattern of ADP inhibition is highly influenced by the concentration of ribose 5-phosphate, which displays substrate inhibition in the presence of ADP (154).

Genetics and Regulation of Expression.

The prs gene, sometimes referred to as prsA, encodes PRPP synthetase in both E. coli and S. typhimurium (63, 123). Mutants with altered PRPP synthetase have been obtained in both E. coli and S. typhimurium (65, 72, 119). One of the E. coli mutants produces a PRPP synthetase with a 27-fold increase in the K_m value for ATP, and the enzyme is no longer inhibited by AMP (9). This mutation results from substitution of aspartate 128 by alanine. One S. typhimurium prs mutant is temperature sensitive and has only 20% of the wild-type PRPP synthetase activity. This mutant enzyme had elevated K_m values for ATP and ribose 5-phosphate and reduced sensitivity to inhibition by ADP (119, 123). The mutation is the result of the replacement of arginine 78 by cysteine (123).

Increased levels of PRPP synthetase have been observed in cultures of S. typhimurium and E. coli partially starved for uracil, and the effector is a uracil nucleotide different from UMP (65, 70, 123, 154). The prs gene is transcribed from two promoters. The major site of transcription initiation from promoter 2 precedes the start of translation by 302 bp in E. coli and 417 bp in S. typhimurium. The pyrimidine regulation of the prs gene results in an increase of a 1.4-kb transcript from promoter 2. The 1.27-kb transcript originating from promoter 1 constitutes only 5% of the total PRPP synthetase mRNA (122). The expression of the prs gene is also subject to PurR control; the PurR protein, plus corepressor, binds in vitro to the control site of the prs gene, and the level of β-galactosidase, expressed from a prs '::' lacZ translational fusion, is increased two- to threefold in a purR mutant (52).

PRPP Pools

Addition of purine bases to exponential-phase cultures of E. coli and S. typhimurium results in an immediate shrinkage of the intracellular PRPP pool (70) (see Table 2). Adenosine and inosine have the same effect as hypoxanthine, whereas guanosine addition has no effect. Purine starvation causes severe depletion of all of the nucleoside triphosphate pools concomitantly with an accumulation of PRPP. In contrast, pyrimidine starvation results in depletion of PRPP and swelling of the purine nucleotide pools. It seems that conditions that lead to high pools of adenine nucleotide result in a reduction of the PRPP pool. The metabolite responsible for these effects could be ADP.

Table 2Effects of purines added to the growth medium of E. coli wild-type cells on levels of enzymes involved in purine interconversion reactions and on pools of ATP, GTP, and PRPP

Mutants with reduced PRPP synthetase activity have low PRPP pools, are stressed, and respond differently from wild-type cells to purine or pyrimidine starvation (70, 72, 119). A PRPP pool could not be detected in a strain with a prs null mutation (63).

DE NOVO SYNTHESIS OF IMP

The pathway for de novo synthesis of IMP is outlined in Fig. 1. There are 11 enzymatic steps for synthesis of IMP, the first nucleotide intermediate in the pathway. Starting from PRPP, which contributes the ribose phosphate moiety of IMP, the purine ring is assembled from two amide nitrogens from glutamine, the amino group of aspartate, the two carbon atoms and amino group of glycine, and three one-carbon units: HCO₃ ^– and the carbon atoms from two molecules of 10-formyltetrahydrofolate (10-formyl-THF). Formate, however, can replace one molecule of 10-formyl-THF. Comparative studies indicate unique patterns of gene organization, unique mechanisms for gene regulation, and distinctive multifunctional enzymes among organisms (178). Since the previous account of the biosynthesis and conversions of purines in 1987 (111), all of the genes for the de novo pathway in E. coli have been cloned, the principal regulatory elements have been characterized, an additional biosynthetic step has been discovered, and a new understanding of several proteins in the pathway has been achieved. The 11 genes that are utilized for synthesis of IMP are organized into nine operons. The genes, operon organization, and gene regulation are described in the following section.

Fig. 1Biosynthesis of the purine nucleotides. (a) Formation of CAIR; (b) conversion of CAIR to IMP; (c) conversion of IMP to GMP and AMP. Alternative GAR transformylase enzymes are used for conversion of GAR to FGAR. The coordinates for the gene positions are those of the E. coli physical map.

Source:

Genes and Gene Regulation

Structural Genes.

Mutations in structural genes for steps 1, 2, and 4 to 11 (Fig. 1) block synthesis of IMP and result in purine auxotrophy. Genes for these steps were cloned by functional complementation of auxotrophic strains and are listed in Table 1. The sequenced genes listed in Table 1, with the exception of purHD, are all from E. coli. There is no reason to suspect significant differences in the corresponding S. typhimurium pur and gua genes. Single mutations that inactivate 5'-phosphoribosylglycinamide (GAR) transformylase and block the conversion of GAR to 5'-phosphoribosyl-N-formylglycinamide (FGAR) (step 3) have not been isolated. The purN gene, which encodes GAR transformylase-N, is contiguous with purM and was initially identified as the distal gene in a purMN operon (148, 149). With the availability of cloned purN, the chromosomal gene was disrupted, and the resulting Pur⁺ mutant was used to isolate a spontaneous purine auxotroph (purN ⁺ purT). This led to isolation of a purN ⁺ purT strain that was Pur⁺. Thus, E. coli has two genes, purN and purT, which encode distinctive GAR transformylases. A mutation in either gene results in a decreased growth rate in minimal medium, but the single mutants are Pur⁺. GAR transformylase-T, the product of purT, uses formate rather than 10-formyl-THF as a C₁ donor for synthesis of FGAR (Fig. 1). For aerobic growth, the formate required for the GAR transformylase-T reaction can be derived from 10-formyl-THF. 10-Formyl-THF hydrolase is encoded by purU (108). purU mutants are Pur⁺, but the purU purN double mutant is a purine auxotroph. Growth is restored by supplementation with either formate or adenine. The purU gene is not coregulated with other pur regulon genes and is therefore not included in Table 1.

Table 1Summary of genes and enzymes for purine nucleotide synthesis

Gene Organization.

Gene organization is given in Fig. 2 together with information about gene regulation. Four of the eleven genes required for synthesis of IMP, purT, purL, purC, and purB, are in monocistronic operons. Map positions are given in Fig. 1. Of the polycistronic operons, only purF is associated with genes not known to have a relationship to purine nucleotide synthesis. purF is provisionally placed in a cvpA purF dedF operon. The gene is transcribed (93) from a promoter that is located 34 bp upstream of the coding sequence for cvpA (colicin V production [35]). There is a 94-bp region between the 3' end of purF and the translation start site for a gene of unknown function designated dedF (downstream [from hisT] E. coli DNA) (113). Nonet et al. (113) assigned dedF to a cvpA purF dedF operon primarily on the basis of the results of an RNA mapping experiment. However, the data for this conclusion are not compelling, and the amount of dedF transcription from the purF operon promoter relative to that from a unique dedF promoter is not known.

Fig. 2Operon organization, regulation, and positions of operator sequences in pur and gua regulon genes. The operator is shown as a thick black line with numbers giving the position of the center of symmetry. +1 indicates the start of transcription. The 5? untranslated mRNA is a thin line followed by the coding sequence (open box). Translation start sites after +60 are indicated by numbers. purR-dependent regulation was determined by using lacZ reporter strains or by direct enzyme assay in the case of GAR transformylase-T (PurT) and IMP dehydrogenase (IMPDH). References are given for regulation.

Source:

In S. typhimurium genes encoding IMP cyclohydrolase (purJ), 5'-phosphoribosyl-4-carboxamide-5-aminoimidazole (AICAR) transformylase (purH) and GAR synthetase (purD) were initially mapped as a purJHD operon linked to the metA gene at min 89 on the genetic map (47, 48). Although genetic complementation analysis in S. typhimurium was interpreted to suggest that AICAR transformylase and IMP cyclohydrolase were products of separate genes (reviewed in reference 47), nucleotide sequence analyses of the loci in E. coli (1, 36) and S. typhimurium (26) demonstrate that both functions are encoded by a single purH gene that is separated from the downstream purD gene by 11 bp.

The purMN operon encoding 5'-phosphoribosylaminoimidazole (AIR) synthetase and GAR transformylase-N was identified in the course of gene cloning and sequence determination (148, 149). The ATG translation start codon for purN has a 1-bp overlap with the purM TAA translation termination codon.

The purE loci in E. coli and S. typhimurium encode activities that were until recently called AIR carboxylase. The locus was initially divided into genes purE ₁ and purE ₂ on the basis of complementation analysis and according to a requirement for high levels of CO₂ for growth of certain mutants (reviewed in reference 47). purE₁ (now called purE) mutants had an absolute growth requirement for purines, whereas purE ₂ (now called purK) mutants were conditional purine auxotrophs, capable of growth in the absence of purines when the CO₂ concentration was elevated. Upon sequence analysis of the cloned genes, the locus was found to encode a two-gene operon that was designated purEK (160, 169). There is a 4-bp overlap between the translation termination codon of purE and the translation initiation codon of purK, indicative of translational coupling. The purK and purE genes encode enzymes recently named 5'-phosphoribosyl-N-carboxyaminoimidazole (NCAIR) synthetase and NCAIR mutase (or isomerase), respectively (103).

Promoters.

Promoters for pur and gua regulon genes have generally well conserved Pribnow boxes around position –10 relative to the start of transcription but have poorly conserved –35 regions. The purF promoter was the first of the pur gene promoters to be identified. It was initially positioned upstream of cvpA by mapping the mRNA 5' end with S1 nuclease and by analysis of in vitro runoff transcripts (93). Promoter identification was later confirmed by examining the effect of site-directed mutations in the control region on purF expression (127). The purF promoter sequence shown in Fig. 3 matches the consensus Pribnow box (131) in five of six positions, but there is correspondence at only three of six positions with the TTGACA –35 consensus sequence. A two-base substitution (5'-TACGCA to 5'-TTGGCA) at positions –35 and –36 that increases the similarity of the purF –35 promoter element to the consensus sequence increased expression 11-fold and supports the –35 promoter assignment for purF that is given in Fig. 3. Likewise, a G-to-T replacement at position –11 that leads to a perfect match with the consensus Pribnow box increased expression twofold.

Fig. 3pur and gua gene promoter sequences. The 3? distal nucleotide represents the transcription start site. In most cases, this position was determined by primer extension or nuclease mapping. In the case of purMN, one of two bases is arbitrarily chosen. The guaBA start site was not determined but is inferred by comparison with the other genes. ?10 promoter elements with a four-of-six-nucleotide or better match with the consensus TATPuAT (131) are underlined. Each sequence starts with a candidate for the ?35 promoter element.

Source:

For the other genes listed in Fig. 3 that are required for synthesis of IMP, the transcription start sites at position +1 were deduced from RNA mapping experiments. These genes, with the exception of purB, have a five- or six-base identity with the consensus –10 promoter element and have poorly conserved –35 regions. The Pribnow box for purB is not apparent. Potential –35 promoter elements are suggested in Fig. 3 on the basis of the best matches to the 5'-TTGACA consensus and the preferred spacing of 17 to 18 bp between the –10 and –35 promoter elements (131). The cvpA purF dedF, purHD, purMN, purT, purL, and purEK promoters have a G+C-rich sequence between the –10 promoter site and the start of transcription, similar to genes that are under stringent control (163). Preliminary evidence was reported for stringent control of purF transcription (93).

Gene Regulation.

Two regulatory elements, a purR-encoded repressor protein and a DNA operator site for repressor binding, are required for regulating the expression of genes involved with synthesis of IMP. The first purR mutations affecting overall regulation of the de novo pathway arose fortuitously in an S. typhimurium purA strain (47). Other purR mutants were subsequently isolated by using resistance to the inhibitory purine analog 6-mercaptopurine (47, 86) and by exploiting the growth inhibition by adenine of pur-lac fusion strains with lactose as a carbon source (47, 86, 87, 158). In these instances, the term purR was used to describe a regulatory phenotype, since the mutations were uncharacterized genetically. Genetic characterization of the phenotype was limited by high spontaneous mutation rates and by the availability of sensitive assays for enzymes in the pathway. Nevertheless, evidence was accumulated for the regulation of purF, purEK, and purHD by purR in S. typhimurium (47) and E. coli (86, 88).

Genetically stable E. coli purR mutants were isolated in two laboratories by different strategies. Rolfes and Zalkin (127) isolated spontaneous purR mutants with defective repression of a purF'-lacZ reporter gene, and Kilstrup et al. (75) selected a purR::Tn10 mutant unable to repress codA, a pyrimidine salvage gene. The purR gene was cloned by functional complementation with these mutants (75, 128), and its nucleotide sequence was determined (128).

PurR binds to a 16-bp palindromic sequence in the purF control region (98, 128). This sequence is shown in Fig. 4. Mutations that abolish binding of PurR in vitro result in defective regulation in vivo, thus defining the PurR binding site as the purF operator (127). As shown in Fig. 4, the 16-bp purF operator sequence is conserved in all genes required for synthesis of IMP. Hypoxanthine and guanine are corepressors in vivo (62) and increase the PurR-operator binding affinity in vitro (99, 129).

Fig. 4pur regulon operator sequences. Bases matching the perfect palindrome are highlighted. Two operators have been identified for purA and purR. The top group of operators includes pur and gua genes; the lower group is from related genes subject to regulation by PurR.

Source:

Expression of all of the pur and gua genes shown in Fig. 2 is repressed by the binding of purine holorepressor to an operator site. Except for purA and guaBA, there is no evidence for additional regulation. The regulation of purA and guaBA is described in the section on synthesis of AMP and GMP from IMP. Operator sites for control by PurR are generally located in the promoter region between position –38 and the start of transcription (Fig. 2). purB, which is needed for synthesis of IMP and also for conversion of IMP to AMP, utilizes an operator that is located in the protein coding region 232 bp downstream from the start of transcription. Binding of repressor to an operator site in the promoter region indicates regulation of transcription initiation. PurR has been shown to inhibit initiation of purF transcription in vitro (54). In the case of purB, binding of repressor appears to block transcription elongation (54), although this effect was not demonstrated in vitro. There is greater than 10-fold repression for all of the genes dedicated to the synthesis of IMP. Differences in the extent of repression could be due to repressor-operator affinity and to the position of the operator relative to the promoter (82). The purB gene, on the other hand, is repressed only about threefold, consistent with the requirement for cells to utilize preformed purine bases when expression of genes for IMP synthesis is shut down.

Properties of Enzymes

Glutamine PRPP Amidotransferase.

Glutamine PRPP amidotransferase, the product of the E. coli purF gene, catalyzes the initial reaction of the de novo pathway for synthesis of IMP: PRPP + glutamine → 5'-phospho-β-d-ribosylamine (PRA) + glutamate + PP_i. The enzyme, a tetramer of identical subunits (100), is one of a family of amidotransferases that utilize the amide of glutamine for biosynthesis (176). As with other enzymes in the glutamine amidotransferase family, NH₃ can replace glutamine as an N donor in vitro and in vivo (94, 100). The present understanding of structure and function is derived from analyses of chemical modifications with affinity analogs of glutamine (100, 164) and nucleotides (180), replacements of amino acid residues by site-directed mutagenesis (94, 96) as well as analyses of deletions (97), and an X-ray structural model of the homologous enzyme from Bacillus subtilis (147).

E. coli glutamine PRPP amidotransferase contains two domains, a C-terminal synthase domain of 264 amino acids containing sites for binding PRPP and for regulation by adenine and guanine nucleotides, connected to an N-terminal glutamine amide transfer (GAT) domain of 240 residues. The GAT domain requires an N-terminal active-site cysteine residue to utilize glutamine as an amide donor. Enzymes of the glutamine amidotransferase family having an N-terminal active-site cysteine have been classified as purF type, or simply F type. In a second, G-type amidotransferase subfamily, the active-site cysteine has an internal position. The role of the active-site cysteine is for covalent catalysis. There are two proposals for the action of Cys-1. In one mechanism, nucleophilic attack of the cysteinyl thiolate on the carboxamide of glutamine releases the amide and yields a covalent γ-glutamyl thioester (176). Release of the amide of glutamine is tightly coupled to synthesis of PRA by an unknown mechanism. According to an alternative mechanism, originally proposed for asparagine synthetase (125, 145), initial formation of a covalent glutamine-ribose 5-phosphate imide is followed by attack of the Cys-1 thiolate to release PRA and the γ-glutamyl thioester covalent intermediate.

E. coli glutamine PRPP amidotransferase is subject to end product inhibition by purine nucleotides (100). There are five main characteristics of the inhibition. (i) Of the adenine and guanine nucleotides, the 5'-monophosphates are most inhibitory, with the effectiveness of GMP being greater than that of AMP. (ii) Inhibition by AMP plus GMP is synergistic. (iii) Inhibition by AMP and GMP is competitive with the substrate PRPP. (iv) In the absence of inhibitor, saturation by PRPP is hyperbolic. PRPP saturation remains hyperbolic with added AMP but is sigmoidal in the presence of GMP. (v) AMP and GMP exhibit cooperativity for inhibition, with Hill coefficients of 2.0 to 4.6. These properties can be accommodated by a model in which there are two nucleotide regulatory sites per subunit, with communication between bound AMP and GMP.

An allosteric A-site and a second nucleotide site that overlaps the catalytic site (C-site) were identified in the X-ray structure (147) and by affinity labeling and site-directed mutagenesis (180). Inhibition assays and equilibrium binding measurements have shown that synergistic inhibition by AMP and GMP results from synergistic binding to two sites per enzyme subunit in the E. coli homotetramer (179). Although each nucleotide can bind to both sites, binding of GMP to the A-site and AMP to the C-site appears to be necessary for synergistic inhibition. Replacements of Lys-326 and Pro-410 perturb nucleotide binding to the A- and C-sites, respectively, with corresponding reductions in end product inhibition. The double mutant is insensitive to inhibition by concentrations of AMP plus GMP that abolish activity of the wild-type enzyme. The structural basis for synergistic binding of nucleotides is not presently known. According to a current model, catalysis is inhibited by binding of nucleotide in the C-site which prevents binding of PRPP to this site and by a conformation change that results from nucleotide bound to the A-site (147).

GAR Synthetase.

GAR synthetase catalyzes the second step in the pathway to IMP: PRA + glycine + ATP ↔ GAR + ADP + P_i. The monofunctional E. coli enzyme is homologous with the corresponding component in the trifunctional invertebrate (57) and vertebrate (2, 28) GAR synthetase-AIR synthetase-GAR transformylase and the bifunctional GAR synthetase-AIR synthetase from Saccharomyces cerevisiae (57). The homogeneous E. coli enzyme is a monomer with a molecular mass of 46 kDa (144). Initial velocity studies and analyses of product and dead-end inhibition are consistent with a sequential ordered mechanism of substrate binding and product release in which PRA binds first, followed by MgATP and then glycine; P_i leaves first, followed by MgADP and GAR. Experiments with [¹⁸O]glycine confirm that the role of ATP in the reaction is to phosphorylate the carboxyl group of glycine. Given the lability of PRA, t _½ of 38 s at pH 7.5 and 37°C (136), it appears unlikely that PRA, produced in the reaction catalyzed by glutamine PRPP amidotransferase, could be a free intermediate. This possibility has led to the proposal for an amidotransferase-GAR synthetase enzyme complex in which the production of PRA could be tightly coupled to synthesis of GAR (144).

GAR Transformylase.

Two GAR transformylases are utilized in E. coli for synthesis of FGAR from GAR. GAR transformylase-N encoded by purN catalyzes the reaction GAR + 10-formyl-THF ↔ FGAR + ADP + P_i. In solution, GAR transformylase-N is a monomeric enzyme with an M _r of 23,200 (67). The enzyme can be assayed with its natural substrate (6R)-10-formyl-THF or more conveniently with 10-formyl-5,8-dideazafolate (68). Initial velocity studies indicate a sequential mechanism.

The X-ray structure of apo-GAR transformylase-N (4, 21), as well as the enzyme complex with GAR and 5-deaza-THF (4), has been solved. The primary structural motif is a central core of smoothly twisting, seven-stranded β-pleated sheet surrounded on both sides by α helices. The fold is a modification of the doubly wound α/β sheet (126). The substrate GAR and the folate analog bind adjacent to the fourth and fifth strands of the β sheet, at the C-terminal end. There are H-bond interactions between the phosphate of GAR and residues Gly-11, Ser-12, Asn-13, Gln-170, and Glu-173. H-bond interactions with the folate analog include Arg-90, Leu-92, Asp-141, and Asp-144. Five residues, Asn-106, His-108, Ser-110, His-137, and Asp-144, are predicted to be in sufficiently close proximity to the formyl group of 10-formyl-THF to participate in catalysis. Experiments with a folate affinity analog and use of site-directed mutagenesis (68) have implicated Asp-144 as a general base for catalysis.

GAR transformylase-T catalyzes the reaction GAR + formate + ATP → FGAR + ADP + P_i (95). Formation of an enzyme- bound formyl-P intermediate is tightly coupled to synthesis of FGAR. The enzyme is a monomeric protein with a molecular mass of 42 kDa. There is no significant sequence similarity between the two GAR transformylases. However, GAR transformylase-T exhibits 27% amino acid sequence identity with purK-encoded NCAIR synthetase. GAR transformylase-T permits E. coli to use formate for synthesis of the purine ring. Approximately, 50% of the C-8 carbon of the purine ring can be derived from formate, with the remainder derived from serine and glycine via 10-formyl-THF (32, 33). E. coli lacks 10-formyl-THF synthetase and therefore cannot convert formate into 10-formyl-THF (31). Since a purN mutant can grow in minimal medium, E. coli has the capacity to synthesize the formate needed for purine nucleotide synthesis. A gene designated purU is required for production of formate under aerobic growth conditions (108). The purU gene encodes 10-formyl-THF hydrolase, which cleaves 10-formyl-THF into formate and THF (107). The hydrolase is inhibited by glycine, thus explaining growth inhibition of a purN mutant by glycine. Glycine does not inhibit the anaerobic growth of a purN mutant (117), indicating an alternative source of formate for purine nucleotide synthesis. The pyruvate formate lyase reaction provides formate for anaerobic growth (135). This enzyme catalyzes a nonoxidative cleavage of pyruvate to acetyl coenzyme A and formate.

FGAM Synthetase.

5'-Phosphoribosyl N-formylglycinamidine (FGAM) synthetase catalyzes step 4 in the de novo pathway: FGAR + glutamine + ATP ↔ FGAM + glutamate + ADP + P_i. The enzyme was first purified and partially characterized from S. typhimurium (37). More recently, the cloned E. coli purL gene was overexpressed, and the enzyme was purified and more fully characterized (137). FGAM synthetase is a single protein chain with a molecular mass 141 kDa. This enzyme is the second glutamine amidotransferase in the de novo pathway. It contains synthetase and GAT domains of approximately 836 and 256 amino acids, respectively, joined to sequences of undetermined function having the arrangement NH₂-160-synthetase-43-GAT-CO₂H (numbers of amino acids are given for regions of nondefined function) (176). The GAT domain is of the G-type subgroup, as determined from the interior position of the active-site cysteine residue and other sequence motifs. A previously isolated active-site peptide from the Salmonella enzyme (29) contains Cys-1135, the inferred GAT domain active-site cysteine.

Cys-1135 is assumed to participate in covalent catalysis, with the resultant formation of a γ-glutamyl enzyme thioester intermediate (137). The unanswered question, however, is whether Cys-1135 attacks the carboxamide of glutamine for delivery of the amide to FGAR or whether the role of Cys-1135 is to cleave a hypothetical glutaminyl-FGAR imide and release glutamate, as proposed for asparagine synthetase (125). The role of ATP in the reaction is to phosphorylate the amide oxygen of FGAR to facilitate its replacement by the amide of glutamine (137). Initial velocity and dead-end inhibition kinetic studies are most consistent with a sequential mechanism in which binding of glutamine is followed by rapid equilibrium binding of MgATP and then FGAR.

FGAM synthetase is a target for inactivation by azaserine and 6-diazo-5-oxonorleucine (14). These compounds are glutamine affinity analogs that also inactivate glutamine PRPP amidotransferase and GMP synthetase. However, FGAM synthetase appears to be uniquely sensitive to inactivation.

AIR Synthetase.

AIR synthetase catalyzes the fifth step in the de novo pathway: FGAM + ATP → AIR + ADP + P_i. Monofunctional AIR synthetase, purified from the overexpressed purM gene, is a dimer of 38.5-kDa subunits (139), in contrast to the trifunctional vertebrate GAR synthetase-GAR transformylase-AIR synthetase (28). The results of initial velocity and product inhibition studies suggest a sequential mechanism in which ATP binds first to the enzyme and ADP is released last. Because of the irreversibility of the overall reaction and the lack of inhibitory analogs of FGAM, further dissection of the kinetic mechanism was not possible (139). Experiments with ¹⁸O-labeled FGAM established that the role of ATP is phosphorylation of the amide carbonyl group. This reaction, involving the conversion of a C=O bond to a C=N bond concomitant with cleavage of ATP to ADP and P_i, is similar to that for synthesis of FGAM from FGAR.

NCAIR Synthetase.

Mutational analyses had indicated that two genes (47), now designated purE and purK in E. coli (160, 169), are required for the carboxylation of AIR to 5'-phosphoribosyl-5-aminoimidazole-4-carboxylate (CAIR). purK mutants are "CO₂ conditional" purine auxotrophs that require a purine for growth at normal CO₂ levels but are prototrophic in a high-CO₂ atmosphere. Gots et al. (47) therefore suggested that the product of purK functions as a CO₂ carrier to lower the K_m of the reaction for CO₂. The idea of an AIR carboxylase hetero-oligomer in which the PurK subunit functions to deliver CO₂ to the PurE carboxylase subunit prevailed for many years (160). Mueller et al. (103) have now established that PurK catalyzes the ATP-dependent carboxylation of AIR to yield NCAIR (step 6 in the pathway) and PurE isomerizes NCAIR to CAIR (step 7).

NCAIR synthetase and NCAIR mutase (isomerase) were overproduced and purified to homogeneity (101). The two proteins were separated by ammonium sulfate fractionation, and direct physical association was not detected by molecular sieving chromatography or by sucrose gradient centrifugation. NCAIR synthetase is a dimer of 39-kDa subunits. Carboxylation of AIR takes place in two steps. In the first step, the ATP-dependent activation of HCO₃ ^– results in synthesis of enzyme-bound carboxyphosphate. One atom of ¹⁸O from [¹⁸O]HCO₃ ^– is transferred to P_i in this step. In the second step, AIR is carboxylated to NCAIR. NCAIR has a half-life of approximately 0.7 min at pH 7.8 and 30°C, with decreasing stability at lower pH. This lability accounts for an AIR-dependent ATPase activity of NCAIR synthetase (101). Incubation of AIR with high concentrations of HCO₃ ^– results in the nonenzymatic synthesis of NCAIR. At 1.0 M HCO₃ ^– and pH 8.0, approximately 75% of the AIR is rapidly converted to NCAIR. Nonenzymatic synthesis of NCAIR explains the capacity of purK mutants to synthesize CAIR at elevated CO₂ levels.

NCAIR Mutase (Isomerase).

Mueller et al. (103) have named the purE gene product NCAIR mutase or isomerase, depending on the chemical mechanism, which remains to be determined. The homogeneous enzyme, an octamer of 17-kDa subunits (101), catalyzes the reversible interconversion of NCAIR and CAIR. A K_m for NCAIR of ∼250 μM and V _max of ∼100 μmol/min in the physiologically relevant direction were determined. The earlier reported PurE-catalyzed conversion of AIR to CAIR in the presence of high concentrations of HCO₃ ^– (180 mM) is explained by the nonenzymatic synthesis of NCAIR followed by its enzymatic conversion to CAIR.

SAICAR Synthetase.

5'-Phosphoribosyl 4-(N-succinocarboxamide)-5-aminoimidazole (SAICAR) synthetase catalyzes the eighth step in the biosynthesis of IMP: CAIR + aspartate + ATP → SAICAR + ADP + P_i. The cloned E. coli purC gene has been overexpressed, and the enzyme has been purified to homogeneity (101). K_m values of 1.3 mM, 40 μM, and 36 μM were reported for MgATP, CAIR, and SAICAR, respectively, along with methods for the synthesis of CAIR and SAICAR. Properties of the enzyme and any relationships to adenylosuccinate synthetase, which catalyzes a similar transfer of the amino group of aspartate to IMP, remain to be reported.

ASL.

Adenylosuccinate lyase (ASL) catalyzes the ninth reaction in the pathway to IMP and also the final step in the synthesis of AMP: SAICAR ↔ AICAR + fumarate; adenylosuccinate ↔ AMP + fumarate. ASL has been purified from human (153) and mammalian (19) sources as well as from Neurospora (175) and yeast (121) cells but not from enteric bacteria. ASL from B. subtilis was copurified with glutamyl-tRNA synthetase (42). The complex functions to increase the affinity about 10-fold for glutamate and ATP and stabilizes the synthetase against heat inactivation. An apparently similar activator was observed earlier for E. coli glutamyl-tRNA synthetase, but the copurification was not reproducible and the activator was not characterized (discussed in reference 42). An interaction between these enzymes, if confirmed, would imply a role of ASL in linking the synthesis of purine nucleotides and proteins.

AICAR Transformylase-IMP Cyclohydrolase.

AICAR transformylase and IMP cyclohydrolase catalyze the penultimate and final steps, respectively, in the de novo pathway to IMP. These activities constitute a bifunctional enzyme encoded by E. coli purH (1, 36). The reactions are AICAR + 10-formyl-THF ↔ 5'-phosphoribosyl-4-carboxamide-5-formamidoimidazole (FAICAR) + THF and FAICAR → IMP + H₂O.

The homogeneous native avian enzyme (104), a recombinant (112) from chicken cells, and a purified human AICAR transformylase (3) have been reported, but the enzyme from E. coli has not been described. Sequences of cloned genes indicate that the enzymes from E. coli (1, 36) B. subtilis (34), chickens (112), and humans (L. Ni, H. Zalkin, and J. E. Dixon, unpublished data) are all bifunctional and likely homologous. A definitive assignment of AICAR transformylase and IMP cyclohydrolase domains in the PurH protein chain has not been made, although Aiba and Mizobuchi (1) have interpreted complementation data in terms of an NH2-IMP cyclohydrolase-AICAR transformylase-CO₂H model.

Relationship of 10-Formyl-THF Hydrolase to One-Carbon Metabolism

PurU-encoded 10-formyl-THF hydrolase provides the formate required by GAR transformylase-T for synthesis of FGAR (see above). However, this enzyme has a more central role in regulating one-carbon metabolism. In the absence of purU, addition of adenine plus methionine to the growth medium starves cells for glycine (108). Starvation for glycine results from limitation of THF that is needed for the reaction catalyzed by serine transhydroxymethylase. 10-Formyl-THF hydrolase functions to balance the C₁-THF/THF ratio and ensure that glycine synthesis can occur when the main biosynthetic pathways that regenerate THF from C₁-THF are turned off. The activity of 10-formyl-THF hydrolase is modulated by the ratio of C₁-THF to THF. Enzyme activity is stimulated by methionine (signal for excess C₁-THF) and inhibited by glycine (signal for excess THF) (107).

SYNTHESIS OF AMP AND GMP FROM IMP

Gene Organization and Regulation

Steps for conversion of IMP to AMP and GMP are used for de novo synthesis as well as in the salvage of purine bases. Two genes, purA and purB, are required for the reactions to AMP, and guaB and guaA are required for synthesis of GMP (Fig. 1). purB is also required for step 9 in the de novo pathway to IMP. purA is monocistronic, whereas guaB and guaA are in a bicistronic guaBA operon (Table 1 and Fig. 2). Promoter sequences conform with those for genes involved in synthesis of IMP, exhibiting similarity to the consensus Pribnow box and rather poor correspondence with the consensus –35 element (Fig. 3). There are two features that distinguish the regulation of purA and guaBA from that of the genes dedicated to synthesis of IMP. First, purA and guaBA are subject to dual regulation. PurR is one component of the regulatory system, but additional regulatory components are required. The approximately two- to threefold PurR-dependent regulation of purA (Fig. 2) is augmented twofold by an uncharacterized adenine-specific control (55). Two purA operators are utilized for the regulation by PurR. For guaBA, 5-fold regulation by PurR (Fig. 2) is further modulated up to 15-fold by DnaA (156). Regulation of guaBA expression by DnaA requires the presence of two DnaA boxes, one about 220 bp downstream of the translation start codon and a second overlapping the –10 promoter site. High levels of DnaA repress guaBA expression. Repression of guaBA by DnaA may serve to coordinate nucleotide biosynthesis with DNA replication.

A second feature that distinguishes purA, purB, and guaBA from genes dedicated to the synthesis of IMP is the extent of regulation by PurR. purA, purB, and guaBA are subject to 2- to 5-fold repression mediated by PurR, in contrast to 11- to 17-fold repression of the other genes (Fig. 2). This regulatory pattern may be designed to permit purines derived from RNA turnover or from the medium to shut down de novo synthesis of IMP and yet allow synthesis of AMP and GMP by salvage reactions.

Enzymes

Adenylosuccinate synthetase and ASL are required for the two- step synthesis of AMP from IMP via succinyl-AMP (Fig. 1). The reaction catalyzed by adenylosuccinate synthetase, IMP + aspartate + GTP ↔ succinyl-AMP + fumarate + GDP + P_i, is formally similar to that of SAICAR synthetase, step 8 in the de novo pathway, although sequence identities indicative of conserved binding sites for aspartate and nucleoside triphosphate have not been noted. E. coli adenylosuccinate synthetase has been purified following overexpression of purA (6), and an X-ray structure has been determined (120). The enzyme is a dimer of identical 48- kDa subunits. The active site is tentatively identified by residues in a crevice that are important for GTP binding and catalysis and by amino acids at the interface between subunits of the dimer. Steady-state kinetic analysis indicates a sequential mechanism with random binding order for substrates and products (151). The suggested mechanism involves the intermediate formation of 6-phosphoryl-IMP, which is subject to nucleophilic attack by the amino group of aspartate. Several nucleotides, including AMP, GMP, and ppGpp, inhibit enzyme activity in vitro. The most potent inhibition is by ppGpp (K_i = 50 μM), suggesting that the stringent response accompanying amino acid starvation might decrease the rate of AMP synthesis, in part, by inhibition of adenylosuccinate synthetase (40, 151).

IMP dehydrogenase and GMP synthetase catalyze the two- step conversion of IMP to GMP according to reactions shown in Fig. 1. IMP dehydrogenase, purified from E. coli (44), is a tetramer of 58-kDa subunits. Inhibition by GMP (K_i = 55 μM) is competitive with IMP (K_m = 11 μM). A regulatory role for this inhibition has not been established.

E. coli GMP synthetase was reported to be a dimer of 58.6-kDa subunits (133), although X-ray diffraction analyses are consistent with a tetramer of D₂ symmetry in the crystallographic asymmetric unit (157). GMP synthetase is the third glutamine amidotransferase in the pathway for synthesis of purine nucleotides (176). It contains a G-type GAT domain, residues 1 to approximately 200, fused to a synthetase domain. As with other amidotransferases, NH₃ can replace glutamine as an N donor. Glutamine affinity analogs such as 6-diazo-5-oxonorleucine inactivate the enzyme by alkylation of active-site cysteine 86. A kinetic mechanism with ordered addition of MgATP, XMP, and NH₃ was derived from initial velocity patterns and positional isotope-exchange experiments (168). The mechanism is consistent with the partial reactions in equations 1 and 2:

XMP + ATP ↔ adenyl-XMP + PP_i(1)

adenyl-XMP + NH₃ → GMP + AMP(2)

GMP synthetase is the target of psicofuranine, an adenine glycoside antibiotic (39, 182). Although earlier work provided evidence that psicofuranine blocks formation of the adenyl-XMP intermediate (39), more recent experiments support the idea that psicofuranine inhibits the reaction of adenyl-XMP with NH₃ (168).

PURINE REPRESSOR

The purine repressor (PurR) is the primary regulatory protein that coordinates expression of pur regulon genes to the availability of purines. There may, however, be other proteins that sense availability of purine bases, nucleosides, or nucleotides for regulating expression of purA (55) and gcv (173). A putative regulatory protein from S. typhimurium (47) and E. coli (78) that exhibited ATP- or GTP-dependent binding to plasmids containing cloned pur genes has been described but not characterized.

Cloning of purR

Plasmids containing E. coli purR were isolated by functional complementation of different purR mutants (98, 128). On a high- copy-number plasmid, expression of purR from its own promoter starves cells for purines and severely slows growth. Expression of purR from a strong promoter in a multicopy plasmid is lethal. The gene is localized at position 1755 on the Kohara restriction map (79), corresponding to min 36 on the E. coli linkage map (75, 128), and encodes a protein of 341 amino acids.

Autoregulation

The purR promoter sequence and transcription start site given in Fig. 3 are based on primer extension mapping (128). Neither the Pribnow box nor the transcription start site is characteristic of typical genes transcribed by σ ⁷⁰ RNA polymerase. Transcription of purR is autoregulated two- to threefold, using operator sites O₁ and O₂, which are situated 103 and 191 bp, respectively, downstream of the transcription start site (99, 128, 130). At low concentrations, PurR binds to operator O₁ in vitro; at higher concentrations, O₂ is saturated. According to a current model, PurR binds independently to O₁ and O₂, and both operators are required for the full two-to threefold autoregulation (130).

Identification of Corepressors

Houlberg and Jensen (62) initially reported that the presence or formation of hypoxanthine and guanine is required for pur gene regulation in S. typhimurium. Identification of corepressors was confirmed by in vitro assay of PurR-operator interaction. Binding of crude (99) or homogeneous (129) PurR to operator DNA is dependent upon hypoxanthine or guanine. Adenine, pyrimidines, and purine nucleosides and nucleotides do not function as corepressors. The requirement for corepressor depends on the assay conditions. Repressor-operator interaction is not dependent on corepressor in a low-ionic-strength buffer, whereas corepressor-dependent binding to operator DNA is seen in a more physiological higher-ionic-strength buffer containing potassium glutamate (129). However, even in the latter case, corepressor-independent specific binding to operator DNA can take place, particularly at higher concentrations of repressor (22).

Properties of PurR

Overexpressed PurR has been purified to homogeneity by two column steps (24, 129). PurR is a dimer of 38-kDa subunits. The 341-amino-acid subunit contains two domains, an N-terminal DNA binding domain of ∼60 amino acids including a 9-amino acid hinge helix joined to a C-terminal domain having functions for corepressor binding and dimerization. The hinge helix is characterized by a protease-sensitive central core sequence. Digestion with trypsin or chymotrypsin leads to specific cleavage after Arg-52 or Leu-54, respectively, and release of the two intact domains (24, 25). The corepressor binding domain has limited amino acid sequence identity (49, 91, 142) and secondary structure similarity (140, 141) with a subgroup of periplasmic sugar binding proteins for arabinose, glucose-galactose, and particularly ribose. The inferred structural similarity of PurR with sugar binding proteins of defined structure (11, 141) is indicative of homology. Furthermore, intact PurR exhibits amino acid sequence similarity and likely homology with the Lac and Gal superfamily of repressors (128, 170). Several amino acid residues that contribute to binding of sugars in the ribose, arabinose, and glucose-galactose binding proteins are conserved in PurR and are also involved in binding purine corepressor, thus further supporting the evolutionary relationship between these proteins (22). Single amino acid replacements that disable corepressor binding abolish the corepressor-dependent PurR-operator interaction, but corepressor-independent binding to operator DNA is retained. These results have led to a model in which there are two routes to form the repressor-corepressor-operator ternary complex that functions in transcriptional regulation: (i) aporepressor-corepressor interaction to form holorepressor, followed by binding to operator, and (ii) low-affinity aporepressor-operator DNA interaction, followed by corepressor binding to generate the high-affinity ternary complex. Apparent binding constants of 11 and 156 nM were obtained for repressor interaction with a 30-mer consensus operator DNA fragment in the presence and absence of corepressor, respectively. Both routes to the ternary complex take place in vitro, although there is no evidence to support the existence in vivo of the binary aporepressor-operator complex needed for route ii.

Guanine and hypoxanthine bind cooperatively (Hill coefficient of 1.5) to single PurR sites in each subunit of the dimer, with K_d values of 1.5 and 9.3 μM, respectively (24). Each of the single amino acid replacements led to defective binding of both corepressors, consistent with a single site to which hypoxanthine or guanine can bind.

The three-dimensional structure of a PurR-hypoxanthine-purF operator ternary complex has been solved at 2.7 Å (0.27-nm) resolution by X-ray crystallography (140). The bipartite structure consists of the N-terminal DNA binding domain and a larger C-terminal corepressor binding/dimerization domain, which is strikingly similar to that of the bacterial periplasmic binding proteins. The DNA binding domain contains a helix-turn-helix motif which contacts the major groove in an orientation reversed from that in other helix-turn-helix proteins. Unexpectedly, PurR binding kinks the central CpG step by 45°. This distortion of the DNA structure is elicited by insertion of the two symmetry-related hinge helices into the minor groove. Key to this novel DNA binding mechanism is the partial intercalation of the symmetry-related residues Leu-54 and Leu-54' from each hinge helix, which act as "leucine levers" to pry open the minor groove.

Global Regulation

PurR not only regulates transcription of the 10 operons required for synthesis of AMP and GMP and autoregulates expression of its own gene but also contributes purine-mediated regulation to other genes connected with nucleotide metabolism. These coregulated genes include pyrC (23, 110, 172) and pyrD (167, 172) for de novo synthesis of pyrimidine nucleotides, codBA (27, 75) for cytosine transport and salvage, prs (52) encoding PRPP synthetase, glyA (152) and the gcv operon (173) involved in glycine synthesis, catabolism, and synthesis of one-carbon units, speA required for polyamine synthesis, and glnB encoding a regulatory protein for gln operon transcription and regulation of glutamine synthetase activity (52). The relationships between these genes and nucleotide metabolism are shown in Fig. 5. All of the genes listed in Fig. 5 are coregulated by PurR. There is, however, no direct evidence that coregulation of prs, glnB, glyA, and gcv by PurR influences the production of PRPP, glutamine, glycine, and N ⁵,N ¹⁰-methylene-THF, respectively. Several of the genes shown in Fig. 5 are subject to pathway-specific control in addition to control by PurR. The regulatory circuit shown in Fig. 5 presumably coordinates related functions.

Fig. 5Outline of the PurR regulatory circuit. Pathways are abbreviated for de novo and salvage synthesis of purine and pyrimidine nucleotides. Of the purine bases used for synthesis of nucleotides, Hyp and Gua are boxed to indicate their role as corepressors for PurR. All genes shown are regulated by PurR. Genes enclosed by parentheses are involved in generating intermediates; there is, however, no direct evidence that regulation by PurR influences synthesis of the intermediates. The purR gene is subject to autoregulation. Interaction of spermidine with DNA (155) may link the regulation of speA to nucleotide metabolism.

Source:

It is important to emphasize that the regulatory system in E. coli coordinates de novo biosynthesis of purine nucleotides to the availability of both purine bases and nucleotides. Purine bases are sensed by PurR to regulate gene expression, whereas adenine and guanine nucleotides provide synergistic allosteric regulation of glutamine PRPP amidotransferase, the key regulatory enzyme in the pathway. The availability of amino acids for protein synthesis as reflected by the ppGpp effect and the capacity for DNA synthesis as reflected by the DnaA effect (see above) may also contribute to controlling the biosynthesis of nucleotides.

PURINE INTERCONVERSIONS

Pathways and Regulation

Adenine and guanine nucleotides can be interconverted through the common precursor, IMP. The conversion occurs via separate pathways, since the reactions leading from IMP to AMP and from IMP to GMP are irreversible. Adenine compounds can be converted to guanine nucleotides via two pathways, whereas a single enzyme catalyzes the conversion of GMP to IMP (Fig. 6). These conversions serve to balance the adenine and guanine nucleotide pools and are particularly important when adenine and guanine compounds are available in the culture medium.

Fig. 6Purine salvage and interconversion. The individual enzymes are identified by their gene symbols: add, adenosine deaminase; amn, AMP glycosylase; apt, adenine phosphoribosyltransferase; deoD, purine nucleoside phosphorylase; gpt, guanine phosphoribosyltransferase; gsk, guanosine kinase; guaA, GMP synthetase; guaB, IMP dihydrogenase; guaC, GMP reductase; hpt, hypoxanthine phosphoribosyltransferase; purA, adenylosuccinate synthetase; purB, ASL. Nucleobases: A, adenine; G, guanine; Hx, hypoxanthine, X, xanthine; dRib-1-P, deoxyribose 1-phosphate; Rib-1-P, ribose 1-phosphate. Ribonucleosides and deoxyribonucleosides are identified by R and dR, respectively.

Source:

The enzymatic regulation of the branching from IMP is achieved in different ways. ATP accelerates the pathway leading to GMP and inhibits the synthesis of IMP from GMP. GTP accelerates the synthesis of AMP from GMP and IMP. Both of the enzymes catalyzing the branching from IMP are inhibited by ppGpp; thus, the interconversion reactions are arrested by the stringent response to amino acid starvation. Adenine supplied to the growth medium causes repression of the synthesis of the enzymes catalyzing the branching from IMP to AMP and the synthesis of IMP from GMP, whereas syntheses of the enzymes catalyzing the formation of GMP from IMP and IMP from adenine are induced. When guanine or guanosine is present in the growth medium, the picture is reversed (Table 2) (98, 111).

Conversions Involving Free Adenine

The conversion of adenine to IMP proceeds via the intermediate formation of adenosine, inosine, and hypoxanthine catalyzed by the successive action of purine nucleoside phosphorylase, adenosine deaminase, and purine nucleoside phosphorylase (Fig. 6). The hypoxanthine formed is phosphoribosylated to IMP. This reaction is feedback inhibited by IMP and inhibited by ppGpp. AMP can be converted to IMP through the intermediate formation of adenine. This conversion is initiated by the hydrolytic cleavage of AMP to adenine and ribose 5-phosphate catalyzed by AMP nucleosidase. The enzyme is activated by ATP and inhibited by P_i and is proposed to regulate the cellular level of AMP (84). Mutants defective in either adenosine deaminase, purine nucleoside phosphorylase, or AMP nucleosidase grow normally on minimal media (85, 116).

Purine nucleoside phosphorylases purified from E. coli and S. typhimurium consist of six identical subunits of 23.7 kDa (71) and catalyze the reaction: purine nucleoside + phosphate

The equilibrium constant favors nucleoside synthesis. Adenosine, guanosine, inosine, and their corresponding deoxyribonucleosides are substrates; K_m values for nucleosides range from 0.1 to 0.2 mM, and the K_m for phosphate is 2 mM. In the reverse reaction, K_m values are 0.4 mM for ribose 1-phosphate and 1 mM for adenine. The enzyme operates by a sequential mechanism in which the nucleoside, P_i, and ribose 1-phosphate bind in random order, and the purine base binds after the ribose 1-phosphate (69). The enzyme is encoded by the deoD gene. The nucleotide sequence of the E. coli enzyme has been determined, and the gene is the distal gene of the deo operon (59). The synthesis of purine nucleoside phosphorylase, discussed in more detail below with respect to nucleoside catabolism, is induced 10- to 20-fold by nucleosides in both E. coli and S. typhimurium and 2-fold by adenine (Table 2) (50, 115).

Adenosine deaminase from E. coli has been purified (114). The natural substrates are adenosine (K_m = 75 μM) and deoxyadenosine (K_m = 40 μM). Coformycin is a potent inhibitor of the enzyme (K_i = 0.03 μM). The nucleotide sequence of the add gene from E. coli, encoding adenosine deaminase of 36.3 kDa, has been determined (20). The synthesis of adenosine deaminase is induced 5- to 50-fold by adenine and hypoxanthine in E. coli, whereas it appears to be synthesized constitutively and in low amounts in S. typhimurium (115). Cyclic AMP affects the level of adenosine deaminase; in crp mutants and cya mutants, a two- to fourfold increase over the basal level is observed (116).

AMP nucleosidase is a highly specific enzyme that hydrolyzes AMP to adenine and ribose 5-phosphate. The E. coli enzyme is an oligomer of identical subunits of 53.8 kDa (83). It is activated by ATP, which decreases S_0.5 for AMP 100-fold, down to 90_μM. P_i is an allosteric inhibitor (K_i = 0.2 mM) reversing the effect of ATP (81, 84). The sequence of the gene for AMP nucleosidase (amn) from E. coli has been determined (83). P_i represses the expression of the gene, which can be induced threefold by cyclicAMP in the presence of low concentrations of phosphate and is eightfold derepressed when phosphate is limiting in the growth medium (83).

Conversion of ATP to IMP

When histidine is synthesized de novo, a molar equivalent of AICAR, and hence IMP, is formed. Histidine inhibits the first enzyme of the histidine pathway, ATP phosphoribosyltransferase, and it represses the synthesis of all of the histidine biosynthetic enzymes. Thus, exogenous histidine prevents the formation of IMP from ATP by this route (115). The pathway may also be subject to stringent control (102).

Conversion of GMP to IMP

To be converted to IMP, guanine and xanthine compounds must first be metabolized to GMP because deamination of GMP is the only reaction leading to IMP (Fig. 6).

GMP reductase catalyzes the reductive deamination of GMP to IMP, with NADPH as the electron donor; the K_m for GMP is 0.1 mM (90). The enzyme, purified from both E. coli and S. typhimurium, consists of four identical subunits of 37.4 kDa. The enzyme is specifically inhibited by ATP in physiological concentrations. The inhibition is completely reversed by GTP (90, 115). The sequence of the structural gene (guaC) has been determined (5). The level of GMP reductase is increased up to 10-fold when guanine or guanosine (Table 2) is added to the culture medium, and this induction is prevented by adenine (74, 115). In S. typhimurium, glutamine starvation leads to increased levels of the enzyme (41), and the enzyme is inactivated under conditions of purine starvation (115).

Conversion of AMP and GMP to ATP (dATP) and GTP (dGTP)

Nucleoside monophosphate kinases specific for AMP (45) and GMP (118) have been characterized. The enzymes are active with the corresponding 2'-deoxyribonucleotides, and ATP (K_m = 0.05 mM) is the preferred phosphate donor. The K_m for AMP is 0.03 mM (45). The genes encoding AMP kinase (adk) and GMP kinase (gmk) from E. coli have been sequenced. Both genes encode a protein of 23.6 kDa (13, 43). Mutations within the structural gene for AMP kinases that are temperature sensitive have been identified (89), and some result in defective phospholipid synthesis (46). Only conditional adk and gmk mutants would be expected because both gene products are essential for growth. A single enzyme, nucleoside diphosphokinase encoded by the ndk gene, catalyzes the transfer of the γ-phosphate of any nucleoside 5'-triphosphate to any nucleoside 5'-diphosphate (see chapter 35).

The deoxynucleoside triphosphates dATP and dGTP are synthesized via ADP and GDP under aerobic conditions and directly from ATP and GTP under anaerobic conditions. A single enzyme reduces all four ribonucleotides in each case (see chapter 35).

SALVAGE PATHWAYS

Physiological Function

In E. coli and S. typhimurium, the salvage pathways fulfill several functions. One is to scavenge exogenous, preformed bases and nucleosides for nucleotide synthesis, and another is to reutilize bases and nucleosides produced endogenously as a result of nucleotide turnover. A third is catabolic, whereby the pentose moieties of exogenous nucleosides and the amino groups of adenine compounds are made available as sources of carbon and nitrogen, respectively.

When free bases or nucleosides present in the growth medium are salvaged, the contribution from the de novo pathways to nucleotide synthesis is greatly suppressed (62). This is accomplished by increased feedback inhibition of the de novo pathway and by repression of the synthesis of the de novo enzymes. Accordingly, the endogenous nucleotide pools and probably also the pool of free bases increase whereas the pool of PRPP decreases. The bases taken up are directly converted to the nucleotide level by a phosphoribosylation step that is regulated by feedback inhibition. The nucleosides are predominantly converted to bases, of which the major part is excreted again; the pentose moiety is further metabolized. The fact that the synthesis of all of the enzymes involved in nucleoside catabolism is inducible by nucleosides emphasizes that catabolism is dominating when nucleosides are abundant in the growth medium. Nucleotides present in the growth medium may also be utilized as nucleic acid precursors. However, their utilization requires prior dephosphorylation to nucleosides by periplasmic enzymes.

Large quantities of ribonucleoside monophosphates are formed endogenously mainly as a result of mRNA turnover and are reutilized either directly or after conversion to nucleosides or bases (115). Free adenine is produced as a result of S-adenosylmethionine turnover. The turnover points to a significant role of the salvage pathways even under conditions in which preformed bases and nucleosides are absent from the growth medium. These reutilization reactions are very efficient because normally there is no detectable endogenous pool of free bases or nucleosides (18).

The purine salvage pathways of E. coli and S. typhimurium are shown in Fig. 6. Purine auxotrophic mutants blocked in the biosynthesis of IMP can satisfy their entire purine requirement with a variety of purine compounds (Table 3).

Table 3Growth rates of purE mutants on various purine compounds

Utilization of Purine Bases

The genes encoding the proteins involved in the transport and phosphoribosylation of purine bases are not linked on the chromosome, and they seem to be expressed constitutively.

Transport of Purine Bases.

Transport of exogenous bases is tightly coupled to the metabolic processes that convert them to nucleotides, and free bases do not accumulate. The apparent K_m values for transport of purine bases range from 0.1 to 0.5 μM (132). Two types of proteins are involved in purine base utilization, a transport protein and a purine phosphoribosyltransferase and/or purine nucleoside phosphorylase. E. coli has a high-affinity transport system for adenine energized by proton motive force, with an apparent K_m of 0.1 to 0.2 μM. A mutation causing a defect in this high-affinity adenine transport, purP, has been located on the chromosomal map (17). Comparison of the nucleotide sequence data (15) and the restriction map of the cloned purP gene indicates that an open reading frame designated f445 is purP; another open reading frame located close to purP, f470, shows 74% amino acid sequence identity to purP and may have a function in purine transport (18). Another adenine transport system operates at adenine concentrations above 10 μM, with an apparent K_m of 20 to 60 μM, and transport by this low-affinity system is apparently not energy dependent.

Metabolism of Adenine.

Adenine is converted to AMP by adenine phosphoribosyltransferase specified by apt (58, 61) and to adenosine by purine nucleoside phosphorylase (Fig. 6). Adenine can satisfy the entire purine requirement of a purine auxotroph (Table 3). The conversion of adenine to IMP via the adenosine deaminase, purine nucleoside phosphorylase-dependent pathway is less efficient in S. typhimurium than in E. coli (115). Mutants of E. coli that have acquired adenine deaminase activity have been isolated (77). The syntheses of adenosine deaminase and purine nucleoside phosphorylase are induced by the addition of adenine to E. coli (Table 2), whereas these enzymes appear to be synthesized constitutively and in low amounts in S. typhimurium.

The sequence of the E. coli apt gene has been determined (58). The enzyme is inhibited by AMP and not by ppGpp. Apparent K_m values for adenine vary from 1.3 to 20 μM and 10 to 125 μM for PRPP (61, 146). 4-Carboxamide-5-aminoimidazole is also a substrate, and the product formed, AICAR, is an intermediary compound in purine biosynthesis (Fig. 1).

Metabolism of Guanine, Hypoxanthine, and Xanthine.

The uptake of guanine, hypoxanthine, and xanthine is specifically coupled to the conversion of the compounds to their corresponding nucleotides. All three bases serve as a purine source but with different efficiencies (Table 3). E. coli and S. typhimurium have two 6-oxopurine phosphoribosyltransferases: hypoxanthine phosphoribosyltransferase and guanine phosphoribosyltransferase, encoded by hpt and gpt, respectively. Both enzymes are inhibited by ppGpp. Hypoxanthine phosphoribosyltransferase is primarily involved in hypoxanthine salvage; its activity with guanine is sufficient to salvage guanine for guanine nucleotide synthesis but not for total purine nucleotide synthesis. Guanine phosphoribosyltransferase, on the other hand, can salvage guanine, hypoxanthine, and xanthine for total purine nucleotide synthesis. hpt gpt double mutants are unable to metabolize hypoxanthine or guanine (115). The sequences of the gpt genes from S. typhimurium (GenBank accession number X63336) and from E. coli (124) have been determined. The purified guanine phosphoribosyltransferases from both bacteria have apparent K_m values of about 2.5, 40, and 170 μM for guanine, xanthine, and hypoxanthine, respectively, and a K_m value of 39 to 95 μM for PRPP with guanine as the second substrate. The enzyme is markedly inhibited by GMP (30, 60). The apparent K_m values for hypoxanthine and PRPP are 120 and 200 μM, respectively, for the hpt-encoded enzyme from E. coli (60).

Utilization of Purine Nucleosides

The genes encoding the purine nucleoside salvage enzymes and proteins involved in transport are located at separate loci on the chromosome and are induced by nucleosides, except for the gsk gene encoding guanosine (inosine) kinase and the S. typhimurium add gene encoding adenosine deaminase.

Transport of Nucleosides.

Nucleosides are rapidly metabolized after their entry into the cells. Two high-affinity transport systems located in the cytoplasmic membrane, designated the G system and the C system, exist in E. coli (106). The G system, which transports all nucleosides, is inactivated by mutations in the nupG locus. The C system transports only pyrimidine and adenine nucleosides and is inactivated by mutations in the nupC locus (106). A double mutant defective in both the nupG and nupC genes is unable to grow on nucleosides as sole carbon and energy sources but is still able to use purine nucleosides as purine source (66, 116). Both systems are energized by the proton motive force (106). The antibiotic showdomycin is transported by the C system, and mutants resistant to showdomycin are defective in the nupC gene (105). The nupG gene has been cloned, and its sequence has been determined (105, 171); nupG has been found to encode a 45.3-kDa protein.

The E. coli outer membrane protein Tsx, encoded by the tsx gene, functions as a nucleoside-specific channel and serves as the receptor for colicin K and a number of T-even bacteriophages, including phage T6 (7). The tsx gene encodes a protein of 31.4 kDa (12). The Tsx protein will also transport the antibiotic albicidin (8). The Tsx protein seems to be particularly important for scavenging <0.1 μM concentrations of nucleosides. All nucleosides except cytidine and deoxycytidine are affected in their transport by a tsx mutation. Mutants resistant to phage T6 are able to transport nucleoside (92, 138); mutants resistant to albicidin are defective in nucleoside transport but still sensitive to bacteriophages, indicating that the binding site for albicidin overlaps the nucleoside-binding site, whereas bacteriophages bind to a different part of the transporter molecule (7, 38). The syntheses of three transport proteins are coregulated with the synthesis of the enzymes involved in nucleoside catabolism (12). A fourth transport system encoded for by the xapB gene, which is the second gene in the xapABR operon, has been found in E. coli. This system is active only when xanthosine serves as a carbon source (143).

Metabolism of Adenosine and Deoxyadenosine.

Adenosine and deoxyadenosine are metabolized by two different routes (Fig. 6) and are excellent purine sources (Table 3). The major pathway involves deamination catalyzed by adenosine deaminase, followed by cleavage of the formed inosine (deoxyinosine) to hypoxanthine, catalyzed by purine nucleoside phosphorylase; direct phosphorolysis of adenosine (deoxyadenosine) to adenine is only of minor quantitative importance (115, 116).

Metabolism of Guanosine and Deoxyguanosine.

Guanosine and deoxyguanosine are rapidly degraded to guanine by purine nucleoside phosphorylase; only small amounts of guanosine escape phosphorolysis by being converted directly to GMP by guanosine kinase. Deoxyguanosine can be metabolized in E. coli and S. typhimurium only through phosphorolysis. Both nucleosides serve as a purine source (Table 3). The E. coli gene encoding guanosine kinase (gsk) has been located on the chromosome, and the nucleotide sequence encodes a protein of 43.8 kDa (51). A purine auxotrophic mutant defective in nucleoside degradation, i.e., a purE deoD double mutant, will not grow on guanosine alone. The requirement for adenine nucleotides must be satisfied by addition of adenine. The lack of growth of such a mutant on guanosine can be partially suppressed by gsk mutations that result in the synthesis of an altered guanosine kinase with a lower K_m for guanosine and increased thermostability(66). Complete suppression can be achieved by further mutations in genes specifying enzymes of the de novo purine biosynthetic pathway before purE (Fig. 1), i.e., mutations in purF, purL, purM, or prs. The basis of this phenomenon is not completely understood (65, 66).

Metabolism of Inosine and Deoxyinosine.

Both inosine and deoxyinosine are rapidly degraded to hypoxanthine by purine nucleoside phosphorylase, which explains why both nucleosides are excellent purine sources (Table 3). However, inosine can also be phosphorylated to IMP. A purine-requiring deoD mutant of E. coli and S. typhimurium will grow slowly on inosine but significantly better than on guanosine (66, 115). Inosine kinase is specified by the same gene that specifies guanosine kinase.

Metabolism of Xanthosine.

E. coli, but not S. typhimurium, contains a gene, xapA, specifying xanthosine phosphorylase, also called inosine-guanosine phosphorylase (80), which can use xanthosine, guanosine, inosine, and their corresponding deoxyribonucleosides as substrates; the K_m range is 44 to 340 μM (50, 80). Because of its very low intracellular levels, this enzyme plays no significant role in purine salvage, and xanthosine will not serve as a purine source in cells grown in glucose medium (Table 3).

Nucleoside Catabolism

The catabolism of purine nucleosides is catalyzed by purine nucleoside phosphorylases and adenosine deaminases, although when adenosine is the substrate for purine nucleoside phosphorylase, it is predominantly deaminated rather than phosphorolyzed (115). The ribose 1-phosphate and deoxyribose 1-phosphate liberated in these processes are further converted to intermediates of the pentose phosphate shunt and of glycolysis, respectively, by the action of phosphopentomutase and deoxyriboaldolase (50).

Wild-type cells and nupC mutants can utilize all nucleosides as carbon sources, whereas a nupG mutant cannot grow on guanosine or deoxyguanosine. The coupling of nucleoside transport and catabolism is further emphasized by the observation that the synthesis of the transport proteins and the synthesis of the nucleoside-catabolizing enzymes are controlled by the same regulatory proteins (12, 106). Thus, expression of the nupG, tsx, and deoD genes is regulated by the deoR repressor, the cytR repressor, and the cyclic AMP receptor protein-cyclic AMP complex (105), whereas control of expression of nupC involves only the latter two regulatory proteins (106).

High exogenous concentrations of xanthosine induce the xapABR operon, encoding xanthosine phosphorylase (xapA), a nucleoside transporter protein (xapB), and a regulatory protein (xapR). This induction is dependent on a functional xapR gene product (76).

E. coli can utilize the amino group of adenine and adenosine as a sole nitrogen source, with adenosine deaminase and purine nucleoside phosphorylase as the key enzymes (116).

Utilization of Nucleotides

E. coli and S. typhimurium can use exogenous nucleotides as purine sources. Since the cytoplasmic membrane is impermeable to nucleotides, they must first be dephosphorylated to the corresponding nucleosides by periplasmic enzymes. A number of periplasmic phosphatases capable of hydrolyzing nucleotides have been identified (181). Of four enzymes found in E. coli, only two were found in S. typhimurium (109). These differences are also reflected by the different growth patterns of the two bacteria on nucleotides as purine sources (Table 3) (165). The ushA gene encoding the E. coli UDP-hydrolase (5'-nucleotidase) has been cloned, and its sequence has been determined (16). Mutants (omp mutants) defective in specific outer membrane proteins involved in pore formation are unable to utilize exogenous nucleotides, although they contain normal levels of the periplasmic phosphatases (166). Thus, specific pores in the outer membrane are required for exogenous nucleotides to reach the hydrolytic enzymes.

Purine Analogs

The purine analogs 2,6-diaminopurine and 6-methylaminopurine are toxic at high concentrations (>1 mM), and 2-fluoroadenine is toxic at 0.1 mM, because they are converted to toxic nucleotides by adenine phosphoribosyltransferase. Mutants resistant to 2-fluoroadenine and 2,6-diaminopurine can be isolated and are adenine phosphoribosyltransferase deficient (apt). However, 2,6-diaminopurine and 6-methylaminopurine may also be converted to guanosine and inosine, respectively, by the consecutive action of purine nucleoside phosphorylase and adenosine deaminase. At low, less toxic concentrations (<0.1 mM), these analogs support growth of purine auxotrophs, although with low efficiency (20, 116). By selecting for fast growth on low concentrations of these analogs as the sole purine sources, mutants with increased levels of purine nucleoside phosphorylase (deoR mutants) and adenosine deaminase (addR mutants) have been obtained (116). Guanine phosphoribosyltransferase mutants (gpt) have been isolated only in S. typhimurium, and 1 mM 8-azaguanine was used for selection (73). In wild-type cells, 6-mercaptopurine either can be phosphoribosylated to the toxic 6-mercaptopurine ribonucleoside monophosphate, which inhibits purine biosynthesis, or can react with the PurR protein (99) and repress purine biosynthesis. By selecting for resistance to 2 mM 6-mercaptopurine, mutants defective in hypoxanthine phosphoribosyltransferase (hpt) or purR have been isolated in E. coli and S. typhimurium (73, 75, 86). This observation indicates that inhibition by 6-mercaptopurine can be overcome either by preventing the formation of the toxic nucleotide or by increasing the level of purine biosynthetic enzymes.

ACKNOWLEDGMENTS

We thank JoAnne Stubbe, Janet Smith, and V. Jo Davisson for kindly providing manuscripts prior to publication. We express gratitude to Rod Kelln, Bente Mygind, Bjarne Hove Jensen, and Peter Nagy for cricital comments on the manuscript. Work from our laboratories has been supported by grants from the National Institutes of Health (GM 24658 to H.Z.) and the Danish Centre for Microbiology (P.N.).