Chapter 48

Biosynthesis and Recycling of NAD

THOMAS PENFOUND and JOHN W. FOSTER

INTRODUCTION

NAD⁺ and the derivative NADP⁺ are ubiquitous and are essential for all living systems. Their importance centers most obviously on their use as cofactors in hundreds of oxidation-reduction reactions throughout the biological world (47, 90). NAD⁺ contains a reactive nicotinamide (NAm) ring that accepts H⁺ and electrons during the oxidation of fuel molecules and delivers hydride ions to the cytochrome system (83). NADP is dedicated to reductive biosyntheses rather than catabolic metabolism. Together, these nucleotides have a direct impact on virtually every metabolic pathway in the cell. However, beyond a role in oxidation-reduction reactions, NAD⁺ can also be used as a substrate for the modification of proteins by ADP-ribosyltransferases and for the synthesis and repair of DNA by prokaryotic DNA ligase. The ADP-ribosylation of proteins has been identified in phages, bacteria, and eukaryotic cells, where it fulfills important cellular regulatory functions (68, 69, 87). For example, bacteriophage T4 utilizes ADP-ribosylation to pirate the host Escherichia coli RNA polymerase, converting it to a form with specificity for T4 genes (30, 69, 73). The ADP-ribosylation of host proteins by many bacterial toxins has been well studied in eukaryotic cells, and although ADP-ribosylation is not known to be a significant regulatory feature in Escherichia coli or Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), there are reports of NAD-dependent protein modifications of unknown function in these organisms (64, 72). It is also significant that the biosynthetic route to NAD serves to supply an intermediate, nicotinic acid mononucleotide (NAMN), required for vitamin B₁₂ (cobalamin) synthesis under anaerobic conditions, further extending the role of NAD metabolism in cellular physiology(49).

The broad manner in which NAD and NADP are used by the cell means that even minor fluctuations in the concentrations of these nucleotides have an enormous impact on many aspects of cellular physiology and growth. Thus, one predicts that the intracellular levels of NAD and NADP must be closely monitored and adjusted when necessary. The concentrations of various pyridine nucleotides within these organisms are 0.8 mM NAD⁺, 0.02 mM NADH, 0.05 mM NADP⁺, and 0.15 mM NADPH, with a total pyridine concentration of about 1 mM (5, 86). Consequently, an important facet of NAD metabolism involves monitoring and balancing the concentrations of the pyridine redox cofactors in response to cellular need.

Several earlier reviews discussed details of NAD metabolism that can be discussed here only briefly (13, 23, 57, 61, 78, 84). Those reviews should be consulted for historical perspective. The various enzymes and genes with known or suspected relevance to NAD metabolism are listed in Table 1. Figures 1 and 2 portray the various pathways and intermediates important to the biosynthesis and recycling of NAD.

Fig. 1Pathway for biosynthesis of NAD. The gene locations are those on the E. coli physical map. FAD, flavin adenine dinucleotide; FADH2, reduced flavin adenine dinucleotide; NAAD, NA adenine dinucleotide.

Source:

Fig. 2PNCs. NAmR, NAm riboside.

Source:

Table 1Enzymes involved in NAD biosynthesis and cycling and in pyridine nucleotide transporta

DE NOVO BIOSYNTHESIS

The Aspartic Acid-Dihydroxyacetone Phosphate Pathway

Early in the study of NAD biosynthesis, the pathways used by eukaryotic and prokaryotic organisms were found to be very different. The pathway in eukaryotes was characterized first; it involves a series of steps emanating from tryptophan. Nutritional studies from the 1940s suggested that tryptophan was not a precursor for niacin (nicotinic acid [NA]) in E. coli. By 1954, investigations with NA-requiring mutants of E. coli and Bacillus subtilis demonstrated clearly that tryptophan intermediates would not support growth of these NAD auxotrophs. A classic study by Yanofsky (89) did suggest that quinolinic acid (QA) could support growth of an NAD auxotroph and would represent a common intermediate between the two systems. Since these early studies, many biochemical and genetic details of prokaryotic NAD metabolism have been revealed by using both E. coli and S. typhimurium.

NAD can be synthesized de novo from aspartate through a series of five reactions or can be scavenged from environmental sources. The scavenging or recycling pathways are collectively known as the pyridine nucleotide cycles (PNCs) and are discussed later. The de novo pathway starts with the synthesis of QA by the nadA-nadB enzyme complex (76). The first enzyme in this process, l-aspartate oxidase (the product of nadB), produces iminoaspartic acid from l-aspartate (7, 58, 59, 60). This unstable intermediate decomposes to oxaloacetate and ammonia if quinolinate synthetase, the product of nadA, is unavailable to catalyze the condensation between the iminoaspartate and dihydroxyacetone phosphate (85). The intracellular efficiency of this system predicts that these enzymes will form a complex to allow optimal use of iminoaspartate. Synthesis of the NadAB enzymes is tightly controlled, with the transcription of both nadA and nadB being subject to repression by NadR (see below).

NadB is a flavoprotein subject to feedback inhibition by NAD and so is a key regulatory focal point of NAD synthesis (59). The nadB sequence of E. coli encodes a 60,306-Da protein with a possible flavin adenine dinucleotide-binding site at amino acid 15 (GXGXXG) that is also a potential NAD-binding site (14). Feedback inhibition of NadB may involve competition between NAD and FAD for this potential binding site. Mutations within nadB that appear to confer a feedback-insensitive phenotype on the gene product have been identified (10, 39). However, neither the identity of the mutations nor the affected amino acid residues have been determined.

The sequence of nadA is known for both S. typhimurium and E. coli (14, 24). The gene from S. typhimurium encodes an open reading frame 40 amino acids longer at both the amino and carboxyl ends of the protein than the open reading frame from E. coli (24, 76, 77). Both NadA proteins contain a sequence motif characteristic of iron-sulfur clusters (Cys-W-X-Cys-Y-Z-Cys) at amino acid positions 249 to 255 for E. coli and 291 to 297 for S. typhimurium. The condensation step catalyzed by NadA requires a dehydration reaction that in several other dehydrases involves (Fe-S)_n groups. It is believed that the presence of this Fe-S group explains why the de novo pathway is sensitive to molecular oxygen (26). Iron-sulfur clusters in several enzymes are reportedly sensitive to oxidative attack, with oxidation causing the release of iron and the formation of a thiocystine bond. A similar reaction may occur with NadA. In terms of the kinetics of QA production, the quantities of NadA and NadB are equally rate limiting (14). Thus, when E. coli is transformed with plasmids containing either nadA or nadB alone, there is little increase in QA production. Increased levels of QA are produced only if both genes are present in high copy number.

The QA produced by NadA and NadB is subsequently decarboxylated and converted to a nucleotide form via QA phosphoribosyltransferase (QAPRTase), which is encoded by nadC (3, 22, 43, 79, 80). The product of this reaction, NAMN, is the convergence point for the de novo and recycling pathways. The nadC locus is not transcriptionally regulated, nor is its product subject to feedback inhibition (35, 71). The gene encodes a 297-amino-acid, 32,428-Da protein that appears as a dimer upon native gel filtration (40).

The final series of reactions in the synthesis of NAD, catalyzed by NadD and NadE, are essential for cell viability. The NAMN produced by recycling or de novo synthesis is adenylylated by NadD (NAMN adenylyltransferase), producing NA adenine dinucleotide, which is then amidated by NadE (NAD synthetase) to form NAD (41, 42, 66, 67, 75). A major difference between NAD synthetases from bacteria and yeast cells is in the donor amide specificity. Bacterial enzymes prefer NH₃, whereas yeast cells can use either NH₃ or glutamine (91). The native NAD synthetases of E. coli and S. typhimurium exhibit distinctly different electrophoretic mobilities in agarose gels, suggesting significant differences in their structures, although the sequences of the genes have yet to be determined. It is thought that NAD synthetase functions as a multimer, since the mixing of extracts from both organisms produces an active hybrid NAD synthetase of intermediate mobility (42).

Mutant Phenotypes

Mutations in any member of the de novo biosynthetic branch of the pathway (nadA, nadB, or nadC) cause an auxotrophy that can be satisfied by NA, NAm, or NAm mononucleotide (NMN). Mutants defective in nadA or nadB but not nadC can also grow on QA. Identification of the nadD and nadE loci of the essential pathway required a more clever selection strategy (39, 41, 42). Temperature-sensitive mutations that at the permissive temperature (30°C) produced products of low activity were selected. Since only minimal levels of NAD were produced, these mutants were derepressed for a nadB-lacZ operon fusion even in the presence of excess NA. These mutations were subsequently shown to be lethal at the nonpermissive temperature (42°C), indicating a defect in an essential metabolic pathway. Another strategy used to isolate nadD mutants relied on the accumulation of NAMN to dilute the toxic effects of the analog 6-amino NAm (6-ANAm). This analog inhibits growth following its conversion to 6-amino NAMN and ultimately 6-amino NAD, which is toxic to the cell.

PNCs

NAD Turnover

NAD-dependent oxidation-reduction reactions do not consume NAD. Nevertheless, the half-life of cellular NAD under aerobic growth conditions is 90 min regardless of the generation time (64). The reason for this turnover is not entirely clear, but once NAD is degraded, pathways designed to recoup the pyridine moiety would certainly be advantageous. The existence of a PNC was first proposed for a variety of species by Gholson in 1966 (28). Subsequently, detailed characterizations carried out with S. typhimurium and E. coli demonstrated multiple forms of this cycle. All of the PNCs in these organisms start with the enzymatic degradation of NAD and proceed through different pathways to the synthesis of NAMN, the pyridine nucleotide precursor common to the NAD salvage and biosynthetic pathways. The cycle is complete when NAMN is converted into NAD via the two final steps of the de novo pathway (66, 67). An important feature of the salvage pathways of E. coli and S. typhimurium is the energetically favorable conservation of the pyridine ring. When recycling is possible, it takes precedence over de novo synthesis (56). The existence of various NAD-recycling pathways consisting of four-, five-, and six-membered cycles, referred to as PNCs IV, V, and VI, respectively, has been supported by demonstrating the required enzymatic activities. PNCs IV and V function predominantly for intracellular recycling, while PNC VI plays an important role in scavenging preformed pyridines from the environment.

The majority of intracellular NAD recycling begins with hydrolysis of the PP_i bond, releasing NMN and AMP (34, 55). Of the two cycles that include NMN (PNCs IV and VI), in vivo recycling experiments indicate that PNC IV is the major intracellular recycling pathway, with 80 to 90% of intracellular recycled pyridine traveling this route (2, 16, 34, 55). One enzyme that can degrade NAD to NMN is prokaryotic DNA ligase, which utilizes the energy released from the PP_i bond to reseal nicks in double-stranded DNA. This reaction is important in the repair, recombination, and replication of DNA. Predictably, the rate of NAD turnover decreases in a ts DNA ligase (lig) mutant at the nonpermissive temperature (55). The lig gene, however, is not subject to any known transcriptional control (64). The degradation of NAD by DNA ligase was once thought to be one of the primary causes of intracellular NAD recycling, but this possibility has been disproven. Cells in which NAD-dependent DNA ligase is replaced with T4 phage ATP-dependent DNA ligase do not exhibit any significant decrease in NAD turnover (64). However, the NAD turnover rate is much higher (three- to fourfold) under aerobic growth conditions than under anaerobic conditions. The faster aerobic rate of NAD recycling may be related to some unknown NAD-dependent enzyme(s) that protects membranes or DNA from oxygen damage (64). Consequently, the enzyme(s) responsible for initiating most intracellular NAD degradation has not been identified. The membrane-associated NAD pyrophosphatase (pnuE), which is required for extracellular degradation of NAD, does not appear to play a role in intracellular NAD turnover, as demonstrated by pnuE deletion studies (65). Clearly, NAD utilization as a substrate in other "unknown" reactions must initiate the recycling events. Park et al. (64) and Skorko and Kur (72) note that aside from DNA ligase, which is adenylated, several proteins are covalently modified by NAD and may be responsible for initiating turnover.

PNC V begins with NAD glycohydrolase degrading NAD to NAm and ADP-ribose. Although this activity has not yet been identified directly in E. coli, two forms have been detected in partially purified extracts of wild-type S. typhimurium and in crude extracts of lig mutants (17, 64). The two forms have molecular weights between 53,000 and 58,000. It is unclear whether these enzymes bear any relationship to the ADP-ribosyltransferase activity reported for E. coli (72). If the glycohydrolases are also ADP-ribosyltransferases, then turnover via PNC V could be the result of unrecognized cellular functions mediated by these proteins.

The second step of PNC IV, NMN amidohydrolase (pncC), deamidates NMN to generate NAMN (34, 50). The enzyme from E. coli has a molecular size of 33 kDa, a pH optimum of 9.0, a K_m of 1.35 × 10^–5, and linear kinetics and is specific for NMN (34). The S. typhimurium enzyme has a pH optimum of 8.7 and is present in the cell at 20 to 30 times the level of NMN glycohydrolase (50). This reaction as well as the final two steps of all PNCs, NadD and NadE, are not known to be under any regulation. Therefore, the constitutive nature of PNC IV could explain its prominent contribution to NAD turnover relative to that of PNC VI (15, 16, 34).

The second step of PNC VI produces NAm through the action of NMN glycohydrolase. This enzyme is found in both cytoplasmic and membrane-associated forms (4). NMN glycohydrolase from S. typhimurium has a molecular mass of 67 kDa with a pH optimum of 8.0 and may serve as a control point directing the flow of pyridine between PNC IV and PNC VI. This proposed control function is based on its feedback inhibition by NAD(H), GTP, GMP, AMP, and ADP-ribose (15).

NAm is converted to NA by the constitutively expressed NAm amidohydrolase (NAm deamidase), which is encoded by pncA (11, 20, 45, 63). The E. coli and S. typhimurium enzymes have estimated molecular sizes of 30 and 35 kDa, respectively (15, 63). The pncA locus appears to reside within an operon downstream of an undefined locus designated pncX (33, 39). Insertions into pncX decrease but do not eliminate expression of pncA, suggesting the existence of a weak internal promoter. The amount of NAm amidohydrolase produced in a pncX mutant is sufficient to support growth on NAm but not to convert 6-ANAm to an inhibitory concentration of 6-amino NA. Thus, pncX mutants are 6-ANAm-resistant.

The next member of both the PNC V and VI pathways is NA phosphoribosyltransferase (NAPRTase; 20, 21, 44), which links the two recycling systems to the de novo pathway through the formation of NAMN. The flow of pyridine through these PNC V and VI systems is also subject to regulation via repression of the pncB locus (20, 46). Thus, when internal NAD levels are too high, the cell can excrete excess pyridine in a form (NA) that can later be scavenged. The pncB gene has been cloned from E. coli and S. typhimurium, and the sequences have been determined (82, 88). The genes encode 46-kDa proteins that are highly homologous. It is interesting that the reactions catalyzed by QAPRTase and NAPRTase are similar, with both resulting in NAMN formation, yet the enzymes show very little amino acid homology (40, 82). A small stretch of homology was found between amino acids 13 through 48 of QAPRTase and 264 through 298 of NAPRTase, although the role of this sequence is unknown. Furthermore, even though both enzymes utilize 5-phosphoribosyl-1-pyrophosphate (PRPP) as a substrate, neither has a consensus PRPP-binding site such as those identified in other phosphoribosyltransferases (38). NAPRTase also differs from QAPRTase and many other phosphoribosyltransferases in its requirement for ATP (47). Vinitsky and Grubmeyer (81) proposed that this enzyme utilizes a novel mechanism for energy coupling in which NAMN formation is linked to ATP hydrolysis. In this model, the enzyme becomes phosphorylated by an intrinsic ATPase activity. The phosphorylated enzyme then binds PRPP and NA. After the phosphoribosyltransferase event, NAMN and PP_i and then P_i are released. The presence of ATP dramatically changes the interaction of NAPRTase with its substrates. The K_ms for both PRPP and NA are lowered 200-fold and V _max catalysis is increased 10-fold in the presence of ATP. Vinitsky and Grubmeyer’s studies also showed that the NAPRTase reaction is reversible when ATP levels are exhausted. This confirms a predicted in vivo reversibility suggested earlier by NAD-recycling experiments in which S. typhimurium was starved for carbon source (16). The physiological significance of energy coupling in NAPRTase is not apparent, but during periods of energy stress, competition for PRPP could be high, and the need to synthesize NAD could be low. Situations leading to the depletion of ATP and thus slow growth would spare PRPP from entering an unnecessary pathway. The use of ATP to pull PRPP into synthesis of a key energy-acquiring molecule like NAD would be valuable to a rapidly growing cell.

Synthesis of NADP

The interconversion of NAD and NADP is not related to pyridine scavenging but forms a cycle nonetheless. NAD is phosphorylated to NADP via NAD kinase, with the reverse reaction catalyzed by NADP phosphatase. Two NAD kinase genes have been reported for S. typhimurium; either one is sufficient for growth on complex media, but both are required on minimal media (7a). During exponential growth, E. coli maintains an NADP/NAD ratio of 0.33 by balancing de novo NAD synthesis with the synthesis and degradation of NADP (53). Starvation for NA, however, drastically alters that ratio to >2.0. Although there is a decrease in total pyridine concentration (NAD + NADP) down to 2% of prestarvation levels, it seems that NAD is sacrificed in order to preserve NADP levels (54). This finding suggests some form of control over NADP phosphatase that inhibits its activity during starvation. Similar results have been noted for S. typhimurium (36).

Transhydrogenase

As noted above, NAD(H) is used for the catabolic activities of the cell, whereas anabolic metabolism specifically requires NADP(H). Reducing equivalents can be exchanged between NAD and NADP via a membrane-bound transhydrogenase encoded by two genes, pntA and pntB (8). The enzyme translocates protons across the bacterial membrane according to the following equation:

nH⁺ _in + NADPH + NAD ↔ nH⁺ _out + NADP + NADH

The enzyme is composed of two subunits, α (M _r, 54,000) and β (M _r, 48,700), organized as an α ₂ β ₂ tetramer(8, 37). This transhydrogenase must contribute significantly to maintaining the proper balance between anabolic reduction charge (NADPH/NADPH + NADP⁺⁾ and catabolic reduction charge (NADH/NADH + NAD⁺)(1). These reduction charges are important for optimizing the biosynthetic and catabolic activities of the cell. The anabolic reduction charge has also been implicated in the transcriptional control of redox-sensitive genes (e.g., sodA, which encodes superoxide dismutase [27]).

Pyridine Analog Resistance and Sensitivity

Various analogs of pyridine nucleotides have been extremely useful in the analysis of NAD metabolism. Mutants defective in NAm amidohydrolase (pncA or pncX) can be identified on the basis of their resistance to 6-ANAm and sensitivity to 6-amino NA. NAPRTase mutants can be selected by resistance to both 6-ANAm and 6-amino NA (9, 21, 39, 51). As noted above, certain mutations in nadB and nadD can also lead to analog resistance either by the diminishing of feedback inhibition (nadB) or through the accumulation of NAMN (nadD). Pyridine analog supersensitivity (pas) mutations have also been identified. These mutations are proposed to affect gene products that utilize NAD as a cofactor or that supply precursors for NAD synthesis (18). One mutation conferring a Pas phenotype occurs within the gene for glucosephosphate isomerase (pgi). The defective enzyme presumably diminishes the supply of dihydroxyacetone phosphate available for NAD synthesis, thereby reducing the amount of analog needed to inhibit growth (48). In support of this theory, mutations that suppress the pas mutant phenotype were isolated at NAD synthesis loci pncB, nadB, and nadD (18).

TRANSPORT

E. coli and S. typhimurium can scavenge from their environments a variety of exogenous pyridines, including QA, NA, NAm, NMN, NAm riboside, and NAD (50, 52, 56). Coupled with the recycling pathways, the preferential uptake of exogenous pyridines over their de novo synthesis provides a significant energy savings for the cell (45, 71). NAD is not transported directly by these organisms but is first degraded to NMN and AMP by a membrane-associated NAD pyrophosphatase, which is encoded by pnuE (12, 21, 29, 65). The active site for this enzyme is on the outer face of the cytoplasmic membrane and does not contribute to internal recycling (12, 29, 65). The NMN produced from extracellular NAD can be taken up by the cell in two demonstrable ways: it may be transported intact via the NadR-PnuC transporter complex, or it can be degraded by a membrane-bound NMN glycohydrolase, with the resulting NAm entering the cell (4, 24, 52, 95). An NAm-binding protein has been demonstrated in E. coli (32), but true active transport has not been proven for either NAm or NA. There are no significant intracellular pools of PNC intermediates. Therefore, transport of preformed pyridines is dependent on their intracellular conversion to NAD, giving the appearance of active transport. Another transport mutant defective in the PNC VI-dependent NMN uptake system required elevated levels of NMN but not NAm for growth (74). The locus, designated pnuD (pyridine nucleotide uptake), did not appear to affect NMN glycohydrolase activity and so presumably acts at another stage of PNC VI-dependent NMN transport. PNC IV-dependent uptake is discussed below.

In general, the preferred route of pyridine uptake is through PNC VI. Ten times more NMN is required to achieve growth of a pncA mutant, in which PNC VI is blocked, than is needed for a pncA ⁺ strain(21, 74). This finding suggests that PNC IV-dependent transport of NMN or NAD occurs only at high external pyridine nucleotide concentrations.

Like NAm, NA is also exclusively scavenged by PNC VI via the pathway originally defined by Preiss and Handler (66, 67). NA uptake by E. coli is most likely a non-energy-dependent diffusion process (56, 62). An energy requirement for NA transport demonstrated by Rowe et al. (70) probably reflected the sequential action of the NadD and NadE proteins required to convert internalized NA to NAD. Rowe et al. (70) did demonstrate that NA, a weak acid, is taken up best at mildly acidic pH₀, a condition that would increase the concentration of protonated NA. It is generally assumed that protonated weak acids can penetrate lipid bilayers, whereas ionized forms cannot.

The only specific pyridine transport system with known protein components consists of a complex containing an integral membrane protein, PnuC (74), and a peripheral accessory protein, NadR (10, 20, 21, 36). This system transports the phosphorylated compound NMN intact across the cell membrane, as shown by NMN double-labeling experiments in which the ³²P- and ³H-NAm moieties of NMN remained linked during transport (52). It is this system that is required for the PNC IV-dependent utilization of NMN. The pnuC structural gene lies downstream of nadA within an operon (24, 93). Although the nadA-pnuC operon is repressed by elevated levels of NAD, an internal promoter allows for basal expression of pnuC (24, 74, 93). The PnuC and NadR proteins are both necessary for the transport of NMN (50, 74, 93, 95). However, the identification of PnuC point mutations (pnuC*) that exhibit NadR-independent NMN transport supports a central role for PnuC in the process (95; T. Penfound and J. W. Foster, unpublished observation). The normally inactive PnuC transporter appears to be activated in an undefined way by NadR when NAD levels are low (95). The PnuC^* protein does not require activation by NadR and will mediate NAD-independent transport of NMN. A membrane location for PnuC is supported by the predicted amino acid sequence and by the isolation of active PnuC-PhoA fusions(95; J. W. Foster and B. L. Bearson, unpublished data). Another locus involved with pyridine nucleotide transport, pnuB, was identified as a mutant displaying increased NMN uptake (52). NMN transport in a pnuB mutant is still PnuC dependent but does not require the NadR activator. The function of wild-type pnuB is still unknown, although the locus lies near nadR (74).

REGULATION OF NAD BIOSYNTHESIS AND RECYCLING

The nad Regulon

Two pathways for NAD synthesis must be controlled for effective regulation of NAD synthesis: the de novo route and the PNC scavenging system. The cell accomplishes this goal in several ways. Key enzymes of both pathways are subject to allosteric control, the transcription of specific nad and pnc genes is regulated, and the transport of preformed compounds is modulated. Allosteric or feedback control has been shown for NMN glycohydrolase (15) and l-aspartate oxidase (nadB) (7, 31).

Transcriptional control has been demonstrated for three NAD-related genes through the construction of lacZ operon fusions. Two of the regulated genes, nadA and nadB, encode QA synthetase in the de novo pathway. The other member of this regulon, pncB, encodes NAPRTase of the PNC. The nadA and nadB loci are more tightly controlled than pncB (19, 71). The fact that nadA and nadB are repressed to a greater degree than pncB is consistent with a cellular preference for utilizing preformed pyridine compounds over expending energy to synthesize NAD de novo. Data correlating total cellular pyridine nucleotide content with the induction of nadA and nadB revealed that these genes become active as total NAD(P) levels fall below 1.0 mM (36).

Sensing the intracellular NAD concentration is a crucial prerequisite for regulating NAD synthesis. As stated above, the system must sense NAD(P) levels around 1 mM. A regulatory gene designated nadR (nadI by others) that encodes a negative regulator of the genes nadA, nadB, and pncB has been discovered (10, 19, 36, 74, 92). A null mutation in nadR causes the constitutive overexpression of these three target genes. Consequently, nadA, nadB, and pncB form what is called the nad regulon. The developing story of how NadR controls NAD biosynthesis has revealed an intriguing protein with unique regulatory features.

The nadR Locus Encodes a Bifunctional Regulator

An important insight into the regulation of NAD biosynthesis was first obtained when the transport of the preformed pyridine nucleotide NMN was examined. As noted above, NMN is among the few phosphorylated compounds that can be transported intact across the membrane (52). Mutations that prevent the uptake of NMN were discovered in two genes, originally called pnuA and pnuC. A curiosity of the pnuA mutations was that some, but not all, also displayed an NadR^– (regulation-defective) phenotype. Likewise, many nadR mutants displayed a PnuA^– (transport-defective) phenotype. This finding suggested two possibilities. Either nadR and pnuA were separate members of one operon, or there was only one gene whose product was both a transcriptional regulator and a component of a nucleotide transport system. The identification in nadR (NadR^s phenotype) of superrepressor mutations that also reduced NMN transport supported the idea that one protein fulfilled both functions (19, 95). Four main nadR phenotypes are now described as R^–T^– (regulation and transport defective), R⁺T^–, R^–T⁺, and R^sT^+/–. The discovery of R^sT^– mutants was the first indication that the NadR product might assume two conformations, one suitable to its role as a repressor and the other more appropriate to fulfilling a transport requirement. The developing model states that superrepressor mutations would lock NadR into a repressor conformation and out of a transport activator form.

Zhu et al. (94) performed an elegant series of genetic and in vivo transport studies that further support the bifunctional repressor-transporter roles of NadR (designated nadI by them) and suggest that NadR directly modulates the NMN transport activity of PnuC rather than being a direct participant in the transport process (94, 95). A key finding was that the modulation of NMN transport by NadR in response to NAD concentration was independent of protein synthesis. Thus, NadR does not regulate the expression of some other, unidentified transport gene. It must participate as a part of the transport process. However, the identification of pnuC mutations that render PnuC transport activity independent of NadR (94; J. W. Foster and T. Penfound, unpublished data) indicates that NadR is not an essential component of NMN transport.

Definitive evidence for the bifunctional model was obtained once the nadR locus was cloned and its sequence was determined (19, 24). The nadR locus encodes a 47,022-Da protein that alone will complement both the repressor and transport defects of Δ nadR mutants. Subsequently, the technique of single-strand conformational polymorphism (SSCP) was used to map chromosomal point mutations specifically in different areas within the nadR open reading frame (25). Labeled PCR-amplified fragments from chromosomal DNA containing nadR mutations were denatured into single strands and then separated on nondenaturing polyacrylamide gels. With SSCP, single-stranded fragments containing mutations often exhibit altered mobilities relative to those of the control fragments. Mutations causing R^–T^– phenotypes were scattered in all three fragments. However, mutations causing the R^–T⁺ phenotype clustered toward the 5' end of the gene, whereas the one R⁺T^– mutation tested lay near the 3' end. These results agree with earlier work in which a fusion protein missing the first 31 amino acids of NadR lacked repressor function but retained transport activity (25). This C-terminal fragment has a fusion joint that deletes a potential DNA-binding site predicted to be in a β-pleated sheet conformation similar to that observed for MetJ and the P22 repressors Arc and Mnt (6). NadR lacks a classic helix-turn-helix motif (Foster and Penfound, unpublished data).

It is interesting that mutations causing the superrepressor phenotype lie primarily in the central portion of this gene. These mutations generally result in diminished transport function in addition to enhanced repressor activity. The SSCP results therefore suggest that this central region of the protein is involved with intramolecular communications between the N-terminal repressor and the C-terminal transporter regions of NadR. This central region also contains a consensus binding site for mononucleotides such as ATP (see below). This binding site may be important for NadR functional transitions.

Characterization of NadR Repression Function

Several laboratories have examined NadR target genes from both S. typhimurium and E. coli. These genes include nadA and pncB from both organisms and nadB from E. coli (14, 24, 82, 88). Subsequent comparisons of these genes revealed a consensus sequence (NAD box) of probable importance to NadR regulation (T. Penfound and J. W. Foster, submitted for publication). The consensus NAD box is ttGTTTAnnnnntaAAACaa. The more tightly controlled genes, nadA and nadB, contain two NAD boxes that overlap a potential –10 RNA polymerase-binding site, whereas the weakly regulated pncB contains only one full NAD box.

Proof that NadR binds to this region centers on gel retardation assays and DNA footprinting with purified NadR (Penfound and Foster, submitted). A significant finding from these studies is that not one but two different nucleotides modulate the repressor activity of NadR. NAD serves as the corepressor for this system. No other pyridine nucleotide functions effectively in this capacity. What is surprising is that ATP also affects the DNA-binding properties of NadR but in the opposite direction. NAD-free NadR has some DNA-binding activity that is eliminated by the addition of ATP. This fact suggests that ATP places NadR in a nonrepressor conformation that can be reversed by the addition of physiological levels of NAD. In this model, ATP may place NadR in a transport-facilitating conformation that is ineffective for DNA binding. When sufficient NAD is present, NadR undergoes a conformation shift to a DNA-binding repressor protein that simultaneously becomes ineffective in NMN transport. Nucleotide sequence analysis supports the presence of an ATP-binding site within NadR. A consensus sequence, Gly X X X X Gly (Lys/Arg) Ser, for mononucleotide-binding sites occurs at Gly-237. An ATP-binding site could enable NadR, and thus NAD synthesis, to sense changes in cellular energy status. The more energy-rich cells will be growing faster and thus need more NAD than energy-poor cells. Alternatively, NadR may require ATP for its transport function, so that ATP bound to NadR shifts the conformational state toward transport and away from repressor configurations. Curiously, the consensus NAD-binding site for dehydrogenases (Gly X Gly X X Gly) does not occur within NadR (24), leaving open the questions of how and where NAD interacts with this repressor.

QUESTIONS THAT REMAIN

The last few years have seen some exciting studies of the regulation of NAD biosynthesis but only a few advances in understanding the biochemistry of NAD metabolism. Several important aspects of NAD metabolism in E. coli and S. typhimurium remain to be explored. These areas include determining how and why intracellular NAD turnover occurs. What are the genes involved, and are they regulated? The genes for NAD kinase and NADP phosphatase have not been identified, nor has the proposed control mechanism for NADP phosphatase been defined. The confirmation of NAD glycohydrolase activity and the possible role in ADP-ribosylation raise intriguing possibilities of further roles for NAD in the cell.

The NadR regulator also presents many interesting questions. The data obtained thus far reveal a unique regulatory molecule with several fascinating molecular interactions. These include protein-DNA interaction as a repressor, protein-protein interaction with PnuC as a transporter, and several nucleotide-binding interactions with NAD and ATP. The current challenge is to characterize each of these interactions at the molecular level. Also, crystal structure analyses with NadR complexed to various ligands will reveal much about the conformational shifts that NadR undergoes while performing its various functions. Finally, mutational mapping analysis of nadR will provide insight into the intramolecular communications required to coordinate the transition between DNA repressor and nucleotide transport activities.

ACKNOWLEDGMENT

Studies conducted by J.W.F. and presented in this article were supported by awards from the National Institutes of Health (GM 39018 and GM 32595).