Genomics-driven Reconstruction of Acinetobacter NAD Metabolism

Enzymes involved in the last steps of NAD biogenesis, nicotinate mononucleotide adenylyltransferase (NadD) and NAD synthetase (NadE), are conserved and essential in most bacterial species and are established targets for antibacterial drug development. Our genomics-based reconstruction of NAD metabolism in the emerging pathogen Acinetobacter baumannii revealed unique features suggesting an alternative targeting strategy. Indeed, genomes of all analyzed Acinetobacter species do not encode NadD, which is functionally replaced by its distant homolog NadM. We combined bioinformatics with genetic and biochemical techniques to elucidate this and other important features of Acinetobacter NAD metabolism using a model (nonpathogenic) strain Acinetobacter baylyi sp. ADP1. Thus, a comparative kinetic characterization of PncA, PncB, and NadV enzymes allowed us to suggest distinct physiological roles for the two alternative, deamidating and nondeamidating, routes of nicotinamide salvage/recycling. The role of the NiaP transporter in both nicotinate and nicotinamide salvage was confirmed. The nondeamidating route was shown to be transcriptionally regulated by an ADP-ribose-responsive repressor NrtR. The NadM enzyme was shown to possess dual substrate specificity toward both nicotinate and nicotinamide mononucleotide substrates, which is consistent with its essential role in all three routes of NAD biogenesis, de novo synthesis as well as the two salvage pathways. The experimentally confirmed unconditional essentiality of nadM provided support for the choice of the respective enzyme as a drug target. In contrast, nadE, encoding a glutamine-dependent NAD synthetase, proved to be dispensable when the nondeamidating salvage pathway functioned as the only route of NAD biogenesis.

Acinetobacter baumannii is an emerging pathogen that belongs to a relatively underexplored branch of ␥-proteobacteria. It can cause severe pneumonia and infections of the urinary tract, bloodstream, and other parts of the body. Some isolates of A. baumannii display resistance to many known antibiotics (1,2), emphasizing the importance of pursuing new therapeutic targets for drug development. Biogenesis of nicotinamide adenine nucleotide (NAD), an indispensable cofactor involved in a multitude of biochemical transformations in metabolic networks of all species, was recently established as a target pathway for the development of new antibiotics (3)(4)(5)(6)(7). Beyond its main function as a redox cofactor, NAD is consumed as a co-substrate by a number of nonredox enzymes such as bacterial DNA ligase and protein deacetylase of the CobB/Sir2 family (8,9). A degradative consumption of NAD by these and, likely, other (not fully elucidated) enzymes demands continuous replenishing of the NAD pool, providing further rationale for targeting essential enzymes involved in its biogenesis and recycling. Among these enzymes, nicotinate mononucleotide adenylyltransferase (NaMNAT) 4 of the NadD family and NAD synthetase of the NadE family are widely recognized as the most promising drug targets (6,7). First, small-molecule inhibitors targeting bacterial NadD and NadE enzymes and showing antibacterial activity were recently reported by several research groups (4, 5, 10 -14). Second, our recent genomic survey of NAD metabolism (captured as "NAD(P) biosynthesis" subsystem in the SEED data base (15)) confirmed that these two enzymes are conserved in the overwhelming majority of Ͼ800 diverse bacterial species with completely sequenced genomes (9,16). NadD and NadE enzymes together comprise a two-step conversion of the committed precursor nicotinic acid mononucleotide (NaMN) to NAD. Alternative de novo and salvage routes leading to NaMN synthesis converge at this nearly universal downstream pathway, explaining the conservation and essentiality of nadD and nadE genes. Nevertheless, a few groups of bacteria (including Acinetobacter spp.) appear to deviate from a common pattern lacking orthologs for one or both of these genes and pointing to the existence of alternative routes of NAD biogenesis as well as to the necessity of alternative targeting strategies. Thus, obligate intracellular pathogens Chlamydia and Rickettsia have lost the entire NAD biosynthetic machinery, replacing it by a unique capability to salvage NAD from the host cell (9). Haemophilus influenzae lacking both nadD and nadE as well as most other common genes of NAD biosynthesis is entirely dependent on a relatively rare PnuC-NadR pathway of nicotinamide riboside (so-called Vfactor) salvage (17). Another alternative route of NAD biosynthesis was recently discovered in Francisella tularensis, the causative agent of tularemia or rabbit fever (16). In this species lacking nadD, the NaMN intermediate is first amidated to NMN by NMN synthetase, a divergent member of the NadE family. In the second step, NMN is converted to NAD by NMNAT of the NadM family (16,18). Members of this family, which also belongs to a large superfamily of HIGH nucleotidyltransferases (19), are commonly present in Archaea as a functional replacement of NadD (20). They are scarcely distributed in bacteria, typically as fusion proteins with ADP-ribose pyrophosphatase. In most cases they are present in addition to (rather than instead of) NadD in the context of a dispensable nicotinamide salvage/recycling pathway, which is initiated by the conversion of nicotinamide to NMN by nicotinamide phosphoribosyltransferase (NMPRT) of the NadV family (21). The recently established essential role of NadM in F. tularensis first highlighted this family as an alternative drug target in NAD biogenesis of bacterial pathogens. This choice is additionally supported by the fact that members of NadM family are only distantly homologous to their human counterparts (22).
To evaluate the last enzymatic steps of NAD biosynthesis as antibacterial targets in A. baumannii, we combined bioinformatics with biochemical and genetic techniques to systematically analyze the complete NAD metabolic subnetwork in the non-pathogenic model organism Acinetobacter baylyi (previously called Acinetobacter sp. ADP1) (23). This analysis led to the discovery of unique aspects of NAD metabolism in the Acinetobacter group providing new guidelines for antibacterial discovery efforts.

EXPERIMENTAL PROCEDURES
Bioinformatics Tools and Resources-Genome analysis and metabolic reconstruction tools implemented in the SEED data base (15) were used to analyze NAD metabolism in A. baumannii ATCC 17978 and A. baylyi ADP1 in comparison with ϳ800 other completely sequenced bacterial genomes. Full results are available at The SEED Viewer in the "NAD and NADP cofactor biosynthesis global" and "NAD regulation" subsystems. Genomes of additional (recently sequenced) strains of A. baumannii strains were analyzed using NCBI and MicrobesOnline data base (24).
Strains, Growth, and Media-Escherichia coli DH5␣ (Invitrogen) and BL21(DE3) (Stratagene) were used as general strains for gene cloning and protein overexpression, respectively. All other strains used in this study are listed and briefly described in supplemental Table S1. A. baylyi ADP1 and its mutant derivatives were grown at 30°C either in rich (Luria broth) or defined minimal Acinetobacter medium (25) (26) at 37°C (i.e. for purposes of niaP in trans experiment). The growth rates of bacterial cultures were monitored using Bioscreen C apparatus (Growth Curves USA, minimal Acinetobacter). Strains to be measured were streaked from cryostorage. Three single colonies were picked from each plate and used to inoculate 5-ml overnight cultures. The following day cultures were diluted to A 600 nm of ϳ0.01 in a final volume of 300 l, and 250 l were transferred to a Bioscreen 100-well plate. Cultures were grown in the Bioscreen at 30°C with continuous shaking with readings taken every 30 -45 min.
Mutant Construction-Strain construction was performed as described in Metzgar et al. (27). Briefly, genes or markers were deleted from the chromosome by a double crossover recombination event after the transformation of a linear PCR product using positive or negative selection markers as previously extensively described (27). Oligonucleotides used for gene deletion and insertion protocols are listed in supplemental Table S2.
Cloning-Primers and plasmids used for molecular cloning are listed in supplemental Tables S1 and S2. A. baylyi pncB, nadV, nadM, and nadE were amplified by polymerase chain reaction from genomic DNA and cloned into the NheI and EcoRI sites of pET28a (Novagen). A. baylyi niaP was cloned into the NcoI and PstI sites of pBAD24 (Invitrogen). A. baylyi pncA was amplified by polymerase chain reaction from genomic DNA and cloned into the NcoI and PstI sites of pPROEX-HTb (Invitrogen). AbpncA was also amplified by PCR from pPROEX-HTb-abpncA construct and cloned into a pETderived vector (28) containing the T7 promoter, His 6 tag, and tobacco etch virus protease cleavage site. AbnrtR was amplified by polymerase chain reaction from genomic DNA and cloned into the BamH1 and Bpu1102I sites of pET15b (Novagen). All constructs were sequence-verified for accuracy.
Enzyme Assays and Kinetics-Nicotinate phosphoribosyltransferase activity for abPncB enzyme was measured with an end-point HPLC-based assay relying on direct quantitation of NaMN product. The reaction mixtures were incubated for 20 min at 30°C and analyzed on an ion-paired analytical Supelco C18 column (3 M, 4.6 ϫ 150 mm) as described (16). Reaction mixtures contained 50 mM Hepes, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 2 mM ATP, 5 mM PRPP, and 0.5 mM nicotinic acid in the presence of 1.2 nM abPncB. For initial rate measurements, nicotinic acid concentration was 0.1-300 M. To test abPncB specificity for nicotinamide, higher concentrations of enzyme (25 M) and substrate (0.03-20 mM) were used. Nicotinamide phosphoribosyltransferase activity for abNadV enzyme was measured as above. Reaction mixtures contained 50 mM Hepes, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 2 mM ATP, 5 mM phosphoribosyl pyrophosphate (PRPP), and 0.5 mM nicotinamide in the presence of 10 nM abNadV. For initial rate measurements, nicotinamide concentration was 0.003-1 mM, whereas a higher concentration of enzyme (15 M) and substrate (0.03-100 mM) was used to test abNadV specificity toward nicotinic acid. Nicotinamide deamidase activity was assayed in 100 mM Hepes, pH 7.5, and 1-300 M nicotinamide in the presence of 2 nM abPncA. The same end-point HPLC method as above was used to monitor nicotinic acid formation. In an alternative continuous assay modified from Boshoff and Mizrahi (31) ammonia release by PncA was coupled to the conversion of ␣-ketoglutarate to l-glutamate by glutamate dehydrogenase in the presence of NADPH. The coupled reaction was monitored by the decrease of NADPH fluorescence at 465 nm (excitation at 340 nm). The assay mixture contained 20 mM potassium phosphate, pH 7.5, 0.2 mM NADPH, 3.3 mM ␣-ketoglutarate, 2 units of bovine liver glutamate dehydrogenase, 1-300 M nicotinamide (Nm), and appropriate amounts of abPncA. NaMN/NMN adenylyltransferase-coupled assays were performed as described (16). The NaMN adenylyltransferase activity of abNadM (15 nM) was assessed by in situ conversion of the NaAD product to NAD (by excess of Bacillus anthracis NAD synthetase) coupled to the reduction of NAD to NADH (by alcohol dehydrogenase) and monitored at 340 nm. NMN adenylyltransferase activity measurements were performed under the same conditions without adding NAD synthetase. NAD synthetase activity was tested using a similar coupled assay in mixtures containing 50 mM Hepes, pH 7.5, 2 mM ATP, 1 mM NaAD, 5 mM ammonia (or glutamine), 10 mM MgCl 2 , 20 mM semicarbazide (or 2 mM Na 2 SO 4 ), 6 units/ml alcohol dehydrogenase, 60 mM ethanol, and 60 nM abNadE. For kinetic measurements, concentration of NaAD was varied in the range 0.005-2 mM. NMN synthetase activity was measured with a HPLC-based end-point assay as described (16). Reaction mixtures consisted of 50 mM Hepes, pH 7.5, 2 mM ATP, 1 mM NaMN, 5 mM ammonia, 10 mM MgCl 2 , and 22 M abNadE. NaMN was varied from 0.1 to 25 mM for initial rate studies. Steady-state kinetic parameters were determined using Prism software (GraphPad) by fitting to standard Michaelis-Menten equation or its modification for substrate (nicotinamide) inhibition (32) as observed for abNadV.
Electrophoretic Mobility Shift Assay (EMSA)-A 90-bp region upstream of prs-nadV operon containing predicted tandem regulatory sites was PCR-amplified from genomic DNA using primers shown in supplemental Table S2. One of the primers was 5Ј-biotinylated (Sigma). The labeled DNA fragment (1 nM) was incubated for 20 min at 25°C with 2 nM purified abNrtR in 10 mM Tris, pH 7.5, 50 mM KCl, 1 mM DTT, 2.5% glycerol, 5 mM MgCl 2 , 0.05% Nonidet P-40, and 0.50 mg/ml bovine serum albumin. The complexes were analyzed by native PAGE (5%) in 0.5ϫ Tris borate-EDTA followed by electrophoretic transfer onto a nylon membrane (Pierce) and UV cross-linking. Biotin-labeled DNA was detected with the LightShift Chemiluminescent EMSA kit (Pierce). To control the specificity of abNrtR binding, a competitive EMSA assay was performed with a 100-fold molar excess of non-biotinylated target DNA (specific competitor) or 1 g of poly(dI-dC) (nonspecific competitor). To evaluate ADP-ribose and other NAD-related metabolites effects on the protein-DNA binding, the protein and metabolite were preincubated at the indicated concentration for 10 min at 25°C before biotinylated DNA was added.
Quantitative RT-PCR Analysis-total RNA was isolated from A. baylyi cells grown in LB and collected at midlog phase using the SV Total RNA Isolation System (Promega). The RNA was subjected to DNase I digestion to remove genomic DNA and reverse-transcribed using SuperScript II reverse transcriptase (Invitrogen). The reaction mix was then subjected to quantitative real-time PCR to detect levels of the corresponding nadV and nrtR cDNAs using primers listed in supplemental Table S2. The levels of expression (relative to 16 S rRNA) were analyzed using MJ Research Opticon TM 2 software (MJ Research, Watertown, MA).

Bioinformatics Predictions-The results of a comparative genomic reconstruction of NAD metabolism in A. baylyi and
A. baumannii performed using a subsystems-based approach (15) implemented in the SEED genomic platform are illustrated in Fig. 1 and Table 1. Although most of the reconstructed sub-network could be seamlessly inferred by analogy with previously studied species, certain key aspects appeared to be unique and, thus, required additional experimental analysis. Most notably, an apparent functional replacement of NadD by NadM could have two alternative interpretations. Based on similarity with previously characterized bacterial representatives of the NadM family (16,18,33), abNadM could be assigned a functional role of NMNAT acting on the NMN intermediate. This assignment would be consistent with its role in the nondeamidating salvage of nicotinamide via NMPRT enzyme (NadV). However, utilization of a canonical NMNAT for the processing of NaMN intermediate generated by de novo (NadB-NadA-NadC) or deamidating salvage (PncA-PncB) routes ( Fig. 1) would be extremely inefficient or require the presence of an additional enzyme, NMN synthetase, as previously shown for F. tularensis (ft-NadEЈ) (16). The amino acid sequence of abNadE did not show appreciable similarity with the NMN synthetase subfamily but was rather characteristic of NAD synthetase with the additional glutamine transferase domain responsible for the utilization of glutamine as in eukaryotes (34) and some bacteria (e.g. Mycobacterium tuberculosis) (35). Nevertheless, this possibility could not be excluded without additional experimental analysis (see below).
A possible alternative interpretation would be that abNadM possesses dual NMNAT/NaMNAT activity. Such dual specificity is characteristic of eukaryotic enzymes from a distinct (albeit evolutionarily and structurally related) PNAT family (36,37). Indeed, both PNAT activities are physiologically relevant for NAD biosynthesis in human cells, which includes de novo and salvage routes proceeding via NaMN or NMN intermediates (36). Moreover, enzymes of NadM, and not NadD, family are universally present in Archaea where they are expected to play the NaMNAT role in the de novo synthesis of NAD (38). 5 To distinguish between these two possibilities, in this study we set to experimentally determine the actual substrate preference of the purified recombinant abNadM and to use genetic techniques to address yet another (albeit unlikely) possibility; that is, the exis-tence of an alternative NaMNAT nonhomologous to previously characterized enzymes.
Another unprecedented feature of Acinetobacter spp. is the simultaneous presence of two alternative routes of nicotinamide salvage inferred from genome analysis and raising questions about their actual physiological roles and regulation. Our approach to addressing these questions combined the enzymatic characterization of respective gene products with the phenotype analysis in a series of engineered A. baylyi mutants.
Two additional bioinformatics predictions related to niacin salvage/recycling were deemed a subject of experimental testing. Gene abNiaP was tentatively assigned a role of Nm and/or nicotinic acid (Na) transporter based on homology with the previously characterized distant member of this family from Bacillus subtilis (39). Gene deletion and phenotype analysis experiments were designed to verify this functional assignment and to test whether abNiaP is essential for Nm and/or Na salvage or (as in B. subtilis) there are additional mechanisms of their uptake.
Expression of nadV gene was predicted to be under control of a transcriptional regulator from the NrtR family (29). Regulons controlled by the members of NrtR family, which typically include genes involved in various aspects of NAD metabolism, have been recently described in many diverse bacteria as captured in RegPrecise data base (40). Two representatives of this family from Shewanella oneidensis and Synechocystis sp. have been characterized in vitro, and their DNA binding activity was shown to be negatively regulated by ADP-ribose (ADPR), a byproduct of NAD degradative utilization. However, at the time of this study, NrtR activity has not yet been demonstrated in vivo in any species. 6 Therefore, in this study we aimed to test the predicted abNrtR function both by characterization of the purified recombinant abNrtR and by the analysis of nadV gene expression in ⌬nrtR mutant. The brief description and interpretation of experimental results addressing all of the questions formulated above is provided in the following sections.
De Novo Synthesis of NaMN-Acinetobacter spp. possess readily detectable orthologs of all three genes, nadB, nadA, and nadC, of the de novo synthesis of NaMN from L-aspartate (Table 1). All three respective enzymes have been characterized from a variety of species (for a recent overview, see Ref. 9). We have confirmed the in vivo functional activity of this pathway by the comparison of growth phenotypes of A. baylyi wild-type and ⌬nadB mutant ( Table 2). As expected, the nadB strain could not grow on minimal medium without added Nm or Na (see also supplemental Fig. S1).

Conversion of Nm (or Na) to NaMN via Deamidating Sal-
vage Route-Although the orthology-based functional assignment of genes encoding both enzymes of this common pathway, nicotinamidase (gene pncA) and nicotinic acid phosphoribosyl transferase (gene pncB), was straightforward (Table 1), their experimental characterization was essential for the assessment of physiological roles of the two alternative salvage routes of Nm (Fig. 1). The analysis of steady-state kinetic parameters revealed a high (physiologically relevant) catalytic efficiency of both purified recombinant enzymes (k cat /K m values of ϳ10 5 -10 6 M Ϫ1 s Ϫ1 , see Table 3). Notably, abPncA enzyme displayed a very high affinity for Nm substrate (K m ϳ 2 M) pointing to an efficiency of the deamidating route even at very low concentrations of Nm. The abPncB enzyme displayed an expected strict preference for Na over Nm substrate (Ͼ3,000,000-fold based on the comparison of respective k cat /K m values, see Table 3). The physiological role of this enzyme and the functional in vivo activity of the entire pathway were directly confirmed by genetic analysis ( Table 2). The A. baylyi ⌬nadB⌬pncB strain lost the ability to grow on minimal medium supplemented by 0.1 mM Na. However, this strain could grow normally in the presence of 0.1 mM Nm, confirming the physiological activity of the alternative nondeamidating route (see below). The physiological activity of ab-PncA is supported by the normal growth of the A. baylyi ⌬nadB⌬nadV strain in the presence of 0.1 mM Nm.   Conversion of Nm to NMN via Nondeamidating Salvage Route-As already mentioned, nondeamidating salvage of Nm is initiated by the NMPRT enzyme encoded by gene nadV (Fig. 1). Previously described members of NadV family share a fold and a general catalytic mechanism of phosphoribosyl transfer with the PncB family (41) but display a distinct substrate specificity for Nm (21,42). The same strict substrate preference for Nm over Na (ϳ5,000-fold) was observed by us for the recombinant purified abNadV enzyme (Table 3), confirming its role in the nondeamidating route. Notably, both phosphoribosyltransferases, abPncB and abNadV, are strongly activated in the presence of 2 mM ATP (ϳ100-and ϳ10-fold, respectively) as previously reported for other representatives of these families (43,44). Comparison of catalytic efficiencies of abNadV and abPncA (Table 3) potentially competing for the Nm substrate suggests that the deamidating route is likely to prevail, especially at low concentrations of Nm (as the K m value of abNadV is ϳ 20-fold higher than of abPncA). The biological significance of the coexistence of two alternative Nm salvage routes in Acinetobacter spp. is not entirely clear. The observed difference in catalytic efficiency of the respective enzymes together with precedents in other species (21) points to a likely role of nondeamidating route in recycling of endogenous Nm (product of intracellular NAD degradation). However, our genetic data (Table 2) proved that A. baylyi can sustain a normal growth with the nondeamidating (NadV-mediated) salvage of Nm as the only route of pyridine nucleotide synthesis.
Uptake of Nm and Na-Two complementary approaches were used to assess the physiological role of the predicted abNiaP transporter similar to how it was previously described for NiaP from B. subtilis (39). In the first "knock-in" approach the overexpression of the niaP gene from A. baylyi in E. coli BW25113 ⌬nadA strain led to a notable increase in its growth rate on minimal medium supplemented with a limiting concentration of either Nm or Na (0.16 M) (Fig. 2, A and B). This strain cannot produce NAD via the de novo route, and its growth is entirely dependent on the uptake and salvage of exogenous niacin. The observed increase in growth rate provided the first experimental verification of the abNiaP functional assignment as high affinity niacin (Nm and Na) transporter. In the second "knock-out" approach, the growth of an A. baylyi ⌬nadB strain (with blocked de novo route) was com-pared with a double-mutant ⌬nadB⌬niaP derivative in minimal medium supplemented with varying concentrations of either Nm or Na (0 -100 M). Whereas the supplementation of 0.5 M Nm or Na was sufficient for the growth of the ⌬nadB strain (Fig. 2, C and D), at least 200-fold higher niacin concentration was required to maintain the normal growth of the ⌬nadB⌬niaP strain (Fig. 2, E and F). This observation suggests that NiaP substantially contributes to niacin transport when its concentration in the medium is very low (Ͻ0.5 M, see supplemental Fig. S3 for dose dependence data). As in case of B. subtilis (39), an alternative (NiaP-independent) uptake mechanism is operational at higher niacin concentration (Ͼ100 M).
Transcriptional Regulation of Nondeamidating Salvage/ Recycling Route-In addition to the regulation of flux distribution between the two salvage/recycling routes through differential catalytic efficiency of the respective enzymes (as described above), the nondeaminating salvage route is regulated at the level of transcription by a predicted ADPR responsive transcriptional repressor abNrtR (29). The results of in vitro characterization of the recombinant purified abNrtR protein are illustrated in Fig. 3. EMSA confirmed the specific binding of abNrtR to the DNA region containing previously predicted DNA motif sites located upstream of the prs-nadV operon (Fig. 3A). The observed binding was efficiently suppressed in the presence of 100 M ADPR or phospho-ADPR (Fig. 3B), confirming a proposed role of these NAD(P) degradation products as antirepressors. None of the intermediary metabolites associated with NAD biosynthesis, Nm, Na, quinolinate, NMN, NaMN, NAD, NaADP, NaAD, NADP, tested at 100 M interfered with abNrtR-DNA binding (some of them are shown in Fig. 3B). The in vivo activity of abNrtR was confirmed by the comparison of nadV expression levels in wild-type A. baylyi and in a ⌬nrtR mutant using RT-PCR. The basal level of nadV expression in the wild-type strain grown in rich medium was very low, whereas it increased ϳ150-fold upon the deletion of nrtR gene (supplemental Fig.  S2). These results are consistent with abNrtR acting as a negative regulator of nadV expression providing in vivo functional verification for a member of NrtR family. 6 Adenylylation of NaMN and NMN by NadM-A steadystate kinetic analysis of the purified recombinant abNadM confirmed its hypothesized dual specificity of NaMNAT/

Enzyme
Nm (1) PRPP, ATP 0.04 Ϯ 0.01 0.12 Ϯ 0.02 3 ϳ10 4 (Nm Ͼ Ͼ Na) Na (1) PRPP, ATP 30 Ϯ 4 0.020 Ϯ 0.001 6.7 ϫ 10 Ϫ4 NadM NaMN (2) ATP 0.013 Ϯ 0.002 3.0 Ϯ 0.1 2.3 ϫ 10 2 ϳ1 (NMN Ϸ NaMN) NMN (2) ATP 0.027 Ϯ 0.002 4.9 Ϯ 0.1 1.8 ϫ 10 2 NadE NaAD (2) ATP, NH 3 0.017 Ϯ 0.002 0.57 Ϯ 0.02 3.4 ϫ 10 ϳ10 5 (NaAD Ͼ Ͼ NaMN) NaMN (1) ATP, NH 3 4.3 Ϯ 1.0 0.0032 Ϯ 0.0002 7.4 ϫ 10 Ϫ4 NMNAT type with nearly identical catalytic efficiency (k cat / K m ) for both physiological substrates (Table 3). To our knowledge it is the first documented example of a dual specificity NaMNAT/NMNAT enzyme in bacteria. Most importantly, the dual specificity of abNadM is fully consistent with its universal and essential role in all routes of NAD biogenesis via each of the two pyridine nucleotide intermediates. The essentiality of abNadM is supported by the fact that the deletion of the respective gene in A. baylyi could be accomplished only with complementation by another NaMNAT or NMNAT encoding gene. Thus, we were able to obtain a viable A. baylyi ⌬nadM knock-out mutant only after introduction of E. coli nadD or the F. tularensis nadM genes into an ectopic site on the chromosome under the control of a constitutive promoter (see supplemental Fig. S4 and Table  S1). As of today, Acinetobacter is the second example of bacterial species (after F. tularensis (16)) where the member of NadM family was shown to be an indispensable enzyme functionally replacing NadD in NAD biogenesis and homeostasis.
Conversion of NaAD to NAD by NadE-The enzymatic characterization of the purified recombinant abNadE confirmed its proposed functional role of NAD synthetase in the last step of a two-step conversion of NaMN to NAD (Fig. 1). As expected from the presence of an additional glutamine transferase domain, abNadE can efficiently utilize l-glutamine (as well as ammonia) for the amidation of NaAD precursor (specific activities are quite similar for both ammonia donors; not shown). No appreciable amidation was observed using NaMN instead of NaAD as a substrate, which effectively excluded a possibility of abNadE having an alternative NMN synthetase activity (Table 3). Based on the confirmed physiological activity of the abNadV-NadM nondeamidating salvage route (Fig. 1), abNadE enzyme was expected to be dispensable in Acinetobacter in the presence of Nm in the growth me- . C0-F, knock-out approach is shown. Supplementation of minimal medium by 0.5 M Nm (C) or Na (D) is sufficient to support the growth of a single knock-out mutant A. baylyi ADP1 ⌬nadB (dashed lines) at the same rate as wild type (in the absence of vitamin (solid line)). Substantially higher concentrations of ϳ100 M Nm (E) or Na (F) are required to restore the normal growth of the double knock-out mutant ⌬niaP⌬nadB. All data points represent the mean of three independent repeats. Final yields and errors are shown in supplemental Table S3. dium. Remarkably, the nadE gene could be indeed successfully deleted but only in the genetic background of the double-knock-out mutant A. baylyi ⌬nadB⌬pncB where the nondeamidating (NadE-independent) salvage of Nm is the only active route of NAD synthesis. The observed conditional essentiality of abNadE enzyme may be explained by toxicity of its substrate NaAD, which would accumulate as a result of nadE gene deletion in a strain with an active route(s) proceeding via NaMN intermediate. Nevertheless, a principal possibility to bypass NadE in NAD synthesis is another unusual feature of Acinetobacter, which was previously demonstrated only in F. tularensis (16).

DISCUSSION
Targeting essential enzymes involved in the biosynthesis of NAD(P), the indispensable redox cofactor, has been recognized as a promising strategy for the development of novel antibiotics (4,5,10,12). However, the rational choice of actual drug targets requires better understanding of pathways that comprise NAD metabolic subnetworks, especially in divergent and poorly explored organisms such as pathogenic species of Acinetobacter. Indeed this study revealed that among the two nearly universal target enzymes, NadD and NadE, the former is functionally replaced in Acinetobacter by a member of a distinct NadM family, whereas the latter is potentially dispensable.
Our approach to the reconstruction of NAD metabolism in Acinetobacter was based on comparative genomics analysis combined with focused experimental testing and refinement of the key bionformatics predictions. The bioinformatics analysis included all Acinetobacter species with completely sequenced genomes (7 genomes were publicly available), whereas the experimental testing was performed using a nonpathogenic strain, A. baylyi ADP1. This strain provided us with an ideal model system due to its remarkable genetic versatility (27,45) and availability of a genome-wide collection of knock-out mutants by Genoscope (25). Moreover, under-standing of NAD metabolism in A. baylyi is of additional importance for its biotechnological applications in biodegradation and manufacturing of biopolymers (46). Indeed, maintenance and regeneration of the NAD cofactor pool is a known bottleneck in many microbial fermentations (47,48).
In silico genomic reconstruction performed in this study revealed an extensive network of pathways leading to NAD, which is conserved in all analyzed Acinetobacter spp. and includes de novo synthesis and two alternative salvage/recycling routes ( Fig. 1 and Table 1). This analysis pointed us to several unique and previously unexplored features, and targeted biochemical and genetic experiments were designed to test the key bioinformatics inferences and reconcile ambiguities and possible alternative scenarios.
In the upstream part of Acinetobacter NAD subnetwork, the most unusual feature is the simultaneous presence of the two alternative (and potentially competing) Nm salvage routes, deamidating (via PncA-PncB) and nondeamidating (via NadV). A widely different phylogenomic distribution of these two alternative salvage pathways and possible evolutionary implications have been previously discussed (8,21), and they are further illustrated in supplemental Table S4.
Through in vitro characterization of all respective enzymes (Table 3) and growth phenotype analysis of engineered knock-out mutants (Table 2), we have confirmed that both salvage routes are operational in Acinetobacter. Indeed, each of them could support the normal growth of A. baylyi as a sole source of pyridine nucleotides on the minimal medium supplemented with Nm. Although the biological significance of this duality is not completely clear, some experimental data obtained here allow us to speculate that the first route is predominantly responsible for the salvage of exogenous Nm (as well as Na), especially when its concentration in the environment is very low ( Table 2 and Fig. 2). On the other hand, the second, abNadV-mediated route is likely to contribute largely to the recycling of endogenous Nm, which may accumulate due to degradative utilization of NAD by CobB/Sir2 NAD-dependent protein deacetylase and, possibly, by other (yet unknown) NAD-glycohydrolases. Indirect evidence in favor of this interpretation was provided by the experimentally confirmed in vivo transcriptional regulation of nadV gene by ADPR-responsive repressor abNrtR. The ability of ADPR to disrupt binding of abNrtR to the predicted operator site in the upstream of prs-nadV operon was confirmed in vitro by EMSA analysis (Fig. 3). These observations are consistent with the hypothesis that accumulation of ADPR resulting from NAD degradation may be interpreted by the cell as a signal to activate recycling of Nm. The latter is released as a byproduct of this degradation, and the abNadV-mediated nondeamidating route appears to be the most efficient way of its conversion to NAD with the help of NMNAT activity of abNadM (see below). Notably, the first gene of the prs-nadV operon encoding a paralog of PRPP synthetase may also directly contribute to Nm recycling by feeding PRPP as a second substrate for NadV reaction (Fig. 1). It is tempting to speculate that ribosyl 5-phosphate required for the synthesis of PRPP may be at least partially supplied by the hydrolysis of ADPR (supplemental Fig. S5) by the product of nudF gene (ACIAD0275 in A. baylyi) tentatively assigned an ADPR pyrophosphatase function based on homology with other characterized members of Nudix family (49).
The remarkable plurality of the upstream pathways in the NAD subnetwork leading to the formation of pyridine nucleotide precursors (NaMN and NMN) emphasizes the importance of NAD biogenesis and homeostasis in Acinetobacter. At the same time, a demonstrated functional redundancy of these pathways (at least in laboratory conditions) argues that none of the respective enzymes would constitute a sustainable drug target. Moreover, of the two enzymes abNadM and ab-NadE involved in downstream conversion of these precursors to NAD, only the former is truly indispensable, as supported by combined genetic and kinetic evidences (Table 1-3). The abNadM is the first documented example of dual NMNAT/ NaMNAT specificity among bacterial members of the NadM family. As noticed before, this family is only sparsely represented in bacteria where NadM is usually present in addition to (and not instead of) the housekeeping NadD enzyme and contributes to an optional Nm salvage/recycling together with NadV enzyme (21). In most cases these two genes, nadV and nadM, are encoded in one operon. Importantly, all previously characterized bacterial NadM enzymes have two features distinguishing them from abNadM; (i) strong (at least 50 -100fold) substrate preference for NMN over NaMN and (ii) the presence of an additional C-terminal domain harboring ADPR pyrophosphatase of Nudix family. Both enzymatic activities are relevant for Nm salvage as discussed above for A. baylyi where the ADPR pyrophosphatase activity is encoded by a remotely located nudF gene. The structural and enzymatic analysis of the two representative bacterial NadM-NudF fusion proteins did not reveal any functional interactions between the two domains that would affect the NadM substrate preference (18). At the same time, the NadM (and not NadD) family is universally present in nearly all Archaea where it appears to function as the housekeeping NaMNAT/ NMNAT enzyme in NAD biogenesis. 5 None of the analyzed archaeal genomes revealed a presence of the bacterial-like NadM-NudF fusion protein. These observations as well as characteristic variations in gene patterns associated with deamidating and nondeamidating Nm salvage/recycling routes point to a likely archaeal origin of bacterial NadM (see supplemental Table S4 and the accompanying comments suggesting a possible evolutionary scenario).
Overall, the most important practical implication of this study is the identification of abNadM as an attractive target for the development of novel antibiotics against pathogenic strains of A. baumannii. This is the second case (after F. tularensis (16)) where a member of NadM family functionally replaces NadD (which is otherwise present in nearly all pathogenic bacteria) and constitutes an alternative drug target. In both cases, members of NadE family were shown to be dispensable (at least in laboratory conditions). The latter observation can be tentatively projected to some other bacterial species (including pathogenic strains of Burkholderia) containing the complete nondeamidating salvage pathway comprised of NMPRT (NadV family) and NMNAT (NadM or NadR family). More generally, it reflects a fundamental differ-ence between the two types of biochemical transformations associated with the downstream pathways of NAD biosynthesis. At least one of the two adenylyl transferase reactions converting pyridine mononucleotides (NaMN or/and NMN) to respective dinucleotides (NaAD or/and NAD) is present and unconditionally essential in all life forms (including all bacterial pathogens) except for some intracellular endosymbionts. On the other hand, the ATP-dependent amidation of the carboxylate moiety of NaAD by NAD synthetase (or, rarely, of NaMN by NMN synthetase) in some species may be bypassed (as in Francisella) or even replaced (as in Haemophilus) by the nondeamidating salvage of amidated precursors (Nm or Nmribose, respectively). NAD metabolism of Acinetobacter analyzed in this study is another example that illustrates this fundamental principle providing important guidelines for drug targets selection and prioritization.