Crystal Structure of Haemophilus influenzae NadR Protein

Haemophilus influenzae NadR protein (hiNadR) has been shown to be a bifunctional enzyme possessing both NMN adenylytransferase (NMNAT; EC 2.7.7.1) and ribosylnicotinamide kinase (RNK; EC 2.7.1.22) activities. Its function is essential for the growth and survival of H. influenzae and thus may present a new highly specific anti-infectious drug target. We have solved the crystal structure ofhiNadR complexed with NAD using the selenomethionine MAD phasing method. The structure reveals the presence of two distinct domains. The N-terminal domain that hosts the NMNAT activity is closely related to archaeal NMNAT, whereas the C-terminal domain, which has been experimentally demonstrated to possess ribosylnicotinamide kinase activity, is structurally similar to yeast thymidylate kinase and several other P-loop-containing kinases. There appears to be no cross-talk between the two active sites. The bound NAD at the active site of the NMNAT domain reveals several critical interactions between NAD and the protein. There is also a second non-active-site NAD molecule associated with the C-terminal RNK domain that adopts a highly folded conformation with the nicotinamide ring stacking over the adenine base. Whereas the RNK domain of the hiNadR structure presented here is the first structural characterization of a ribosylnicotinamide kinase from any organism, the NMNAT domain ofhiNadR defines yet another member of the pyridine nucleotide adenylyltransferase family.

Haemophilus influenzae, a small nonmotile Gram-negative bacterium, resides in the upper respiratory mucosa in humans and causes otitis media and respiratory tract infections, mostly in children. This organism lacks almost all of the enzymes necessary for the synthesis of NAD (1,2), and it requires the presence of the so-called V-factors (NADP, NAD, NMN, 1 or N-ribosylnicotinamide) in the growth medium (3,4). Phosphorylated V-factors (NADP, NAD, and NMN) are degraded by recently identified extracellular and periplasmic hydrolases (5) to N-ribosylnicotinamide, which is likely to be the only V-factor transported across the inner membrane as an ultimate NAD precursor (6). Two enzymatic steps are required to convert ribosylnicotinamide to NAD in the cytoplasm: a ribosylnicotinamide kinase (RNK; EC 2.7.1.22) to catalyze the phosphorylation of nicotinamide riboside to produce NMN and a NMN adenylyltransferase (NMNAT; EC 2.7.7.1) to link NMN and the AMP moiety of ATP to generate NAD. Whereas the gene encoding RNK has not been identified in any organism until very recently, 2 several NMNATs and functionally related NaMNATs (EC. 2.7.7.18) have been characterized at the molecular level from many species during last few years. We will use pyridine nucleotide adenylyltransferase (PNAT) as a generic name for both NMNAT and NaMNAT as well as for enzymes with dual specificities, such as in the case of human NMN/NaMNAT. The three-dimensional structures of PNATs and their complexes with substrate/product have been solved from several sources, including Methanococcus jannaschii (7), Methanothermobacter thermautotrophicum (8), Escherichia coli (9), Bacillus subtilis (10), and humans (11)(12)(13). Since PNATs catalyze the indispensable central step in NAD biosynthesis and recycling, genes encoding these activities have been found in almost all organisms with completely sequenced genomes. In H. influenzae, the ortholog of bacterial NaMNAT (NadD gene product) is lacking, and the alternative housekeeping PNAT function is encoded in the N-terminal domain of the bifunctional NadR protein, 2 which shares 54% identity to the E. coli and Salmonella typhimurium NadR proteins. The multifunctional NadR proteins contain the signature nucleotidyltransferase (H/T)IGH motif, and its NMNAT activity has been confirmed experimentally in the E. coli enzyme (14) and, more recently, in H. influenzae and S. typhimurium as well. 2 In addition to the NMNAT activity, NadR in S. typhimurium and E. coli contain a DNA-binding helix-turn-helix motif N-terminal to the NMNAT domain. The S. typhimurium NadR has long been demonstrated to be a NAD-dependent repressor for the transcription of genes involved in both de novo biosynthesis (nadB and nadA) and niacin salvage (pncB) (15,16). It was also suggested that NadR may interact with an integral membrane transporter PnuC protein and directly participate in the uptake of exogenous NAD precursors (17). H. influenzae NadR lacks the helix-turn-helix DNA binding domain, which correlates with the absence of any genes of de novo NAD biosynthesis or niacin salvage in this organism.
Recently, we have predicted and experimentally verified that previously uncharacterized RNK activity resides within the C-terminal domain of NadR. The essentiality of NadR for the growth and survival of H. influenzae has also been established. 2 Here we report the crystal structure of H. influenzae NadR protein (hiNadR) complexed with NAD. This structure reveals that whereas the NMNAT domain is mostly similar to the archaeal NMNAT, the C-terminal RNK domain is structurally similar to the yeast thymidylate kinase and several other P-loop-containing nucleotide and nucleoside kinases.

EXPERIMENTAL PROCEDURES
Preparation of H. influenzae NadR Protein-The cloning, expression, and purification of hiNadR protein is described elsewhere. 2 Briefly, the gene encoding residues 52-421 of sequence NADR_HAEIN (gi͉1171638) was PCR-amplified from H. influenzae genomic DNA and was cloned into a pET-derived vector containing a T7 promoter, His 6 tag, and TEV protease cleavage site (gift from Dr. Meg Philips, UT Southwestern Medical Center). The resulting plasmid was transformed into the E. coli strain BL21(DE3) (Invitrogen) for expression. The overexpressed hi-NadR protein was first purified with a Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) column, followed by a Superdex 200 gel filtration column (Amersham Biosciences). The selenomethionine hiNadR was expressed in the met Ϫ auxotrophic strain B834(DE3) (Novagen), grown in minimal medium supplemented with selenomethionine and other nutrients (18), and purified using the same procedure as the native protein.
Crystallization and Data Collection-The hiNadR crystals were grown at 20°C using the hanging drop vapor diffusion method. 10 mg/ml hiNadR in 100 mM Hepes, pH 7.2, 0.3 M NaCl, and 1 mM dithiothreitol was first incubated with 3 mM of NAD and 3 mM of ATP and then mixed with an equal volume of the reservoir solution (0.1 M MES, pH 6.0, 1.0 M ammonia sulfate) and equilibrated against the reservoir. Large diamond-shaped crystals appeared after 2-7 days. These hiNadR co-crystals belong to a hexagonal crystal system with cell dimensions a ϭ b ϭ 106.9 Å, c ϭ 174.9 Å. ␣ ϭ ␤ ϭ 90°, and ␥ ϭ 120°. The exact space group of these crystals was later determined to be P6 4 22. There is one hiNadR molecule in the asymmetric unit. These crystals diffract to 3.0-Å resolution on a rotating anode x-ray generator and about 2.7 Å at synchrotron. The selenomethionine hiNadR crystals were grown at similar conditions and are of same quality as the native crystals.
For data collection at 100 K, crystals were transferred stepwise to a cryoprotection solution containing all components of the reservoir and additional 30% ethylene glycol. A 3.3-Å resolution multiwavelength anomalous diffraction data set and a 2.9-Å resolution native data set were collected at beamline X12B (National Synchrotron Light Source, Brookhaven National Laboratory). All diffraction data were processed and scaled with the HKL2000 package (19). The statistics for all data sets are listed in Table I.
Phasing and Refinement of hiNadR-The initial phases of the hi-NadR crystal structure were solved by the multiwavelength anomalous diffraction phasing method using SOLVE (20). Six selenium sites out of a total of eight were located, and the resulting phases have a figure of merit of 0.64. Density modification was performed with RESOLVE (21), which resulted in a clearly interpretable electron density map. The hiNadR polypeptide chain was manually built into this map using O (22).
The refinement of hiNadR structure was carried out using CNS (23) with simulated annealing protocol. The first round of refinement resulted in an R work of 32.5% and R free of 38.2%. Several rounds of manual rebuilding and refinements improved the R factors to a value of 27.3% for the R work and 33.9% for the R free . At this point, densities for two NAD molecules became evident: one at the active site of the NMNAT domain and the other associated with the C-terminal RNK domain but not in its active site. The starting coordinates for the NAD at the active site of the NMNAT domain were derived from the coordinates of NAD in the M. thermoautotrophicum NMNAT⅐NAD complex structure (8). The second NAD molecule adopts a highly compact conformation with the nicotinamide ring stacking over the adenine base. The model for this NAD molecule was built manually to fit the density, and its conformation was optimized with CNS (23). Several putative sulfate molecules were also included in the subsequent rounds of refinement. Further rounds of refinement with these ligands included in the model resulted in a decrease in R work to 23.6% and R free to 29.8%. The current refinement statistics are listed in Table I.
Analytical Ultracentrifugation Studies-Sedimentation equilibrium experiments were performed with a Beckman XL-I analytical ultracentrifuge and using both scanning absorption and Rayleigh interference optics. Wavelengths used were 295 nm in the absorbance mode and 675 nm for Rayleigh interference. We used an An-60Ti rotor with double sector cells of 1.2-cm path length at 4°C for all experiments. The sedimentation equilibrium was carried out at 9,000 r.p.m. and 13,000 r.p.m. Two different concentrations of the protein (1.2 and 0.4 mg/ml) where F o and F c are the observed and calculated structure factors, respectively. c R free is the R factor calculated for a randomly selected 10% of the reflections that were omitted from the refinement. dissolved in 100 mM Hepes buffer, pH 7.2, containing 0.3 M NaCl, 2 mM dithiothreitol, 1 mM EDTA were used in the study. All centrifugation data were analyzed with the Optima TM data analysis software.

RESULTS AND DISCUSSION
Description of hiNadR Monomer Structure-The overexpressed recombinant hiNadR protein contains residues 52-421 of the NADR_HAEIN sequence (total of 370 amino acids), which includes the NMNAT and RNK domains. We chose to carry out crystallographic study of this segment, since the multiple alignment of all NadR homologs has shown that the reliable sequence similarity between hiNadR and other NadR homologs starts only from the NMNAT domain. Overexpression of the extended version of hiNadR containing an N-terminal His 6 tag fused at position Met 15 results in two protein products, suggesting an alternative translation initiation at Met 52 . 2 This possibility is further strengthened by the fact that it is more consistent with the predicted starts of NadR homologs in several other species, such as Mycobacterium tuberculosis (gi͉92220269), Nostoc punctiforme (gi͉91112614), and Lactococcus lactis (gi͉15673969). The N-terminally truncated version of hiNadR (residues 52-421) possesses full NMNAT and RNK activities and is functionally indistinguishable from the full-length protein. 2 The current crystal structure model of hiNadR contains residues 57-345 and 357-411 (among the expressed segment 52-421). The first 5 residues along with the N-terminal His 6 tag region, a loop containing residues 346 -356, and the last 10 residues at the C terminus are disordered in the crystal structure. Additionally, there are two NAD and five sulfate molecules that are included in the model. Although 3 mM ATP was present in the crystallization solution, no ATP molecule could be located in the density map; therefore, the current structure must be discussed as a hiNadR⅐NAD complex.
The structure of the hiNadR monomer revealed the presence of two distinct domains connected by a loop (Fig. 1). The Nterminal NMNAT domain (residues 52-224) adopts a Rossmann-like three-layered ␣/␤/␣ fold with the central fivestranded parallel ␤-sheet having the strand order 32145. The structure of this domain closely resembles archaeal NMNATs (7,8) with r.m.s. deviation of 2.6 Å for 147 C-␣ atoms when superimposed with the M. janaschii NMNAT (Protein Data Bank code 1f9a). The C-terminal RNK domain (residues 225-421) also adopts a three-layered ␣/␤/␣ fold with a five-stranded central parallel ␤-sheet of strand order 23145. Note that the strand order in the central ␤-sheet of the C-terminal RNK domain differs from that in the NMNAT domain. Search of the protein structure data base (PDB) using DALI (24) Fig. 2.
Oligomerization of hiNadR-Although there is only one hi-NadR molecule in the asymmetric unit, inspection of crystal packing indicated that four symmetry-related hiNadR monomers are tightly associated, and the protein is likely to exist as tetramer (Fig. 3). To investigate the oligomeric status of the protein in solution, we performed analytical ultracentrifugation studies at two different hiNadR concentrations (1.2 mg/ml and 0.4 mg/ml) and two different speeds (9,000 and 13,000 r.p.m.). The apparent molecular masses obtained are 170 kDa (1.2 mg/ml) and 150 kDa (0.4 mg/ml), which are somewhat lower than the calculated molecular mass of hiNadR tetramer (180 kDa), indicating the presence of lower molecular weight species. The data collected at the two different protein concentrations and two different speeds allow us to fit either a dimer/ tetramer or a monomer/tetramer model for hiNadR, both of which yielded dissociation constants in the low micromolar range (K D of 1-10 M for the dimer-tetramer model, data not shown). This indicates that tetramer is the major oligomerization species of hiNadR in solution at higher than micromolar concentrations.
The hiNadR tetramer in crystal displays a 222 symmetry. Two distinct types of interfaces between individual monomers are observed in the tetramer. Monomers A and B (or C and D) form a "handshake-like" association that buries 1384 Å 2 of surface area per monomer at the interface. The C-terminal RNK domain from one monomer sits snugly in a hinge grove between the two domains of the second monomer, making extensive contacts with both domains of the second monomer (Fig. 3a). The other type of dimer interface is between monomers A and D (or B and C) and is less extensive, burying a surface area of only 894 Å 2 per monomer. The resulting tet- ramer has the appearance of a closed barrel with the hinge between the two domains of each monomer located at the middle of the barrel. This creates a cavity at the center of the barrel with a diameter of ϳ18 Å. A narrow channel extends from this central cavity and runs through the whole length of the tetramer, exiting between the less densely packed dimer (Fig. 3b). The electrostatic potential mapped on the surface of the tetramer revealed a strikingly positively charged cluster in the middle and along the central channel of the barrel (Fig. 3c). Amino acids Lys 215 , Lys 219 , Arg 222 , His 296 , His 298 , and Lys 299 from each of the four monomers contribute to this positively charged patch, which is located at the interface between monomers in the tetramer. It was thus not surprising that several sulfate molecules were found in this region. The active sites of both NMNAT and RNK domains are located on the outside of the barrel, facing away from the central cavity (Fig. 3).
The "Non-active-site"-Bound NAD-Four non-active-sitebound NAD molecules associated with the C-terminal RNK domains of the four hiNadR monomers were found aggregated inside the central cavity of the tetramer (Fig. 3). Each of these NAD molecules adopts a highly folded conformation with the nicotinamide ring stacked over the adenine base in a nearly parallel fashion (Fig. 4a). The distance between the adenine C-6 and nicotinamide C-2 atoms, which has been used in the literature as a general measure of compactness of the bound NAD(P) (25), is 4.2 Å for this NAD molecule. Whereas most of the enzyme-bound NAD molecules adopt extended conformation such as the one bound in the active site of NMNAT domain of hiNadR, there is only one other example of enzyme-bound NAD that adopts a similarly highly folded conformation. In the crystal structure of flavin reductase P complexed with its inhibitor NAD, the NAD molecule adopts a compact conformation with a distance between adenine C-6 and nicotinamide C-2 atoms of 3.9 Å (26). However, the conformation of the flavin reductase P-bound NAD differs drastically from the second NAD in the hiNadR⅐NAD complex structure (Fig. 4b), reflecting the remarkable flexibility of NAD molecule.
Notably, nuclear magnetic resonance studies and molecular dynamics simulation have indicated that, in contrast to the protein-bound forms, NAD in aqueous and some organic solutions adopts compact folded conformations in which the distance between the nicotinamide and adenine rings is 4 -5 Å (27,  dark red). b, superposition of the C-␣ trace of hiNadR RNK domain (blue) with yeast TMK (3TMK) (red). c, structure-based alignment of hi-NadR with mthNMNAT and yeast TMK. The color and letter coding of secondary structural elements in the sequence is the same as in Fig. 1. The (H/T)IGH motif of NMNAT, the P-loop (Walker-A motif), Walker-B motif, and the "LID" of the RNK domain are boxed. The sequences in regions that are not superimposed are in lowercase type, and the disordered segments are in italic lowercase type. 28). There are several specific interactions between the nonactive-site NAD molecule and hiNadR tetramer. The adenine and nicotinamide rings are sandwiched between the side chains of Tyr 292 and Trp 256 (Fig. 4a). The pyrophosphate moiety of the NAD points outward and is in contact with the Lys 126 side chain of an adjacent monomer in the tetramer. Additionally, the oxygen of the nicotinamide carboxyamide group is within the hydrogen bonding range of the main chain amide of Trp 256 , and the N-6 of the adenine appears to be in contact with main chain carbonyl of Tyr 289 . Although we cannot rule out the possibility that binding of the non-active-site NAD molecule is a crystallization artifact, the specific interactions between this NAD molecule and hiNadR tetramer indicate that it may have biological implications. In E. coli and S. typhimurium, the NadR protein represses three genes involved in NAD biosynthesis in a NAD-dependent manner (15,16). It is possible that the NAD molecule as an effector may bind to a site different from the active site of NMNAT domain. Binding of the NAD effector at this regulatory site presumably induces a conformational change and leads to the binding of NadR to operator DNA. It is thus tempting to speculate that the second nonactive NAD binding site in hiNadR may represent the NAD effector binding site in E. coli and S. typhimurium NadR. Although hiNadR do not possess repressor function, the sequence identities between the NMNAT and RNK domains of hiNadR and that of E. coli NadR and S. typhimurium NadR are high (ϳ54%). Therefore, hiNadR must have retained many structural and functional features of E. coli NadR and S. typhimurium NadR. More biochemical and structural studies are needed to investigate this hypothesis.
The Active Site of the NMNAT Domain of hiNadR-The NAD molecule that was found in the active site of the N-terminal NMNAT domain of hiNadR adopts an extended conformation similar to that bound to the mthNMNAT (8) (Fig. 5). The adenine nucleotide binding site includes the (H/T)IGH motif, the preceding loop between the (H/T)IGH motif, and the end of the first ␤-stand, which is also highly conserved, the N terminus of ␣-helix E that contains the ISSTXXR motif, and a loop connecting ␤-strand e and helix E that interacts with the adenine base. A sulfate molecule found in this active site is superimposable with that in the mthNMNAT⅐NAD complex (8) and appears to occupy the site that would normally be occupied by the ␤ and ␥ phosphates of ATP, as was shown in the structure of the M. janaschii NMNAT⅐ATP complex (7). Similar to the archaeal NMNATs, in hiNadR the loop connecting strand c to helix C (residues 138 -149) plays a central role in the binding of the nicotinamide portion of the substrate. We therefore term this loop the "nicotinamide recognition loop." The exocyclic carboxyamide group of the nicotinamide interacts with several main chain groups from this nicotinamide recognition loop (Fig. 5); the carboxyl group lies in close proximity to the main chain amide of Trp 149 , whereas the amide group is close to the main chain carbonyls of Pro 143 and Ser 144 . Trp 152 , which is aligned with Trp 84 in MthNMNAT, appears to play the same role in forming a stacking interaction with the nicotinamide ring. Additionally, the indole ring of Trp 149 appears to interact with the nicotinamide ring in a face-to-edge manner. This tryptophan residue is absent in the archaeal NMNAT, where two leucine residues occupy the same site and form hydrophobic interactions with the nicotinamide ring. Notably, the conformation of this loop shows a large difference between hiNadR and archaeal NMNAT (Fig. 2a). Because of this large difference at the nicotinamide binding site, the conformation of the bound NAD differs from that in mthNMNAT⅐NAD complex primarily at the glycosyl bond N and the nicotinamide ring is oriented at slightly different angles in the two structures (ϳ30°apart) (Fig. 2a).
Comparision of the C-terminal RNK Domain with TMK-The C-terminal RNK domain of hiNadR described here is the first structurally characterized RNK from any organism. Its structural similarity to the thymidylate kinase and other nucleotide/nucleoside kinases places it in the "nucleotide and nucleoside kinases" family of the "P-loop-containing nucleotide triphosphate hydrolase" fold in SCOP (29). Whereas the current hiNadR structure does not contain any bound ligand at the active site of the RNK domain, the location of the RNK active site can be inferred from the superposition with yeast TMK complexed with a bisubstrate inhibitor P 1 -(5Ј-adenosyl) P 5 -(5Јthymidyl) pentaphosphate (30) (Fig. 2b, 6). The Walker-A or P-loop ( 233 GGESSGKS 240 ) and Walker-B ( 300 IAFID 304 ) motifs in the RNK domain of hiNadR have the similar spatial arrangement as those in TMK. The extremely conserved Lys 239 of Walker-A and the Asp 304 of the Walker-B motif in hiNadR superimpose well with the corresponding residues Lys 18 and Asp 93 of TMK. Their location indicates potential ATP and nicotinamide ribose substrate binding site in the RNK domain of hiNadR (Fig. 6). The loop connecting strand i and helix I is disordered in the hiNadR structure. This disordered segment corresponds to the so called "LID" region in TMK and other structurally related kinases, such as uridylate kinase (31), and adenylate kinase (32,33). This LID has been shown to undergo drastic movements upon binding of substrate (33,34). It has been proposed that three distinct conformations of this loop exist for TMK: an open state in the absence of substrates, a partially closed state when one substrate is bound, and a closed state when both substrates are present (30). It is likely that this apparently flexible loop in hiNadR-RNK may also undergo Inspection of the superposition of hiNadR-RNK with yeast TMK at their active site also reveals substantial structural variations (Fig. 6). In particular, helix GЈ in the two structures shows a shift of ϳ7 Å, whereas the preceding ␤-strand g and the following helix G in the two structures superimposed relatively well. Helix GЈ composes part of the nucleoside binding site and corresponds to the NMP bind region in adenylate kinase (32). It has been shown that in addition to the LID region, the NMP bind region of adenylate kinase also undergoes large "opento-close" conformational changes upon substrate binding (32,33). In hiNadR, helix I following the LID loop also shows ϳ20°d ifference in orientation relative to that in TMK structures. Whether this is due to the presence of the bisubstrate P 1 -(5Јadenosyl) P 5 -(5Ј-thymidyl) pentaphosphate in the TMK structure or merely reflects the intrinsic differences between the two remotely related proteins is not clear. Considering the likely conformational changes upon binding of the substrates, it is difficult to speculate at this point about the detailed interaction between hiNadR-RNK and its substrates.
In the hiNadR monomer, the active site of the NMNAT domain is well separated from that of the RNK domain as indicated by the location of the P-loop and Walker-B motifs. The distance between the two active sites in the monomer is more than 40 Å. The adjacent active sites of neighboring monomers in hiNadR tetramer are also distant from each other (ϳ35 Å apart). Therefore, although the two domains of hiNadR catalyze sequential reactions, there appear to be no direct interactions (or channeling) between the two active sites within either the monomer or tetramer. Mutagenesis and kinetic measurements of the two enzymatic activities encoded by hi-NadR also indicated no such interactions. 2 Comparison and Classification of PNAT Family Proteins-Because of the central role of PNAT family enzymes (NMNATs and NaMNATs) at the convergent point of NAD biosynthesis and recycling, the variations in the enzymatic properties and structures of PNATs from different organisms are indicative of distinct NAD metabolic fluxes in these organisms. So far, several PNATs from species of all three kingdoms of life have been characterized kinetically and structurally. The first is the ar-chaeal NMNAT (nadM gene product), which strongly prefers NMN over NaMN as substrate (35). Its structure closely resembles bacterial phosphopantetheine adenylyltransferase (PPAT) (36), contains a five-stranded central parallel ␤-sheet and associated helices and forms hexamers. There is also a three-helical bundle subdomain at the C terminus of the protein. The first helix in this bundle contains the second conserved motif ISSTXXR that interacts with the bound ATP phosphates (7). The bacterial version of NadM-like NMNATs are encoded in the bifunctional NMN adenylyltransferase/ NUDIX hydrolase (ADP-ribose pyrophosphotase) in Synechocystis sp.(NADM_SYNY3) (37), Deinococcus radiodurans (gi͉7471334), and Ralstonia solanacearum (gi͉17549056). The N-terminal NMNAT domain of these proteins is closely related to archaeal NMNAT both in sequence and kinetic properties and has a strong preference for NMN over NaMN (38).
All of the characterized representatives of the second subfamily of PNAT are the NadD gene product conserved in the majority of bacteria. They display a strong preference for the deamidated NaMN substrate over NMN and therefore should be termed NaMNAT (10,39). The structures of bacterial NaMNAT have an overall nucleotidyltransferase fold but with a larger central ␤-sheet of either six or seven strands (9,10). The complex structure of E. coli NaMNAT and deamido-NAD reveals that the active site of E. coli NaMNAT is substantially different from that of the archaeal NMNAT, and the interactions between protein and the bound NAD strongly favor a carboxylate over carboxyamide group on the pyridine ring (9). The high degree of preference for the deamidated substrate NaMN suggests that these bacterial NaMNATs are predominantly involved with the NAD metabolic fluxes via NaMN intermediate, such as the three-stepped aspartate de novo pathway and a two-stepped niacin salvage pathway (39,40). Although E. coli NaMNAT is a monomer (9), B. subtilis NaMNAT was shown to be a dimer (10), indicating that there are variations in the oligomerization states among bacterial NaMNATs.
The third PNAT subfamily includes all of the eukaryotic representatives identified so far. As exemplified by the most studied human nuclear NMN/NaMNAT (hsnPNAT), eukaryotic enzymes have dual substrate specificity and recognize both NMN and NaMN almost equally well (40,41). Therefore, they are capable of participating in de novo and salvage/recycling routes of NAD biosynthesis, via both NMN and NaMN intermediates (13). The structures of hsnPNAT complexed with NAD and with deamido-NAD, respectively, revealed that an active site water molecule appears to participate in the dual substrate recognition (13). Human hsnPNAT was shown to be a hexamer in the crystal structures (11)(12)(13). Analytical ultracentrifugation studies indicated that lower molecular weight species (most likely dimer) co-exist with the hexamer in solution (13). Interestingly, the dimerization of hsnPNAT appears to be similar to that of B. subtilis NaMNAT (10,13).
The NMNAT domain of the NadR proteins represents yet another subfamily of PNATs. In concert with its C-terminal RNK domain, NadR protein alone can convert the imported ribosylnicotinamide directly to NAD, thus playing a critical role in the salvage of the NAD precursors. Kinetic measurements have shown that NadR strongly prefers NMN over NaMN (14), 2 consistent with its biological role.
Since the representative structures of most PNAT subfamilies, including hiNadR, have now become available, the structural basis for the different enzymatic properties of these enzymes can be analyzed. These structures also allow us to construct a reliable structure-based multiple sequence alignment of all currently identified PNAT representatives and to compare them with the related phosphopantetheine adenylyltransferases and glycerol cytidyltransferases for the analysis of evolutionary relationships between these protein families. A recently developed program EESG (42) was used to group these sequences according to their evolutionary distances (Fig. 7). This program uses the global multiple sequence alignment to calculate identity fractions q ij between sequence pairs i and j and uses the formula d ij ϭ 1/((q ij Ϫ q ij ran )/(1 Ϫ q ij ran ) Ϫ 1) in conversions of identity fractions to evolutionary distances, with q ij ran representing the expected identity between two random sequences with the same amino acid composition as the sequences i and j. Each sequence was represented as a point in a multidimensional Euclidian space in such a way that Euclidian distances d ij between the points optimally approximated the estimated evolutionary distances d ij between the sequences by minimizing the following function (42). It is clear from the distribution of PNATs in the sequence space that two major groups emerge well separated. Group I contains archaeal NMNAT (NadM gene product), NadM/NUDIX, and NadR proteins, and group II contains bacterial NaMNAT (NadD gene product) and the eukaryotic subfamily of PNATs. As discussed above, these two groups of PNATs are distinct in sequence conservation patterns, enzymatic properties, and active site architectures. Three structures are now available for the Group I PNATs: the NMNAT from M. jannaschii (7), M. thermoautotrophicum (8), and the NadR from H. influenzae presented in the current study. In all of these structures, a primary nicotinamide recognition loop located between strand c and helix C makes most of the interactions with the nicotinamide ribose portion of the substrate. Several main chain interactions with the carboxyamide and the stacking interaction of a tryptophan with the pyridine ring all come from this nicotinamide recognition loop. Currently, three structures are available for the group II PNATs: the NaMNAT from E. coli (9), B. subtilis (10), and human nuclear PNAT (11)(12)(13). In these structures, two loops (one connecting strand c and helix C and the other between strand d and helix D) and the N terminus of helix C are involved in the binding of nicotinic acid/nicotinamide. The critical tryptophan that forms a stacking interaction with the pyridine ring comes from the second loop, and the carboxylate/carboxyamide interacts with main chain atoms from both loops (13). It has been noted that the Group I PNAT appears more closely related to bacterial phosphopantetheine adenylyltransferases than to the Group II PNATs (9). Interestingly, our sequence analysis also reveals that there are two distinct groups of phosphopantetheine adenylyltransferases: bacterial phosphopantetheine adenylyltransferases (encoded by the gene CoaD) that are relatively close to Group I PNAT and the recently identified archaeal/eukaryotic phosphopantetheine adenylyltransferases (43) that are closer to glycerol cytidyltransferases than to their bacterial counterparts (Fig. 7).
In H. influenzae, the NadR protein encodes the only PNAT activity in the organism. The RNK activity encoded in NadR was also shown to be of central importance for the NAD generation in H. influenzae (6,44). The essentiality of hiNadR for the growth and survival of H. influenzae has recently been asserted and was confirmed experimentally. 2 Therefore, hi-NadR may present a promising narrow spectrum anti-infectious drug target specific for H. influenzae and several other V-factor-dependent pasteurellaceae. The structural studies that delineate the active site architectures and protein-ligand interactions of hiNadR will be critical for the design and improvement of potent inhibitors for this essential protein.