Nicotinamidase (EC 220.127.116.11) catalyzes hydrolysis of nicotinamide to nicotinic acid and ammonia, an important reaction in the NAD(+) salvage pathway. This activity has a fortuitous medical benefit since the Mycobacterium tuberculosis enzyme converts the nicotinamide analog prodrug pyrazinamide into the bacteriostatic pyrazinoic acid,[2-4] hence the alternative name, pyrazinamidase (PncA). Pyrazinoic acid inhibits M. tuberculosis type I fatty acid synthase, represses mycolic acid biosynthesis, and appears to affect membrane energetics and acidification of the cytoplasm. It is active against semi-dormant tubercle bacilli and with rifampicin and isoniazid, forms the front line tuberculosis treatment.[2,3] Studies of PncA have revealed aspects of structure and biochemical activity[6-9] there are no structural data on how the enzyme binds and processes physiological ligands. High-resolution crystal structures of Acinetobacter baumanii PncA (AbPncA) complexed with nicotinic acid and pyrazinoic acid now provide direct evidence for the interactions that govern specificity and mechanism and of how a valued antibacterial agent is activated.
Recombinant AbPncA was prepared, the dimeric, colorless enzyme purified in high yield and kinetic properties determined. With pyrazinamide as the substrate the following values were obtained; KM 106.9 μM, Vmax 62.8 nMol min−1, kcat 3.1 min−1, specific activity 132 μM min−1 mg−1. These values are comparable to literature values, e.g. the specific activity of M. tuberculosis PncA (MtPncA) with pyrazinamide is 82 μM min−1 mg−1.
Two crystal forms (I and II) were obtained with nicotinic and pyrazinoic acid respectively and the structures determined. PncA is a divalent cation-dependent enzyme and activity has been reported with Fe2+, Mn2+, and Zn2+ . As expected, metal ions were observed in the structures. Inductively coupled plasma-atomic emission spectrometry (ICP-OES) identified that recombinant AbPncA contained Fe2+ and Zn2+ in an approximate 1:1 ratio with a trace of Mn2+ present. However, anomalous dispersion measurements are consistent with a higher occupancy of Zn2+ at the active site and the crystallographic models contain that cation. We refer to Zn2+ in discussion but judge it likely that AbPncA functions in the presence of different divalent cations. Experimental details, including enzyme activity and metal ion identification, together with sequence alignments and additional figures are given as supplementary material (Figure S1-S8).
Crystals were obtained in the presence of cacodylate buffer and form II shows dimethylarsinoyl-modified Cys159 in the active site, an artefact of crystallization (Figure S1, S2). The steric hindrance of this modification precludes full occupancy of pyrazinoic acid such that the final refinement was performed with occupancy 0.8 for pyrazinoic acid, 0.2 for the modified cysteine. Crystal form I has two molecules, form II a single molecule in the asymmetric unit respectively with an r.m.s.d. (root-mean-square deviation) derived from least-squares fit of Cα atoms of these three molecules of 0.2 Å. The structures and the interactions formed by ligands within the active sites are essentially identical and we concentrate on form I, a 1.65 Å resolution structure with full occupancy ligand (Figure S3). About 60 % of residues form elements of secondary structure, eight α-helices and nine β-strands (Figure 1, S4). The core of the subunit is a parallel β-sheet of strands 1, 2, 5–9. Three helices (α5, α6, α7) lie on one side of the sheet, with α2 placed against the other. A sub-domain is placed at one end of the β-sheet and includes β3 and β4 and a single turn of helix, α3.
Gel filtration and analytical ultracentrifugation indicate that AbPncA is a stable dimer in solution, approximate mass 47 kDa. The asymmetric unit of form I is a dimer (Figure S5), with an interface of 900 Å2, about 10% of the surface area of a subunit. In form II the crystallographic two-fold axis generates the same dimer. The interactions that stabilize the dimer mainly involve residues on α5 and β5. The AbPncA subunit resembles orthologues from Pyrococcus horikoshii (PhPncA) and Saccharomyces cerevisae (ScPncA).[6,7] Superpositions give r.m.s.d. values of 1.2 Å for the overlay of an AbPncA subunit on either PhPncA (44 % sequence identity, 167 Cα atoms) or ScPncA (33 % identity, 191 Cα atoms).
The AbPncA active site is between the core and the sub-domain (Figure 1), formed by residues on β1, β2, the β4-α4 loop, β5, and the β6-α6 turn. It is buried and completely occluded from solvent by four polypeptide segments; the loops linking α3-β3, β5-α5 and β8-α9 together with strand β4. Numerous hydrophobic residues (discussed below) surround the active site and a gross conformational change is likely required to permit substrate binding or release of products.
The active site Cys159 is located at the N-terminus of α6 at one side of the active site with Zn2+ positioned on the other side. The metal ion is held tightly in the AbPncA active site since activity was retained in the presence of 10 mM EDTA (data not shown). Octahedral coordination of Zn2+ involves Asp54 OD2, His56 NE2 and His89 NE2, two water molecules and nicotinic acid N5 (Figure 2). Cation ligand distances range from 2.11 to 2.28 Å, consistent with data on Zn2+ ligand geometry. A network of hydrogen bonds position the coordinating groups. One water molecule forms hydrogen bonds with Ser62 OG and Asp121 OD2, the other with Asp54 OD1 and Asp121 OD1. The coordinating His56 and His89 donate hydrogen bonds from ND1 to carbonyls of Gly115 and Pro87 respectively (data not shown). Nicotinic acid is tethered to the cation and positioned between five hydrophobic residues; Phe21, Leu27, Trp86, Tyr123 and Cys159 (Figure S6). Trp86 NE1 and Tyr123 OH form a hydrogen bond to hold these residues in place over the ligand. A further six residues (Val29, Ile154, Ala155, Phe158, Ile184 and Leu186) stabilize the hydrophobic environment around the nicotinic acid and occlude the active site (Figure S6). In the absence of a ligand, Zn2+ coordination is completed with a water as indicated in structures of PhPncA and ScPncA.[6,7] Nicotinic acid O8 and O9 are 2.5 and 2.6 Å respectively from Cys159 SG suggesting the presence of a bifurcated hydrogen bond. The next nearest functional group to Cys159 SG is Asp16 OD2 at a distance of 3.4 Å. O9 accepts hydrogen bonds donated from main chain amides of cis-Ala155 and Cys159. Interaction with the cis-Ala155 carbonyl suggests that O8 is a hydroxyl and that protonation of Asp16 OD2 may facilitate a second hydrogen bond (Figure S7). Alternatively, O8 as hydroxyl group or if protonated may participate in a bifurcated hydrogen bond with Asp16 and Ala155. Asp16, together with Asp54 and Lys114 form a cluster of interacting hydrophilic residues on one side of the ligand-binding site. Lys114 NZ donates hydrogen bonds to Asp16 OD1, Asp54 OD2 (Figure 2) and the main chain carbonyl of Tyr123 (not shown). The close proximity of Lys114 to Asp16 is likely to influence the pKa. The Asp16 carboxylate and main chain amide form hydrogen bonds with Thr52 OG1 (Figure S7), a pairing strictly conserved in PncA (Figure S8).
Amidation involves either acidic or basic hydrolysis and the structures of the enzyme:product complex suggest that the latter applies in PncA. Acid hydrolysis would involve nucleophilic water attacking the carbon of a protonated amide. The hydrophobic environment on one side of the amide and close interactions with functional groups on the other renders it difficult to envisage how water could be placed to attack C7. Nicotinic acid O9 accepts hydrogen bonds from two amides so protonation at O9 would destabilize the complex due to the proximity of the amides.
A more likely mechanism is indicated by a strictly conserved and essential Cys159,[6,8] on the polar side of the active site, ideally placed to attack the carbonyl carbon in a manner similar to that proposed for other enzymes, for example trypanothione synthetase amidase and the nitrilase enzyme superfamily. Nitrilases exploit a catalytic triad consisting of a reactive cysteine, a lysine and a glutamate. The triad of PncA has a conservative difference with aspartate replacing glutamate.
We now propose a four-stage mechanism (Figure 3). In stage I substrate binds in the axial position displacing a Zn2+ coordinating water. The equatorial metal ion coordinating waters are also held in position by hydrogen bonding interactions to the enzyme (Figure 2), whereas the axial water lacks such a restraint. It is this water that vacates the coordination sphere as substrate binds. This water may be prevented from exiting the active site due to the hydrophobic lid covering the active site (Figure S7). The displaced water, or an incoming water would preferentially bind to the hydrophilic side of the active site near Lys114. Proton abstraction by Asp16 would generate a Cys159 thiolate facilitating nucleophilic attack at C7. Two main chain amides (cis-Ala155 and Cys159) form an oxyanion hole to support thiolate attack by stabilizing the resulting tetrahedral intermediate in a similar fashion to cysteine proteases. Proton donation from Asp16 to N8 would promote C-N bond cleavage and release of NH3 as the tetrahedral-intermediate collapses in stage II; being converted to an acyl-intermediate. Water activation by Asp16, where the amino acid is protonated, generates a nucleophilic hydroxyl to attack the acyl-intermediate producing nicotinic acid (or pyrazinoic acid) and a thiolate in stage III. In stage IV, products are released and the thiolate accepts a proton from Asp16 to regenerate a thiol.
A previously proposed mechanism invoked a Zn2+-coordinated hydroxyl and an incoming water participating in catalysis. Firstly a hydroxyl group directly attacks the acyl-intermediate to replace the amino group, and then another incoming water coordinates Zn2+ as a hydroxyl with a proton removed by the nearby aspartate to enable nucleophilic attack on the acyl-intermediate. This mechanism appears unduly complicated with two rounds of metal ion directed water activation and is incompatible with our new structures. Activation of water to hydroxyl groups by the strong Lewis acid that is four-coordinate Zn2+ is a commonly invoked feature of zinc-dependent enzymes. In PncA, the metal ion is six-coordinate and this would reduce Lewis acid strength. In addition, that PncA catalysis is supported by different divalent cations suggests that Lewis acid strength is less important than the structural role provided by octahedral coordination, precise placement of the substrate and an appropriate ligand exchange rate to support substrate binding with release of a water molecule. We therefore propose a simpler mechanism (Figure 3) with no requirement for the metal ion to activate water.
Recombinant MtPncA produced in Escherichia coli is reported to carry a mixture of Fe2+ and Mn2+ and reconstitution of apoenzyme with these ions resulted in almost full recovery of activity, surprisingly reconstitution with Zn2+ did not. Moreover, mutation of MtPncA His56 to alanine led to almost complete loss of metal ion binding and enzyme activity. These observations were taken to imply that His56 was directly involved in metal ion binding and that MtPncA binds ions in a different manner from PhPncA. MtPncA His56 corresponds to AbPncA His60, a residue that contributes significantly to the formation of the cation binding site by forming a hydrogen bond to Ser62 OG and positioning the serine to stabilize one of the waters that coordinates Zn2+ (Figure 2). His60 also stabilizes the coordinating His56 by van der Waals interactions and this histidine pair is strictly conserved in PncA sequences. Indeed, 16 of the 20 amino acids involved in cation coordination, substrate recognition and catalysis are strictly conserved between AbPncA and MtPncA, with a further three involving conservative substitutions (Figure S8). This suggests that all PncA enzymes share metal ion binding properties and mechanism of action. Structural data on recombinant enzyme and metal ion analysis of native MtPncA would further clarify this.
Over 60 M. tuberculosis strains display pyrazinamide resistance due to pnca gene mutations and more than 50% of these localize changes to three short polypeptide sections.[4,8] In AbPncA these correspond to β1-α1, α3-β3 segments and α6, near the active site and with residues important for substrate binding and catalysis. Of note is the strictly conserved Lys114. Pyrazinamide-resistant mutants of M. tuberculosis harbouring a lysine-threonine change in this position have been identified. This change may disrupt the hydrophilic environment required to allow for the exploitation of water in the proposed mechanism and/or would fail to align Asp16 for catalysis.
In summary, the precise nature of how nicotinic acid, and pyrazinoic acid bind AbPncA casts doubt on a previously proposed mechanism, which was based on modelling substrate in a manner that did not include direct interaction with Zn2+. Our data reveal that substrate recognition involves interaction between Zn2+ and the pyridyl nitrogen and the position of functional groups then allows us to propose a new mechanism for nicotinamidase activity and one which likely applies to the activation of the anti-tuberculosis drug pyrazinamide.