Crystal Structure of Mycobacterium tuberculosis MenB, a Key Enzyme in Vitamin K 2 Biosynthesis*

Bacterial enzymes of the menaquinone (Vitamin K 2 ) pathway are potential drug targets because they lack human homologs. MenB, 1,4-dihydroxy-2-naphthoyl-CoA synthase, the fourth enzyme in the biosynthetic pathway leading from chorismate to menaquinone, catalyzes the conversion of O -succinylbenzoyl-CoA (OSB-CoA) to 1,4-dihydroxy-2-naphthoyl-CoA (DHNA-CoA). Based on our interest in developing novel tuberculosis chemotherapeutics, we have solved the structures of MenB from Mycobacterium tuberculosis and its complex with acetoacetyl-coenzyme A at 1.8 and 2.3 Å resolution, respectively. Like other members of the crotonase superfamily, MenB folds as an ( (cid:1) 3 ) 2 hexamer, but its fold is distinct in that the C terminus crosses the trimer-trimer interface, forming a flexible part of the active site within the opposing trimer. The highly conserved active site of MenB contains a deep pocket lined by Asp-192, Tyr-287, and hydrophobic residues. Mutagenesis shows that Asp-192 and Tyr-287 are essential for enzymatic catalysis. We postulate a catalytic mechanism in which MenB

Critical issues in the treatment and control of tuberculosis, a disease that kills more than two million people world-wide every year (1,2), include the role of this disease as a major opportunistic pathogen in patients with HIV-AIDS and the emergence of multidrug resistance strains of M. tuberculosis (MDRTB) (3)(4)(5). Based on the knowledge that menaquinone is the sole quinone in M. tuberculosis and that humans must obtain this vitamin (K 2 ) either in their diet or from intestinal bacteria, menaquinone biosynthesis may be an attractive novel target for anti-TB drug discovery. Consequently, we have initiated studies on the putative biosynthetic enzymes in this pathway from M. tuberculosis.
Quinones are lipid-soluble molecules that shuttle electrons between the membrane-bound protein complexes in the electron transport chain. In mammalian cells, the electron transport chain is located in the inner mitochondrial membrane where the membrane-soluble quinone is ubiquinone, a benzoquinone (coenzyme Q, Fig. 1). In prokaryotes, the electron transport chain sometimes utilizes menaquinone, a naphthoquinone (vitamin K 2 , Fig. 1), in addition to or instead of ubiquinone. The principal structural variations within the two main classes of quinones occur from the length of the isoprenoid chain. Thus, in mitochondria the ubiquinone has a side chain of 10 isoprene units (CoQ-10) while in Escherichia coli the ubiquinone has a side chain of 8 isoprene units (CoQ-8) (6). E. coli, a facultative anaerobe, utilizes ubiquinone (CoQ-8) under aerobic conditions, but uses menaquinone (MK-8) when grown anaerobically (6,7). Bacillus subtilis, a Gram-positive aerobe, contains only menaquinone (MK-7) (7). Thus, inactivation of genes encoding the menaquinone biosynthetic enzymes in B. subtilis produces an absolute requirement for menaquinone or a downstream metabolite in the growth media (8,9).
Several pieces of data support the notion that M. tuberculosis, like B. subtilis, utilizes only menaquinone in the electron transport chain. Firstly, studies classifying bacteria based on their quinone content identified only menaquinone and not ubiquinone in M. tuberculosis (6,10,11). Secondly, Goldman and co-workers (12,13) isolated the NADH oxidase system from M. tuberculosis and demonstrated that it catalyzed the reduction of naphthoquinones. Lastly, the M. tuberculosis genome contains homologs of most of the E. coli men genes (see below and genolist.pasteur.fr/TubercuList/, Ref. 14). Conversely, homologs of some of the key ubiquinone biosynthetic genes are absent. For example, there is no homolog of chorismate pyruvate-lyase (UbiC), the enzyme that converts chorismate into 4-hydroxybenzoate (15,16), the first committed step in ubiquinone biosynthesis.
The biosynthesis of menaquinone has been most heavily studied in E. coli (reviewed in Ref. 16) and, to a lesser extent, in Bacillus subtilis (9,17,18) and Mycobacterium phlei (19 -29). The proposed pathway from E. coli, where the primary menaquinone has a 40 carbon isoprene chain , is shown in Fig. 2. Chorismate, derived from the shikimate pathway, is initially converted into isochorismate by MenF, an isochorismate synthase (30) and then to 2-succinyl-6-hydroxy-2,4-cyclohexadiene-1-carboxylate (SHCHC) by the action of MenD, a thiamin-dependent enzyme (31,32). SHCHC is then dehydrated by MenC to give the aromatic compound O-succinylben-zoate (OSB) 1 (33). MenC, which is a member of the enolase superfamily, is perhaps the best studied enzyme from the E. coli pathway and an x-ray structure has been determined by Gerlt and co-workers (34) and Rayment and co-workers (35). Following the MenC reaction, OSB is then converted into 1,4dihydroxynaphthoic acid (DHNA) by the actions of at least two enzymes. Following the ATP-dependent synthesis of OSB-CoA by MenE (36), a member of the acyl-CoA ligase superfamily (37), the naphthoquinol skeleton of DHNA is generated via a cyclization reaction catalyzed by 1,4-dihydroxynaphthoyl-CoA synthase, MenB (27,38,39). It was previously thought that MenB also catalyzed the hydrolysis of DHNA-CoA to DHNA. However Meganathan (43) has recently proposed that yfbB, an open reading frame that follows menD in the E. coli operon, codes for a thioesterase that fulfills this function. Finally, in the last two steps of the pathway, menaquinone is synthesized from DHNA by the actions of MenA, which catalyzes the prenylation of DHNA, and by MenG, an S-adenosylmethionine (SAM)-dependent methyl transferase. In another example of divergence between mammals and prokaryotes, the polyprenyl pyrophosphate required by MenA is synthesized via a nonmevalonate isoprenoid pathway (40 -44).
Although electron transport in humans does not involve menaquinone, this compound plays an essential role in blood clotting, for example being involved in the ␥-carboxylation of glutamic acid residues in prothrombin (45). However, humans cannot synthesize menaquinone (vitamin K 2 ) and must obtain it in the diet or from bacteria present in the gut. Consequently, the menaquinone biosynthetic pathway in M. tuberculosis may be an attractive target for the development of novel chemotherapeutics.
Based on the premise that menaquinone biosynthesis is essential for survival of M. tuberculosis, we have initiated a series of studies to identify and characterize the mycobacterial enzymes involved in this pathway. Although not organized into a distinct operon as they are in E. coli, the M. tuberculosis homologs of menA, menB, menC, menD, menE, and menH are clustered in one region of the genome (Fig. 3). Devolving from our long standing interest in the enzymology of enoyl-CoA hydratase and other members of the crotonase superfamily (46), we have initially focused on the putative 1,4-dihydroxy-2naphthoyl-CoA synthase, MenB, from M. tuberculosis (Rv0548c) which is 48% identical and 62% similar (over 285 out of 314 residues) to the E. coli homolog and is a member of the crotonase superfamily (47)(48)(49). Here we report the 1.8 Å x-ray structure of the M. tuberculosis MenB, as well as the 2.3 Å structure of a complex with acetoacetyl-CoA (AA-CoA), which reveals the active site geometry and confirms that this enzyme is indeed a member of the crotonase superfamily as previously suggested (47)(48)(49). We also demonstrate that MenB synthesizes 1,4-dihydroxynaphthoyl-CoA (DHNA-CoA) from O-succinylbenzoyl-CoA (OSB-CoA), as expected, and identify active site residues required for catalysis. Based on these studies, we propose a catalytic mechanism that will guide the search for inhibitors of this key enzyme of the menaquinone pathway.

EXPERIMENTAL PROCEDURES
Materials-M. tuberculosis genomic DNA was obtained from the TB Research Materials Facility at Colorado State University.
Cloning, Overexpression, and Purification of MenB from M. Tuberculosis-Rv0548c, the gene encoding the putative 1,4-dihydroxynaphthoyl-CoA synthase (MenB) from M. tuberculosis, was obtained by PCR from genomic DNA and cloned into the pET-15b plasmid (Novagen). Use of the NdeI and XhoI restriction sites placed the menB gene inframe with an N-terminal His tag sequence. Protein expression was performed using BL21(DE3)pLysS cells. Following growth in 800 ml of LB media containing 0.2 mg/ml ampicillin and induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 12 h at 37°C, cells were harvested by centrifugation (5,000 rpm for 5 min at 4°C), resuspended in 10 ml of His-binding buffer (5 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9) and lysed by 5 passages through a French Press cell (12,000 psi). Cell debris was removed by centrifugation at 33,000 rpm for 90 min at 4°C, and the clarified supernatant was passed through a column containing 3 ml of His-bind resin (Novagen). The column was washed with 50 ml of His-binding buffer and 30 ml of wash buffer (60 mM imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH7.9) before MenB was subsequently eluted using 20 ml of elution buffer (0.5 M imidazole, 0.5 M NaCl, 20 mM Tris-HCl, pH 7.9). Fractions containing MenB were pooled and the imidazole immediately removed by chromatography on G-25 resin using 20 mM NaH 2 PO 4 , 0.1 M NaCl, pH 7.0, as the elution buffer. The concentration of MenB was determined by measuring the absorption at 280 nm using an extinction coefficient of 41,370 M Ϫ1 cm Ϫ1 calculated from the primary sequence.
Coupled Assay for MenB-OSB-CoA, the substrate for MenB, is unstable and decomposes relatively rapidly to OSB spirodilactone (Fig.  4). Consequently, we used a coupled assay with MenE, the preceding enzyme in the pathway, to synthesize OSB-CoA in situ. The menE gene from E. coli was amplified by PCR and ligated into a pET-15b plasmid in-frame with an N-terminal His tag sequence using the NdeI and XhoI restriction sites. Following overexpression in BL21(DE3)pLysS cells, MenE was purified by His tag affinity purification following a similar protocol as described for MenB. The concentration of MenE was determined as described for MenB, using an extinction coefficient of 104,770 M Ϫ1 cm Ϫ1 . Coupled assays contained OSB (130 M), ATP (130 M), CoA (2-130 M), MenE (2 M), and MenB (0.14 M). The formation of the MenB product was monitored at 392 nm using a CARY-100 spectrophotometer. Reactions were initiated by the addition of MenB following preincubation of the other reactants for 10 min to allow complete conversion of OSB to OSB-CoA. All kinetics were performed at 25°C. Product formation was also monitored using reversed-phase HPLC. Reaction mixtures were analyzed using a Vydac C18 analytical column and running a gradient of 0 -40% buffer B over 40 min at 1 ml/min. Buffer A was 50 mM NH 4 ϩ CH 3 COO Ϫ , pH 5.9 while buffer B was 100% methanol.
Crystallization and Structure Solution-Wild-type MenB was crystallized using the hanging drop vapor diffusion method, equilibrating a mixture of 1l of protein solution and 1l of reservoir solution containing 0.5-2% dioxane, 1.2-1.3 M (NH 4 ) 2 SO 4 , and 0.1 M MES pH 6.5 against an excess of reservoir solution. Crystals were soaked in reservoir solution containing 30% glycerol and subsequently cryo-cooled in liquid nitrogen. Data were collected at beamline X26C at the National Synchrotron Light Source at Brookhaven National Laboratory, equipped with an ADSC Quantum 4 detector. Diffraction data were indexed, integrated, and scaled using the HKL software (50). Crystals belong to space group P2 1 with a ϭ 90.4 Å, b ϭ 139.4 Å, c ϭ 142.0 Å, and ␤ ϭ 97.3°, and contain 2 (␣ 3 ) 2 hexamers per asymmetric unit (asu). Dynamic light scattering suggests that MenB is a hexamer in solution. The self-rotation function of the diffraction data is compatible with two hexamers of 32 point symmetry in the asu. The structure was solved by 1 The abbreviations used are: OSB-CoA, O-succinylbenzoyl-coenzyme A; AA-CoA, acetoacetyl-coenzyme A; DHNA-CoA, 1,4-dihydroxy-2naphthoyl-CoA; rms, root mean square; asu, asymmetric unit; NCS, non-crystallographic symmetry; MES, 4-morpholineethanesulfonic acid; HPLC, high performance liquid chromatography. Structure of MenB molecular replacement using the programs BEAST (51) and COMO (52) for the rotation and translation search, respectively and the structure of the hexamer of enoyl-CoA hydratase (PDB code 1DUB) as a search model. The presence of two hexamers per asu allowed 12-fold density averaging with the program DM (53) to extend phases from 4 Å to 1.8 Å resolution. The main chain of various loops and the C-terminal residues 270 -314 were rebuilt using the program O (54). Side chains from enoyl-CoA hydratase were replaced with those from MenB using Swiss-Pdb Viewer (55) and fitted into the electron density using the program O. 12-fold non-crystallographic symmetry (NCS) restraints were maintained throughout the entire refinement process using REF-MAC (56,57). The tightness of constraints was chosen to minimize the free R-value. The average root mean square deviation of equivalent main chain and side chain atoms of NCS-related subunits is 0.05 Å and 0.30 Å, respectively.
MenB crystals were soaked with AA-CoA at a concentration of 2.2 mM. Crystals were cryo-protected and data were collected as described above. The soaked crystals were isomorphous to the unsoaked MenB crystals, allowing difference Fourier methods to be used for the calculation of electron density maps. AA-CoA was modeled into the density and refined at an occupancy of 0.75.
Preparation and Kinetic Assays of MenB Mutants-QuikChange mutagenesis (Stratagene) was used to generate the R133A, Y287F, D185N and D192N MenB mutants. Following DNA sequencing, the mutant enzymes were overexpressed and purified using the method described for wild-type MenB. Subsequently, the coupled assay with MenE was used to assay each of the mutant MenB proteins.

Enzymatic Activity of M. tuberculosis MenB-Wild-type
MenB was assayed using a coupled assay in which OSB-CoA, the substrate for MenB, was generated in situ by the action of MenE on OSB, ATP and CoA (Fig. 2). Sufficient MenE was included in reaction mixtures so that OSB-CoA was formed from OSB and CoA rapidly enough to preclude significant decomposition of OSB-CoA to the spirodilactone prior to the addition of MenB. In the presence of MenB, UV-visible absorption scans revealed the time-dependent formation of a species with max at 392 nm that corresponded to a new peak upon HPLC analysis eluting at 39.7 min, while OSB, ATP, and CoA eluted at 6.7, 5.1, and 21.6 min, respectively. The product peak obtained by HPLC was characterized by two absorption bands with max values at 260 and 392 nm, consistent with the presumption that it was 1,4-dihydroxy-2-naphthoyl-CoA (DHNA-CoA). Assuming that only CoA absorbed at 260 nm, the observed product absorbance ratio A 260 /A 392 of 4 allowed the estimation of an extinction coefficient at 392 nm of 4,000 M Ϫ1 cm Ϫ1 for the product. Using this extinction coefficient, wild type MenB had k cat and K m values of 14.9 Ϯ 0.4 min Ϫ1 and 7.3 Ϯ 0.9 M, respectively, assuming that the concentration of OSB-CoA in the reaction mixture was equal to the limiting MenE substrate (CoA). To confirm that the product of the MenB-catalyzed reaction was indeed DHNA-CoA, a reaction mixture in 20 mM ammonium acetate (pH 7) was analyzed using ESI-MSMS before and after the addition of MenB. Mass spectrometry (M ϩ H) revealed the MenB-dependent formation of a single species with a mass of 954.1 Da consistent with the  Table I.
MenB forms a 208kDa (␣ 3 ) 2 hexamer (a dimer of trimers) with two hexamers in the asymmetric unit ( The fold of each of these enzymes is fundamentally identical except for domain swapping of the C-terminal residues (60). In three-dimensional domain swapping, one domain of a subunit is replaced by the same domain from an identical subunit, resulting in an intertwined oligomer (66). The two C-terminal ␣-helices of the crotonase fold (corresponding to ␣9 and ␣10 spanning residues 271-288 and 290 -301 in MenB) are oriented in different directions such that they are part of different active sites in the hexameric structures (Fig. 6). In ⌬ 3 -⌬ 2 -enoyl-CoA isomerase, 6-oxo camphor hydrolase, and methylmalonyl CoA decarboxylase, the C terminus folds back onto the core domain such that it covers the active site of its own subunit (Fig. 6D). This is in contrast to enoyl-CoA hydratase, 4-chlorobenzoyl CoA dehalogenase, dienoyl CoA isomerase, and the human AUH protein, in which the C terminus protrudes away from the core domain and covers the active site of an adjacent subunit within the trimer (Fig. 6C). MenB is the first structurally characterized superfamily member that uses its C terminus to cross over the trimer-trimer interface, forming part of the active site of a subunit on the opposing trimer (Fig. 6, A and B). Regardless of the way the C termini of these proteins fold, helices ␣9 and ␣10 are in equivalent positions with respect to the active site, and contain residues that are conserved across the superfamily and interact with the CoA moiety of their respective substrates. A minor variation is a helix in methylmalonyl CoA decarboxylase (equivalent to ␣9 in MenB) that is pushed out toward the trimer interface.
A smaller difference between MenB and other crotonase superfamily structures concerns the disordered region spanning residues 108 -125 in MenB located near its active site. Throughout the crotonase superfamily, this region shows variability in sequence and structure (Fig. 7). In enoyl-CoA hydratase, this region has been predicted to be important in controlling the relative reaction rate of varying substrates by reorienting itself depending on the size of the substrate's alkyl chain (67). In MenB the substrate does not contain an alkyl chain, and the function of this region is unclear.
MenB in Complex with Acetoacetyl-CoA-Initially, MenB  is the ith measurement and ͗I͘ is the weighted mean of all measurements of I. ͗I͘/͗I͘ indicates the average of the intensity divided by its average standard deviation. Numbers in parentheses refer to the respective highest resolution data shell in each data set. R cryst ϭ ¥ʈF o ͉ Ϫ ͉F c ʈ/¥͉F o ͉ where F o and F c are the observed and calculated structure factor amplitudes. R free same as R cryst for 5% of the data randomly omitted from the refinement. Ramachandran statistics indicate the fraction of residues in the most favored, additionally allowed, generously allowed, and disallowed regions of the Ramachandran diagram, as defined by the program PROCHECK (59 crystals were soaked with DHNA-CoA, the product of the MenB reaction (Fig. 2). Difference density was observed only for the CoA moiety in the active site, but not for the DHNA moiety. We then soaked MenB crystals with AA-CoA and solved the structure of the complex at 2.3 Å resolution. Both DHNA-CoA and AA-CoA contain a ␤-keto-thioester (the (CϭO)CH 2 (CϭO)S fragment), but AA-CoA lacks the bicyclic aromatic ring of DHNA. The structure reveals that AA-CoA binds to MenB in a similar orientation as seen in enoyl CoA hydratase and other crotonase family members. The binding cleft is lined on one side by the N-terminal part of one subunit and on the other side by the residues C-terminal of Asp-277 of a subunit located on the opposing trimer (Fig. 8).
MenB binds to AA-CoA through hydrogen bonds, salt bridges, and hydrophobic interactions (Fig. 8). The adenine moiety is sandwiched between Phe-299 and Ala-60. In the unliganded structure, the distance between these residues varies significantly in the 12 copies of the asu (the average C␣-C␣ distance is 15.0 Å, with values ranging from 14.5 Å to 15.7 Å). In the AA-CoA MenB complex, this distance is smaller and less variable (the average distance is 14.5 Å, with values ranging from 14.2 Å to 14.7 Å). Superposition of the unliganded and AA-CoA bound structure shows that the C-terminal helix ␣10 (residues 290 -301) is displaced by up to 1.1 Å in the direction of the substrate binding pocket (Fig. 9A). In both the AA-CoAbound and the unliganded structure, the C-terminal residues have the highest mobility as indicated by high B-factors (in the unliganded structure, the average B-factor for all atoms in residues 271 through 314 is 53.4 Å 2 , compared with 30.9 Å 2 in the remainder of the molecule). A comparison of the MenB/AA-CoA structure with the enoyl CoA hydratase/AA-CoA structure reveals that the adenine is sandwiched tighter in the enoyl CoA hydratase with a distance between the C␣ atoms of Phe-279 and Ala-60 (corresponding to Phe-299 and Ala-60 in MenB) of 13.2 Å and a distance of 4.1 Å between Phe-279 and C5 of the adenine base. In MenB, Phe-299 is positioned too far away from the adenine (the distance varies from 4.2 to 5.3 Å in the 12 active sites present in the asu, with an average value of 4.8 Å) to interact productively. However, the phenylalanine is strictly conserved, suggesting that in the MenB complex with OSB-CoA, Phe-299 moves closer to the adenine.
The hydrogen bonds to AA-CoA involve only main chain atoms of MenB. The thioester oxygen of the mercaptoethylamine unit hydrogen bonds to the nitrogen atoms of the strictly conserved residues Gly-105 and Gly-161, whose equivalent residues Ala-98 and Gly-141 in enoyl CoA hydratase form what is referred to as the oxyanion hole (49). Furthermore, Gly-105 O and Gln-107 N interact with N6 and N1 of the adenine ring, respectively. The conserved positively charged residues Arg-58 and Lys-302 are in close proximity to the negatively charged phosphates of CoA. The indole ring of Trp-57 engages in hydrophobic interactions with the dimethyl group of the pantothenate unit. The electron density observed for the acetoacetyl moiety is weak except for the thioester carbonyl. This apparent flexibility of the acetoacetyl moiety might be related to the differences between the real substrate/product and AA-CoA.
Apart from the movement of the C-terminal helix described above, the main chain of MenB undergoes no significant change upon binding of AA-CoA; the rms deviation of all C␣ atoms is 0.21 Å. Nevertheless, there is a notable change in the side chain conformation of Tyr-287, which points away from the binding pocket in the structure of unliganded MenB. Upon binding of AA-CoA, the side chain of Tyr-287 swings toward Ser-190 in the active site pocket in 2 out of 12 subunits, maintains the conformation observed in the unliganded structure in 3 subunits and becomes disordered (with difference density for the side chain observed in both locations) in the remaining 7 subunits. The two conformations are shown in Fig. 8. A conformational change of the Tyr-287 side chain toward Ser-190 places the phenol oxygen group in closer proximity to the active site.
Location and Properties of the Active Site Pocket-Where does the aromatic ring of MenB's substrate OSB-CoA bind? We have mapped the conservation of MenB sequences of different organisms on the structure of M. tuberculosis MenB (Fig. 9B). The highest degree of conservation on MenB's surface is observed in a deep pocket next to Gly-161. This pocket lies in between a loop in extended conformation (residues 190 -192) and three helices (Fig. 8). Residues Leu-134, Ile-136, and Leu-137 of one helix as well as Gly-280, Thr-283, and Tyr-287 from a helix of the neighboring subunit form a hydrophobic lining on one side of the pocket, while the other side of the pocket contains the charged residue Asp-192. A calculation of the electrostatic surface potential of MenB shows that the bottom of the pocket is negatively charged, while the sides are positively charged (Fig. 9C). The positive charge is not due to basic residues, but rather due to the dipole of the helix spanning residues 161 to 168. This positive charge is located near the strictly conserved Gly-161 of the oxyanion hole. The loop that is disordered in the MenB crystals is located above the active site pocket. If ordered upon substrate binding, this loop could potentially obstruct access to the pocket while the reaction proceeds. If residues of this loop come into close proximity with the substrate, they could also play a role in binding or catalysis. However, only Arg-110, Gly-111, and Gly-114 of this region are strictly conserved. Finally, comparison of the MenB structure FIG. 6. Domain swapping of the C terminus in MenB and its structural neighbors enoyl CoA hydratase (1DUB) and methylmalonyl CoA decarboxylase (1EF8). For each structure, the three C-terminal helices are shown as cylinders. One subunit is shown in light blue, with the swapping helices ␣9 and ␣10 highlighted in dark blue, while a second subunit is shown in light green with the swapping helices in dark green. Helices ␣9 and ␣10 form part of the active site, which binds to the acyl-CoA derivatives shown in ball-and-stick representation. The loop between ␣8 and ␣9, which acts as a hinge in the domain swapping, is shown in red. MenB is shown looking down the 3-fold axis (A) and rotated by 90 degrees (B), illustrating how the green subunit and the C-terminal helices of the blue subunit and vice versa form the active sites by crossing the trimer-trimer interface indicated by horizontal black lines. C in enoyl CoA hydratase, the C-terminal helices of the blue subunit form a functional active site with the adjacent green subunit in the same trimer. D, in methylmalonyl CoA decarboxylase, the C-terminal helices fold back on its own subunit to form an independent active site.
with that of enoyl-CoA hydratase reveals that Asp-185 in MenB occupies a similar position to Glu-164 in enoyl-CoA hydratase. This is significant since Glu-164 is a key catalytic residue in the reaction catalyzed by enoyl-CoA hydratase (68). While it is intriguing to speculate that Asp-185 in MenB plays a similar role to that of Glu-164, we note that Asp-185 is not conserved in the MenB family, and is sometimes replaced by a glycine (Fig. 7).
Active Site Mutants Demonstrate Residues Important for Catalysis-The reaction catalyzed by MenB involves the formation of a carbon-carbon bond through an intramolecular Claisen-like condensation (formally a Dieckmann condensation). Enzymatically catalyzed Claisen condensations play an important role in fatty acid biosynthesis (69), and require that at least one of the reacting carboxylates is activated by formation of a thioester with CoA. For the reaction to proceed, the thioester has to be deprotonated, leading to an enolate intermediate. This reactive species acts as a nucleophile in a substitution at the second carboxylate, yielding a ␤-keto ester. Enzymatic catalysis proceeds by stabilizing the enolate anion (through hydrogen-bond donors or charged groups) and providing a good leaving group for the carboxylate (through prior reaction to a thioester with another CoA or with an active site cysteine).
The MenB active site contains only a few potential proton donors and acceptors that are strictly conserved and sufficiently close to Gly-161 (i.e. close to the enolate anion) to directly interact with the reactants. The obvious candidates are Ser-190 and Asp-192, in addition to Asp-185, although the latter residue is not conserved in the MenB family. Interestingly, there are no nearby histidines or cysteines, residues that are implicated in the condensation reactions of fatty acid biosynthesis (69). An additional candidate is Tyr-287. Its hydroxyl group points away from the active site in the unliganded structure of MenB. Upon ligand binding, however, the side chain of Tyr-287 becomes disordered in the majority of the subunits and is positioned close to the active site Ser-190 in two others (Fig. 8).
We have prepared the following point mutants of MenB to study the effects of active site residues on structure and catalytic activity of MenB: R133A, D185N, D192N, and Y287F. All mutants expressed well and were soluble. We obtained crystals of Y287F and R133A. Preliminary diffraction data (results not shown) indicate that the overall fold of the mutants has not changed in comparison to the wild type enzyme. All four MenB mutants, R133A, Y287F, D185N, and D192N, showed no activity when assayed under the same conditions as those used for wild-type MenB. R133A exhibited 20% wild-type activity when used at a concentration 30-fold higher than that used for the wild-type enzyme. However, no enzymatic activity was observed for D192N and Y287F, even at 5 M enzyme concentra-tions. The conservative exchanges in the mutants make it highly unlikely that the global structure of MenB is influenced by the mutations; our preliminary crystallographic data supports this statement. Therefore, our mutational data indicate that Asp-192 and Tyr-287 are either necessary for substrate binding or for enzymatic catalysis. Because the mutants are completely inactive rather than impaired, kinetic assays fail to distinguish between these two possibilities. DISCUSSION The structural homology of MenB with the crotonase family suggests that catalysis by MenB involves formation of a CoA thioester enolate, which is stabilized by the oxyanion hole characteristic of this class of enzymes. To understand how MenB catalyzes the ring closure of OSB, we have to identify the proton acceptor responsible for enolate formation and also consider how the benzoate is modified to contain a good leaving group. In addition, a mechanism should address how MenB prevents formation of the spirolactone to which OSB reacts in solution, but rather guides OSB into a conformation that brings the two carboxylates into close proximity.
To study the constraints enforced by the active site of MenB on the conformation of its substrate OSB-CoA, we modeled OSB-CoA into the MenB structure. We placed OSB-CoA such that the thioester of OSB-CoA hydrogen-bonds to Gly-105 N and Gly-161 N as observed in the MenB-AA-CoA complex. With this constraint, OSB is forced to position itself in its productive conformation with the benzoate carboxylate close to the thioester in order to accommodate the aromatic ring into the active site pocket. This positions the hydrophobic ring into favorable contact with Leu-134, Ile-136, and Leu-137 (Fig. 10). It is conceivable that the CoA moiety of the substrate binds first, allowing the OSB moiety to sample different conformations until the productive one is obtained. In this conformation the benzene ring would bind tightly into the hydrophobic pocket. The flexibility of the loop (residues 108 -125) above the active site cleft observed in the crystal structure would allow the OSB moiety of the substrate to approach from above after its CoA To show the narrowing of the pocket, corresponding subunits of the unliganded and bound structure were superimposed and the transformation applied to the neighboring subunits, of which residues 277Ј through 314Ј are shown. Cyan, unliganded; yellow, AA-CoA-bound structure. The particular active site shown (one of the 12 active sites in the asu) narrows by 1.1 Å, which is more than the average value of 0.6 Å (calculated for the 12 active sites in the asu). B and C, conserved residues (B) and electrostatic potential (C) on the surface of the active site cleft. AA-CoA is shown in all-bonds representation. Gln-107, which bridges over the substrate analog and divides the active site cleft into two halves, was omitted from the surface calculation for clarity. Residues strictly conserved in 9 MenB sequences (see Fig. 7) are shown in green. The calculation of the electrostatic potential at 100 mM salt included partial charges of the backbone peptides; red, negative, blue, positive potential with a range of Ϯ 10kT. The reaction scheme is similar to that described in Ref. 16. The initial rapid reversible abstraction of the pro-2R succinyl proton by the benzoic carboxylate (Step 1) results in the formation of a resonance-stabilized carbanion that subsequently attacks the protonated carboxyl group (Step 2). The tetrahedral intermediate that is formed eliminates H 2 O to generate the keto form of the product (Step 3) which then enolizes (Step 4). In Step 3, a proton (shown in the scheme as originating from an unidentified general acid AH) is required to assist in the elimination of H 2 O. In addition, since the pro-3R succinyl proton is stereospecifically retained in the product, we suggest that keto-enol tautomerism (Step 4) occurs on the enzyme and involves the participation of Tyr-287. moiety has already been tethered to MenB.
How might Asp-192 and Tyr-287 function in MenB to help formation of the enolate anion and provide a good leaving group for the benzoate? It is possible that these residues interact with the keto group of the substrate making them necessary for substrate binding rather than catalysis. In this respect, it is interesting to note that both Tyr-287 and Asp-192 are conserved in the ring cleavage enzyme BadI. This 2-ketocyclohexanecarboxyl CoA hydrolase is part of the benzoate degradative pathway in Rhodopseudomonas palustris (70). It essentially catalyzes the back-reaction of the ring closure catalyzed by MenB; however, its product does not have an additional keto functionality like the substrate of MenB.
Based on our findings, we suggest a plausible mechanism that will be investigated in future experiments (Fig. 10). In the reaction catalyzed by MenB, the aliphatic OSB carboxyl is activated by formation of a CoA ester. However, unlike the standard enzymatic Claisen reaction, the electrophile in the reaction is presumably the carbonyl group of the unactivated aromatic OSB carboxylate. Based on model building, we propose a mechanism in which the active site positions the OSB succinyl ␣-carbon in close proximity to the aromatic carboxyl as a prelude to nucleophilic attack. The carboxylate group is then ideally positioned to abstract the ␣-proton resulting in a protonated carboxyl and the required resonance-stabilized carbanion. This step is rapid given the observation that the enzyme stereospecifically exchanges the pro2(R) OSB proton with solvent deuterium (29). Nucleophilic attack of the ␣-carbon on the carboxyl then leads to the formation of a tetrahedral intermediate that collapses with the elimination of water.
Several challenges must be met by the enzyme in order to catalyze the reaction. The main chain amides of Gly-105 and Gly-161 provide hydrogen bonds to stabilize the carbanion thereby lowering the pK a of the OSB ␣-protons. Tyr-287 and Ser-190 hydrogen bond to the OSB keto group in order to position the OSB in the active site, while Asp-192 and Asp-185 are in close proximity to the OSB carboxylate, thereby raising the pK a of this group to promote proton abstraction from the OSB ␣-carbon. Since the OSB pro3(R) proton is selectively retained in the product, we believe that enolization of the product is guided by the enzyme and suggest that Tyr-287 acts as a proton shuttle during product enolization. The x-ray structure of 1,4-dihydroxynaphthoyl-CoA synthase (MenB) represents a first glance at this key enzyme in the menaquinone (Vitamin K 2 ) biosynthetic pathway of the pathogen M. tuberculosis. Description of the active site geometry, identification of residues that are not common to the crotonase superfamily, but are crucial for the activity of MenB, and a proposal of a mechanism for the carbon-carbon bond formation catalyzed by MenB are steps toward designing potent inhibitors of MenB. Thus the present results, which are derived from structural data on one hand and biochemical characterization of novel mutants on the other, may aid in the development of new drugs for the treatment of tuberculosis.