Crystal structure of the thioesterification conformation of Bacillus subtilis o-succinylbenzoyl-CoA synthetase reveals a distinct substrate-binding mode

o-Succinylbenzoyl-CoA (OSB-CoA) synthetase (MenE) is an essential enzyme in bacterial vitamin K biosynthesis and an important target in the development of new antibiotics. It is a member of the adenylating enzymes (ANL) family, which reconfigure their active site in two different active conformations, one for the adenylation half-reaction and the other for a thioesterification half-reaction, in a domain-alternation catalytic mechanism. Although several aspects of the adenylating mechanism in MenE have recently been uncovered, its thioesterification conformation remains elusive. Here, using a catalytically competent Bacillus subtilis mutant protein complexed with an OSB-CoA analogue, we determined MenE high-resolution structures to 1.76 and 1.90 Å resolution in a thioester-forming conformation. By comparison with the adenylation conformation, we found that MenE's C-domain rotates around the Ser-384 hinge by 139.5° during domain-alternation catalysis. The structures also revealed a thioesterification active site specifically conserved among MenE orthologues and a substrate-binding mode distinct from those of many other acyl/aryl-CoA synthetases. Of note, using site-directed mutagenesis, we identified several residues that specifically contribute to the thioesterification half-reaction without affecting the adenylation half-reaction. Moreover, we observed a substantial movement of the activated succinyl group in the thioesterification half-reaction. These findings provide new insights into the domain-alternation catalysis of a bacterial enzyme essential for vitamin K biosynthesis and of its adenylating homologues in the ANL enzyme family.

which operates only in microorganisms and is an attractive target for development of novel antibiotics (4 -6). It catalyzes ATP-dependent activation of o-succinylbenzoic acid (OSB) like other acyl-CoA synthetases (Fig. 1) and is an adenylating enzyme of the ANL family (7), which encompasses acyl/aryl-CoA synthetases (ACS), adenylation domains of non-ribosomal peptide synthetases, and firefly luciferases. The adenylating enzymes in the ANL family are highly diverse with low sequence homology (ϳ20% identity), and their representatives have been found to reconfigure their active site in two different active conformations, one for the adenylation half-reaction and the other for the thioesterification half-reaction, in a domainalternation catalytic mechanism (7). Formation of the two active conformations involve rotation of the small C-terminal domain (C-domain) relative to the large N-terminal domain (N-domain) by a large angle, as much as 140°as observed for bacterial acetyl-CoA synthetase (8) and 4-chlorobenzoate: coenzyme A ligase (4CBL) (9). Supporting evidence for this catalytic mechanism has been accumulated in the last 2 decades (10 -13), and a few ANL members have been structurally characterized for both active conformations (9, 14 -20).
Because of its medicinal value as an antibiotic target, MenE has been scrutinized for its catalytic mechanism using kinetic and structural methods (5,21). Consistent with its expected use of the domain-alternation catalytic strategy, Bacillus anthracis MenE has been kinetically analyzed to adopt an ordered Bi Uni Uni Bi Ping-Pong catalytic mechanism in which the enzyme sequentially binds ATP and OSB (Bi), releases pyrophosphate (Uni) after adenylation, binds CoA (Uni) as the third substrate, and then releases OSB-CoA and AMP (Bi) after thioesterification (22). Using crystallographic methods, ligand-free MenE orthologues from various sources have been shown to adopt an open conformation with their C-domains rotated at a widely variable angle relative to the N-domains (PDB code 3ipl) (23,24), which is closed by binding of ATP through an open-closed conformational change to form the compact conformation for the first half-adenylation reaction as captured in the crystal structure of Bacillus subtilis MenE (bsMenE) in complex with ATP (24). The bsMenE-ATP structure confirms the role of transition state stabilization for a strictly conserved lysine residue (Lys-471 in bsMenE) and reveals the structural details of the carboxylate-binding site created by ATP binding. In addition, this structure shows that the strongest interaction between N-and C-domains occurs at the ␤-phosphate in the triphosphate moiety of ATP to allow easy disassembly of the compact adenylation conformation in an encrypted C-domain release mechanism, which is required after adenylation to allow transition to the thioester-forming conformation in the domain-alternation mechanism. Importantly, this complex structure solidifies a general mode of gripping interactions between the P-loop of ANL enzymes with ATP, which forms the basis for the critical structural role of ATP in configuring the adenylation-active conformation (24).
More recently, we have solved a new bsMenE structure in complex with the o-succinylbenzoyl-adenosine monophosphate (OSB-AMP) (26), which is subtly different from a previously determined one, also in complex with OSB-AMP (23). This new structure represents the state of the enzyme with the intermediate in a strained conformation formed right after adenylation and provides direct evidence for the in-line backside nucleophilic attack mechanism for the adenylation halfreaction by comparison with the pre-adenylation bsMenE-ATP structure (24). Structural comparison also revealed structural changes in both the reactants and the active site that contribute to catalysis of the adenylation reaction. Moreover, with the help of site-directed mutagenesis and kinetic analysis, the conserved His-196 was found to likely desolvate the adenylation active site instead of stabilizing the adenylation transition state. Furthermore, comparison of the new structure with the previous OSB-AMP-bound structure allows us to propose a mechanism of the post-adenylation C-domain release, which may be driven by the relaxation of the strained succinyl group from an eclipsed-to anti-conformation. These results not only allow us to better understand the adenylation reaction mechanism, but also provide clues on how the enzyme changes from its active adenylation conformation to the thioesterification conformation in the domain-alternation mechanism.
To further understand the MenE catalytic mechanism, it is important to know the structure of the active thioesterification conformation. However, this thioesterification conformation remains elusive despite a previous attempt to determine its atomic structure (23). In this study, we synthesized a non-hydrolyzable analogue of the OSB-CoA product, o-succinylbenzoyl-amino coenzyme A (OSB-NCoA), in an attempt to trap the enzyme in the second active conformation. Co-crystallization of the product analogue with the enzyme was successful, but the resulting crystals gave rise to uninterpretable diffraction patterns. To overcome this problem, two non-conserved residues were mutated to obtain a catalytically active variant I454R/ A456K (IRAK) and successfully co-crystallized with OSB-NCoA with or without AMP for determination of the targeted thioesterification conformation structure at 1.76 and 1.90 Å resolution, respectively. These structures reveal the unique thioesterification active site of the adenylating enzyme and provide fresh insights into its domain-alternation catalysis.

Crystallization
To capture bsMenE in a thioesterification state, we synthesized a non-hydrolyzable analogue of the OSB-CoA product, o-succinylbenzoyl-amino coenzyme A (OSB-NCoA), following a reported chemoenzymatic method (27) and used it to co-crystallize with the wild-type enzyme. The co-crystallization was successful, affording large crystals that diffracted to ϳ1.7 Å in in-house tests. However, most of the reflection spots, either at high or low diffraction angles, suffered severe twinning, which led to uninterpretable data sets. This diffraction problem was not overcome by optimizing the cryoprotectant or the crystallization conditions but was probably due to the intrinsic disorder of the protein complex. Suspecting that this disorder may be caused by loose ligand binding, two non-conserved residues Ile-454 and Ala-456 were mutated into arginine and lysine, respectively, to make a double mutant IRAK (I454R/A456K) to strengthen the binding between OSB-NCoA and the enzyme. This was based on the fact that these two residues correspond to Arg-475 and Lys-477, respectively, of 4CBL in sequence alignment, which form hydrogen bonding or ionic interaction with the adenine ring and 3Ј-phosphate of the enzyme-bound product analogue 4-chlorobenzoyl-CoA (9). As shown in Table  1, the catalytic efficiency (k cat /K m ) of the double mutant is moderately decreased by 15-26-fold toward ATP or CoA but remains essentially unchanged toward the OSB substrate. In comparison, the single mutant I454R exhibits a very similar level of activity changes as the double mutant, indicating that the other mutation A456K has a negligible effect on catalytic function. The activity results clearly showed that the double mutant is fully competent in catalysis without changing the catalytic mechanism.
Co-crystallization of the IRAK double mutant with OSB-CoA in the presence or absence of AMP was very successful,

Thioester-forming structure in MenE domain alternation
affording large cubic single crystals. These single crystals diffracted to high resolution without the twinning problem observed for the wild-type enzyme and were used to determine the structures of the binary complex (IRAK-OSB-NCoA) and the ternary complex (IRAK-OSB-NCoA-AMP) at a resolution of 1.90 and 1.76 Å, respectively, by molecular replacement using the N-domain of ligand-free bsMenE (PDB code 5BUQ) (24) as the template. The statistics of data collection, data processing, and structural refinement are summarized in Table 2. In a quick examination of the structures, the mutated residues were found to locate on the protein surface at a far distance from the ligand without forming the anticipated interactions or other interactions with other amino acid residues. The improved crystallization properties of the double mutant may be a result of the improved hydrophilicity of the protein surface caused by the mutations.

Overall structure
There are four full-length protein molecules in the unit cell of both the binary and tertiary IRAK complexes in the triclinic P1 space group. These protein molecules take essentially the same conformation and are superimposable with an r.m.s.d. of Ͻ0.20 Å over 420 C␣ atoms. Their structures are very well defined by continuous electron densities from the N to C terminus with few rotameric or torsional errors, except for Ser-293 that remains as a Ramachandran outlier as observed in all the previous bsMenE structures (PDB code 3ipl) (23,24). In each unit cell, two homodimers (chains A and C form one dimer and chains B and D form another dimer) are formed by back-toback contacts of the N-domains (residues 1-379) of the monomers, whereas an active site is formed far from the dimer interface between the N-and C-domains (residues 393-486) with participation of the residues from the linker (residues 380 -392) and binds an OSB-NCoA ligand in each monomer (Fig. 2).
Interestingly, the two OSB-NCoA ligands in any homodimer exhibit significant differences in the ribose ring together with the associated 3Ј-phosphate, whereas their acyl group and the amino pantetheinyl group are superimposable with exactly the same orientations (Fig. 2, D-F). Take one dimer of either complex as an example: the ribose ring of OSB-NCoA takes a normal endo-pucker conformation with the C3Ј-phosphate at the axial position in chain C (Fig. 2E), and the 3Ј-phosphate of chain A in the same dimer occupies the equatorial position in an almost flat ribose ring (Fig. 2D). This difference is caused by  three hydrogen bonds between the chain A 3Ј-phosphate with the backbone or side chain of Ser-220Ј, Val-221Ј, and Ser-222Ј from chain CЈ of the neighboring unit cell (Fig. 2D), whereas such crystal packing interactions are not found for the chain C ligand (Fig. 2E). Moreover, the two dimers in the unit cell of the ternary complex are different, of which the dimer including chain A and chain C contains AMP in each chain (Fig. 2, D and E) and the other dimer contains no AMP similar to those in the binary complex. AMP in the former dimer has a low occupancy (0.60 in chain A and 0.58 in chain C) and is much poorer in electron density compared with OSB-NCoA. For simplicity, chain D of the binary complex and chain C of the ternary complex without the crystal packing interactions are used as the representatives of the two complexes.

Thioester-forming conformation
Although both IRAK complexes maintain the same topology and di-domain folding as the previously determined bsMenE structures (24, 26), their C-domains are oriented very differ- ently relative to the large N-domains. These new structures are, however, closely similar to that of 4CBL (9), ACS (8), and ACSM2A (19) in a thioesterification state. This comparison clearly shows that the IRAK complexes take a thioesterification conformation dedicated for the second half-reaction. In these IRAK complexes, a compact domain interface with a 1021.8 Å 2 surface is formed from many N-domain residues and residues on one side of the C-domain. In comparison, a smaller interdomain interface (ϳ751.5 Å 2 ) is formed between the same set of N-domain residues and residues on the other side of the C-domain of bsMenE in the pre-and post-adenylation states (23,26). As calculated by the DynDom server (29), the C-domain rotates by 139.5°from its position in the adenylation conformation to that in the ligand-bound IRAK structures, hinging on Ser-384 in the linker region (Fig. 3A). To accommodate this C-domain rotation, the Ser-384 dihedral angles ⌽ and ⌿ drastically change from Ϫ62.4 to Ϫ110.5°and from Ϫ9.3 to Ϫ24.4°, respectively.
The large C-domain rotation significantly reconfigures the enzyme active site, which binds the AMP product and OSB-NCoA product analogue in two interconnected, perpendicular tunnels (Fig. 3A). One tunnel binds the AMP and the benzoic acid moiety of the OSB-NCoA ligand in the same positions as the corresponding components of OSB-AMP in the adenyla-tion conformation (26) and is thus referred to as the adenylate tunnel. The other tunnel is linear and binds the amino pantetheinyl portion of the OSB-NCoA analogue, which is thus referred to as the CoA tunnel. The adenylate and CoA tunnels meet at the mid-point of the former, where the succinyl group is twisted at a nearly vertical angle with its SC2-SC3 ethylene group taking an energetically unfavorable gauche conformation and links the amino-CoA moiety to the benzoic acid moiety in the OSB-NCoA ligand (Fig. 2G). The product analogue is thus bound as a crowbar and is unable to dissociate from the protein if the interdomain interface is not disassembled (Fig. 3A).
The adenylate tunnel contains AMP close to the protein surface and exposes it to the solvent, whereas the CoA tunnel is also open to solvent at one end and exposes the 3Ј-phosphorylribosyl group of the adenosine diphosphate moiety of the OSB-NCoA product analogue in the bulk solution (Fig. 3B). As noted earlier, OSB-NCoA is unable to dissociate from the protein in the thioesterification conformation due to its unique binding mode, whereas AMP is easy to escape to the bulk solvent due to the absence of blocking barriers in the exit channel. From this difference, it is obvious that AMP is released from the enzyme active site ahead of OSB-CoA after the two products are formed from the thioesterification reaction. This is consistent with the absence of the AMP product in one dimer and the considerably poorer electron density for AMP relative to OSB-NCoA in the other dimer in the unit cell of the ternary IRAK complex (Fig. 2, D and E).

Adenylate tunnel
In the adenylate tunnel (Fig. 3A), AMP in the ternary IRAK structure forms the same interactions with the N-domain as the AMP moiety of the OSB-AMP intermediate in the bsMenE-OSB-AMP structure; its adenine ring is sandwiched between the phenolic side chain of Tyr-286 and the backbone of Gly-263-Gly-264 -Gly-265, although its 6-amino group forms a hydrogen bond with the backbone carbonyl group of Ser-285, and its 2Ј-and 3Ј-hydroxyl groups form hydrogen bonds with the Asp-367 side-chain carboxylate. However, AMP has no interactions with the C-domain or the linker region. In contrast, the AMP moiety of OSB-AMP forms hydrogen bonds with Arg-382 of the linker region and Lys-471 of the C-domain in the bsMenE-OSB-AMP structure (26). Thus, AMP is significantly weakened in its interaction with the enzyme in the thioesterification conformation after remodeling of the adenylate tunnel, thus facilitating its release from the thioesterification active site. Another important change in the remodeled adenylate tunnel is the drastic rearrangement of the His-196 side chain, which forms a hydrogen bond with the bridging oxygen atom of OSB-AMP intermediate in the bsMenE-OSB-AMP structure but is reoriented by 37°to form an inter-domain hydrogen bond with Glu-392 of the C-domain in the thioesterification conformation. This movement of His-196 connects the adenylate tunnel and CoA tunnel together and exposes the OSB-AMP intermediate to the attack of the CoA substrate. This gate-keeping function of His-196 is consistent with that of the equivalent residues in other ANL enzymes (9,20).
The OSB-binding pocket is little changed in the adenylate tunnel remodeling by the large C-domain rotation because it is formed entirely by N-domain residues and is buried deep inside the protein. As shown in Fig. 4, it is composed of amino acid residues specifically conserved among MenE orthologues and forms the same binding interactions for the benzoic acid moiety of OSB-NCoA as those found in the bsMenE-OSB-AMP structure (26). Nonetheless, the newly solved IRAK-OSB-NCoA complex structure at 1.76 Å resolution reveals that water molecules are involved in the binding interaction. Specifically, the aromatic carboxylate of the ligand is stabilized through hydrogen bonding with three well-defined water molecules O1, O2, and O3 (Fig. 4A), of which O1 forms an additional hydrogen bond with the Ser-237 side chain and O2 further hydrogenbonds with the side chains of Ser-293 and Gln-294. These water-mediated interactions were not identified in the previously solved structure at a lower resolution (26).

CoA-binding tunnel
The large C-domain rotation in reconfiguration of the enzyme active site creates a new tunnel at the domain interface for binding of the amino-CoA moiety of the OSB-NCoA ligand. Its major portion is linear and extends 18.2 Å from the protein surface to the adenylate tunnel at a vertical angle (Fig. 3A), whereas its minor part is non-continuous to the linear portion and allows the amino CoA moiety of OSB-NCoA to bind in a hairpin conformation. The adenine ring at one end of the ligand hairpin is inserted between the aromatic side chains of Tyr-452 and Trp-422 to form a strongstacking interaction. This adenine ring is further stabilized by polar interactions, including hydrogen bonds with the side chain hydroxyl of Ser-389, ⑀-NH 2 of Lys-86, and the carbonyl oxygen in the pantoic moiety at 6Ј-NH 2 (Fig. 4A).
The 3Ј-phospho-ADP moiety of the ligand, which forms the "U" turn and links the pantetheinyl group to the adenine ring, is exposed in the solvent (Fig. 3B). The diphosphate group in the ADP moiety forms hydrogen bonds with four water molecules (O5, O9, O10, and O12) and chelates a Mg 2ϩ ion, which is further coordinated with the backbone carboxyl group of Phe-219 and three additional water molecules (O6, O7, and O8) to form a clearly defined octahedral coordination sphere (Fig. 4A). These interactions with water and the metal ion help to position the diphosphate in a conformation to make ϳ90°turn in the formation of the hairpin structure. Similar Mg 2ϩ chelation to the diphosphate group of the CoA ester ligand is found in 4CBL, which, however, only interacts with the ␣-phosphate of the CoA moiety and forms a square-planar coordination system with three additional water molecules (9). In contrast, no metal is chelated to the diphosphate group of the CoA ligand in all other known crystal structures of acyl/aryl-CoA synthetases in the thioesterification conformation.
The 18.2 Å linear binding tunnel of the pantetheinyl group is composed of residues from both N-and C-domains and binds the ligand with extensive van der Waals interactions and a few polar interactions. As shown in Fig. 4A, a tripeptide fragment, Leu-192-Pro-193-Leu-194, and a ␤-hairpin motif consisting of residues 237-243 from the N-domain form extensive van der Waals contacts with the ligand. In addition, a ␤-hairpin loop (Ser-389 -Gly-390 -Gly-391-Glu-392 from the linker region, designated as the SGGE motif) forms multivalent interactions with the ligand. Specifically, the backbone carbonyl oxygen atoms of Gly-390 and Gly-391 form hydrogen bonds with the nitrogen atoms of the two pantetheine amide groups, whereas the side chain ethylene group of Glu-392 forms van der Waals contacts with the ␤-alaninyl moiety of the pantetheinyl group. Besides these direct binding interactions, two water molecules, O4 and O13, are also involved in the ligand binding via hydrogen bonds (Fig. 4A).

Mutation of active-site residues
We mutated the conserved amino acid residues underlined in Fig. 4A to understand their contribution to the formation of the thioesterification conformation and the enzymatic catalysis. The resulting site-directed mutants, including S198A, S384P, S389A, E392A, W422A, and Y452A, were readily expressed and purified to homogeneity like the wild-type bsMenE. The pure mutant proteins were found similar to the wild-type enzyme in circular dichroism spectrometry without significant changes of the secondary structure.
The single-substrate steady-state kinetic constants (k cat and K m ) were determined by varying the concentration of one substrate while keeping the other two substrates at a saturated concentration for the wild-type bsMenE and its mutant proteins (Table 1). Interestingly, no detectable activity was found

Thioester-forming structure in MenE domain alternation
for the mutants S384P, W422A, and Y452A, demonstrating the absolute requirement of these residues for the enzymatic catalysis. S384P loses the activity likely due to the fact that a proline residue is unable to serve as a hinge residue, whereas W422A and Y452A lose their activity likely due to elimination of the stacking interaction between the aromatic residues and the adenine ring of the CoA ligand (Fig. 4A). In contrast, the S198A and S389A mutations cause no more than 14-fold decrease in catalytic efficiency (k cat /K m ) for all substrates, although both Ser-198 and Ser-389 form a hydrogen bond with the panteth-einyl group of OSB-NCoA (Fig. 4A). However, the alanine substitution of Glu-392, which has no direct interaction with the bound CoA ligand but forms a hydrogen bond with His-196 to open the adenylate tunnel for access of the CoA substrate to the OSB-AMP intermediate, reduces the catalytic efficiency by 3 orders of magnitude for CoA-SH and 2 orders of magnitude for OSB and ATP. This significant activity loss sharply contrasts the results for the H196F mutant, which also eliminates the Glu-392-His-196 interaction but essentially retains all the activity of the wild-type enzyme (26). One possible expla-  (380 -392), and C-domain (393-486) are respectively colored pale-cyan, green, and pale-yellow, except the two mutated residues I454R and A456K are highlighted in red. The water molecules that directly interact with OSB-NCoA or active-site residues are shown as red dots, and the magnesium ion is shown as a green sphere. The yellow dashed lines denote hydrogen bonds with a distance shorter than 3.5 Å. The oxygen, nitrogen, and phosphorus atoms in the whole panel are colored red, blue, and orange, respectively. The water molecules O1-O13 correspond to HOH14, HOH35, HOH1472,  HOH791, HOH412, HOH238, HOH1513, HOH1482, HOH1055, HOH1821, HOH101, HOH875, and HOH239 of the ternary structure (chain C), respectively. The underlined residues have been mutated for kinetic studies. B, alignment of sequence fragments contributing to binding of the OSB moiety and the pantetheinyl group of OSB-NCoA. Secondary structural elements are shown for the bsMenE complex containing OSB-NCoA and AMP in the alignment. Residues conserved in 100, 100 to 95, and 95 to 80% sequences are represented with blue triangles, marked with red boxes, and shown in yellow background, respectively. nation for this difference is that the unconstrained His-196 side chain in the E392A mutant may interact with the pantetheinyl group or residues in the CoA tunnel to affect CoA binding and thus significantly reduces the catalytic activity, whereas the unconstrained Glu-392 or Phe-196 side chain in the H196F mutant has no such adverse effects and causes much less activity loss. In this possible scenario, Glu-392 plays the role of restricting the His-196 side chain from forming adverse interactions with the substrate or other active-site residues.
We also determined the reaction rate of the first adenylation half-reaction for the site-directed mutants and the crystallization-enhancing mutants in comparison with the wild-type enzyme. As shown in Table 3, except that S384P possesses 13% of the adenylation activity of the wild-type enzyme, all other mutants are comparable with the wild-type enzyme (Ն59%) in adenylation rate. Because S384P loses the activity to form the final product OSB-CoA in the steady-state experiment (Table  1), its retention of 13% adenylation activity shows the importance of the Ser-384 backbone flexibility to the domain movement in forming the thioesterification and adenylation conformations, of which the former is affected much more than the latter. For all other mutants, the comparable adenylation rates clearly show that activity decrease in the steady-state kinetic experiments (Table 1) is mainly due to the mutational effects on the second half-reaction, thioesterification, via the loss of the binding interactions specific in the thioesterification conformation (Fig. 4A). Because these mutated residues are not involved in forming the adenylation-active site (23,24,26), the kinetic data unambiguously demonstrate the thioesterificationspecific effects of the mutated residues and provide strong support for re-configuration of the active site for thioesterification half-reaction through domain alternation. Most clear-cut results were obtained for the W422A and Y452A mutant proteins, of which the adenylation reaction rates are essentially unaffected and yet the activities to form the thioesterification products are completely eliminated. Although similar thioesterification-specific mutational effects have been provided previously for thioesterification active-site residues in homologous enzymes with steady-state kinetics (10,11), the current mutational results provide the most direct evidence.

Discussion
As a member of the ANL enzyme family, OSB-CoA synthetase (MenE) is a valuable drug target in the vitamin K biosynthetic pathway. In this study, we have determined high-resolution crystal structures of a catalytically competent double mutant (IRAK) of bsMenE in complex with a stable product analogue, OSB-NCoA, and a combination of OSB-NCoA and AMP. They are found to take a thioester-forming conformation, which results from a 139.5°rotation of the small C-domain around the hinging residue Ser-384 from its position in the previously determined bsMenE adenylation conformation (24,26). Interestingly, the product analogue takes a crowbar-like shape and tightly binds to the protein in an energetically unfavorable conformation, whereas AMP loosely binds to the protein with easy access to the bulk solvent. These features of the ligand binding strongly support a sequential product release in which AMP is released before OSB-CoA, in contrast to the results from previous kinetic studies that suggest that OSB-CoA is released before AMP (22). In addition, these crystal structures revealed the amino acid residues contributing to recognition and binding of the substrates and products in the thioesterification reaction. Site-directed mutagenesis showed that these residues significantly affect the thioesterification half-reaction but have negligible effects on the adenylation half-reaction. These results have not only provided unambiguous support for the domain-alternation catalytic mechanism of ANL enzymes in general, they have also revealed the unique thioesterification active site of OSB-CoA synthetases and provided the structural basis for their catalysis of the thioesterification reaction to facilitate their utility as a drug target.
The bsMenE thioester-forming structure is very similar to those of other ANL family members in folding and C-domain orientation. Its closest structural orthologue is Nicotiana tabacum 4-coumarate-CoA ligase (4-CCL) (20) with r.m.s.d. of 1.66 Å and a sequence identity of 30.08%. Despite this structural similarity, these homologous enzymes contain only a limited set of conserved amino acid residues such as His-196 and Glu-392 that form a hydrogen-bonding pair in bsMenE (Fig. 4A). In contrast, most active-site residues are conserved specifically among the MenE family, including those responsible for binding the OSB moiety of OSB-NCoA in the previously identified the OSB-binding site (24). The residues responsible for binding the pantetheinyl group of OSB-NCoA in bsMenE are also found to be specifically conserved among 200 unique MenE sequences with sequence identity from 25 to 50% identified by BLAST from the UniProtKB database (30) using bsMenE as query and setting the E-threshold at 0.01 (Fig. 4B).
Interestingly, the IRAK complex structures reveal a distinct binding mode for CoA and its derivatives, which recognizes CoA as a hairpin with its polar 3Ј-phospho-ADP head inserted firmly into the C-domain of the thioesterification conformation (Fig. 5A). This CoA-binding mode is similar to that observed in 4CBL (9) but different from all other known ANL enzymes with a CoA or CoA ester in the thioesterification conformation, including human ACSM2A (19), Salmonella enterica acetyl-CoA synthetase (ACS) (8), and 4-CCL (20), in which the polar head of the CoA hairpin is turned by ϳ180°and bound to the N-domain (Fig. 5, B-D). This different CoA-binding mode is due to the structural difference in the 3Ј-phospho-ADP-binding motifs of the enzymes. In bsMenE (Fig. 5A) and 4CBL (9), the polar CoA headgroup is bound by theinteraction with two conserved aromatic residues and the electrostatic interaction with a positively charged groove on the C-domain surface (Fig. 5A). In comparison, other acyl/aryl-CoA synthetases con-  5A) and 4CBL (Arg-87). The metal binding found in bsMenE (Fig. 5A) and 4CBL, which is absent in other known structures of acyl/aryl-CoA synthetases containing a CoA or CoA derivative (Fig. 5, B-D), changes the surface charge on the ADP group of CoA and may also prevent it from binding to the N-domain.
According to the mode of CoA binding and the associated binding motifs, two subtypes of acyl/aryl-CoA synthetases are found in 22 representatives selected from acyl-CoA synthetases in the CATH6 superfamily 3.30.300.30 (31) and structurally solved acyl/aryl-CoA synthetases after excluding redundant sequences and sequences unreviewed in the UniProtKB database. The type I synthetases bind CoA with its polar headgroup inserted into the C-domain as found for bsMenE and 4CBL and are characterized by the two conserved aromatic residues and an N-domain-blocking bulky residue corresponding to Trp-422, Tyr-452, and Lys-86 in bsMenE, respectively (Fig. 5E). This group of enzymes is also characterized by Mg 2ϩ binding to the ADP group and by the residues contributing to the positive electrostatic potential groove on the C-domain (Fig. 5A). They include the majority of MenE orthologues and other acyl/aryl-CoA synthetases such as YhfT from B. subtilis (UniProtKB O07619) or FadD13 from Mycobacterium tuberculosis (32). In contrast, the type II synthetases bind CoA with its polar headgroup bound to the N-domain and without metal chelation and are characterized by the absence of the conserved sequence motifs for the type I synthetases (Fig. 5E). Besides the three enzymes structurally determined for their thioesterification conformations in complex with a CoA or its derivative (8,19,20), the type II enzymes include a few MenE orthologues such as MenE from Arabidopsis thaliana (33) and other acyl/aryl-CoA synthetases such as PaaK1 from Burkholderia cenocepacia (34). Noticeably, all MenE orthologues from pathogenic bacteria are type I enzymes, and their distinct CoA-binding mode as found in bsMenE may be explored for development of new antibiotics in the future.
Besides shedding light on the unique CoA recognition mode, the IRAK complex structures also offer the template for modeling of OSB-AMP, the other substrate of the thioesterification half-reaction, into the thioesterification active site. This modeling is achieved manually by superimposing the OSB-AMPbound N-domain of the post-adenylation bsMenE structure to the N-domain of the IRAK complex structure and leave the OSB-AMP ligand in the IRAK structure after removing the post-adenylation protein structure. As shown in Fig. 6, the AMP product and the benzoic acid moiety of the OSB-NCoA overlap very well with the corresponding structural components of the modeled OSB-AMP intermediate. Interestingly, the succinyl carbonyl carbon atom (SC1) of the OSB-AMP intermediate is located 2.1 Å from the nitrogen atom of OSB-NCoA, which corresponds to the nucleophilic thiol group of free CoA-SH when it is bound to the CoA-binding tunnel like the CoA moiety of the product analogue. This close distance should allow OSB-AMP and CoA-SH to react readily to form the S-C thioester bond (ϳ1.82 Å) in the OSB-CoA product. Thus, this modeled structure offers an opportunity to speculate on how the two substrates are arranged and positioned for the thioesterification reaction in the domainalternated conformation.
The succinyl group is basically extended and linear in the modeled OSB-AMP (26), whereas it is twisted in a gauche conformation in the OSB-NCoA product analogue (Fig. 2G). This strained ligand structure is unlikely due to the mutations in the protein because both mutated residues are far away from the active site and make no contact with the ligand (Fig. 4A). This difference in the ligands indicates that the reacting groups, particularly the carbonyl group (SC1-SCO) in the adenylate intermediate, must undergo a long-distance movement of ϳ4.0 Å as suggested by the positional difference of the oxygen atom before and after the thioesterification reaction (Fig. 6). It is not clear at present how this large movement is accommodated in the thioesterification process. Noticeably, there is no suitable polar group nearby to interact with the carbonyl oxygen and activate the acyl group in the modeled OSB-AMP for the thiolation reaction with CoA-SH. This activating group may be hidden due to deviation of the modeled structure from the genuine structure. Alternatively, its absence may be due to the fact that OSB-AMP is active enough for the thiolation reaction and needs no further activation.
After the thioesterification reaction, the AMP product is released first, but OSB-CoA or its analogue cannot be released before disassembly of the thioesterification conformation, due to the different binding modes noted earlier for the products (Figs. 2, D and E, and 4A). This difficulty in product release is fully consistent with the finding that product release is the ratelimiting step in the MenE-catalyzed reaction in a previous kinetic study (22). In this connection, it should be noted that the succinyl group is strained in a gauche conformation in the OSB-NCoA product analogue. The release of this strained conformation is thus proposed to assist or even drive the disassembly of the interdomain interface and eventually the whole thioes- terification conformation for the enzyme to release the OSB-CoA product and return to a ligand-free state to complete a catalytic cycle.
In summary, we have for the first time determined the crystal structure of a catalytically competent MenE mutant in complex with the OSB-NCoA product analogue or both AMP and OSB-NCoA in the thioesterification conformation. These crystal structures define a 139.5°C-domain rotation in configuration of the thioesterification conformation from the adenylation conformation containing the adenylate intermediate OSB-AMP, and they provide strong support for the domain-alternation mechanism of the ANL enzymes. They have revealed a distinct mode for interaction with the CoA substrate in comparison with other acyl/aryl-CoA synthetases and have enabled identification of active-site residues discretely conserved among MenE orthologues. In addition, these structures have also revealed that the product is dead-locked in the thioesterification conformation with a strained conformation, raising the possibility that release of the strained energy in the product ligand may be a preinstalled molecular mechanism to release the product and disassemble the thioesterification conformation to complete a catalytic cycle. Moreover, the product analogue-bound structures allow modeling of the adenylate intermediate OSB-AMP in the thioesterification conformation to illuminate how the second half-reaction occurs in members of the MenE enzyme family. These results have provided fresh insights into the domain-alternation mechanism of the essential vitamin K biosynthetic enzyme and its homologues in the ANL enzyme family.

Expression and purification of bsMenE and its mutants
The bsMenE used in this study contained only the wild-type sequence, which is different from the protein used in previous studies with a hexahistidine tag at either the N or C terminus (23,24,26). The menE gene was amplified from the genomic DNA of B. subtilis strain 168 using primers listed in Table 4 and cloned into the pET28a expression vector using the restriction sites of NcoI and XhoI. After the gene sequence was confirmed by DNA sequencing (Beijing Genomics Institute, Shenzhen, China), the plasmid construct was transformed into Escherichia coli strain BL21 (DE3). For protein expression, cells were grown in 2 liters of Luria-Bertani broth (LB) at 37°C to reach an A 600 of 0.8 before induction with 0.2 mM isopropyl ␤-D-thiogalactoside for protein overexpression at 18°C for 20 h. For purification, the harvested cells were homogenized by sonication in 25 mM Tris buffer (pH 8.0). After high-speed centrifugation at 18,000 rpm for 50 min, the supernatant was fractionated by ammonium sulfate, and proteins precipitated by the salt at a concentration from 50 to 60% saturation were collected. Following desalting, bsMenE was purified from the precipitated protein fraction on a HiPrep DEAE FF 16/10 column (GE Healthcare) using a slow salt gradient from 0.11 to 0.20 M NaCl in 25 mM Tris buffer (pH 8.0) at a flow rate of 2 ml/min over 90 min. The collected bsMenE fractions were combined, concentrated, and further purified by size-exclusion chromatography. The obtained protein was greater than 95% pure as assessed by SDS-PAGE and was concentrated and stored until use at Ϫ80°C in the buffer containing 25 mM Tris, 200 mM NaCl, and 10% glycerol (pH 8.0). The protein concentration was measured using Pierce TM Coomassie protein assay kit. All the buffers used in the purification process contained 2.0 mM 2-mercaptoethanol.
Point mutations were introduced into bsMenE to obtain mutant proteins S198A, R382P, S384P, S389A, E392A, W422A, Y452A, I454R, and I454R/A456K using the bsMenE-expressing plasmid in the pET28a vector as template and primers listed in Table 4 Synthetic primers for cloning and site-specific mutation of bsMenE

Protein
Oligonucleotide sequence (the mutated codons are underlined) Table 4 for mutagenic reactions using the QuikChange TM mutagenesis kit (Stratagene). The genes of the mutant proteins were fully sequenced by Beijing Genomics Institute to confirm that only the desired point mutations were introduced. These mutant proteins were expressed and purified with the procedures described above for the wild-type protein and were all shown to present a secondary structure similar to that of the wild-type protein by circular dichroism spectroscopy.

Steady-state and adenylation kinetics
The single substrate steady-state kinetics of bsMenE and its mutants was determined as described previously (22,24,26). The assay was coupled to excessive MenB obtained from previous studies (35)(36)(37), which converts the bsMenE product OSB-CoA to DHNA-CoA for measurement of the DHNA-CoA characteristic absorbance at 392 nm (molar extinction coefficient ⑀ ϭ 4000 M Ϫ1 ⅐cm Ϫ1 ) (38). The substrate OSB was chemoenzymatically prepared from chorismate using the enzymes EntC, MenC, MenD, and MenH obtained from previous studies (39 -45) and purified by reverse-phase HPLC. Chorismate was prepared from an engineered bacterial strain as described previously (46).
The single-turnover adenylation half-reaction rate was determined using a previously described method with minor modifications (22). The assay was carried out in the presence of OSB and ATP without coenzyme A, and its rate was determined by coupling to inorganic pyrophosphatase (IPP) and purine nucleoside phosphorylase (PNP) (47), which hydrolyze the pyrophosphate product to phosphate and convert phosphate into 2-amino-6-mercapto-7-methylpurine (⑀ ϭ 11,000 M Ϫ1 ⅐cm Ϫ1 at 360 nm) in the presence of 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG), respectively (23). A typical reaction contained 1 mM OSB, 1 mM ATP, 0.2 unit of PNP, 0.08 unit of IPP, 400 M MESG, 10 mM MgCl 2 , and 20 mM NaCl in 50 mM Tris-HCl (pH 7.5) and was initiated by addition of 10 M wild-type bsMenE or its mutants after a 1-min incubation. The slopes of the initial linear region of the progress curves were used to calculate the initial reaction rates, which were normalized by setting the rate of the wild-type bsMenE to 100%. All the assays were repeated three times at 20°C.

Crystallization, data collection, structure determination, and refinement
The initial crystallization screen was carried out for the wildtype non-tagged bsMenE at 10, 15, and 20 mg/ml supplemented with 3 mM OSB-NCoA and 20 mM MgCl 2 , mixed with the reservoir solution at a 1:1 ratio, and subjected to sitting drop vapor diffusion at 16°C using commercial kits from Hampton Research (Index, Crystal Screen, and PEG-Rx), Rigaku (Wizard 3&4), and Molecular Dimension (JCSG1&2). Small bar-shaped crystals were observed in one condition (#21, Wizard 4) after 3 days. The crystals were easily reproduced in optimization experiments and grown to a size of ϳ300 m in the longest dimension in subsequent additive screening. The resulting crystals diffracted to a high resolution of ϳ1.7 Å, but each diffraction point was severely split, giving rise to uninterpretable diffraction datasets obtained from either in-house diffractometers or synchrotron radiation facilities. This twinning problem remained after seeding, optimization of cryoprotectants, or additional additive screenings at multiple temperatures (8, 12, 16, and 20°C).
Crystallization screens for the double mutant I454R/A456K (IRAK) of bsMenE were conducted using essentially the same protocol as for the wild-type protein. In a parallel screen, 0.8 mM AMP was added to the protein solution to capture a ternary complex. Several rectangular single crystals were observed within 3 days in a JCSG condition (JCSG1 #46) containing 40% v/v PEG300, 0.1 M sodium cacodylate (pH 6.5), and 0.2 M calcium acetate. Further optimization gave rise to more regular cubic/rectangular crystals using the reservoir solution of 32% v/v PEG300 containing 0.15 M sodium cacodylate (pH 6.5) and 0.2 M calcium acetate. In-house tests showed excellent diffraction quality and low mosaicity in a P1 space group for the crystals of both the binary (IRAK-OSB-NCoA) and ternary (IRAK-OSB-NCoA-AMP) complexes, which were grown under the same conditions. X-ray diffraction data were collected at 100 K for the two bsMenE complexes at BL17U with an ADSC Quantum 315R charge-coupled device detector at the Shanghai Synchrotron Radiation Facility (SSRF) (50) or BL19U1 with a PILATUS3 6M detector at the National Facility for Protein Science Shanghai, China. To achieve high completeness and redundancy for the ternary complex crystal in the P1 space group, 720 diffraction frames were collected with the oscillation angle set at 0.5°per frame and the ⌽ angle rotated from 0 to 360°. All the images were indexed, integrated by iMosflm (51), and scaled by Aimless (52) in the CCP4 Suite. The highest resolution was set to 1.76 Å, and its Matthews coefficient indicated four protein molecules in the unit cell (53). For the binary complex, 440 diffraction frames were collected where the oscillation angle was set to 1.0°per frame and the ⌽ angle rotated from 0 to 440°, and the images were similarly processed as for the ternary complex. The statistics of data collection and processing are summarized in Table 2.
The ternary structure was solved by molecular replacement (MR) by Phaser-MR in PHENIX (54), where the N-domain (residues 1-379) of the ligand-free bsMenE structure (PDB code 5BUQ, chain A) was used as search model. The unit cell contained four homologous subunits that were nicely fitted to the initial electron density map as checked in COOT (55). Rigidbody refinement was performed by PHENIX-Refine (56) with 5% reflection taken for R free calculation. The resulting R work / R free reduced from 0.4042/0.4040 to 0.3122/0.3470, indicating a correct MR solution. After the preliminary refinement, the C-domain of chain A was built manually by COOT based on the extra electron density, and non-crystal symmetry was applied to generate the C-domains of the other three subunits. Subsequently, the coordinates and geometry constraints of the two ligand molecules were prepared by eLBOW (57). OSB-NCoA was unambiguously modeled to the extra density at the interdomain channel of all four subunits, although we could only observe a relatively poor density for AMP in chain A and chain C. As the refinement converged, TLS anisotropic refinement was incorporated in which each subunit was partitioned into four segments (residues 1-162, 163-305, 306 -381, and 382-486) as analyzed by TLSMD (58). Chain C of the finalized ternary structure was used as the MR template to solve the binary structure (IRAK-OSB-NCoA), which was refined and examined with the same procedure as the ternary structure. The overall data quality was assessed by PROCHECK (59) and Mol-Probity (60). The statistics for the structural refinement are given in Table 2.

Structural analysis and sequence alignment
All graphics were generated by PyMOL (61), which was also used to perform structural analysis of the protein crystal structures. The C-domain rotation angle and the involved bending residues were calculated by the DynDom server (29). The interfaces were analyzed; and the quaternary structure was determined using PISA (28), and the electrostatic potential surface was calculated using PDB2PQR plus APBS (62). Multiple sequence alignment was performed with ClustalW 2.0 (63) or Clustal Omega (64) for structure-based presentation using ESPript 3.0 (25).
Author contributions-Z. G. conceived and coordinated the study. Z. G. and Y. C. wrote the paper. Y. C. designed, performed, and analyzed the experiments in all figures. Y. C. and L. T. L. participated in protein purification, protein crystallization, data collection, and structural determination. Y. C., L. T. L., X. B. L., X. L., and X. D. L. carried out the chemoenzymatic synthesis of the product analogue and prepared all relevant enzymes. All authors reviewed the results and approved the final version of the manuscript.