The Structure of 4-Hydroxybenzoyl-CoA Thioesterase from Arthrobacter sp. strain SU*

The 4-chlorobenzoyl-CoA dehalogenation pathway in certain Arthrobacter and Pseudomonas bacterial species contains three enzymes: a ligase, a dehalogenase, and a thioesterase. Here we describe the high resolution x-ray crystallographic structure of the 4-hydroxybenzoyl-CoA thioesterase from Arthrobacter sp. strain SU. The tetrameric enzyme is a dimer of dimers with each subunit adopting the so-called “hot dog fold” composed of six strands of anti-parallel β-sheet flanked on one side by a rather long α-helix. The dimers come together to form the tetramer with their α-helices facing outwards. This quaternary structure is in sharp contrast to that previously observed for the 4-hydroxybenzoyl-CoA thioesterase from Pseudomonas species strain CBS-3, whereby the dimers forming the tetramer pack with their α-helices projecting toward the interfacial region. In the Arthrobacter thioesterase, each of the four active sites is formed by three of the subunits of the tetramer. On the basis of both structural and kinetic data, it appears that Glu73 is the active site base in the Arthrobacter thioesterase. Remarkably, this residue is located on the opposite side of the substrate-binding pocket compared with that observed for the Pseudomonas enzyme. Although these two bacterial thioesterases demonstrate equivalent catalytic efficiencies, substrate specificities, and metabolic functions, their quaternary structures, CoA-binding sites, and catalytic platforms are decidedly different.

During the last century, large quantities of 4-chlorobenzoate or related herbicides and polychlorinated biphenyl pesticides were released into the environment because of commercial production and careless waste disposal (1)(2)(3)(4). Strikingly, a variety of soil-dwelling bacteria capable of employing 4-chlorobenzoate as their principal source of carbon have been discovered in richly contaminated areas (5)(6)(7). In these microorganisms, 4-chlorobenzoate is first converted to 4-hydroxybenzoate, which is subsequently metabolized via the ortho-or meta-cleavage pathways (5). The 4-chlorobenzoate dehalogenation pathway, as outlined in Scheme 1, consists of three reaction steps catalyzed by 4-chlorobenzoyl-CoA ligase, 4-chlorobenzoyl-CoA dehalogenase, and 4-hydroxybenzoyl-CoA thioesterase (8). Genes encoding these enzymes are organized within an operon that is under the regulatory control of 4-chlorobenzoate (5). In some bacteria, the gene cluster is located within the chromosomal DNA (12- 14), whereas in others it is plasmid-encoded (15,16).
The 4-chlorobenzoyl-CoA pathway operons of certain Arthrobacter and Pseudomonas bacterial strains display significant differences in both gene order and sequences (17). At the primary structural level, the amino acid sequence identity between paired Arthrobacter and Pseudomonas ligases is ϳ38%, whereas that between paired dehalogenases is ϳ50%. Remarkably, however, there is no significant amino acid sequence identity shared between the thioesterases from these species.
The three-dimensional structure of the thioesterase from Pseudomonas sp. strain CBS-3 was solved several years ago in this laboratory and was shown to have a "hot dog fold" motif (18,19). This type of molecular architecture was first observed in the x-ray structure of ␤-hydroxydecanoyl thiol ester dehydrase from Escherichia coli (20) and has since been found in the (R)-specific enoyl-CoA hydratase from Aeromonas caviae (21). The structures of these enzymes are dominated by a fivestranded anti-parallel ␤-sheet that cradles a rather long ␣-helix of approximately five turns.
Here we report a high resolution x-ray crystallographic analysis of the Arthrobacter sp. strain SU 4-hydroxybenzoyl-CoA thioesterase complexed with its products (4-hydroxybenzoate and CoA) or with the substrate analogs, 4-hydroxyphenacyl-CoA or 4-hydroxybenzyl-CoA (Scheme 2). Although the overall topology of the Arthrobacter thioesterase subunit is similar to that described for the Pseudomonas enzyme, its quaternary structure, CoA-binding site, and catalytic platform are different.
A search for crystallization conditions was conducted utilizing a sparse matrix screen (designed "in-house") composed of 144 conditions at both room temperature and at 4°C via the hanging drop method of vapor diffusion. The conditions were tested with both the apo enzyme and that complexed with 1 mM 4-hydroxyphenacyl-CoA. The protein solution, at a concentration of 18 mg/ml, contained 10 mM HEPES (pH 7.5), 150 mM KCl, and 1 mM 1,4-dithio-D,L-threitol. The best crystals were observed growing at room temperature from poly(ethylene glycol) 3400 at pH 7.0 in the presence of 4-hydroxyphenacyl-CoA. Large single crystals were subsequently obtained via hanging drops with precipitant solutions containing 17-20% poly(ethylene glycol) 3400, 100 mM MOPS 1 (pH 7.0), and 200 mM LiCl. The crystals achieved maximum dimensions of 0.7 mm ϫ 0.7 mm ϫ 0.3 mm in ϳ1-2 weeks. They belonged to the trigonal space group P3 2 21 with unit cell dimensions of a ϭ b ϭ 113.2 Å and c ϭ 62.5 Å and contained one dimer in the asymmetric unit. The crystals of the enzyme complexed with either 4-hydroxybenzoyl-CoA or 4-hydroxybenzyl-CoA were prepared in a similar manner. It was never possible to grow crystals of the apo enzyme in a form suitable for a high resolution x-ray analysis.
Structural Analysis of the Thioesterase/4-Hydroxyphenacyl-CoA Complex-An initial x-ray data set was collected to 2.6 Å resolution at 4°C with a Bruker HISTAR area detector system equipped with Supper long mirrors. The x-ray source was CuK␣ radiation from a Rigaku RU200 x-ray generator operated at 50 kV and 90 mA. The x-ray data were processed with XDS (23,24) and internally scaled with XSCALI-BRE. 2 The x-ray data collection statistics are presented in Table I. One isomorphous heavy atom derivative was prepared by soaking a crystal in 1 mM methylmercury acetate for 1 day. The x-ray data (including Friedel mates) were collected to 2.6 Å. The R-factor between the native and mercury derivative x-ray data sets was 26.8% (where R ϭ ⌺͉F N Ϫ F h ͉/⌺͉F N ͉ ϫ 100, F N is the native structure factor amplitude and F h is the heavy-atom derivative structure factor amplitude). Four heavy atom-binding sites were determined with CNS (25). The positions, occupancies, and temperature factors for these sites were refined with CNS, yielding an overall figure of merit of 0.44 and a phasing power of 1.58. Protein phases were calculated to 2.6 Å resolution with CNS and improved by the method of solvent flipping (as implemented in CNS) to yield a figure of merit of 0.94. Greater than 95% of the residues were built into the solvent flattened electron density map with the graphics software TURBO (26). This "partial" model was subsequently refined via least squares with the program package TNT (27). The resulting electron density maps calculated with coefficients of the form (2F o Ϫ F c ) or (F o Ϫ F c ) allowed for the placement of the remainder of the amino acid residues and the 4-hydroxyphenacyl-CoA ligand.
High Resolution X-ray Data Collection and Least Squares Refinement-Thioesterase crystals were harvested from hanging drop experiments and equilibrated in a synthetic mother liquor composed of 20% poly(ethylene glycol) 3400, 150 mM KCl, 100 mM LiCl, 1 mM 4-hydroxyphenacyl-CoA, and 100 mM MOPS (pH 7.0). They were then serially transferred to a cryoprotectant solution containing 30% poly(ethylene glycol) 3400, 200 mM KCl, 250 mM LiCl, 15% ethylene glycol, 1 mM 4-hydroxyphenacyl-CoA, and 100 mM MOPS (pH 7.0). The crystals were suspended in a loop of 20-m nylon and flash frozen in a stream of nitrogen gas. Unit cell dimensions changed to a ϭ b ϭ 112.5 Å and c ϭ 60.6 Å upon cooling to 120 K. A native x-ray data set was collected to 1.6 Å resolution, processed with SAINT (Bruker AXS, Inc.), and scaled as previously described. This structure was solved via molecular replacement with the program AMORE (28) employing as the search model the refined structure determined at 2.6 Å resolution. Iterative cycles of least squares refinement and manual model building reduced the Rfactor to 18.0% for all measured x-ray data from 30.0 to 1.6 Å resolution. The least squares refinement statistics are presented in Table II.
Structure Determination of the Thioesterase Complexed with 4-Hydroxybenzoyl-CoA (Substrate) or 4-Hydroxybenzyl-CoA-These two structures were solved by difference Fourier techniques with the initial models lacking ordered solvent molecules and ligands. The complex of the enzyme with bound products was prepared by crystallizing the enzyme in the presence of its substrate, 4-hydroxybenzoyl-CoA that was hydrolyzed to 4-hydroxybenzoate (or 4-hydroxybenzoic acid) and CoA. For the sake of simplicity, the ligand will be referred to as 4-hydroxybenzoate, but this does not imply that its protonation state is known. All of the x-ray data for these two complexes were collected in a manner identical to that employed for the thioesterase/4-hydroxyphenacyl-CoA complex as described above. Relevant x-ray data collection and least squares refinement statistics are given in Tables I and II, respectively. In all three complexes, the first 10 and 11 residues were disordered in subunits I and II, respectively. Apart from these disordered N-terminal residues, the electron densities corresponding to both polypeptide chains in the asymmetric unit were continuous throughout the map. The only significant outlier in the Ramachandran plot was Asp 39 (in both subunits). This residue is located ϳ14 Å from the active site. The dihedral angles adopted by Asp 39 produce a bulge in the first ␤-strand of the sheet. All of the figures were prepared with the software package MOLSCRIPT (29).
where V ϭ initial velocity, V max ϭ maximum velocity, [S] ϭ substrate concentration, K m ϭ Michaelis constant, [I] ϭ inhibitor concentration, and K i ϭ the inhibition constant.

RESULTS AND DISCUSSION
Inhibitor Binding-The binding constants of the substrate analogs were evaluated by measuring their competitive inhibition constants. Both analogs displayed linear competitive inhibition versus 4-hydroxybenzoyl-CoA. The K i of 4-hydroxybenzyl-CoA was 0.6 Ϯ 0.1 M, and the K i of 4-hydroxyphenacyl-CoA was 0.003 Ϯ 0.0003 M. These results indicate that these ligands bind tightly to the substrate-binding site.
Overall Structure of the Thioesterase/Product Complex-All of the crystals employed in this investigation contained two subunits/asymmetric unit. For the sake of simplicity, the following discussions will refer only to Subunit II of the x-ray coordinate file unless otherwise indicated. A ribbon representation of the monomer with the bound products, 4-hydroxybenzoate and CoA, is depicted in Fig. 1. The monomer contains 151 amino acid residues, and its topology, referred to as the hot dog fold, is dominated by a six-stranded anti-parallel ␤-sheet formed by Val 37 -Met 41 , Ala 46 -Val 52 , Met 89 -Phe 100 , His 108 -Ala 118 , Trp 123 -Arg 130 , and Arg 135 -Arg 149 . These ␤-strands are labeled A-F in Fig. 1. Four of the six ␤-strands (A, C, D, and F) contain ␤-bulges at Asp 39 ( ϭ Ϫ101°, ϭ Ϫ73°), Gln 94 ( ϭ Ϫ115°, ϭ Ϫ30°), Ile 116 ( ϭ Ϫ113°, ϭ Ϫ20°), and Cys 137 ( ϭ Ϫ94°, ϭ Ϫ29°), respectively. In addition to the six ␤-strands, there are two ␣-helices delineated by Leu 30 -Val 34 and Gly 65 -Val 84 , six Type I turns, and one Type III turn. 1 The abbreviation used is: MOPS, 3-(N-morpholino)propanesulfonic acid. 2  Nearly 80% of the amino acid residues lie within classical secondary structural elements. The monomer is compact with overall dimensions of ϳ38 Å ϫ 49 Å ϫ 35 Å.
In that the asymmetric unit contained only two subunits, ultracentrifugation experiments were subsequently performed to define the quaternary structure of the Arthrobacter thioesterase. These experiments were consistent with a tetrameric quaternary structure. The tetramer thus packed in the crystalline lattice with one of its 2-fold rotational axes coincident to a crystallographic dyad. Shown in Fig. 2a is a ribbon represen-   tation of the two monomers contained within the asymmetric unit. The tetrameric form of the enzyme, as depicted in Fig. 2b, was generated by rotation of the dimer in the asymmetric unit about the crystallographic 2-fold rotational axis. As can be seen from Fig. 2 (a and b), the quaternary structure of the enzyme can be aptly described as a dimer of dimers with overall dimensions of ϳ72 Å ϫ 70 Å ϫ 52 Å.
The subunit-subunit interface of the dimer is quite extensive with a total buried surface area of ϳ3100 Å 2 as calculated according to the method of Lee and Richards (30). There are three major regions of subunit-subunit interactions between the two monomers. These intermolecular contacts are formed by Tyr 22 -Val 34 (␣-helix), Asp 54 -Met 74 (␣-helix), and Gly 93 -Pro 103 (␤-strand C) in one subunit and the symmetry-related residues in the second monomer. Indeed, the six-stranded sheets in each monomer come together, through ␤-strands C, to form a 12-stranded anti-parallel ␤-sheet in the dimer. The active sites are wedged between the two subunits of the dimer and are separated by ϳ24 Å (Fig. 2a). The two major ␣-helices of the dimer run anti-parallel to one another.
The 12-stranded ␤-sheet of one dimer abuts the ␤-sheet motif in the second dimer to form the tetramer (Fig. 2b). This packing arrangement is reminiscent to that observed for the dimeric thioesterase II from E. coli (31). In this enzyme, each monomer contains two hot dog folds referred to as the "double hot dog." The subunit-subunit interface for the E. coli thioesterase II is likewise built with the ␤-sheets back to back and the major ␣-helices facing outwards.
Approximately 2125 Å 2 of surface area is buried per monomer upon tetramer formation in the Arthrobacter thioesterase. The dimer-dimer interface is lined with water molecules and various side chains from each subunit including His 97 , Phe 101 , His 117 , and Phe 124 . As can be seen in Fig. 2b, the histidines at position 97 form a hydrogen bonding ring at the middle of the dimer-dimer interface. The pyrophosphate groups of the CoA moieties project into this interface. Ser 120 and Thr 121 , which lie in a Type I turn defined by Gly 119 -Thr 122 , provide hydrogen bonds to the phosphoryl oxygens of the CoA. Specifically, O ␥ and the peptidic NH group of Ser 120 lie within hydrogen bonding distance to the two ␣-phosphoryl oxygens, respectively, whereas O ␥ and the peptidic NH group of Thr 121 are situated within ϳ2.5 Å of one of the ␤-phosphoryl oxygens. Each of the binding sites for CoA is thus formed by three of the four subunits comprising the tetramer.
Active Site with Bound Products, 4-Hydroxybenzoate and CoA-As can be seen from Fig. 2a, the 4-hydroxybenzoate is primarily wedged between the two subunits of the dimer pair in the tetramer and specifically between the two major ␣-helices. A close-up view of the active site is presented in Fig. 3. Key side chains involved in binding the 4-hydroxybenzoate moiety to the protein include Glu 73 , Thr 77 , and Glu 78 from Subunit II. In addition, the 4-hydroxyl group lies within hydrogen bonding distance to an ordered water molecule, whereas one of the carboxylate oxygens hydrogen bonds to the peptidic NH group of Gly 65 in Subunit I. One of the carboxylate oxygens of the 4-hydroxybenzoate ligand is positioned within 2.8 Å of the sulfhydryl group of CoA, indicating that the substrate has indeed been hydrolyzed. The CoA binds to the protein in a quite curved conformation with its ribose adopting the C 3Ј -endo con- formation. Only solvent molecules and backbone carbonyl or peptidic NH groups lie within hydrogen bonding distance to the oxygens and nitrogens of the CoA ␤-mercaptoethylamine and pantothenate units. The pyrophosphate group oxygens of CoA, however, form hydrogen bonds with the side chains of Ser 120 and Thr 121 situated in the third subunit of the tetramer. Two arginine residues, 102 from Subunit I and 150 from Subunit II, point toward the 3Ј-phosphate group of the CoA ribose. Additionally, the side chain of Arg 150 (Subunit II) runs nearly parallel to the plane of the adenine ring. The amino group at position 6 of the adenine ring forms hydrogen bonds with a water molecule and the carbonyl oxygen of Pro 148 from Subunit II.
Structure of the Arthrobacter Thioesterase Complexed with Either 4-Hydroxyphenacyl-CoA or 4-Hydroxybenzyl-CoA-The binding of either 4-hydroxyphenacyl-CoA or 4-hydroxybenzyl-CoA within the thioesterase active site resulted in little perturbation of the polypeptide chain backbone compared with that were observed with bound products. Indeed, the ␣-carbons for these two complexes both superimpose onto the protein/ product complex with a root mean square deviation of 0.17 Å. For all atoms, the three models presented here superimpose upon one another with typical root mean square deviations of 0.45 Å or less. In all three of the complexes, the conformations of the side chains lining the active sites are identical within experimental error. A superposition of the three ligands bound to the protein is presented in Fig. 4.
The K i of 4-hydroxybenzyl-CoA is 0.6 Ϯ 0.1 M. As can be seen in Scheme 2, the carbonyl functional group of the substrate has been replaced by a methylene bridge in 4-hydroxybenzyl-CoA. The electron density for this inhibitor is indicative of two conformations with the sulfur atoms of the CoA differing by ϳ1.8 Å. Interestingly, the K i of 4-hydroxyphenacyl-CoA is considerably lower at 0.003 Ϯ 0.0003 M. In this inhibitor, an additional methylene group has been inserted between the 4-hydroxybenzoyl and the CoA moieties (Scheme 2). It is possible that this inhibitor is a mimic for the transition state as the carbon-sulfur bond lengthens during substrate hydrolysis.
Structure Comparison of the Arthrobacter Enzyme with the Thioesterase Isolated from Pseudomonas sp. Strain CBS3-The first thioesterase structure to be solved in the 4-chlorobenzoyl-CoA degrading pathway was that isolated from Pseudomonas sp. strain CBS3 (18,19). Each subunit of this enzyme is characterized by a five-stranded anti-parallel ␤-sheet and three major ␣-helices. The Arthrobacter thioesterase monomer, with 151 amino acids, is slightly larger than the Pseudomonas enzyme (141 amino acids). A superposition of these enzymes, both complexed with 4-hydroxyphenacyl-CoA, is given in Fig. 5. In the Arthrobacter enzyme, residues Gly 12 -Pro 43 fold into a short ␣-helix and the first ␤-strand of the ␤-sheet. There are no structural counterparts in the Pseudomonas protein. Indeed, the two enzymes start to superimpose at Glu 44 and Arg 3 for the Arthrobacter and Pseudomonas enzymes, respectively. Those regions of structural correspondence between these two thioesterases superimpose with a root mean square deviation of 1.4 Å for 73 equivalent ␣-carbons.
In both thioesterases, the two subunits orient in an antiparallel fashion to form the dimers (Fig. 2). In the case of the Arthrobacter thioesterase, the third ␤-strands in each monomer lie at the dimeric interface and form backbone hydrogen bonds via Gln 94 , Asn 96 , Thr 98 , and Phe 100 in each subunit. In the Pseudomonas thioesterase, the second ␤-strands of the monomers interact at the dimeric interface through hydrogen bonds formed by the backbone atoms of Val 62 , Cys 64 , Ala 66 , and Phe 68 in each subunit.
From Figs. 2 and 5, it can be seen that the 4-hydroxyphenacyl ligands bind to the two thioesterases in distinctively different manners. In the Pseudomonas thioesterase, the 4Јphosphopantetheine portion of the CoA winds to the right as it reaches the solvent, thereby allowing a phosphoryl oxygen to interact with the backbone nitrogen of Ala 129 in its extra Cterminal appendage and the 3Ј-phosphate group to hydrogen bond with the backbone peptidic NH groups positioned in the reverse turn connecting the third and fourth ␤-strands. In the Arthrobacter thioesterase, the 4Ј-phosphopantetheine moiety winds to the left as it reaches the solvent, thereby positioning the nucleotide to interact with Arg 102 in Subunit I and Pro 148 and Arg 150 in Subunit II. The nucleotide is thus directed toward the center of the dimer-dimer interface where it interacts with residues contributed by a third monomer in the tetramer (Fig. 2b). Despite the common hot dog scaffold in these thioes- terases, the substrate-binding sites are not similar, and the differences are apparently a function of the differing N-and C-terminal regions.
In both thioesterases, the 4-hydroxyphenacyl moieties are positioned into the core of the protein such that the thioester CϭO interacts with the N terminus of the major ␣-helix via a hydrogen bond to the backbone amide NH (Tyr 24 in the Pseudomonas thioesterase and Gly 65 in the Arthrobacter enzyme). This interaction, which leads to polarization of the CϭO for nucleophilic attack (18,19) is thought to be an important component of the catalytic mechanism. The other essential component is believed to be an active site carboxylate residue that mediates the hydrolysis reaction via either nucleophilic or general base catalysis. In the Pseudomonas thioesterase, this residue is Asp 17 (Fig. 6a). It is positioned within the same monomer as the hydrogen bond donor to the thioester CϭO, on the loop connecting the first ␤-strand to the major ␣-helix. The catalytic role of Asp 17 is indeed supported by site-directed mutagenesis experiments, demonstrating that when this residue is replaced with an asparagine, the k cat at 25°C and pH 7.5 is reduced from 15 s Ϫ1 to 5 ϫ 10 Ϫ4 s Ϫ1 (32). The situation is markedly different in the Arthrobacter thioesterase. The residue corresponding to Asp 17 is Gln 58 (Fig. 6b). Inspection of the active site for the Arthrobacter thioesterase demonstrates that the only possible candidate for the catalytic base (or nucleophile) is Glu 73 , which is on the opposite side of the substrate binding pocket. Site-directed mutation of this glutamate to an alanine residue reduces the k cat at 25°C and pH 7.5 from 6.7 s Ϫ1 to 0.0001 s Ϫ1 . 2 It thus appears that in the Arthrobacter protein, the active site base is Glu 73 , which is located on the major ␣-helix of the one monomer of the dimer, whereas the hydrogen bond donor to the thioester CϭO is located at the N terminus of the ␣-helix contributed by the second monomer. The remarkable conclusion drawn from this observation is that these two thioesterases (of equivalent catalytic efficiency, substrate specificity, and metabolic function) are apparently employing different regions of the active site scaffold to position the key catalytic carboxylate residue.
Comparison of the Quaternary Structures of the Two Thioesterases-The differences in active site geometries between these two enzymes extends to their quaternary structures as well. Recent ultracentrifugation experiments demonstrate that the Pseudomonas thioesterase is also a tetramer. As can be seen in Fig. 2c, in the Pseudomonas thioesterase, the dimers come together to form the tetramer with their long ␣-helices facing inwards. The pyrophosphate moieties of the CoA ligands project outwards, and the active site in one dimer is ϳ 21 Å from the symmetry related active site in the second dimer. This arrangement is in sharp contrast to that observed for the Arthrobacter enzyme, whereby the dimers combine to form the tetramer with their long ␣-helices facing outwards (Fig. 2b). In the Arthrobacter enzyme, the active sites between symmetryrelated dimers are ϳ27 Å apart, and the pyrophosphate moieties of the CoA ligands project into the dimer-dimer interfacial region. The question arises as to why these two enzymes form tetramers in such remarkably different manners. It appears that the extra N-terminal motif observed in the Arthrobacter thioesterase, positioned before the major ␣-helix, prevents the helix-to-helix face packing as observed in the Pseudomonas thioesterase. This additional N-terminal region in the Arthrobacter enzyme faces outwards and is not involved in dimerdimer interactions but is involved in the formation of the active site.
The question of how two bacterial thioesterases, with identical substrates, nearly equal catalytic efficiencies, and similar molecular scaffolds evolved with such different quaternary structures and active site geometries is, indeed, intriguing. Are these thioesterases examples of convergent or divergent evolution? Perhaps the answer lies somewhere between. Regardless of the manner in which these proteins arose, this investigation underscores the importance of a combined kinetic and x-ray crystallographic approach in exploring protein structure/function relationships.