Structure, Function, and Mechanism of the Phenylacetate Pathway Hot Dog-fold Thioesterase PaaI*

The structure and biochemical function of the hot dog-fold thioesterase PaaI operative in the aerobic phenylacetate degradation pathway are examined. PaaI showed modest activity with phenylacetyl-coenzyme A, suggestive of a role in coenzyme A release from this pathway intermediate in the event of limiting downstream pathway enzymes. Minimal activity was observed with aliphatic acyl-coenzyme A thioesters, which ruled out PaaI function in the lower phenylacetate pathway. PaaI was most active with ring-hydroxylated phenylacetyl-coenzyme A thioesters. The x-ray crystal structure of the Escherichia coli thioesterase is reported and analyzed to define the structural basis of substrate recognition and catalysis. The contributions of catalytic and substrate binding residues, thus, identified were examined through steady-state kinetic analysis of site-directed mutant proteins.

Aromatic compounds serve as a rich source of carbon and energy for a wide variety of microorganisms and plants (1). The enzymes that make up the aromatic catabolic pathways are useful for bioremediation of environmental aromatic pollutants (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11) and for chemical synthesis (3,(12)(13)(14)(15). Studies of the aromatic pathways operative in common bacteria such as Escherichia coli and Pseudomonas putida (16,17) have provided a framework against which novel pathways of environmental bacteria can be understood (18,19). In this paper, we focus on a pathway for aerobic phenylacetate degradation. Phenylacetate derives primarily from phenylalanine, but it is also formed in degradation pathways that target a variety of environmental aromatics of natural and synthetic origin and, thus, is the central element of an important catabolon (20 -22).
The gene clusters that encode the phenylacetate pathway enzymes typically contain structural genes, regulatory genes, and transport genes (17,19,20,(22)(23)(24)(25)(26)(27). In E. coli, the 14 open reading frame cluster includes the structural genes paaABCDEFGHIJKXYZ (23). Only one of these genes, paaK, has been assigned function through isolation and kinetic characterization of its protein product, phenylacetate-CoA ligase (20,23,28,29). This ligase catalyzes the first committed step of the pathway, which is the conversion of the ring acetate group to the corresponding CoA thioester (28,29) (Fig. 1). Phenylacetyl-CoA then induces the synthesis of the other pathway enzymes (30,31). The reactions catalyzed by these enzymes have been investigated but not firmly demonstrated (32). Nevertheless, a working model that incorporates the findings from the studies of both the phenylacetate pathway (32) and the analogous benzoate pathway of Azoarcus evansii (26,33,34) can be inferred (Fig. 1). According to this model, the PaaABCD enzyme complex catalyzes the addition of oxygen across the ring C(1)AC(2) bond of phenylacetyl-CoA followed by reduction to the dihydrodiol (32). The next hypothetical step is ring opening to the C-8 aldehyde, which is converted to the corresponding acid. The ensuing CoA thiolytic cleavage forms a C-6 CoA ester (32), which is processed via ␤-oxidation by enzymes of the crotonase superfamily (35,36).
The one gene of the pathway gene cluster, paaABCDEFGHI-JKXYZ, that does not have a defined role in phenylacetate degradation is paaI. The protein product PaaI is a member of the hot dogfold enzyme superfamily (37,38). Moreover, it belongs to the acyl-CoA thioesterase subfamily and, in particular, to the same clade of acyl-CoA thioesterases as does the Arthrobacter 4-hydroxybenzoyl-coenzyme A thioesterase (4-HBA-CoA thioesterase) 3 (39,40). Acyl-CoA thioesterases generally function in the cell to release the carboxylate unit for degradation, export (39,(41)(42)(43)(44)(45), or regulation (43, 46 -48) and/or to release the CoA in thioester metabolites for participation in alternate metabolic pathways (26).
In this paper we report and interpret the results from substrate screens of A. evansii and E. coli PaaI to define the biochemical role of PaaI in the context of the phenylacetate pathway. The x-ray structure of the E. coli apo PaaI is also reported. The E. coli PaaI and Thermus thermophilus PaaI (49) structures are analyzed in the context of the PaaI substrate specificity profile and the kinetic properties of the PaaI active site mutants. A mechanism for PaaI substrate recognition and catalysis is proposed.
Crystallographic Analysis-C-Terminal His-tagged E. coli PaaI was prepared for crystallization as detailed in the supplemental material. The N-terminal Met was replaced by Ser-Leu. Crystals of the selenomethionine (54) PaaI were grown in 13% polyethylene glycol (average molecularweight,4000),0.1M[bis(2-hydroxyethyl)imino]-tris(hydroxymethyl)methane (pH 6.0) and 0.2 M lithium sulfate at 18°C using the hanging drop vapor diffusion method. They were incubated in the presence of 30% glycerol before being flash-frozen in liquid nitrogen. The crystals belong to trigonal space group P3 1 21 with a ϭ b ϭ 69.9 Å and c ϭ 117.2 Å. X-ray diffraction data were collected at APS beamline 31ID at wavelength 0.9798 Å. Data were reduced with DENZO, SCALEPACK (55), and CCP4 (56). The structure of PaaI was determined using SAD technique employing the anomalous signal from selenium atoms as implemented in the software SOLVE (57). Ten selenium atoms were located, and phases derived from these anomalous centers were used for density modification as implemented in the software RESOLVE (58). Manual model building was accomplished using O (59). Refinement of the model was accomplished using CNS (60) and Refmac as implemented in CCP4 (56). Water molecules were located and refined. The final model contains 2016 protein atoms, 79 solvent atoms, and 15 sulfate atoms and was refined to an R value of 18.7% and an R free value of 23.0%. Crystallographic details are in Table 3.
Preparation of Wild-type and Mutant A. evansii and E. coli PaaI for Kinetic Study-A. evansii (accession number Q9F9VO) and E. coli paaI (accession number P76084) genes were amplified by PCR (61) using genomic DNA prepared from A. evansii DSM 6898 (DSMZ) and E. coli strain ATCC 11105 (Manassas, VA) as templates, respectively, commercial oligonucleotides as primers, and Pfu Turbo as the DNA polymerase. The PCR products were digested with the restriction enzymes NdeI and HindIII and then purified by agarose gel chromatography before T4 DNA ligase catalyzed the ligation to pET-23b (ϩ) vector (Novagen) digested by the NdeI and HindIII. The resulting clones named EcWT-PaaI/pET-23b and AeWT-PaaI/pET-23b were verified by DNA sequencing. Mutagenesis was carried out using a PCR-based strategy with commercial primers and the WT-PaaI/pET-23b (ϩ) plasmid serving as template. The purified PCR products were used to transform to E. coli JM109-competent cells (Stratagene). Mutant genes were verified by DNA sequencing. For protein production PaaI/pET-23b transformants of E. coli BL21(DE3) were grown aerobically at 28°C in LB media containing 50 g/ml carbenicillin for ϳ10 h (A 600 1.0) then induced with 0.4 mM isopropyl-␤-D-galactopyranoside for 5 h. Cells were harvested by centrifugation (5000 ϫ g for 10 min), suspended in 100 ml of 50 mM K ϩ Hepes, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM DTT (pH 7.5, 0°C) and passed through a French press at 1200 p.s.i. twice before centrifugation at 48,000 ϫ g for 30 min at 4°C. The supernatant was applied to a 45 ϫ 3.5-cm DEAE-Sepharose (Amersham Biosciences) column and eluted at 4°C with a 2-liter gradient of 0 -0.5 M KCl in 50 mM K ϩ Hepes, 1 mM DTT (pH 7.5). The thioesterase-containing fractions (eluted at ϳ 0.3 M KCl) were combined and then precipitated with ammonium sulfate (40 -60% cut for A. evansii and 0 -40% cut for E. coli PaaI). The A. evansii PaaI was dissolved in 2 ml of 10 mM K ϩ Hepes (pH 7.5) buffer with 0.15 M KCl and 1 mM DTT, and E. coli PaaI was dissolved in 2 ml of 10 mM Tris (pH 8.1) buffer with 0.05 M KCl and 1 mM DTT. The protein solution (1 ml) was applied to a 100 ϫ 2-cm Sephacryl S-200 column (Amersham Biosciences) and eluted at 4°C with the same buffer at 0.5 ml/min. The thioesterase-containing fractions were pooled and concentrated with a 10-kDa Macrosep centricon (Pall Filtron). The activity of the A. evansii PaaI was stable for storage in 10 mM K ϩ Hepes (pH 7.5) buffer or in 10 mM Tris (pH 8.1) containing 10 mM KCl at Ϫ80°C for several months. The E. coli PaaI could be stored in 10 mM Tris (pH 8.1) at 0°C (not frozen) ice for several weeks without significant loss of activity. The yields of homogeneous (based on SDS-PAGE analysis) A. evansii PaaI proteins in mg/g of wet cells were WT (11), D75A (11), D75N (9), N60A (9) and N60D (7), whereas those for the E. coli PaaI proteins were WT (50), D61A (4), N46A (9), E14A (22), N15A (10), D16A (7), H52A (6), and N46Q/D61E (8). Fast protein liquid chromatography gel filtration chromatography (Hiload 60 ϫ 16cm Amersham Biosciences Superdex S-200) was used to determine whether the A. evansii PaaI mutant proteins were contaminated with E. coli PaaI. A solution of 10 mM K ϩ Hepes (pH 7.5) containing 0.15 M KCl and 1 mM DTT was used to elute the column at a flow rate of 1 ml/min. The retention time of the E. coli PaaI was 52.3 min versus 47.0 min for the A. evansii PaaI. The activities of the A. evansii PaaI mutants were measured before and after chromatography and found to be identical, thereby demonstrating that the protein purification protocol described above removes E. coli PaaI from the A. evansii PaaI.
Molecular Size Determination-The molecular mass was calculated from the amino acid composition, derived from the gene sequence, by using the EXPASY Molecular Biology Server program Compute pI/Mw. The molecular mass was measured by ESI mass spectrometry and by SDS-PAGE (4% stacking gel and 18% separating gel). The mass of native PaaI was estimated by fast protein liquid chromatography gel filtration column chromatography carried out as described in the previous section. Commercial protein molecular weight standards (Amersham Biosciences) were used to generate a plot of log M r versus elution volume from the column.
The initial velocity data, measured as a function of substrate concentration, were analyzed using Equation 1,  (62)). The inhibition constant K i was obtained by fitting the initial rates to Equation 2, where [I] is the concentration of the inhibitor, and K i is the inhibition constant. The log k cat and log(k cat /K m ) values obtained from the pH rate profile analysis were fitted using Equation 3, where Y is k cat or k cat /K m , [H] is the hydrogen ion concentration, C is the pH independent value of k cat or k cat /K m , K a is the acid dissociation constant, and K b is the base dissociation constant. Data analysis was carried out using the computer program KinetAsyst (IntelliKinetics). The reported error was computed for the data fitting. evansii and E. coli PaaIs migrate on SDS-PAGE gels as 15-and 14-kDa proteins, whereas the native molecular masses determined by gel filtration chromatography correspond to 67 and 58 kDa. Both PaaIs are, therefore, homotetramers.

RESULTS AND DISCUSSION
The pH rate profiles of PaaI-catalyzed 3,4-dihydroxyphenylacetyl-CoA hydrolysis (identified as the most active substrate; see below) were measured using initial-velocity techniques (Fig. 2). Both PaaIs are stable at pH 6 -10 ( Fig. 2, insets) and most active at pH 6 -9. The bell-shaped curves were fitted to define the apparent pK a values for ionization of essential residues in the enzyme-substrate complex (log(k cat )) and in the uncomplexed enzyme and substrate (log(k cat /K m )). The A. evansii PaaI log(k cat ) profile defines a pK a of 5.5 Ϯ 0.2 (7.4 Ϯ 0.2 for E. coli PaaI) for ionization of an essential base and an apparent pK a of 9.6 Ϯ 0.2 (9.1 Ϯ 0.2 for E. coli PaaI) for ionization of an essential acid. The A. evansii PaaI log(k cat /K m ) defines an apparent pK a of 6.5 Ϯ 0.3 (6.5 Ϯ 0.3 for E. coli PaaI) for ionization of an essential base and an apparent pK a of 8.2 Ϯ 0.2 (9.0 Ϯ 0.23 for E. coli PaaI) for ionization of an essential acid. The substrate ring hydroxyl group ionizes with a pK a ϭ 9.8, as determined by pH titration monitored at 300 nm (data not shown) and, therefore, does not contribute to the log(k cat /K m ) pH profiles. The pH profiles define the optimum pH for catalytic function, and in well defined systems they can be used to identify substrate binding or catalytic residues. Under the section "Catalytic Site of the E. coli PaaI" (below) the active site residues that might contribute to the observed pH profiles are identified.

PaaI Substrate Screen for Determination of Biochemical Function-The
object of this substrate screen was to identify acyl-CoA thioesters that have k cat /K m values in a range that would be considered to be "physiologically relevant." The k cat /K m values of metabolic enzymes that have reached catalytic perfection are as large as 1 ϫ 10 8 M Ϫ1 ⅐s Ϫ1 , but for most enzymes the value is ϳ1 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 , and for those enzymes that function in secondary metabolic pathways, the value can be as low as 1 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 (63). The steady-state kinetic constants measured for the aryl-CoA and aliphatic acyl-CoA thioesters tested as substrates for A. evansii and E. coli PaaIs are summarized in Tables 1 and 2, respec-TABLE 1 Steady-state kinetic constants for A. evansii DSM6898 PaaI wild-type and its mutant-catalyzed hydrolysis of acyl-CoA thioesters at 25°C, pH 7.5 The reported error limits were computed for the S.E. in the data fitting. The reproducibility of the measured value in repeated trials is within 10% error, whereas in independent experiments the values may vary on average of 50%.

Enzyme
Substrate  The aryl-CoA thioesters 4-chlorobenzoyl-CoA, 4-methoxybenzoyl-CoA, 4-hydroxybenzoyl-CoA, and 3-hydroxybenzoyl-CoA (structures shown in Structure 1) were tested to determine whether PaaI hydrolyzes aromatic thioesters in which the thioester group is directly attached to the aromatic ring. Such is the case with the Arthrobacter sp. strain SU 4-HBA-CoA thioesterase, which is the closest homolog to PaaI, having known catalytic function (39). None of these substituted benzoyl-CoA thioesters were, however, hydrolyzed by PaaI. Together, the results suggested that the PaaI does not target aliphatic acyl-CoA thioesters nor are they functionally related to the benzoyl-CoA thioesterases that participate in benzoate-based metabolic pathways (26,39,64) including the genisate pathway (42).
The next set of substrates screened contained the phenylacetyl-CoA structural core (Structure 1). Phenylacetyl-CoA is a fair substrate (k cat /K m ϭ 1 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 for A. evansii PaaI and 4 ϫ 10 4 M Ϫ1 ⅐s Ϫ1 for E. coli PaaI) (Tables 1 and 2). This result shows that the phenyl ring and its spacing from the thioester group are important determinants of PaaI substrate recognition. 2-Hydroxyphenylacetate had been reported to accumulate as a dead-end product in a mutant lacking the ring-cleaving enzyme (32) (see Fig. 1). The precursor, 2-hydroxyphenylacetyl-CoA, might therefore be targeted by PaaI. Despite the numerous strategies tried (including the procedure reported in Alonso et al. (65)), we were not successful at the chemical synthesis of 2-hydroxyphenylacetyl-CoA. We were, however, able to generate the target compound transiently using E. coli phenylacetate-CoA ligase (which was isolated for this pur-pose) to catalyze the reaction between 2-hydroxyphenylacetate, MgATP, and CoA (in 50 mM K ϩ Hepes buffer (pH 7.5), 25°C). The reaction was monitored by HPLC to show that the 2-hydroxyphenylacetyl-CoA was formed but that it was quickly hydrolyzed to 2-hydroxyphenylacetate and CoA. The chemical instability of 2-hydroxyphenylacetyl-CoA might be attributed to nucleophilic catalysis by the ortho ring-hydroxyl group, which is positioned to attack the carbonyl carbon of the acetyl-CoA unit, forming a lactone that undergoes subsequent hydrolysis. Insofar as the reaction conditions used here might mimic those in the cell, it would appear that the spontaneous hydrolysis of the 2-hydroxyphenylacetyl-CoA obviates the need for a dedicated thioesterase. On the other hand, we cannot rule out the possibility that the time frame of cellular chemistry might demand such a catalyst.
Although we were unable to evaluate the substrate activity of the 2-hydroxyphenylacetyl-CoA, the mono-and dihydroxylated compounds 4-hydroxyphenylacetyl-CoA, 3-hydroxyphenylacetyl-CoA, 3,4-dihydroxyphenylacetyl-CoA, and 3,5-dihydroxyphenylacetyl-CoA were found to be excellent substrates for both PaaIs (Tables 1 and 2), with k cat /K m ϳ 1 ϫ 10 5 M Ϫ1 ⅐s Ϫ1 for A. evansii PaaI and ϳ1 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 for E. coli PaaI. The C(3) and/or C(4)-ring hydroxylation increases the reactivity by 2 orders of magnitude. For the E. coli PaaI, this is primarily a k cat effect, whereas in the case of the A. evansii PaaI, the k cat and K m are equally impacted. The PaaIs, although not identical in catalytic efficiency, appear to be tailored for catalysis of hydroxyphenylacetyl-CoA hydrolysis.
The results from the present substrate screen provide insights into PaaI biochemical function. First, given that PaaI is only minimally active toward aliphatic acyl-CoA substrates, it is unlikely that it participates in the phenylacetate lower pathway (Fig. 1). Second, the rate of catalyzed hydrolysis of phenylacetyl-CoA is adequate for CoA release in the event that the pathway ring dioxygenase-reductase becomes limiting (Figs. 1  and 3). An analogous function has been suggested for the hot dog-fold STRUCTURE 1 Phenylacetate Pathway Thioesterase thioesterase of the benzoate metabolic pathway (26). Third, the high catalytic efficiency exhibited toward ring-hydroxylated phenylacetyl-CoA indicates that PaaI is poised to liberate CoA from dead-end products. Specifically, if the phenylacetate pathway proceeds via the novel 1,2-dihydrodiol intermediate as proposed (26,32) and if catalysis by the ring opening enzyme must compete with spontaneous dehydration to the dead-end product 2-hydroxyphenylacetyl-CoA as suggested in Mohamed (29), then the PaaI thioesterase might serve in a salvage capacity to release CoA and 2-hydoxyphenylacetate for alternate reaction pathways (Fig. 3). Similarly, if the phenylacetate pathway proceeds via a 2,3-dihydrodiol rather than the putative 1,2-dihydrodiol (this pathway alternative has not been ruled out) and if the 2,3-dihydrodiol undergoes spontaneous dehydration before undergoing ring cleavage, the PaaI might function to release CoA from 2-and 3-hydroxyphenylacetyl-CoA (Fig. 3). Clearly, additional work on the phenylacetate pathway and on its integration with the hydroxyphenylacetate degradation pathways that are also operative is needed to define the full range of PaaI biochemical function. Here we have merely demonstrated that the ring-hydroxylated phenylacetyl-CoA is the preferred substrate.
Quaternary and Tertiary Structure of the Apo E. coli PaaI-The structure of apo E. coli PaaI was determined to 2.0 Å of resolution (statistics are provided in Table 3, and electron density obtained from a simulated annealing omit map of the active site is represented in Fig. 4A) using crystals grown from the Seleno-Met protein at pH 6.0 in 13% polyethylene glycol, ammonium sulfate solution. The overall structure of the E. coli PaaI protomer resembles that of several other proteins in the PDB as revealed by a search with the program DALI (66). E. coli PaaI has the most structural and sequence similarity to PDB 1j1y, with a 1.3-Å root mean square deviation on 112 C␣ atoms with 36% sequence identity (PaaI protein from T. thermophilus (49)). E. coli PaaI is also similar to several other hot dog-fold thioesterases, although none exhibited sequence identities of greater than 24%. Among these is the Arthrobacter sp. strain SU 4-HBA-CoA thioesterase that will be the focus of the structure-function comparisons described below.
E. coli PaaI, which is a tetramer in solution, crystallized with a dimer in the crystallographic asymmetric unit. Although the protomer structures appear similar, both the N-and C-terminal regions adopt different structures. In protomer A, the N-terminal ␣-helix (amino acids 1-23) is continuous (Fig. 4B), whereas in protomer B the N-terminal ␣-helix is kinked at amino acid 16 (Fig. 4C). In addition, six additional amino acids are ordered in the electron density for protomer B.
The protomer topology is dominated by a 6-stranded anti-parallel ␤-sheet formed by amino acid residues 25-30, 33-39, 76 -84, 93-103, 107-115, and 121-130 (Fig. 4, B and C). Residues 2-22 of the N terminus form an ␣-helix (helix 1 in Fig. 4, B and C) that runs anti-parallel to ␤-strand A. The ␤-sheet wraps over one face of the second central ␣-helix (helix 2 in Fig. 4, B and C); hence, the name "hot dog-fold." Two protomers associate to form a continuous 12-stranded anti-parallel ␤-sheet, and this dimer associates back-to-back with a second dimer (Fig. 4D). The protomer-protomer interface is formed through interactions between respective ␤-strands (F) which orient anti-parallel (contacts exist between residues 78 and 84 and 80 and 82) and through interactions between the first two turns of the respective ␣-helices (2), also oriented anti-parallel (contact is between residues 18 and 52).
The locations of the active sites in the E. coli PaaI dimer are identified in Fig. 5A by the superposition of the structure of T. thermophilus PaaI bound with hexanoyl-CoA at one active site (49) and in Fig. 5B by the superposition of the structure of Arthrobacter sp. strain SU 4-HBA-CoA thioesterase complexed with the inhibitor 4-hydroxyphenacyl-CoA at both active sites (40). The acyl-CoA ligand is bound at the protomer-protomer interface with the acyl-thioester moieties inserted into the core of the dimer. The ligand CoA units rest on the dimer surface. The 4-hydroxyphenacyl-CoA ligand (Structure 1) was shown to be a competitive inhibitor of the A. evansii PaaI with a K i ϭ 39 M (Table 1).
Catalytic Site of the E. coli PaaI-The residues forming the catalytic site of E. coli PaaI are shown in Fig. 5C. The N-terminal ␣-helix, which is linear in protomer A (Fig. 4B) and kinked in protomer B (Fig. 4C), forms one side of the active site. The helix conformation, thus, impacts on the constitution of the active site. In protomer A, the Glu-14 and Asn-15 side chains project into the catalytic site, whereas in protomer B they are directed to solvent. In contrast, in protomer B the Asp-16 is directed at the active site, whereas in protomer A it is solvated. Which protomer A or B is the best representative of the catalytically active conformation of the enzyme? To address this question, the three residues Glu-14, Asn-15, and Asp-16 were separately replaced with Ala by site-directed mutagenesis. The kinetic properties of the mutants were tested with the three substrates 3,4-dihydroxyphenylacetyl-CoA, 3-hydroxyphenylac-  APRIL 21, 2006 • VOLUME 281 • NUMBER 16 etyl-CoA, and 4-hydroxyphenylacetyl-CoA (Table 2). Whereas the D16A mutant showed significantly reduced activity, the E14A and N15A mutants did not. Thus, protomer B appears to be the active conformer.

Phenylacetate Pathway Thioesterase
There are several ionizable residues located in the catalytic site of protomer B that might contribute to the pH rate profiles shown in Fig. 2. These include Asp-61, His-48, His-52, and Asp-16. Each of these residues is conserved in the A. evansii and T. thermophilus PaaIs.
To gain insight into the possible functions of the active site residues in the E. coli PaaI, the catalytic site of protomer B was superpositioned with the catalytic site of the Arthrobacter sp. strain SU 4-HBA-CoA thioesterase complexed with the inhibitor 4-hydroxyphenacyl-CoA (40) (Fig. 5D). The E. coli PaaI residues that surround the reaction center are Asn-46, His-52, Gly-53, Asp-61, and Thr-62. It is known from earlier studies that two key catalytic residues in the 4-HBA-CoA thioesterase are Glu-73, which functions as a nucleophile or as a general base, and Gly-65, which engages in hydrogen-bond formation with the thioester CAO (40,50,67). The corresponding residues in the E. coli PaaI are Asp-61 and Gly-53. To test the importance of Asp-61 to catalysis, it was replaced with Ala. A small but detectable level of catalytic activity was observed ( Table 2) that was suspected to derive from the contamination of the mutant protein with the native E. coli PaaI. To avoid potential contamination issues, the A. evansii PaaI was utilized because it could be separated from the native E. coli PaaI by fast protein liquid chromatography gel filtration chromatography (see "Materials and Methods"). Accordingly, the A. evansii PaaI D75A mutant was found to have no detectable activity. Thus, Asp-61, and by analogy Asp-75, are essential to E. coli and A. evansii PaaI catalysis, respectively. The A. evansii PaaI D75N mutant retained a small amount of activity (Table 1), which might be a reflection of the rate of attack by an oriented water molecule at the polarized thioester CAO.
The E. coli PaaI His52 corresponds to the Arthrobacter 4-HBA-CoA thioesterase active site His-64 (Fig. 5D), which is positioned for hydrogen-bond interaction with the catalytic Glu-73. The His-52 is seen in two different rotomer conformations in the E. coli PaaI (Fig. 5C), indicating some degree of conformational freedom in the unliganded active site. In unpublished work (68), catalysis by the Arthrobacter 4-HBA-CoA thioesterase H64A mutant was shown to be ϳ10 4 less efficient than that of the wild-type enzyme (k cat /K m ϭ 5 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 versus k cat /K m ϭ 3 ϫ 10 2 M Ϫ1 ⅐s Ϫ1 ). In the present study, E. coli PaaI H52Acatalyzed hydrolysis of the hydroxylated phenylacetyl-CoAs was evaluated ( Table 2). A 10 3 -fold decrease in the k cat /K m was observed with the 4-hydroxyphenylacetyl-CoA, and a 10 2 -fold decrease in the k cat /K m value was observed with the 3-hydroxyphenylacetyl-CoA or 3,4-dihydroxyphenylacetyl-CoA.
The E. coli PaaI Asn-46 is positioned opposite to the Asp-61 (Fig. 5D), just as the Gln-58 -Glu-73 pair is positioned in the Arthrobacter 4-HBA-CoA thioesterase (40) (Fig. 5D). To determine the contribution of the active site Asn-46 to PaaI catalysis, it was replaced with Ala by site-directed mutagenesis. In addition, the corresponding residue Asn-60 in the A. evansii PaaI was replaced with Ala and with Asp. The kinetic properties of the mutants (Tables 1 and 2) indicate that the Asn side chain contributes significantly to the efficiency of catalytic turnover.
Structural Basis for PaaI Substrate Specificity-Although 4-hydoxybenzoyl-CoA binds to the A. evansii PaaI active site (competitive inhibition constant K i ϭ 260 M; Table 1), it is not hydrolyzed. The PaaI utilizes an Asp for general base (or nucleophilic) catalysis, whereas the 4-HBA-CoA thioesterase uses a Glu. Likewise, the PaaI employs an Asn opposite to the catalytic carboxylate, whereas the 4-HBA-CoA thioesterase uses Gln (Fig. 5D). We asked the question, Does the Asn/Asp versus Gln/Glu control whether 3,4-dihydroxyphenylacetyl-CoA or 4-hydroxybenzoyl-CoA is the substrate? To address this question, the catalytic activity of the E. coli PaaI N46Q/D61E double mutant was evaluated with these two thioesters serving as substrates. No activity was observed. We conclude from this result that the substrate specificity cannot be simply switched by swapping the paired active site residues. Indeed, the catalytic sites of the two enzymes differ in other aspects. Most notably, the active site Thr residues (Thr-62 in PaaI and Thr-77 4-HBA-CoA thioesterase) are stationed differently; His-48 in PaaI replaces Trp-60 in the 4-HBA-CoA thioesterase, and the residue flanking the catalytic carboxylate in the E. coli PaaI is Ala-77 and Gly-93 in the 4-HBA-CoA thioesterase (Fig. 5D). Thus, substrate specificity derives from the composite of the active site residues. This is consistent with the observation that whereas the 4-hydroxyphenacyl-CoA (K i ϭ 39 M; K i ϭ 29 M for the 3-hydroxyphenacyl-CoA) and 4-hydroxybenzyl-CoA (K i ϭ 3.4 mM) analogs bind to PaaI (Table 1), they do so with much reduced affinity than is observed for binding to the Arthrobacter 4-HBA-CoA thioesterase (K i ϭ 0.003 M and K i ϭ 0.6 M, respectively).
The preference that PaaI shows for the 3-and 4-hydroxyphenylacetyl-CoA versus phenylacetyl-CoA (Tables 1 and 2) is striking. Inspection of the surface picture of the E. coli PaaI (Fig. 6) with the 4-hydroxy-  Residues are labeled and numbered with asterisks indicating side chain positions emanating from the adjacent protomer in the dimer interface. Structural representations were prepared with PyMol unless otherwise indicated (69). B and C, the stereo pictures of E. coli PaaI monomers, generated the coordinates deposited under PDB ID 1PSU and the graphics program Pymol (69). The ␣-helices 1 and 2 are colored yellow and cyan. The ␤-strands A to F and the loops are colored green. The N and C termini are labeled. D, the stereo pictures of E. coli PaaI tetramer, generated with the coordinates deposited under PDB code 1PSU and the graphics program Pymol (69). The ␣-helices 1 and 2 are colored yellow and cyan. The ␤-strands A to F and the loops are colored green. FIGURE 5. A, the stereo picture of the E. coli PaaI (cyan) dimer overlaid with the T. thermophilus PaaI dimer (yellow) bound with the ligand hexanoyl-CoA the type 2 orientation (blue) (49). The E. coli PaaI N and C termini are cyan, and the E. coli PaaI ␣-helices 1 and 2 and the T. thermophilus PaaI N and C termini are black. The figure was generated from the E. coli PaaI (PDB code 1PSU) and T. thermophilus PaaI (PDB ID 1WN3) using the graphics program Pymol (69). B, the stereo picture of the E. coli PaaI (cyan) dimer overlaid with the Arthrobacter 4-HBA-CoA thioesterase dimer (green) bound with the ligand 4-hydroxyphenacyl-CoA. The E. coli PaaI N and C termini are cyan, and the E. coli PaaI ␣-helices 1 and 2 and the Arthrobacter 4-HBA-CoA thioesterase N and C termini are as shown as black. The figure was generated from the E. coli PaaI (PDB code 1PSU) and Arthrobacter 4-HBA-CoA thioesterase (PDB code 1Q4T) using the graphics program Pymol (69). C, stereo picture of the active site regions of E. coli PaaI protomer A (cyan) and protomer B (blue). The residues E14, N15, D16, C18, D61, T62, A65, and A77 are from one subunit, whereas residues N46*, H48*, H52*, and G53* are from an adjacent subunit. The residues E14, N15, D16, and C18 are located on (N-terminal) ␣-helix 1, and D61, T62, and A65 are located on (central) ␣-helix 2. The residues N46, H48, and H52 are located on the connecting loop between ␤-strand B and ␣-helix 2, and G53 is located at the N-terminal of ␣-helix 2 (see Fig. 4, B and C). D, a stereo picture of E. coli PaaI active site residues (blue) (protomer B) overlaid with Arthrobacter 4-HBA-CoA thioesterase active site residues (green) with truncated ligand 4-hydroxyphenacyl-CoA (red). The H-bond and the trajectories between the Arthrobacter 4-HBA-CoA thioesterase Glu-73 carboxylate oxygen and ligand benzoyl carbonyl carbon are indicated with black dashed lines. The E. coli PaaI residues N46, H48, H52, G53, D61, T62, A65, and A77 correspond in their locations to the Arthrobacter 4-HBA-CoA thioesterase residues Q58, W60, H64, G65, E73, M74, T77, and G93, respectively.
phenacyl-CoA ligand (placed by the superimposition of liganded Arthrobacter 4-HBA-CoA thioesterase structure depicted in Figs. 5, B and D (40)) provides insight into why this might the case. The aromatic ring of the ligand sits in a binding pocket that opens to solvent. The ring hydroxyl substituents might engage in interaction with nearby water molecules.
Conclusions-PaaI shares the same branch of the evolutionary tree of the hot dog-fold thioesterase superfamily with the 4-HBA-CoA thioesterase from Arthrobacter. PaaI has diverged to recognize the phenylacetyl-CoA core structure rather than the benzoyl-CoA core structure. PaaI is most active with hydroxyphenylacetyl-CoA substrates. The partial exposure of the aromatic ring to solvent might account for this preference. Thus, by design or by accident, PaaI can liberate CoA from phenylacetyl-CoA and from hydroxyphenylacetyl-CoA substrates that form from the phenylacetate pathway.