Large Scale Structural Rearrangement of a Serine Hydrolase from Francisella tularensis Facilitates Catalysis*

Background: Acyl protein thioesterases control protein S-acylation at cellular membranes. Results: FTT258 is a serine hydrolase with broad substrate specificity that binds to bacterial membranes and exists in two distinct conformations. Conclusion: Conformational changes in FTT258 are correlated with catalytic activity. Significance: Structural rearrangement dually regulates the membrane binding and catalytic activity of acyl protein thioesterases. Tularemia is a deadly, febrile disease caused by infection by the Gram-negative bacterium, Francisella tularensis. Members of the ubiquitous serine hydrolase protein family are among current targets to treat diverse bacterial infections. Herein we present a structural and functional study of a novel bacterial carboxylesterase (FTT258) from F. tularensis, a homologue of human acyl protein thioesterase (hAPT1). The structure of FTT258 has been determined in multiple forms, and unexpectedly large conformational changes of a peripheral flexible loop occur in the presence of a mechanistic cyclobutanone ligand. The concomitant changes in this hydrophobic loop and the newly exposed hydrophobic substrate binding pocket suggest that the observed structural changes are essential to the biological function and catalytic activity of FTT258. Using diverse substrate libraries, site-directed mutagenesis, and liposome binding assays, we determined the importance of these structural changes to the catalytic activity and membrane binding activity of FTT258. Residues within the newly exposed hydrophobic binding pocket and within the peripheral flexible loop proved essential to the hydrolytic activity of FTT258, indicating that structural rearrangement is required for catalytic activity. Both FTT258 and hAPT1 also showed significant association with liposomes designed to mimic bacterial or human membranes, respectively, even though similar structural rearrangements for hAPT1 have not been reported. The necessity for acyl protein thioesterases to have maximal catalytic activity near the membrane surface suggests that these conformational changes in the protein may dually regulate catalytic activity and membrane association in bacterial and human homologues.


Tularemia is a deadly, febrile disease caused by infection by the Gram-negative bacterium, Francisella tularensis.
Members of the ubiquitous serine hydrolase protein family are among current targets to treat diverse bacterial infections. Herein we present a structural and functional study of a novel bacterial carboxylesterase (FTT258) from F. tularensis, a homologue of human acyl protein thioesterase (hAPT1). The structure of FTT258 has been determined in multiple forms, and unexpectedly large conformational changes of a peripheral flexible loop occur in the presence of a mechanistic cyclobutanone ligand. The concomitant changes in this hydrophobic loop and the newly exposed hydrophobic substrate binding pocket suggest that the observed structural changes are essential to the biological function and catalytic activity of FTT258. Using diverse substrate libraries, site-directed mutagenesis, and liposome binding assays, we determined the importance of these structural changes to the catalytic activity and membrane binding activity of FTT258. Residues within the newly exposed hydrophobic binding pocket and within the peripheral flexible loop proved essential to the hydrolytic activity of FTT258, indicating that structural rearrangement is required for catalytic activity. Both FTT258 and hAPT1 also showed significant association with liposomes designed to mimic bacterial or human membranes, respectively, even though similar structural rearrangements for hAPT1 have not been reported. The necessity for acyl protein thioesterases to have maximal catalytic activity near the membrane surface suggests that these conformational changes in the protein may dually regulate catalytic activity and membrane association in bacterial and human homologues.
Francisella tularensis is an endemic, zoonotic Gram-negative bacterium and the causative agent for the febrile disease, tularemia (1,2). Although F. tularensis infection is uncommon in the general population, the extreme virulence and relatively high mortality rate of F. tularensis infection have led to the classification of F. tularensis as a probable biological warfare agent (3,4). As evidence of its extreme virulence, F. tularensis infection can be initiated with as few as 10 organisms and if left untreated has mortality rates of 30 -60% (5).
Serine hydrolases are an emerging class of therapeutic targets (6 -8), including novel antibiotic targets (9,10). Serine hydrolases (EC 3.1) are members of the highly conserved ␣/␤-hydrolase enzyme superfamily and are involved in a variety of key biological processes in both eukaryotes (neural signaling, signal transduction, and metabolism) and prokaryotes (quorum sensing and virulence) (6,8,11). Evidence of their important biological functions in mammals is shown by the existence of over 110 identified members of this family (6,11). To accomplish their diverse biological roles, serine hydrolases target a variety of compounds that include ester, thioester, and epoxide bonds (12). Members of the superfamily can also catalyze unusual chemical transformations for various commercial applications (12)(13)(14). The defining characteristics of this diverse family of enzymes are their conserved ␣/␤-hydrolase protein fold, conserved catalytic triad, and conserved catalytic mechanism (12).
One subgroup of serine hydrolases of recent pharmacological interest is eukaryotic acyl protein thioesterases (EC 3.1.2.22), which control dynamic cycles of S-acylation on a subclass of palmitoylated and membrane-bound proteins (15). The archetype of an acyl protein thioesterase is human acyl protein thioesterase 1 (hAPT1), 5 which controls essential cellular functions including signal transduction, ion transport, and neural development (16 -18). hAPT1 was recently confirmed as a target for cancer therapeutics, where inhibition of hAPT1 led to partial phenotypic reversion of Ras-dependent cancers (18). Bacterial lipoproteins are also S-acylated at N-terminal cysteine residues with a growing number of lipoproteins from F. tularensis having been directly connected to the virulence and pathogenicity of F. tularensis (19 -22). These lipoproteins from F. tularensis localize to the outer bacterial membrane and activate the mammalian inflammatory response through toll-like receptor 2 (TLR2) signaling (19 -22). The central importance of bacterial lipoproteins in the virulence of F. tularensis parallels other Gram-negative pathogens, including Neisseria meningitidis, Borrelia burgdorferi, and Escherichia coli, where lipoproteins are essential to their infectivity and their recognition by the host immune system (23,24).
Although no bacterial deacylases have been identified, homologues of hAPT1 are found in a variety of bacterial species, including F. tularensis (1). Two bacterial homologues of hAPT1 from Pseudomonas aeruginosa and Pseudomonas fluorescens have been previously characterized as ␣/␤ hydrolases with broad substrate specificity and high structural and sequence similarity to hAPT1 (25)(26)(27). Despite the important biological functions and potential therapeutic significance of acyl protein thioesterases, structural data are only available for hAPT1 (Protein Data Bank (PDB) ID 1FJ2), PA3859 from P. aeruginosa (CPa) (PDB IDs 3CN9 and 3CN7), and carboxylesterase II from P. fluorescens (CPf) (PDB IDs 1AUO and 1AUR) (26 -28). Each of these enzymes displays a canonical ␣/␤ hydrolase fold with a seven-stranded central ␤-sheet surrounded by six ␣-helices and a classic catalytic triad of serine, histidine, and aspartate. Significant variations between the structures were observed in the loop insertions on the edge of the ␤ sheet and in the binding pocket (26 -28). These loop insertions in ␣/␤ hydrolases often confer their unique substrate specificity or are responsible for their conformational changes during adsorption to the lipid phase (29). The major difference between the structures of the human and bacterial homologues is that the putative substrate-binding site in hAPT1 is narrower and has limited access compared with the binding pockets of the Pseudomonas esterases that can accommodate a range of substrates (26 -28). This divergence in structure suggested differential substrate specificities and biological functions between the human and bacterial enzymes (26).
The different conformations of the structures of hAPT1 and bacterial homologues have made comparison of biological functions, catalytic activity, and substrate specificity difficult. Herein, we describe the three-dimensional structure of FTT258, a novel hAPT1 homologue from F. tularensis, in two different structural forms and bound to a mechanistic ligand. We determine the catalytic activity and substrate specificity of FTT258 and directly compare this activity to hAPT1. Additionally, we measure the membrane binding ability of FTT258 and hAPT1 and the importance of the conformationally flexible ␤ 2 and ␤ 3 loops on the catalytic activity of FTT258. Overall, the structural and functional characterization of FTT258 suggests a unique mechanism for linking the regulation of catalytic activity and membrane binding that could control the spatiotemporal activity of acyl protein thioesterases.

FTT258 Protein Preparation and Crystallization
The FTT258 gene was subcloned into the pMCSG7 vector, heterologously expressed in E. coli, and purified as previously described (30). Crystallization trials were conducted using a variety of Qiagen crystallization screens, including Classics II, JCSGϩ, and PACT. The FTT258 protein (7.4 mg/ml) in 10 mM Tris-HCl buffer, pH 8.3, 500 mM NaCl, 5 mM ␤-mercaptoethanol, and 1 mM inhibitor (compound 14, Fig. 1C) was used for the vapor diffusion sitting drop crystallization method with 1 l of protein and 1 l of reservoir solution. Crystals composed of large plates grew in 0.1 M ammonium acetate, 0.1 M Bis-Tris, pH 5.5, and 17% PEG 10,000 at room temperature after 3 months. No crystals grew in the absence of cyclobutanone compound.

Data Collection and Structure Determination
A single thin plate crystal of FTT258 protein was cryo-cooled in the reservoir solution. Data were collected at 100 K at the Life Sciences Collaborative Access Team at the Advance Photon Source (Argonne, IL) using a CCD MarMosaic 225 detector. The unit cell was identified with DENZO as P1, a ϭ 51.0 Å, b ϭ 64.4 Å, c ϭ 139.0 Å, ␣Å, b ϭ 9°Å, b ϭ 64.4 Å, ϭ 89.9°. Data were processed with HKL2000 for integration and scaling (31). The structure of FTT258 was solved by molecular replacement with PHASER (32) from the CCP4 suite (33) using the apo-form structure of hAPT1 (PDB ID 1FJ2) as the starting model. Gaps, turns, and side chains for the solved structure were fit manually using the program COOT (34).

Model Refinement and Quality of the Structure
In an effort to further improve the map, the free atom refinement procedure and automatic chain tracing as implemented in ARP/wARP were performed (35). Upon completion of model building, the structure was refined by the phenix.refine program (36) using XYZ coordinates, group B-factors, TLS parameters, occupancies, and twinning refinement strategies. The automatic recognition of water molecules was resolved in ARP/ wARP (35). Details of data collection, structure determination, and refinement are provided in Table 1.
The final model of FTT258 consists of eight refined protein molecules (A, B, C, D, E, F, G, H), 459 water molecules, and the cyclobutanone ligand 14 covalently bound as the hemiacetal to Ser-116 of molecule E. The protein used in this study contains an N-terminal polyhistidine tag (His tag) plus a tobacco etch virus cleavage site (24 residues long) followed by the FTT258 sequence starting at methionine 1. The last two residues of the His tag are resolved in the electron density map in molecules A, B, C, and D, and the last three residues of the His tag are resolved in molecules E, F, G, and H of the structure. Disordered regions were observed in each of the eight 222-amino acid polypeptide chains in the asymmetric unit and indicate the localized flexibility of the FTT258 protein structure. The residues that were not well defined include 22-23 and (37)(38)(39)(40) and shows 94% of the residues in the most favored / regions and no outliers, with the exception of the catalytic residue Ser-116 in four molecules (A, D, F, and G) of the structure. This unique conformation "a nucleophile elbow" around the catalytic serine has been observed in other ␣/␤ hydrolases (41). The atomic coordinates and structure factor data for FTT258 are available from the PDB with the accession code 4F21.

Synthesis of Cyclobutanone Compounds
All reactions were performed under an atmosphere of nitrogen, and all solvents and reagents were used without further purification unless otherwise noted. Merck silica gel 60 (230 -400 mesh) was used for flash chromatography. Merck Kieselgel 60 F254 DC-Fertigplatten (0.25 mm, Art. 5719) was used for TLC. 1 H NMR spectra were obtained from either a Varian INOVA 300 or Varian Gemini 2000 300 MHz spectrometer with tetramethylsilane as an internal standard. Melting points were determined on a Thomas-Hoover melting point apparatus and are uncorrected. Noise-decoupled and 13 C NMR spectra were recorded at 75 MHz on either the Varian INOVA 300 or the Varian Gemini 2000 spectrometer. IR spectra were recorded on a Thermo Nicolet Nexus 670 FT-IR. Mass spectra were run on a Thermo Finnigan LCQ Advantage instrument.
Building upon earlier work describing the preparation of variously substituted 2-amino cyclobutanones (42) and an extension of the work by Vederas and co-coworkers (43), dimethyl succinate was alkylated (Scheme 1) with allyl bromide after deprotonation with LDA to yield dimethyl 2-allylsuccinate i, which was then subjected to the general acyloin method of Frahm and co-workers (44) to afford allyl-substituted (bistrimethylsilyloxy)cyclobutene ii. Treatment of cyclobutene derivative ii with p-toluenesulfonamide in the presence of HCl afforded allyl-cyclobutanone stereoisomers iii, iv, and 14, which were separable by normal-phase column chromatography on a gram scale (Scheme 1). Regiochemical and stereochemical assignments were made by comparison of DEPT spectra of the three compounds with calculated DEPT spectra generated by SpartanPro'10 (Wavefunction) from ab initio density functional level of theory using the RB3LYP functional and the 6-31-G(D) basis sets. Compounds were identified as the (Ϯ)-trans-3-allyl stereoisomer for the first-eluted isomer iii, (Ϯ)-cis-2-allyl stereoisomer as the second-eluted isomer iv, and (Ϯ)-trans-2-allyl stereoisomer as the third-eluted isomer 14 (see the supplemental DEPT spectrum).

Dimethyl 2-(Prop-2-en-1-yl)butanedioate (i)
n-Butyl lithium (64.7 ml, 0.107 mol) was added dropwise to a solution of anhydrous THF (450 ml) and freshly distilled diisopropyl amine (14.7 ml, 0.105 mol) at Ϫ78°C. The reaction was allowed to stir at Ϫ78°C for 45 min. Dimethylsuccinate (13.4 ml, 0.103 mol, 1 eq) was added dropwise to the reaction maintaining a temperature of Ϫ78°C. The reaction was allowed to stir at Ϫ78°C for 30 min. Freshly distilled allyl bromide (10.65 ml, 0.123 mol, 1.2 eq) was slowly added dropwise, and the reaction was allowed to warm to ambient temperature producing a cloudy yellow solution. The reaction was then quenched with cold water (100 ml) and extracted with diethyl ether (100 ml ϫ3). The combined organic layers were washed once with brine (100 ml) and dried over sodium sulfate. The solvent was then removed under vacuum. Column chromatography on silica gel eluting with petroleum ether/ethyl acetate (9/1) provided dimethyl 2-allylsuccinate (i) as a colorless oil (11.9 g, 0.064 mol, 62%). R f ϭ 0.15 (PE-EA, 9:1

Di-substituted Cyclobutanones
3-Allyl-1,2-bis(trimethylsilyloxy)cyclobutene ii (5.0 g, 0.019 mol) was added dropwise to a solution of 4-methylbenzenesulfonamide (3.90 g, 0.023 mol, 1.2 eq) in a 1.0 M HCl solution in diethyl ether (90 ml) cooled to 0°C over 15 min. The reaction was then heated under reflux for 4 h at 80°C. Concentration gave a brown residue that was purified on silica gel using a high loading ratio (200:1), eluting with EtOAc/DCM (EA-DCM, 0.5/ 99.5) to afford the desired di-substituted cyclobutanones, in order of elution from the column as follows.

Inhibition Assays
The inhibitory properties of the cyclobutanone compounds on FTT258 were tested using the enzyme purified for crystallization trials. Activity was measured against p-nitrophenyl acetate in the presence or absence of different cyclobutanone compounds (iii, iv, 14, Scheme 1). The substrate and cyclobutanone compounds were solubilized in 100% ethanol and 100% DMSO, respectively, with 1% final concentration of each solvent in the reaction. The reaction mixture was composed of 100 mM Tris-HCl, pH 7.5, 1 mM p-nitrophenyl acetate and varying concentrations of cyclobutanone compound in a 0.3-ml volume for 10 min at room temperature with 23 nM enzyme. The enzyme was incubated with the cyclobutanone compound in buffer for 10 min before starting the reaction. Reactions were initiated with the substrate p-nitrophenyl acetate, and the initial rate of p-nitrophenol production was determined by measuring its absorbance at 415 nm in a BioTek ELx808 microplate reader equipped with filters (Biotek Instruments; Winooski, VT). All reactions were performed in triplicate.

FTT258 Purification for Kinetic Characterization
A bacterial expression plasmid (pDest17-FTT258) containing the FTT258 gene from F. tularensis (Target GenBank TM : SCHEME 1. Synthesis of cyclobutanone ligands of FTT258. Synthesis shown is of cyclobutanone ligands of FTT258 via alkylation of dimethyl succinate, acyloin condensation, and acid-catalyzed addition of p-toluenesulfonamide followed by normal-phase chromatographic separation of isomers. Each of the three compounds contains a conserved cyclobutanone ring, a sulfonamide, and an allyl substituent, differing only in the stereochemistry and attachment point of the allyl substituent (iii, (Ϯ)-trans-3-allyl; iv, (Ϯ)-cis-2-allyl; 14, (Ϯ)-trans-2-allyl). Compound 14 was the most potent inhibitor of FTT258 and was used for co-crystallization trials. This ligand was bound to FTT258 in the crystal structure.
YP_169310; protein name FTT258) was obtained from the Harvard Plasmid Repository (clone ID FtCD00063611). This bacterial plasmid was transformed into E. coli BL21 (DE3) RIPL cells (Agilent). A saturated overnight culture was used to inoculate LB-media (1.0 liter) containing ampicillin (100 g/ml) and chloramphenicol (30 g/ml), and the bacterial culture was grown with constant shaking (225 rpm) at 37°C. Once the A 600 reached 0.6 -0.8, the temperature was dropped to 16°C, and isopropyl ␤-D-1-thiogalactopyranoside (0.5 mM; Gold Biotechnology) was added to induce protein expression for 12-16 h. Bacterial cultures were collected by centrifugation at 6000 ϫ g for 10 min at 4°C. The bacterial cell pellet was resuspended in PBS (10 ml) and stored at Ϫ20°C.
To disrupt the bacterial cell wall, lysozyme (50 mg; Sigma) and BugBuster detergent (1ϫ, EMD Chemicals) were added and incubated with the cell pellet for 60 min at 25°C on a rotating shaker. Precipitated proteins and cell material were removed by centrifugation at 18,000 ϫ g for 10 min at 4°C. Ni-nitrilotriacetic acid-agarose (1 ml; Qiagen) was added to the soluble fraction and incubated at 25°C for 15 min on a rotating shaker. The resin was transferred onto a gravity flow column and washed 1 time each with PBS containing increasing concentrations of imidazole (40 ml each of PBS containing 10 mM imidazole, 25 mM imidazole, or 50 mM imidazole). FTT258 was eluted in PBS containing 250 mM imidazole (600 l) and dialyzed against PBS overnight at 4°C with constant stirring (10,000 molecular weight cut-off; Pierce).
The purity of FTT258 was confirmed by SDS-PAGE on a 4 -20% gradient gel, and the purity was shown to be Ͼ95%. The concentration of FTT258 was determined by measuring the absorbance at 280 nm and by calculating the extinction coefficient (⑀ 280 ϭ 29,910 M Ϫ1 s Ϫ1 with all free cysteines) using the modified Edelhoch and Gill/Von Hippel methods on Expasy (45)(46)(47).

hAPT1 Purification
A bacterial plasmid (pDONR221) containing the LYPLA1 gene from Homo sapiens (Target GenBank TM CB008652; protein name hAPT1) was obtained from the Arizona State University Plasmid Repository (Clone ID HsCD00043610). The LYPLA1 gene was subcloned into the pDest17 bacterial expression vector using the Gateway cloning system (Invitrogen). The hAPT1 protein was then heterologously expressed in E. coli and purified to homogeneity using identical conditions to the purification of FTT258. The concentration of hAPT1 was determined by measuring the absorbance at 280 nm using the extinction coefficient (⑀ 280 ϭ 20,970 M Ϫ1 s Ϫ1 ).

Site-directed Mutagenesis and Purification
Variants of FTT258 were created by QuikChange II site-directed mutagenesis using the manufacturer's suggested procedure (Agilent). Proper mutations in the FTT258 DNA sequence were confirmed by DNA sequencing (Genewiz). Plasmids with FTT258 variants were transformed, expressed, and purified using the same procedure as the wild-type FTT258.

Enzymatic Activity against Fluorogenic Enzyme Substrates
The enzymatic activity of FTT258 was measured against fluorogenic enzyme substrates (Fig. 1A) using a 96-well microplate assay (48). Fluorogenic substrates were diluted into PBS containing acetylated BSA (PBS-BSA; 0.1 mg/ml) to a starting concentration of 10 M from 10 mM stock solutions in DMSO. Eight serial dilutions (1:2; 55 l into 165 l total volume) of each substrate (10 M-4.6 nM final concentrations) were made using PBS-BSA. Fluorogenic enzyme dilutions (95 l) were then transferred to a black 96-well microplate. To initiate the reaction, wild-type or variants of FTT258 (5 l of 125 g/ml) were added to the diluted fluorogenic enzyme substrates in the 96-well microplate. The FTT258 concentration in each reaction was 6.25 g/ml, and the fluorescence change ( ex ϭ 485 nm, em ϭ 528 nm) was measured for 4 min at 25°C on a Biotek Synergy 2 fluorescent plate reader (Biotek Instruments). The fluorescence change was converted to molar concentrations using a fluorescein standard curve (30 nM-0.23 nM), which was produced for each experiment. The initial rates of the reactions were measured in triplicate and plotted versus fluorogenic enzyme substrate concentration. The saturation enzyme kinetic traces were fitted to a standard Michaelis-Menten equation, and values for k cat , K m , and k cat /K m were calculated using Origin 6.1 (OriginLab Corp.). The background fluorescence of substrates 1-9 was determined by measuring the kinetics of fluorogenic enzyme substrate hydrolysis using an active site knock-out of FTT258 (S116A).

Enzymatic Activity against p-Nitrophenyl Substrates
The enzymatic activity of FTT258 was measured against p-nitrophenyl acetate, p-nitrophenyl butyrate, p-nitrophenyl octanoate, and p-nitrophenyl laurate (Sigma) (Fig. 1B) using a 96-well microplate assay (49). All four substrates, p-nitrophenyl acetate (2 M), p-nitrophenyl butyrate (2 M), p-nitrophenyl octanoate (200 mM), and p-nitrophenyl laurate (200 mM), were prepared as stock solutions in acetonitrile and diluted into PBS containing acetylated BSA (PBS-BSA; 0.1 mg/ml) to reduce the nonspecific adsorption to the plasticware common in microplate analysis (50). The starting concentration for p-nitrophenyl acetate was 20 mM and for p-nitrophenyl butyrate, p-nitrophenyl octanoate, and p-nitrophenyl laurate was 2 mM. Eight serial dilutions (1:1; 110 l into 220 l of total volume; 20 mM-156 M final concentrations for acetate; 2 mM-15.6 M for butyrate, octanoate, and laurate) were made using PBS-BSA containing 1% acetonitrile. Substrate dilutions (95 l) were transferred to a clear 96-well microplate, and FTT258 (5 l) was added to start the reaction. FTT258 concentrations in each reaction were 12.5 g/ml for acetate and butyrate and 37.5 g/ml for octanoate and laurate. The absorbance change at 412 nm was measured on a Biotek Synergy 2 fluorescent plate reader for 4 min at 25°C. The change in absorbance was converted to molar concentrations using the extinction coefficient of p-nitrophenol (⌬⑀ 412 ϭ 1.034 mM Ϫ1 cm Ϫ1 ) (51). Initial rates of the reactions were measured in triplicate, and kinetic parameters were determined in the same manner as the fluorogenic assay.

Liposome Binding Assay
Liposomes were prepared with lipids purchased from Avanti Polar Lipids as described previously (52). Bacterial liposomes contained molar ratios of 75% phosphatidylethanolamine, 20% phosphatidylglycerol, and 5% cardiolipin. Eukaryotic liposomes contained molar ratios of 60% phosphatidylcholine, 20% phosphatidylethanolamine, and 20% cholesterol. Purified recombinant proteins (3 M) were incubated alone or with liposomes (3 mM) at 37°C for 5 min then precipitated by ultracentrifugation at 200,000 ϫ g for 120 min at 25°C. Proteins in the soluble and pellet fractions were separated by SDS-PAGE and visualized by Coomassie Blue staining.

RESULTS
Kinetic Characterization of FTT258 and hAPT1-Acyl protein thioesterases are a highly conserved subclass of serine hydrolases with diverse intracellular substrates and significant therapeutic applications (16). Based on the role of bacterial lipoproteins in the virulence of F. tularensis and the confirmed pharmacological importance of hAPT1 (18), we sought to identify and study an APT1 homologue from F. tularensis and compare it to hAPT1. The F. tularensis subsp. tularensis SCHU S4 genome was searched for bacterial homologues to hAPT1, and the gene FTT258 was identified as a predicted bacterial esterase and potential acyl protein thioesterase. Sequence alignment of the FTT258 protein to diverse APT1 homologues gave high sequence identity to both hAPT1 (39%) and hAPT2 (34%) and to previously characterized bacterial homologues from P. aeruginosa (40%) and P. fluorescens (38%) ( Table 2).
The FTT258 and hAPT1 proteins were heterologously expressed in E. coli, purified to homogeneity, and subjected to detailed kinetic analysis to confirm that FTT258 was a serine hydrolase and compare its activity to previously characterized Pseudomonas homologues. Two distinct libraries of esterase substrates were used to determine the enzymatic activity and substrate specificity of the recombinant FTT258 and hAPT1 proteins. The first library consisted of a synthesized set of nine fluorogenic enzyme substrates that contain ester substituents of varying alkyl length, steric bulk, and stereoelectronic properties (Fig. 1A) (53,54). A second esterase library was composed of commercially available p-nitrophenyl esterase substrates with acyl ester substituents of various chain lengths (C2, C4, C8, and C12; standard lipid numbering is used where the carbonyl carbon is C1) (Fig. 1B). To confirm that FTT258 is a bacterial esterase, the predicted serine nucleophile (Ser-116) was mutated to alanine, which ablated all catalytic activity and showed that FTT258 is indeed a classic serine hydrolase with a GXSXG catalytic motif (Table 2) (12,55).
Precise enzyme kinetic measurements of FTT258 and hAPT1 against all nine fluorogenic enzyme substrates produced classic Michaelis-Menten saturation kinetics. The highest catalytic activity of FTT258 was against straight-chain alkyl substrates (substrates 1-5) with maximal activity against valeryl (C5) and butyl (C4) esters ( Fig. 2A). The catalytic activity increased incrementally from C2-C5 esters but then dropped off against the C6 ester (substrate 5). When looking at absolute k cat /K m values, FTT258 maintained catalytic activity greater than 1.0 ϫ 10 3 M Ϫ1 s Ϫ1 toward all fluorogenic enzyme substrates, except substrates 7 and 8, highlighting the broad substrate specificity of FTT258 toward a wide range of esters. Similar broad substrate specificity was observed against p-nitrophenyl esters, as FTT258 had k cat /K m values within a factor of two toward alkyl chains ranging from C2 to C8 (Tables 3 and  4). The substrate specificity profile of FTT258 closely parallels the specificity of bacterial homologues from P. aeruginosa and P. fluorescens, as the Pseudomonas homologues also showed

binding pocket, and membrane binding loop
The amino acid sequence of FTT258 was aligned using ClustalW (63). The amino acid numbering corresponds to the amino acid numbering in FTT258. The number in parentheses refers to whether the protein has been designated as a closer homologue of human APT1 (1) or APT2 (2 APRIL 12, 2013 • VOLUME 288 • NUMBER 15

Structural Rearrangement of a Hydrolase from F. tularensis
broad substrate specificity across a range of shorter alkyl chain (Ͻ10 carbons) with maximal activity toward C8 (25)(26)(27). Recombinant hAPT1 and FTT258 had nearly identical relative and absolute k cat /K m values for p-nitrophenyl substrates of C4, C8, and C12 (Tables 3 and 4). The enzymatic similarity between FTT258 and hAPT1 does not, however, transfer to the fluorogenic enzyme substrates (Fig. 2C). hAPT1 does not display a recognizable trend in catalytic activity toward the library of fluorogenic enzyme substrates (Fig. 2C), even toward straight-chain alkyl substrates, whereas FTT258 had a nearly linear increase in catalytic activity from C2 to C5 (substrate 1-4; Fig. 2A). The comparison of k cat /K m values between FTT258 and hAPT1 accentuates the broad substrate specificity of FTT258.
Structure of FTT258-To determine the level of structural conservation between FTT258 and bacterial and human homologues, crystallization trials were conducted with FTT258. Initial crystallization trials of apoFTT258 did not yield suitable results, so co-crystallization was conducted in the presence of a cyclobutanone ligand (Fig. 1C). Cyclobutanones have previously been explored as reversible ␤-lactamase inhibitors with modest, but broad-spectrum ␤-lactamase inhibition. Cyclobutanones also exhibit covalent binding as the hemiketal to the active-site serine of serine hydrolases as seen crystallographically (56). The three different isomers of a mechanistic cyclobutanone ligand (Scheme 1) were screened for relative activity against FTT258 before attempting co-crystallization. Compounds iii (3.10 Ϯ 0.12 mol/min/mg), iv (4.50 Ϯ 0.08 mol/ min/mg), and 14 (1.95 Ϯ 0.21 mol/min/mg) inhibited the catalytic activity of FTT258 (4.72 Ϯ 0.02 mol/min/mg) to varying degrees with compound 14 as the tightest ligand (IC 50 ϭ 1.12 mM). As the ring substituents on the cyclobutanone rings remain consistent, the increased binding affinity for cyclobutanones iii and 14 may be due to the conserved trans stereochemistry in these two derivatives. Co-crystallization of FTT258 in the presence of the cyclobutanone ligand (14) at slightly below its IC 50 value was successful, and the three-dimensional structure was refined to 2.5 Å resolution.
In the final structural model, FTT258 is a monomer but has eight molecules per asymmetric unit. Each molecule shows a canonical ␣/␤ hydrolase fold with a classic catalytic triad of  (Fig. 1A) is shown. B, shown is a comparison of the enzymatic efficiency (k cat /K m ) of FTT258 variants against substrates 2 and 9. All enzymatic efficiency values for FTT258 variants were normalized based on the enzymatic efficiency of wild-type FTT258 against each substrate. Kinetics values for each FTT258 variant and substrate are given in Tables 3 and 4. C, substrate specificity of hAPT1 is shown. Enzymatic efficiency of fluorogenic enzyme substrate hydrolysis by hAPT1 against all nine members of the substrate library (Fig. 1A) is shown. Kinetic values are given in Tables 3 and 4.  The order of the strands within the ␤-sheet is linear with respect to the sequence except at the N terminus (Fig. 3A), where ␤ 3 is positioned between ␤ 1 and ␤ 2 . The ␤-sheet lacks the first ␤-strand of the prototypic ␣/␤ hydrolase fold (57). Among the seven strands in the central ␤ sheet, six parallel ␤ strands, ␤ 2 -␤ 7 , provide the framework onto which the catalytic residues reside (Fig. 3A).
Large Scale Structural Rearrangement-Superposition of the FTT258 subunits within the asymmetric unit reveals two different conformations of the polypeptide chain (Fig. 3A). Four molecules of the protein structure (A, B, C, and D) show a closed conformation, and four molecules (E, F, G, and H) display an open conformation, with one molecule (chain E) covalently bound to a cyclobutanone ligand. The transition between the closed and open forms of FTT258 requires a large conformational change that is centered around the active site and the binding pocket of FTT258 ( Fig. 3A and 4, A, B, and D). The shift between the closed and open forms involves four key changes in the FTT258 structure 1) reorienting the active site His-Asp dyad, 2) removing the inhibitory action of Asp-25, which blocks the substrate binding pocket in the closed form, 3) rearranging the ␤ 2 and ␤ 3 loops, which exposes the hydrophobic substrate binding pocket, and 4) flipping out a potential membrane binding loop containing Trp-66 and Tyr-67. For the ligand-bound form, the electron density shows the cyclobutanone moiety of compound 14 covalently bound to Ser-116 (Fig. 4D), but the remaining substituents on the ligand are unresolved. The lack of strong electron density for the cyclobutanone ligand likely reflects its weak affinity for FTT258.
These large scale rearrangements between the closed and open forms of FTT258 recapitulate many features of a classic serine hydrolase mechanism where a His-Asp dyad (His-202 and Asp-170) activates the nucleophilic serine (Ser-116) for direct attack on the ester carbonyl (12). In the transition between the closed and open forms of FTT258, the catalytic His-Asp pair shifts with respect to the core of the structure and the nucleophilic serine (Ser-116) (Fig. 4C). The Ser-116 also forms a direct H-bond with the NE2 of His-202 in the closed form, which disappears upon rearrangement to the open form (Fig. 4D). In the canonical serine hydrolase mechanism, the transient oxyanion formed from the ester carbonyl upon reaction with the catalytic serine is also stabilized by a positively charged oxyanion hole (12). In FTT258, the backbone NH on Leu-22 and the NE2 NH of Gln-117 form the oxyanion hole in the closed form, and the oxyanion hole becomes occupied by the cyclobutanone hemiacetal oxygen in the ligand bound open state (Fig. 4, C and D). The structural rearrangement also affects the substrate binding pocket, as Asp-25 is positioned across the binding pocket in the closed form occluding the hydrophobic binding pocket. Upon shifting to the open form, the hydrophobic binding pocket is exposed (Fig. 4, A and B), similar to the induced fit binding observed in lipases (29,58).
The conformational rearrangement in the structure of FTT258 also shifts the position of a conserved ␤ 3 loop (Figs. [3][4][5]. The movement of the ␤ 3 loop is centered around newly exposed tryptophan (Trp-66) and tyrosine (Tyr-67) residues, which are bracketed by two positively charged residues (Arg-64 and Lys-70). The exposure of Trp-66 and Tyr-67 in the FTT258 structure is stabilized by extensive hydrophobic burial of both amino acids in neighboring FTT258 subunits, and this interaction is conserved across all subunits in the asymmetric unit (Fig.  5). This motif in the ␤ 3 loop of hydrophobic residues surrounded by positively charged residues closely match with membrane attachment domains observed in bacterial toxins and predicted in bacterial lipases (52,59). Based on the location, composition, and conservation of residues involved in the conformational change, we hypothesized that the structural rearrangement observed in the crystal structure of FTT258 has direct relevance to its catalytic activity, substrate specificity, and potentially membrane binding.
Structural Rearrangement in Bacterial and Human Homologues-To determine if similar structural plasticity was observed in the structures of bacterial and human homologues of FTT258, the open and closed conformations of FTT258 were aligned with the structures of hAPT1, CPa, and CPf. For these structures, the root mean squared difference calculated for the C␣ atoms between the closed structure of FTT258 and the structures of hAPT1, CPa, and CPf are only 1.53, 1.48, and 1.45 Å, respectively (Fig. 3C). Major differences between the structure of FTT258 and homologous enzymes revolve around Structural Rearrangement of a Hydrolase from F. tularensis APRIL 12, 2013 • VOLUME 288 • NUMBER 15 the variable movement of the conformationally flexible ␤ 2 and ␤ 3 loops and proximal ␤-sheets from the FTT258 structure (Fig. 3C).  (26,27). Similar to the substrate specificities of the Pseudomonas homologues, the hydrophobic binding pocket on FTT258 can accommodate a range of hydrophobic substrates using the side chains from putative binding pocket residues Leu-22, Leu-72, Phe-115, and Val-172 (26,27). The transition of the ␤ 2 and ␤ 3 loops in the crystal structure of FTT258 reorients these hydrophobic residues in the active site, suggesting that they play an essential role in binding of FTT258 substrates. Conversely, the closed form of the FTT258 structure depicts a narrower hydrophobic binding pocket similar to the hAPT1 structure (Fig. 3C) (28). The structural resemblance of the two conformations of FTT258 to the corresponding structural homologues reinforces the hypothesis that the two forms of FTT258 depict snapshots of a structural rearrangement required for catalysis and/or membrane binding. To examine the importance of the structural rearrangement to   catalysis and membrane binding, each of these potential functions was analyzed independently.
Structural Rearrangement and Catalytic Activity-To determine the importance of the structural rearrangement to catalytic activity, key residues positioned in the buried hydrophobic binding pocket and in the ␤ 2 and ␤ 3 loops were substituted with alanine, and the kinetic activity of the variants was determined against two fluorogenic substrates (substrates 2 and 9). Among amino acids located in the buried hydrophobic binding pocket (Leu-22, Leu-72, Phe-115, and Val-172), each one of these hydrophobic residues proved essential for FTT258 catalysis as the enzymatic activity of the alanine variants dropped to near background levels against both substrates 2 and 9 (Fig. 2B). Drastic shifts in enzymatic activity with substitution of each of the binding pocket residues suggests that this hydrophobic network of substrate binding residues is essential to the activity and substrate binding specificity of FTT258. These results also suggest that the exposure of these hydrophobic amino acids upon structural rearrangement is necessary for catalysis, as each of these residues is only accessible in the open conformation (Fig. 4). In comparison, a limited sampling of residues located in the hinge regions (Ile-57) suggests that hinge residues may not be required for conformational rearrangement, as the I57A variant retained near wild-type activity. The structural rearrangement is instead likely driven by exposure of ␤ 3 loop residues like Trp-66 and Tyr-67. Reaffirming this hypothesis, substitution of the Trp-66 residue alone (W66A) or in concert with Tyr-67 (W66A/Y67A) nearly ablated catalytic activity with nearly 4 orders of magnitude difference in catalytic activity against substrate 2 compared with the wild-type FTT258 protein (Fig. 2B). Conversely, Y67A FTT258 showed only a single order of magnitude decrease in activity, indicating that this residue is not critical for catalysis (Fig. 2B). Thus, the movement and hydrophobic burial of Trp-66 likely controls the structural rearrangement of the substrate binding pocket observed in FTT258. This structural movement then regulates the catalytic activity of FTT258 by controlling exposure of the substrate binding pocket and makes structural rearrangement necessary for the full catalytic activity of FTT258.
Structural Rearrangement and Membrane Binding-In addition to catalytic activity, the conformational change and exposure of the ␤ 3 loop was hypothesized to be directly involved in membrane binding. To test this hypothesis, a liposome binding assay was used to determine if FTT258 could bind to artificial membranes, where liposome binding was assayed based on the separation of membrane-bound and -soluble fractions upon centrifugation ( Fig. 6) (52). The FTT258 protein was prevalent in the pellet fraction when assayed with liposomes designed to mimic bacterial membranes (70% phosphatidylethanolamine, 20% phosphatidylglycerol, and 5% cardiolipin), indicating that FTT258 binds to the bacterial liposomes in vitro (Fig. 6). The presence of some FTT258 in the soluble fraction suggests that not all of the FTT258 protein was tightly bound to the membrane. Interestingly, the relative percentage of soluble and pellet bound fraction is near 50%, matching with the relative ratio of open and closed conformers in the FTT258 structure (Fig. 6). As an important control, FTT258 protein was found almost exclusively in the soluble fraction when liposomes were absent (Fig. 6), showing that the presence of the mem-brane and not the state of protein aggregation determines the fractionation pattern of FTT258. To determine if the Trp-66 or Tyr-67 residues control membrane association, the single and double alanine-substituted variants were tested for liposome binding. Binding of all three FTT258 mutants (W66A, Y67A, or W66A/Y67A) to liposomes was indistinguishable from the wildtype protein (Fig. 6), indicating that other residues on the ␤ 3 loop or throughout the protein are more important in membrane binding. Thus, FTT258 is capable of associating with the plasma membrane, but the hydrophobic residues in the conformationally flexible ␤ 3 loop do not directly control membrane attachment.
To determine whether the membrane binding observed in FTT258 is conserved in human acyl protein thioesterases, hAPT1 was tested for liposome binding against mammalian specific lipids (Fig. 6). hAPT1 showed significant binding to the mammalian membranes, with all of the protein found in the membrane pellet (Fig. 6). Thus, although the mechanism of membrane binding is currently unknown, FTT258 and hAPT1 both show similar levels of membrane attachment. The divergence in substrate specificity and catalytic activity between hAPT1 and FTT258 suggests broader cellular substrates and biological functions in FTT258, but similar membrane binding between hAPT1 and FTT258 indicates conserved subcellular localizations for activity.

DISCUSSION
Serine hydrolases are broadly split into two different subclasses, esterases and lipases, based on substrate specificity and structural characteristics (8,12). Both esterases and lipases have the same active site residues, catalytic mechanism, and overall structural fold (␣/␤-hydrolase fold), but esterases recognize short substrates (Ͻ10 carbons), and lipases hydrolyze ester bonds in longer substrates (Ͼ10 carbons). To maintain substrate specificity for longer carbon substrates, lipases contain an extra structural domain known as a lid that obscures the active site and binding pocket in the closed form but undergoes large structural rearrangements to expose the hydrophobic binding pocket upon interaction with lipid-water interfaces (29). This interfacial activation at lipid-water boundaries has been proposed to be either thermodynamically driven by the exposure of greater hydrophobic surface at the lipid interface or to be a response to substrate, co-enzyme, or inhibitor binding (29,59).
Based on substrate specificity and the absence of a lid domain, FTT258 is closer to an esterase than a lipase. However, similar to lipases, the structure of FTT258 displayed two distinct conformations based on the large scale movement of the flexible ␤ 2 and ␤ 3 loop (Figs. 3 and 4). Unlike previous structures of lipase conformations, both forms of FTT258 were present in one asymmetric unit where four subunits (A-D) adopted the closed form and four subunits (E-H) adopted the open form (29, 58 -60). One subunit of FTT258 also contained electron density for a cyclobutanone ligand that was required for crystallization and was covalently bound to the catalytic serine residue (Fig. 4, C and D). This diversity of structural conformers is rare in protein crystal structures and, to the best of our knowledge, previously unobserved in serine hydrolase structures.
Similar to lid rearrangements in lipases, the structural rearrangement in FTT258 exposes the hydrophobic binding pocket, as Asp-25 moves away from the narrow hydrophobic binding pocket. Yet, unlike lipases, the nucleophilic serine and oxyanion hole in FTT258 are solvent-exposed before loop rearrangement (29,58). Instead, the flexible movement of the ␤ 2 and ␤ 3 loops on FTT258 mirrors a flexible hydrophobic loop identified on various bacterial lipases, whose movement was independent of lid movement and was proposed to serve as a membrane anchor (59). Similar to this loop on bacterial lipases, the exposed loop on FTT258 is highly hydrophobic, is unlikely to be fully solvent exposed within the cytoplasm, and is highly conserved across FTT258 homologues from mammals to bacteria ( Fig. 3B and Table 2). In the crystal structure of hAPT1, the ␤3 loop occupies an identical position to the closed conformation of FTT258 (28), whereas structures of bacterial homologues from Pseudomonas placed the loop closer to the open conformation of FTT258 (26,27). Thus, the conformational flexibility in the ␤ 2 and ␤ 3 loops is likely conserved across homologues and serves a conserved role in their functions.
The cause of the differential loop shift within the closed and open conformations of FTT258 is unclear. Binding of the cyclobutanone ligand is a potential explanation, but only one of the four subunits of FTT258 crystallized in the open conformation contains a bound cyclobutane ligand. As the cyclobutanone ligand binds only weakly to FTT258 and was present in a concentration (1 mM) around its IC 50 value (1.12 mM), the lack of ligand molecules present in the other open conformation subunits might represent the binding and then eventual release of the inhibitors during or after crystallization. Conversely, the different conformations of the loop could reflect the inherent flexibility of the loop, as computational calculations of similar flexible bacterial loops suggested significant movement in aqueous and organic conditions (59). Structures of lipases in open conformations have also been obtained without inhibitors or substrates present (58,60), and the variation in hydrophobic environment has been proposed to stimulate their interfacial activation through conformational rearrangement of the flexi-ble loops (29,58,59). The exposure of the hydrophobic ␤ 2 and ␤ 3 loops in FTT258 and their association with a nearby hydrophobic pocket in a related FTT258 subunit likely activates FTT258 and represents a biological interaction between the loops and a protein binding partner or cellular membrane.
The connection between structural rearrangement, catalytic activity, and exposure of a potential membrane binding loop suggests a potential link between the subcellular localization and catalytic activity of FTT258. The structural rearrangement could localize the hydrolase activity of FTT258 near bacterial lipoproteins and serve to control the broad substrate specificity of FTT258. We investigated this hypothesis by examining the role of the structural rearrangement and key residues undergoing conformational change on the catalytic activity and membrane binding of FTT258. The effect of structural rearrangement on the catalytic activity originates largely from the emergence of a new hydrophobic binding pocket upon loop rearrangement, as the positioning of the catalytic triad (Ser, His, Asp) remains largely unchanged (Figs. 3 and 5). The hydrophobic pocket is lined by a variety of conserved hydrophobic amino acids whose side chains are indispensable to the enzymatic activity of FTT258 (Fig. 2B). Similar to models of Cpa, FTT258 can also accommodate a variety of hydrophobic substrates, especially straight-chain lipid chains of varying lengths (26). The formation of the hydrophobic binding pocket upon structural rearrangement is largely driven by the burial of Trp-66 within a nearby subunit of FTT258. The Trp-66 residue is required for catalysis ( Fig. 2B) but is distantly located from the catalytic site in the closed conformation, making direct involvement in the hydrolysis mechanism impossible (Figs. [3][4][5]. The importance of structural rearrangement for full catalytic activity is reinforced by the essentiality of buried hydrophobic residues to the catalytic activity of FTT258 (Fig. 2B), as each of these residues is only fully solvent-exposed upon transition to the open form (Fig. 4). Thus, the importance of Trp-66 to the catalytic activity of FTT258 suggests that structural rearrangement stabilized by burial of Trp-66 serves to control the catalytic activity of FTT258.
Although burial of Trp-66 and structural rearrangement is required for full catalytic activity, their role in bacterial membrane binding is less well defined. FTT258 does localize with bacterial liposomes, but the degree of localization is independent of the presence of Trp-66, Tyr-67, or both residues combined, making burial of these two hydrophobic amino acids unlikely to control membrane association (Fig. 6). Instead, these residues are likely involved in substrate recognition, binding, or interactions with other proteins once FTT258 is bound to the membrane. Further biophysical characterization of FTT258 in the presence of lipid substrates and identification of specific regions of FTT258 responsible for membrane association through systematic alanine scanning should clarify the absolute roles of these residues in FTT258 activity.
The membrane association observed for FTT258 is also conserved in the human homologue, hAPT1, where the cellular substrates and cellular compartmentalization are known (15,61). A conserved conformational change for hAPT1 would link the biological role of hAPT1 in recycling proteins such as Ras from the plasma membrane with the cellular location for this role (16). hAPT1 has a conserved tryptophan and a conservatively substituted phenylalanine for tyrosine in its ␤ 3 loop (Table 2). hAPT1 and FTT258, however, diverged in their substrate specificities, as hAPT1 had significantly decreased enzymatic activity and narrower substrate specificity compared with FTT258, Cpa, and Cpf, especially toward diverse ester substrates (Fig. 2). hAPT1 may undergo a similar conformational change to FTT258, but the importance of the structural rearrangement to the catalytic activity of FTT258 and the divergence in activity between FTT258 and hAPT1 will require further investigation.
Conclusions-The observed structural rearrangement of the conserved hydrophobic ␤ 2 and ␤ 3 loops in FTT258 and its affect on catalytic activity and membrane binding provide new hypotheses of how this conserved family of serine hydrolases could potentially control their biological function. Given the confirmed necessity of acyl-protein thioesterases like hAPT1 to depalmitoylate membrane bound proteins (15,18), the structural rearrangement of FTT258 suggests connected regulation of substrate binding and membrane binding or protein-protein interactions. Whether similar protein acylation cycles exist in bacteria and whether structural rearrangement in acyl-protein thioesterases directly controls acylation remain unanswered questions, but the current study provides a link between substrate binding, structural rearrangement of the active site and hydrophobic ␤ 2 and ␤ 3 loops, and membrane attachment. With the important biological roles of acyl-protein thioesterases and the validated pharmacological effects of inhibiting hAPT1, the structural rearrangement of FTT258 could provide novel avenues for targeting inhibitors of bacterial and human homologues for the treatment of tularemia and human cancer.