Reengineering the specificity of a serine active-site enzyme. Two active-site mutations convert a hydrolase to a transferase.

Two residues are known to play important catalytic roles in fatty acyl-thioester hydrolase, thioesterase II: Ser-101, the site of a covalent acyl-enzyme intermediate, and His-237 which is within hydrogen bonding distance of Ser-101 and facilitates catalysis by increasing the nucleophilicity of this residue. In this study we have examined the effect of mutations at these two residues on the ability of the enzyme to function as a hydrolase and, in the presence of a thiol acceptor, as an acyltransferase. In the hydrolase reaction kcat values for the wild-type, H237R, S101C, and S101C, H237R thioesterase enzymes were 0.11, < 0.002, 0.10, and < 0.002 s-1, respectively, and at steady state, the proportion of each enzyme present as the covalent acyl-enzyme intermediate was 11, 91, 71, and 100%, respectively. In the acyltransferase reaction no activity could be detected for the wild-type or H237R enzymes but the specific activities of the S101C and S101C/H237R thioesterases were 170 and 1300 nmol/min/mg of protein, respectively. From this data we conclude the following: the wild-type enzyme functions exclusively as a hydrolase. The H237R mutant acts ineffectively as a hydrolase primarily because the deacylation reaction is drastically retarded. The S101C enzyme functions well as a hydrolase, even though the rate of deacylation is adversely affected, and this enzyme can also perform as an acyltransferase. Mutation of both catalytic residues leads to a complete loss of hydrolase activity and the S101C,H237R mutant functions as an effective acyltransferase exhibiting kcat values higher then those of the wild-type enzyme acting as a hydrolase. This study reveals that, with only two amino acid replacements, an enzyme capable of functioning exclusively as a hydrolase can be converted into an equally active enzyme performing solely as an acyltransferase.

Thioesterases have been implicated as important enzymes in a variety of metabolic pathways including the biosynthesis of fatty acids (1-51, polyketides (6, 7), and peptide antibiotics (8,9). Although only the thioesterases involved in fatty acid synthesis have been studied extensively, it appears likely that the role of thioesterases associated with all of these pathways is to release the fatty acid, polyketide or peptide, from the phosphopantetheine which provides the thiol template for assembly of the product. * This work was supported in part by National Institutes of Health Grants DK-16073 and RR-06505 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Currently 2 residues have been identified as playing a catalytic role in the thioesterases: Ser-101 (numbering for rat mammary gland thioesterase 11) located in the serine esterase motif Gly-Xaa-Ser-Xaa-Gly (3,(10)(11)(12)(13) and His-237 found within Gly-Xaa-His motif (13)(14)(15)(16)(17). Replacement of the active-site histidine of rat mammary gland thioesterase I1 with arginine, leucine (X), or alanine (13) reduced the catalytic activity by 2-3 orders of magnitude. Substitution of the active-site serine with cysteine in rat thioesterase I (16) and rat mammary gland thioesterase I1 (13,171 reduced activity only slightly. In this paper we analyze in detail the effects of mutations at the active-site seryl and histidyl residues of thioesterase I1 and show that a double mutation, S101C,H237R converts this hydrolase into an excellent acyltransferase. EXPERIMENTAL PROCEDURES Materials-Lysyl endoproteinase from Achromobacter lyticus was bought from Wako BioProducts (Dallas, T X ) and [l-14Clpalmitoyl-CoA (54 Ci/mol) was obtained from Amersham Corp. Other chemicals were purchased from Sigma and Aldrich.
Mutant Construction and Expression-The cDNAs coding for the single mutants of thioesterase 11, SlOlC, and H237R, cloned into a modified pJLA502 vector (15,17) were used for construction of a cDNA encoding the double mutant. Standard recombinant DNA techniques (18) were utilized if not otherwise indicated. The H237R and SlOlC single mutant cDNAs were restricted with BamHI and EcoRI enzymes and the 5.2-kilobase pair fragment from the SlOlC mutant cDNA, containing the vector sequence and the mutated SlOlC codon, was ligated to the 0.93-kilobase pair fragment from the H237R mutant cDNA, which contained the mutated H237R codon. The resulting plasmid containing the double mutation, SlOlC and H237R, was cloned into Escherichia coli DH5a cells and expression of the encoded protein was induced at 42 "C in a 2-1' culture of TB medium (24 gA yeast extract, 12 gA tryptone, 16.43 gA K2HP04, 2.3 g/l KH2P04, 1 mM glycerol, 50 mgA carbenicillin). The enzyme was purified as described previously (15,19). A final purification step was added that involved chromatography on a high performance anion-exchange column (TSK-DEAEXPW, 2.15 x 15 cm, 10 p m , Bio-Rad) using a NaCl gradient (04.15 M over 27 min) in 50 m~ Tris-HC1, pH 7.811 m~ EDTA at a flow rate of 7 d m i n . All chromatographic and storage buffers contained dithiothreitol (1 or 2 m~) unless otherwise stated.
Assay of Mutant Actiuity-Thioesterase activity of the mutant was determined spectrophotometrically with acyl-CoA(17) andp-NP-Dec (3) as the substrates. The use of p-NP-Dec as substrate was particularly convenient for assay of activities of thioesterase I1 mutants that contained a cysteinyl residue at the active site, position 101; these mutants are rapidly inactivated by 5,5'-dithiobis(2-nitrobenzoate) and cannot be assayed by the usual procedure that allows direct reaction, with 5.5'dithiobis(2-nitrobenzoate), of the CoA thiol released following formation of the covalent acyl-enzyme intermediate. Prior to assay dithiothreitol was removed from all enzyme preparations by gel filtration. Specific activities were calculated using the amount of unoxidized form of the enzyme (i.e. Cys-101 present exclusively in the free thiol form) that was estimated from the absorbance profile on HPLC (see below). All data were corrected for non-enzymatic release of p-nitrophenol. The kinetic canoate; 2-ME, 2-mercaptoethanol; HPLC, high performance liquid The abbreviations used are: 1, liter(s); p-NP-Dec, p-nitrophenyl dechromatography; ESIMS, electrospray ionization mass spectrometry.
parameters are presented as an average from five different methods of calculation: Lineweaver-Burk, Eadie-Hofstee, Hanes-Woolf, direct linear plot, and non-linear regression.
Reverse-phase HPLC-Proteins, substrates, and products were separated on a Vydac C4 reverse-phase HPLC column (5 pm, 300-A pore size, 0.46 x 25 e m ) using acetonitrile gradients generated from 0.1% trifluoroacetic acid in water (solvent A) and 0.1% trifluoroacetic acid in acetonitrile (solvent B) according to the following programs: system I: 30% solvent B for 5 m i n , followed by a three-step linear gradient (30-42% B over 5 mi n, 4246% B over 28 min, and 56-75% B over 2 min); system 11: five-step linear gradient (10-141 B over 6 min, 14-27% B over 13 min, 2735% B over 16 min, 35-658 B over 30 min, and 6590% B over 10 min); system 111: as the system I except that the first step lasted for 10 min. The flow rate was 0.7 d m i n for all separation systems.
On-line Reverse-phase HPLCIESZMS-Separation of proteins, substrates, and products of the acyltransferase reaction was performed on a microbore HPLC system (Michrom Biosources, Pleasanton, CA) equipped with a reverse-phase C4 column (5 p, 300-A pore size, 0.1 x 15 c m ) using a flow rate of 40 flmin at 40 "C. The gradient systems were similar to those described above. The flow was split directly in the UV cell allowing 4 flmin to be delivered to the mass spectrophotometer through a modified VG BioQ probe.z A sheath liquid (0.2% trifluoroacetic acid in methanol) was added at a flow rate of 2 flmin prior to nebulization. ESIMS was performed on a VG BioQ quadropole mass spectrometer (VG BiotechlFisons, Altrincham, United Kingdom). The instrument was controlled and data analyzed using a LabBasa software (VG Biotech/Fisons, Altrincham, U.K.). A mixture of heart horse myoglobin and PEG 550 was used to calibrate the mass scale within the 300-1,300-Da range.

RESULTS AND DISCUSSION
General Properties of the SlOlC,H237R Thioesterase II-Approximately 30 mg of the double mutant was isolated from the 2-1 culture, and its authenticity was confirmed by ESIMS. The experimentally determined value for the average molecular mass (29,508.8 2 3.2 Da) was in good agreement with that predicted from the primary sequence with a N H 2 terminally unblocked enzyme (29,506.2 Da) that is consistent with our earlier finding that the recombinant wild-type thioesterase I1 (19) and H237R, H237L (151, and SlOlC mutants (15,17) accumulated in E. coli cytosol as deformylated proteins. The double mutant, unlike the recombinant wild-type thioesterase 11, when stored without dithiothreitol, was slowly converted to a +32.0 Da form (Fig. lA). A similar +32.0-Da form was produced on ageing of the single mutant SlOlC thioesterase I1 in the absence of dithiothreitol. We conclude that these +32.0 Da-forms resulted from the oxidation of the Cys-101 thiol to the sulfinic acid for the following reasons: the +32.0-Da protein was not formed from the wild-type or the H237R enzymes upon ageing; the presence of 2 m~ dithiothreitol on storage almost entirely prevented formation of the +32.0-Da forms; the +32.0-Da species did not form covalent acyl-enzyme intermediates (see below) and the increase in the amount of the +32.0-Da form was accompanied by a lowering of the specific activity of the enzyme (details not shown).
On incubation of the double mutant with p-NP-Dec, a brief presteady state burst of p-nitrophenol release was observed and was followed by a very slow liberation ofp-nitrophenol. The slow release ofp-nitrophenol from 20 wp-NP-Dec measured at pH 7, 8.2, and 9.2 was 0.5, 1.8, and 3.4 nmol/(min mg), respectively, that compares to 160 and 150 nmoY(min mg) determined for the wild-type thioesterase I1 and SlOlC mutant, respectively, at pH 8 (optimal pH). These data are in agreement with a small hydrolase activity (~0.3%) that was reported for the S101C,H237A thioesterase I1 mutant (13). Evidently the replacement of His-237 in the SlOlC thioesterase 11, as well as in the wild-type enzyme (13, 151, significantly reduces hydrolase * E Bitsch and C. H. L. Shackleton, unpublished data. The average molecular masses were estimated by on-line HPLCESIMS and the value for unmodified S101C,H237R mutant and differences between a particular form and the unmodified enzyme are shown. The expected average molecular mass for the S101C,H237R mutant is activity confirming an important catalytic role for this histidyl residue. Acylation of the Double Mutant-Examination of the products of the reaction of p-NP-Dec or decanoyl-CoA with S101C,H237R thioesterase I1 (Fig. lZ3) revealed that the double mutant was entirely converted to the +154.3-Da form. The increase in the molecular mass was in an excellent agreement with the value expected for the covalent binding of a single decanoyl residue (+154.4 Da). Similar results were observed when palmitoyl-CoA (+242.0 Da versus expected +239. 4 Da) was used as substrate. The extent of acylation varied less than 5% over the pH range 6.6-9.2 and reached a maximum in the presence of only a slight excess of substrate. For example, 100% of the double mutant was converted to the +154.3-Da form in the presence of only a 1.5-fold molar excess of decanoyl-CoA. These data indicate that whereas formation of the acyl-enzyme intermediate proceeds freely with the double mutant, subsequent hydrolysis is seriously impaired leading to accumulation of the acyl-enzyme species. In contrast, the oxidized form of the double mutant (putative cysteine sulfinic acid derivative) did not generate a +154.3-Da species on incubation with p-NP-Dec indicating that this form cannot generate the covalent acylenzyme intermediate.
The acyl moieties were removed from the double mutant acyl-enzyme by treatment with hydroxylamine in 8 M urea at pH 7 indicating that they were attached to the enzyme via thioester linkage (data not shown). To verify that Cys-101 is the acylated residue, the double mutant was labeled with [14C]pal-  (8 nmol) was treated with a 20 nmol of [14C]palmitoyl-CoA, and the radioactive acyl-enzyme was separated from the remaining [14C]palmitoyl-CoA and the +32-Da form by reverse-phase HPLC. The [14Clacyl-enzyme was digested with lysyl endopeptidase (lO:l, final ratio) in 4 M urea buffered with 0.1 M potassium phosphate, pH 7, and 10 m~ EDTA at 30 "C for 16 h (ll8 of the protease was added at 2-h intervals), and separated on reverse-phase HPLC using solvent system 11. The average molecular mass of the eluted radioactive peptides was determined in parallel by on-line HPLCl ESIMS. The (Table I). Four radioactive peptides were found, and three of them were products of partial digestion. All radioactive peptides contained Cys-101, the cysteine that replaced the active-site serine, while none contained any of the other 3 cysteinyl residues present on thioesterase I1 (positions 18, 31, and 256) confirming Cys-101 as the site of acylation. Similar results were obtained when the double mutant was acylated with decanoyl-CoA (data not shown). The incomplete digestion of the acylated double mutant enzyme was surprising in view of the fact that cleavage of all lysyl peptide bonds was observed in 4 M urea after only 8 h digestion of the carboxymethylated enzyme (no fatty acyl present). We hypothesize that the presence of a n acyl-chain covalently bound to Cys-101 makes Lys-93, Lys-115, and, to some extent, Lys-117 unavailable for cleavage by the endopeptidase. This protection from protease attack could result from interaction of the acylchain with a hydrophobic region of the thioesterase that extends in this peptide from residues 84-89,97-98, and 105-112 (hydrophilicity index 5 -1 in the Kyte-Doolittle scale, window 7). Additional support for localization of the fatty acyl hydrophobic binding site comes from comparison of the rat thioesterase I1 sequence with two structurally related thioesterases, one having a similar substrate specificity and the other having a different substrate specificity, viz a thioester hydrolase involved in medium chain fatty acid synthesis in the duck and one involved in peptide synthesis in Bacillus brevis, encoded by the grsT gene. The grsT protein is involved in the biosynthesis of the cyclic peptide antibiotic gramicidin S which is assembled by a non-ribosomal thiotemplate mechanism; the substrate for the grsT enzyme is believed to be the gramicidin S peptide bound by a thioester linkage to a protein-phosphopantetheine (8,(20)(21)(22). When the sequence of the rat thioesterase is separately compared with each of these thioesterases using a mutation index plot (23), it is evident from the coincident peaks in the two plots, that the same regions have accumulated mutations over most of the sequence (Fig. 2A). These regions most likely represent sequences that are unimportant either for catalysis or for determination of substrate specificity. On the other hand, the plot clearly reveals that only two regions ofthe proteins have diverged significantly between the rat and grsT thioesterases but not between the rat and duck thioesterases: one in the carboxyl-terminal 20 residues, the other between residues 70-95 (Fig. 2% ). We suggest that these two regions are important in determining substrate specificity. Indeed, there is strong experimental evidence indicating that the carboxyl-ter-  2. Comparison of the frequency of non-conservative mutations between thioesterases with different specificity. Rat thioesterase I1 was aligned with duck thioesterase I1 and the grsT protein using the ALIGN program (Protein Identification Resources, Washington, D. C.), and the mutation index was calculated for each panel A. The mutation index is a measure of the frequency of accepted alignment using a 20-amino-acid window (23), smoothed, and shown in non-conservative mutations. Panel B presents the smoothed arithmetical difference between mutation indices for the grsT protein uersus rat thioesterase I1 and rat thioesterase I1 uersw duck thioesterase 11. The residue numbers given are those of the rat enzyme. minal region of rat thioesterase I1 is involved in the interaction between rat thioesterase I1 and the fatty acid synthase that bears the fatty acyl-phosphopantetheine thioester, the target for hydrolysis (13,24,25). The second section of sequence divergence, from residue 70 to 95, includes a predominantly hydrophobic region (residues 84-89) and much of the polypeptide chain that was protected from lysyl endopeptidase digestion by the presence of the fatty acyl chain. We suggest, therefore, that this is the region of thioesterase I1 that encodes the fatty acylchain-binding site.
Deacylation by 2-ME and Other Thiols-Studies on stability of the acyl-enzyme intermediate formed by the double mutant thioesterase I1 revealed that in the presence of certain thiol reagents the acyl chain was rapidly removed from Cys-101. Incubation of the double mutant with p-NP-Dec and 2-ME resulted in a rapid and steady release of p-nitrophenol. The reaction was pH dependent (Fig. 3) and involvement of a moiety with a pK, above 8, perhaps a thiol, was indicated by the pWactivity profile. This thiol could be the one associated with the enzyme active-site cysteine and/or that associated with the 2-ME substrate.
Nucleophilic substitution at decanoyl-Cys-101 thioester by 2-ME would form decanoic acid (2-hydroxy)mercaptoethyl ester releasing free enzyme thiol available for reacylation. To determine whether the double mutant was functioning as a transferase, we identified the products of the reaction. When p-NP-Dec was used as acyl donor and CoA as acceptor, decanoyl-CoA was identified, from its characteristic retention time by HPLC (10 min using solvent system 111) and from its molecular mass, 922. of the catalyzed reaction between 2-ME and p-NP-Dec was established as decanoic acid (2-hydroxy)mercaptoethyl ester by comparison of HPLC retention time with the value determined for the synthesized standard (both 24 min using solvent system 111). Identification of the reaction products confirmed our conclusion that the double mutation of thioesterase I1 converted the acyl-thioester hydrolase into an acyltransferase.
Considering the available data, the reaction catalyzed by the double mutant can be described as two successive reactions with no formation of a ternary complex.
The observation that the enzyme can be fully acylated with a minimal excess of p-NP-Dec indicates that the last reaction of the first equation is functionally irreversible. In this case the apparent dissociation constant for the binding of p-NP-Dec to the enzyme is almost zero and the steady-state equation for the reaction involving two substrate reduces to (26):  Table 11. The kcat values calculated for the two half-reactions are not exactly identical. However, since under experimental conditions full saturation is not attained, the difference between these values seems to be acceptable. The turnover number for the acyltransferase reaction catalyzed by the double mutant is more than 10.0-and 2-fold higher than kcat values calculated for hydrolysis of p-NP-Dec and decanoyl-CoA, respectively, by wild-type thioesterase I1 (17). These results demonstrate that the mutations of SlOlC and H237R in thioesterase I1 converted the acyl-thioester hydrolase into a highly effective acyltransferase.
The K, values calculated for 2-mercaptoethanol, dithiothreitol, cysteamine, and GOA are very similar (Table II), indicating that there is no preferential binding of any of these acceptors. However, all effective acceptors contain the HS-CH2-CH-moiety that may be recognized by the enzyme.

Kinetic parameters for the reaction between p-NP-Dec and different acceptors catalyzed by SlOlC,H237R thioesterase II
Reaction were carried out in 0.2 ml of 50 rn borate buffer, pH 9.2, at 30 "C. The enzyme and substrate were at the following concentrations: (a) donor kinetics: 0.04 enzyme, 30 m 2-ME and p-NP-Dec up to 6 p a ; (b) acceptor kinetics: 0.21 pa enzyme, 20  Butanethiol and pantetheine were also active as acceptors; pante- a Calculated from HPLC absorbance profile. Substrate (70-110 p a ) concentration was at least 10-fold over the enzyme concentration. Up to 25% of substrate was hydrolyzed at the moment of injection. m 2-ME, and 0.1-0.9 pa enzyme (2.6 pa for H237R mutant) at 30 "C The transfer reaction was carried out with 80-100 pap-NP-Dec, 20 for several min and immediately injected onto the reverse-phase HPLC column. Solvent system 111 was used and absorbance monitored at 232 nm. The amount of the synthesized product was calculated from the peak area using authentic standards. No decanoic acid (2-hydroxy)merp-NP-Dec, 20 nm 2-ME, 10 min reaction).
captoethyl ester product was found in the absence of the enzyme (80 pa

5-1
% nmol 1 (mg min) e Data taken from Ref. 17. Comparison of the Wild-type and Mutant Thioesterases ZZ-The properties of the wild-type thioesterase and its double and single mutants were compared according to several criteria: their ability to catalyze the hydrolysis of p-NP-Dec and acyl-CoA esters, the levels of acyl-enzyme formed on incubation with these substrates, and their ability to catalyze acyltransfer between p-NP-Dec and 2-ME (Table 111). A low fraction of acyl-enzyme intermediate, lo%, and no transfer product were found in the case of the wild-type thioesterase I1 indicating that acylation is the rate-limiting step and hydrolysis of the acyl-enzyme occurs rapidly. The role of the His-237 in activating the active-site residue Ser-101 has been substantiated by our demonstration that the 2 residues are within hydrogen-bonding distance of each other (17) and by the finding that mutations at His-237 result in dramatically reduced hydrolase activity (13,15). It seems likely that His-237 may also assist the deacylation reaction by acting as a general base, facilitating the nucleophilic substitution by water on the acylenzyme intermediate as it is generally believed to occur for the serine proteases. In fact, for the H237R mutant, 90% of the enzyme was found to be acylated in the presence of p-NP-Dec suggesting that the decreased hydrolase activity of the mutant results mainly from impaired hydrolysis. The SlOlC mutant can catalyze both the hydrolase and acyltransferase reactions and in this case too, the acyl-enzyme intermediate accumulated to relatively high levels (Table 111). The acyltransferase activity is a novel property of the enzyme acquired by the mutation SlOlC and presumably results from in-

Reengineering a Serine Active-site Enzyme 383
creased nucleophilicity of the acylated amino acid, Cys versus Ser (17), and from reduced enzyme hydrolytic activity. Taking into account that hydrolysis of thioesters (acyl-Cys) requires lower activation energy than hydrolysis of esters (acyl-Ser) and that the water concentration was 55.5 M and the 2-ME only 20 nm, it seems likely that water positioning must be partially obstructed in the SlOlC mutant. Finally, in the case of the double mutant (SlOlC,H237R) replacement of His-237 with arginine leads to the almost complete elimination of hydrolase activity, leaving an enzyme that functions exclusively as an acyltransferase.

Speculation on the Evolutionary Significance of the Simple Hydrolase to Acyltransferase
Conversions-We have previously pointed out (17) that structurally and functionally related serine active-site enzymes in the thioesterase family utilize both types of serine codons, TCN and AGY, that could have been interchanged via codons for cysteine (TGY), since substitution of cysteine for the active-site seryl residue generates functionally active enzymes from thioesterases using both types of codon (16,17). The present study reveals that a two-step transformation of hydrolase to acyltransferase can be made via a catalytically active intermediate. These findings raise the possibility that a primordial cysteine active-site enzyme (exemplified by the thioesterase I1 SlOlC mutant described herein) possessing both hydrolase and acyltransferase activity could have given rise to two families of enzymes, one having exclusively hydrolase activity (exemplified by the thioesterase I1 wild-type), the other exclusively acyltransferase activity (exemplified by the thioesterase I1 S101C,H237R mutant). This evolutionary transition could have occurred via single nucleotide changes, one resulting in replacement of the active-site cysteinyl with a seryl residue forming the hydrolase family, the other resulting in replacement of the activesite histidine forming the acyltransferase family. Although we are unaware of the existence of a naturally occurring acyltransferase that is structurally related to thioesterase 11, one of the genes of the bialophos peptide antibiotic gene cluster encodes a thioesterase 11-like enzyme with putative active-site cysteinyl and histidyl residues (9), analogous to the SlOlC mutant described in our study. The exact catalytic role of this enzyme, be it hydrolase or transferase, has yet to be ascertained.
use of the Mass Spectrometry Facility at the Children's Hospital Oak-