Biochemical characterization of a palmitoyl acyltransferase activity that palmitoylates myristoylated proteins.

Dynamic regulation of signal transduction by reversible palmitoylation-depalmitoylation cycles has been recently described. However, further understanding of fatty acylation reactions has been hampered by our lack of knowledge about the specific transferases and thioesterases involved. Here, we describe an assay for the palmitoyl acyltransferase (PAT) that palmitoylates “myrGlyCys” containing members of the Src family of protein tyrosine kinases (PTKs). Since N-myristoylation of Fyn PTK, a member of the Src family, has been shown to be a prerequisite for palmitoylation, a new single plasmid vector that allows overexpression of myristoylated Fyn substrate in Escherichia coli was developed. Purified myristoylated protein substrates were incubated with iodopalmitoyl CoA, a palmitoyl CoA analog, in the presence of bovine brain lysates. Transfer of radiolabel to the Fyn substrate was detected by SDS-polyacrylamide gel electrophoresis and autoradiography. This assay was used to partially purify and characterize PAT activity from bovine brain. Here, we demonstrate that PAT is a membrane-bound enzyme, which palmitoylates myristoylated Fyn substrates containing a cysteine residue in position three. The PAT activity attached palmitate to Fyn proteins via a thio-ester linkage and exhibited a fatty acyl CoA preference for long chain fatty acids. It is likely that palmitoylation of Fyn and other Src family members by PAT regulates PTK localization and signaling functions.

Modification of proteins by fatty acids greatly alters their structure, function, and subcellular localization and has recently been shown to be involved in several aspects of cellular signaling (see Refs. 1 and 2 for recent reviews). More than 200 proteins are known to be fatty acylated, including viral and cellular proteins (1)(2)(3). Roles for protein fatty acylation range from anchoring proteins to membranes (3,4) and stabilizing protein-protein interactions (5) to regulating enzymatic activities in mitochondria (6). In general, mutations that prevent fatty acylation abolish or greatly alter the biological function of these proteins.
Protein fatty acylation can be divided into two categories: myristoylation and palmitoylation. N-Myristoylation involves the cotranslational attachment of the 14-carbon fatty acid myristate onto an N-terminal glycine residue of a protein via an amide linkage. Due to the high stability of this amide bond, myristoylation is irreversible, with some exceptions (7). The enzymology of N-myristoylation has been well characterized (Ref. 8 and references therein). A methionyl aminopeptidase first removes the initiator methionine. N-Myristoyl transferase (NMT) 1 then catalyzes the transfer of myristate to the glycine residue in position two. This glycine residue is an essential element of the substrate recognition sequence, since its substitution by any other amino acid within a protein prevents myristoylation.
Many proteins contain the 16-carbon fatty acid palmitate attached to specific cysteine residues. In contrast to myristoylation, palmitoylation occurs post-translationally and is readily reversible. The enzymology of palmitoylation is not well characterized, and a palmitoyl acyltransferase (PAT) has not yet been isolated. No apparent consensus sequence has been found at the palmitoylation site, suggesting the presence of more than one type of PAT. However, the N termini of 7 of 9 members of the Src family of PTKs and several of the ␣ subunits of the heterotrimeric G proteins contain the sequence myrGlyCys, where Cys-3 is palmitoylated (2,3). In most of these cases, myristoylation has been shown to be a prerequisite for palmitoylation to occur (9,10).
Many signal-transducing proteins translocate reversibly between plasma membrane and cytosol (1,2). Several have been shown to be reversibly palmitoylated, such as ␣ subunits of G proteins and nitric oxide synthase. These proteins undergo agonist-stimulated palmitate turnover (11,12), suggesting that dynamic palmitoylation of proteins can regulate signal transduction.
Little is known about the enzymes that specifically remove or transfer palmitate onto signal-transducing proteins. Recently, a palmitoyl thioesterase, which deacylates Ras proteins and ␣ subunits of G proteins in vitro, has been purified and the corresponding cDNA cloned (13,14). Upon further analysis, this palmitoyl thioesterase was shown to be a secreted protein.
Ras proteins and G ␣ subunits, which are located on the cytoplasmic layer of the plasma membrane, are therefore unlikely to be physiologic substrates for a secreted palmitoyl thioesterase.
Progress toward the identification of a PAT has been limited 1 The abbreviations used are: NMT, N-myristoyl transferase; FynSH432His 6 , Fyn protein-tyrosine kinase truncated after the SH2 domain to which a hexahistidine tag was appended; IC16, iodohexadecanoic acid (a palmitate analog); myr, myristate; pal, palmitate; PAT, palmitoyl acyltransferase; PTK, protein-tyrosine kinase; SH4, SH3, and SH2, Src homology domains 4, 3, and 2, respectively; PAGE, polyacrylamide gel electrophoresis; PCR, polymerase chain reaction; WT, wild type. by lack of a rapid, sensitive, and specific assay. Here, we report development of such an assay for the PAT that palmitoylates the Src-related PTKs containing the myrGlyCys mini-consensus sequence. Since prior myristoylation of these proteins has been shown to be a prerequisite for palmitoylation, a source of myristoylated protein to be used as substrate in the assay was required. An Escherichia coli overexpression vector for protein myristoylation was used to produce large amounts of myristoylated truncated Fyn PTK proteins. To circumvent long exposure times required to detect incorporation of tritiated palmitate into proteins, a [ 125 I]iodopalmitoyl CoA analog was utilized (15,16). Analysis of the products of the PAT assay by SDS-PAGE allowed visualization of incorporated iodopalmitate in as little as 30 min using phosphorimager technology. This assay was used to partially purify and characterize a PAT activity from bovine brain that palmitoylates myristoylated proteins.  (15)(16)(17). Nickel chelating agarose and Q-Sepharose Fast Flow were from Qiagen and Pharmacia Biotech Inc., respectively. RPS-F TLC plates used for fatty acid analysis were from Analtech.

Reagents-Fatty
Bacterial Strains, Plasmids, and Cloning-E. coli strain DH5␣ was originally purchased from Life Technologies, Inc. Plasmid pET19b and E. coli strain BL21DE3 were from Novagen. Human NMT cDNA and human p59 fyn cDNA were from laboratory stocks (18,19). DNA-modifying and restriction enzymes were purchased from New England Biolabs including the alkaline phosphatase, T4 DNA ligase, and DeepVent thermostable DNA polymerase used for all PCR. DNA manipulation and E. coli transformations were performed as described in Ref. 20. DNA sequencing reactions were carried out using the Sequenase Kit from Amersham as described by the supplier.
Construction of an Artificial Operon-An artificial operon was constructed to allow efficient expression and myristoylation of a variety of FynSH432His 6 proteins on a single plasmid. This engineered vector takes advantage of the polycistronic nature of prokaryotic mRNA and uses the strong T7lac promoter to drive the synthesis of a single mRNA, allowing tandem expression of a truncated version of the Fyn protein tyrosine kinase and human NMT. pETFyn432hNMT was constructed stepwise; the final construct is shown in Fig. 1A. First, hNMT cDNA was subcloned into pET19b as a NdeI-BamHI fragment to yield the pET19bhNMT plasmid. Second, an intercistronic region including a polyhistidine tag, a ribosome binding site (Shine-Dalgarno sequence), and a stop codon was appended after the SH2 domain of the Fyn PTK (after Cys-246) by a single step PCR reaction using appropriate oligonucleotides. The intercistronic sequence was based on the sequence described by Shoner et al. (21) and was utilized to allow the ribosome to start translating the second cistron encoding hNMT. The FynSH432His 6 SD DNA fragment was gel purified (22) and fused to hNMT cDNA by PCR using the splicing by overlap extension methodology (23,24). The FynSH432His 6 SDhNMT DNA fragment was gel purified, digested with NcoI, and subcloned into the digested and dephosphorylated pET19bhNMT plasmid. After restriction analysis, clones exhibiting the proper orientation were sequenced. The resulting plasmid was named pETFynSH432hNMT. Mutations in the N-terminal SH4 region were engineered by PCR using a mutagenic sense primer and an antisense primer at the SacI site. SH4 mutant pieces of DNA were fused to the T7lac promoter by splicing by overlap extension. A 122-base pair BglII-NcoI fragment of the pET19b vector containing the T7lac promoter was used in the splicing by overlap extension reaction with appropriate oligonucleotides. The T7lacSH4 fragment was gel purified and subcloned as a BglII-SacI fragment into the pETFyn432hNMT vector. Fyn SH4 mutants were sequenced, and no mutations other than those engineered were found.
Overexpression and Purification of FynSH432His 6 Proteins-E. coli BL21DE3 transformed with the appropriate pETFyn432hNMT plasmid were grown at 37°C in 50 ml of LB-broth containing 200 g/ml ampicillin until A 600 ϭ 0.6. Induction of FynSH432His 6 protein overexpression was carried out by the addition of 1.0 mM isopropyl-1-thio-␤-Dgalactopyranoside. After 3 h, bacteria were harvested by centrifugation and lysed in 10 ml of buffer A (8 M urea in 100 mM KH 2 PO 4 , 10 mM Tris-HCl, pH 8.0) at room temperature. The suspension was sonicated three times for 1 min, with a 1-min interval on ice. The suspension was centrifuged at 15,000 ϫ g for 20 min at 25°C to remove cellular debris. All following chromatographic steps were carried out at room temperature and were based on the protocol described by Hochuli et al. (25). The supernatant was applied to a 1.0-ml Ni-NTA-agarose equilibrated in buffer A. The column was washed with 5.0 ml of buffer A and 3.0 ml of buffer A pH 6.3. The FynSH432His 6 protein was eluted with buffer A pH 5.7. The A 280 of each fraction was monitored to localize proteincontaining fractions. Typically, 1.0 -1.5 mg of highly purified FynSH432His 6 was obtained using this procedure, which is equivalent to 20 -30 mg/liter of culture. The purified proteins were stored in aliquots at Ϫ80°C. FynSH432His 6 protein substrates diluted in reaction buffer remained soluble, and a 100,000 ϫ g centrifugation for 1 h did not pellet the proteins (data not shown). Attempts to remove urea by dialysis, gel filtration, or ultrafiltration resulted in low recovery of the myristoylated protein.
[ 3 H]Myristate Labeling of FynSH432His 6 Proteins in Vivo-25 Ci of [ 3 H]myristate in ethanol was aliquoted into sterile glass culture tubes, and ethanol was removed using a nitrogen stream. 1.0-ml aliquots of isopropyl-1-thio-␤-D-galactopyranoside-induced E. coli cultures were transferred into the myristate-containing tubes and grown for 3 h at 37°C. Bacteria were harvested, and the resulting pellets were resuspended in 1 ϫ sample buffer containing 100 mM dithiothreitol. The lysate was incubated at 100°C for 2 min and centrifuged at 12,000 ϫ g for 5 min prior to SDS-PAGE analysis. To visualize [ 3 H]myristate incorporation, the 12.5% polyacrylamide gel was stained with Coomassie R-250, destained, soaked in water for 30 min, and soaked in 1 M sodium salicylate for 30 min. The dried salicylated gel was subjected to fluorography at Ϫ80°C.
Bovine Brain Fractionation-Bovine brains were obtained fresh from a local slaughterhouse and kept on ice until delivered. Upon arrival, the main blood vessels and the remaining part of the spinal cord were removed. All of the following steps were performed at 4°C unless otherwise stated. The cortex and cerebellum were diced into 1.0-cm pieces, washed, and frozen as described (26) in a cryo-buffer made of 10% Me 2 SO in H buffer (210 mM mannitol, 70 mM sucrose, 10 mM Tris-HCl, pH 8.5, 1 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride, 1.0 g/ml leupeptin, 1.0 g/ml aprotinin, 1.0 g/ml pepstatin, and 5.0 g/ml trypsin inhibitor. The frozen suspension (66% brain (w/v)) was kept at Ϫ80°C. Brain pieces stored in this fashion maintained their PAT activity for at least 6 months. 120 g of frozen suspension (equivalent to 80 g bovine brain) was crushed with a mortar and pestle and thawed in 500 ml of cold H buffer. The brain tissue was homogenized twice for 30 s in a Waring blender with a 5-min rest interval. The homogenate (H) was centrifuged at 10,000 ϫ g for 15 min to remove the cellular debris and the mitochondria/nuclei containing fraction (P10). The supernatant was carefully transferred to a fresh container, and the pellet was reextracted with 500 ml of H buffer as described above.
The second supernatant was pooled with the first, and the resulting suspension (1000 ml) was centrifuged at 100,000 ϫ g for 45 min. The supernatant (S100) was decanted, and the pelleted membrane fraction (P100) was resuspended in 100 ml of buffer H with five up and down cycles on a motor-driven Potter-Elvehjem homogenizer. The suspension was adjusted to 100 mM Na 2 CO 3 at pH 11.0 and stirred for 30 min. Membranes were pelleted by centrifugation at 100,000 ϫ g for 45 min. The carbonate-washed membranes were resuspended in 50 ml of buffer H using the motor-driven homogenizer as described above, and the pH of suspension was adjusted to 8.5.
The membrane suspension was adjusted to 25% (v/v) glycerol, and membranes were solubilized with 1% (w/v) Triton X-100 at a detergent to protein ratio (w/w) of 3. The suspension was stirred for 30 min and then centrifuged as above for 45 min. The supernatant was transferred to new tubes and kept frozen at Ϫ80°C until needed. Subcellular fractions were assayed for activity as described below.
Q-Sepharose Chromatography-The solubilized PAT was diluted 5-fold in buffer Q (20 mM Tris-HCl, pH 8.5, 25% glycerol, 0.5 mM dithiothreitol, 0.1% Triton X-100) to reduce the Triton X-100 concentration to approximately 0.2%. The proteins (275 mg) were loaded at 2.0 ml/min on a 5.0 ϫ 8.0-cm Q-Sepharose Fast Flow XK50/20 column pre-equilibrated with buffer Q. The column was washed with two bed volumes of buffer Q and eluted with an increasing salt gradient. PAT activity reproducibly eluted between 175-300 mM NaCl.
Palmitoyl Acyltransferase Assays-Typically, 2.5 g of protein samples containing PAT activity were incubated with 1.0 g of purified FynSH432His 6 and 100 nCi of [ 125 I]iodopalmitoyl-CoA analog in 50 l for 30 min at 25°C. The reaction was buffered with 100 mM imidazole-HCl, pH 7.0, and 0.3% Triton X-100 was added to keep PAT soluble. The FynSH432His 6 substrate was from a 1.0 ϫ 2.0 mg/ml stock in 8 M urea, pH 5.7. The products of the reaction were separated by SDS-PAGE on 12.5% polyacrylamide gel (27). Incorporation of the labeled palmitate analog into the 32-kDa FynSH432His 6 protein was visualized by phosphorimager analysis for 1 h or after a 36-h exposure by autoradiography on x-ray film.
To study substrate specificity, 2.5 g of partially purified PAT from the Q-Sepharose pool was assayed as above in the presence of 1.0 g of the purified FynSH432His 6 mutants. Alternatively, PAT activity was assayed with the WT FynSH432His 6 in the presence of myristoylated dodecapeptides (100 M) corresponding to the N termini of Src-related protein tyrosine kinases (19).
To study temperature sensitivity of PAT activity, 2.5 g of partially purified PAT (Q pool) were preincubated for 5 min at 25, 37, 45, 55, 75, and 100°C, chilled for 5 min on ice, and assayed for activity. The fatty acyl CoA specificity was investigated by assaying PAT activity in the presence of fatty acyl CoAs (10 M) of increasing chain length.
In Situ Hydrolysis and TLC Analysis of Fatty Acid Bound to FynSH432His 6 -PAT activity was assayed in the presence of WT and C3,6S FynSH432His 6 proteins. Polyacrylamide gels were soaked in neutral 1 M hydroxylamine or 1 M Tris-HCl, pH 7.0, for 16 h at 25°C. The gels were washed with water three times for 10 min, stained with Coomassie R250, destained, dried, and subjected to autoradiography. PAT activity was assayed with WT FynSH432His 6 as substrate. The wet gel was subjected to autoradiography to localize the labeled FynSH432His 6 . The appropriate gel region was excised, transferred to a glass tube, crushed with a glass rod, and hydrolyzed with 0.5 ml of 1.5 N NaOH for 24 h at 25°C. The hydrolysate was neutralized with 0.75 ml of 1 N HCl and extracted twice with 1 ml of chloroform. The pooled chloroform fractions were dried under a nitrogen stream. The hydrolyzed 125 I-labeled iodofatty acid analog was redissolved in a small volume of acetone and applied to a RPS-F TLC plate next to an 125 Ilabeled IC16 standard. The plate was developed with a mobile phase of a 1:1.75:1.75 mixture of H 2 O:CH 3 CN:CH 3 COOH. After chromatography, the plate was dried and subjected to autoradiography.

Design and Construction of an Artificial Operon for Tandem Expression of Truncated Fyn Proteins and Human NMT on a
Single Plasmid-Previous work by this laboratory established that the Src family member Fyn is palmitoylated and that N-myristoylation is required for efficient palmitoylation in vivo (9). To study the palmitoylation reaction in vitro, an expression system was designed to generate a myristoylated Fyn substrate. Milligram amounts of myristoylated protein can be produced in E. coli provided that the NMT is coexpressed with the target protein (17). Initial experiments, in which Fyn and NMT were maintained on two separate plasmids, resulted in low yields of myristoylated Fyn product. 2 We therefore constructed an artificial operon with four important characteristics. First, the genes for Fyn and NMT were placed within a single plasmid under control of the T7lac promoter (Fig. 1A). Second, human NMT cDNA was used instead of NMT from the yeast Saccharomyces cerevisiae. Yeast NMT exhibits a high K m for peptides containing a cysteine at position six (28) and was inefficient at myristoylating Fyn. Third, a truncated version of Fyn was employed, which terminated after the SH2 domain. We found that the expression level of full-length Fyn in E. coli was very low and that removal of the C-terminal catalytic domain enhanced Fyn expression dramatically. Finally, a hexahistidine tag was appended to the C-terminal end of the truncated Fyn to facilitate purification.
To ensure that the palmitoylation assay was specific for the N-terminal region of the FynSH432His 6 protein, a series of mutations was engineered at the Fyn N terminus. As shown in Fig. 1B, cysteine residues were substituted for serine residues individually (C3S and C6S) and together (C3,6S). In addition, the myristoylation signal was abolished by substituting the essential glycine residue at position two with an alanine residue (G2A). A SrcFyn chimera containing the first 10 amino acids of Src replacing the corresponding Fyn N-terminal region was also engineered as a negative control, since the Src terminus is known not to be palmitoylated (9). All the mutations and the engineered stretch of DNA were fully sequenced, and no mutations other than those engineered were found in the clones utilized in this study.
Overexpression, Myristoylation, and Purification of Recombinant FynSH432His 6 Proteins-Overexpression of myristoylated FynSH432His 6 proteins was achieved using the various pETFyn432hNMT plasmids in E. coli. The appended polyhistidine tag allowed purification of the different FynSH432His 6 proteins to apparent homogeneity by nickel-chelating chromatography ( Fig. 2A). Upon lysozyme/Triton X-100 lysis, overexpressed myristoylated FynSH432His 6 proteins were found exclusively in the insoluble fraction (membranes and bacterial debris), and the FynSH432His 6 proteins were only partially solubilized using detergent or high salt conditions (data not shown). FynSH432His 6 proteins were therefore solubilized in 8 M urea, purified under denaturing conditions, and diluted in reaction buffer. Urea at low concentrations did not interfere with the palmitoylation assay. Upon induction with isopropyl-1-thio-␤-D-galactopyranoside, all the FynSH432His 6 proteins readily incorporated the [ 3 H] myristate label with the exception of the G2A mutant, which lacks the myristoylation signal (Fig. 2B). All the myristoylated FynSH432His 6 constructs were overexpressed to equivalent levels and were myristoylated to a similar extent.
Development of a PAT Assay and Partial Purification of PAT-The substrates for the PAT assay consisted of highly purified FynSH432His 6 proteins and [ 125 I]iodopalmitoyl CoA, a palmitoyl CoA analog previously described by our laboratory (15). The use of the [ 125 I]iodopalmitate analog allows rapid detection of PAT activity, requiring as little as 30 -60 min of exposure when using phosphorimager technology. Incubation of myristoylated FynSH432His 6 , [ 125 I]iodopalmitoyl, CoA and crude homogenates of various mouse organs resulted in incorporation of radiolabel into FynSH432His 6 protein. The amount of label incorporated was linear up to 30 min at 25°C, and the assay exhibited specificity with respect to protein and fatty acyl CoA substrates (see below). Brain homogenates had the most PAT activity (data not shown).
To scale up the purification of PAT and to determine its intracellular localization, a bovine brain homogenate was fractionated into membrane-bound and soluble compartments. Membrane fractions contained the most PAT activity, while the cytosolic fraction was devoid of detectable activity (Fig. 3). The 100,000 ϫ g pellet (P100) was washed with carbonate (pH 11) to remove peripheral membrane proteins. PAT activity was clearly detected in the carbonate-washed membranes and could readily be extracted with detergent (Fig. 3). Solubilization of PAT from the low speed, P10 membrane fraction was inefficient.
Solubilized PAT was fractionated on a Q-Sepharose Fast Flow column at pH 8.5 and eluted between 175 and 300 mM NaCl (Fig. 4). This PAT pool was used in all subsequent assays. The PAT activity was very labile, thereby complicating its purification. Indeed, enzymatic activity could not be measured after storage for more than 10 days at 4°C, and the enzyme did not tolerate multiple freeze/thaw cycles. 3 Substrate Sequence Requirements-The substrate specificity of the Q-Sepharose-purified PAT activity was investigated using the six FynSH432His 6 proteins. When equal amounts of PAT and FynSH432His 6 proteins were incubated in the presence of the [ 125 I]iodopalmitoyl CoA analog, the labeled iodopalmitate was readily transferred to the WT and C6S mutant FynSH432His 6 substrates (Fig. 5A). The C3S and C3,6S mutant substrates were not palmitoylated, nor was the SrcFyn  3. Subcellular fractionation of bovine brain PAT and partial purification of the solubilized PAT. Different subcellular fractions (2.5 g) were prepared and assayed for PAT activity as described under "Materials and Methods." The position of FynSH432His 6 substrates is indicated by the arrow. PAT assays were carried out with WT (odd-numbered lanes) and C3,6S (even-numbered lanes) FynSH432His 6 proteins as substrates. Products of the reactions were separated by SDS-PAGE, and [ 125 I]iodopalmitate analog incorporation into FynSH432His 6 proteins was visualized by autoradiography. H, homogenate; P10, 10,000 ϫ g pellet; P100, 100,000 ϫ g pellet; S100, 100,000 ϫ g supernatant; C.W.P100, carbonate-washed membranes; Sol. PAT, 1.0% Triton X-100-solubilized carbonate-washed P100 membranes; and Q pool, Q-Sepharose pool (see Fig. 4 for chromatographic details). Lanes 15 and 16, substrate alone controls (no enzyme added). Exposure time was 36 h. chimera, which lacks cysteine residues in its N-terminal region. Two additional chimeric constructs, YesFyn and G ␣i0 Fyn, containing the N-terminal sequences (10 amino acids) of the Yes PTK and G ␣io subunit, were also expressed, purified, and assayed. Both substrates contained the sequence myrGlyCys and were shown to be palmitoylated (data not shown). These results strongly suggest that the cysteine residue in position three is required for palmitoylation and is likely the major palmitoylated cysteine residue in vitro.
The G2A Fyn mutant substrate, which is not myristoylated, was not palmitoylated in vitro, despite the presence of cysteines at positions three and six. This result suggests that myristoylation is a prerequisite for palmitoylation. When PAT was assayed in the presence of myristoylated dodecapeptides corresponding to the N termini of Src-related PTKs (Fig. 5B), only the peptides containing the N-terminal sequence myrGly-Cys inhibited the reaction. Indeed, the PAT activity was inhibited greater than 95% by 100 M myrFyn and myrLck, while myrLyn and myrYes peptides inhibited the reaction by 84 and 69%, respectively. Interestingly, peptides containing two cysteines (myrFyn and myrLck) were better inhibitors. No inhibition by the myrSrc peptide (which has no cysteine residue in its sequence) nor by the non-myristoylated Yes peptide was observed. These observations imply that proteins containing the myrGlyCys motif are substrates for palmitoylation by PAT.
Other investigators have reported that non-enzymatic palmitoylation of proteins and peptides can occur at physiological pH (29,30). However, under our assay conditions, the protein substrate (WT) alone and enzyme alone controls did not reveal any significant incorporation of the labeled palmitate analog in the region corresponding to FynSH432His 6 substrate (Figs. 3 and 5A and 5B). In addition, PAT activity was heat labile, as preincubation of the PAT preparation at 45°C for 5 min completely prevented palmitoylation (Fig. 5C). These results indicate that PAT activity is enzymatic.
PAT Is Inhibited by Long Chain Fatty Acyl CoAs-The fatty acyl CoA requirements of PAT were investigated by assaying palmitoylation in the presence of fatty acyl CoAs of increasing chain length (Fig. 6A). Palmitoyl CoA was the best competitor of the PAT activity, inhibiting the palmitoylation reaction by 70%. The myristoyl and stearoyl CoA derivatives also inhibited the reaction by 30 and 55%, respectively. Shorter fatty acyl CoAs did not significantly inhibit the reaction.
Iodopalmitate Analog Is Linked via a Thioester Bond-The chemical nature of the bond between the iodopalmitate analog and the protein substrate was investigated by soaking the polyacrylamide gel containing palmitoylated Fyn substrate in neutral hydroxylamine for 16 h. Thioester bonds are readily hydrolyzed by neutral hydroxylamine or by alkali. As depicted in Fig. 6B, neutral hydroxylamine treatment completely removed the bound radioactive iodopalmitate, thereby confirming the thioester nature of the chemical bond between the palmitate and the FynSH432His 6 protein.
In a parallel experiment, the palmitoylated substrate was excised from the gel and hydrolyzed with alkali. The hydrolysate was neutralized, chloroform extracted, and analyzed by thin layer chromatography. As shown in Fig. 6C, the mobility of the sample derived from the hydrolysate was identical to that of the [ 125 I]iodopalmitate analog standard, thereby identifying the fatty acid bound to the FynSH432His 6 substrate as the iodopalmitate analog. In lane 3 of the same chromatogram, the hydrolysate sample and iodopalmitate standard were spotted together, yielding a single spot of increased intensity after development of the TLC. This latter result further confirmed the identity of the iodopalmitate analog as the transferred fatty acid.

DISCUSSION
In this manuscript, we report a rapid and sensitive assay for the PAT that palmitoylates myrGlyCys containing proteins of Src-related PTKs and its use in the purification and characterization of a PAT activity from bovine brain. The assay employs a high energy radiolabeled [ 125 I]iodopalmitoyl CoA analog and a variety of highly purified myristoylated FynSH432His 6 protein substrates. Detection of PAT activity was accomplished with less than 1 h of exposure of dried SDS-PAGE gels on a phosphorimager screen using 2.5 g of protein or less. This  (19). WT substrate alone and enzyme alone were used as negative controls (last two lanes). C, the PAT preparation was incubated for 5 min at the indicated temperature, assayed as described under "Materials and Methods" with WT FynSH432His 6 substrate, and analyzed by SDS-PAGE and autoradiography. Exposure time, 36 h. represents a significant improvement over previous palmitoylation assays, which required a minimum of 48 h of exposure using tritiated palmitate and 50 g of PAT extract to visualize palmitate incorporation into viral glycoproteins (31).
A series of myristoylated Fyn substrates was overexpressed in E. coli with a single plasmid myristoylation system and used to delineate the substrate specificity of PAT. All FynSH432His 6 substrates were shown to be expressed and myristoylated to equivalent levels (except G2A, which lacks the myristoylation signal) (Fig. 2). N-Myristoylation was shown to be essential for palmitoylation in vitro and is likely to be part of the recognition signal for PAT within the Fyn sequence. Two lines of evidence support this conclusion: 1) the G2A mutant FynSH432His 6 substrate, which is not myristoylated, is also not palmitoylated (Fig. 5A) and 2) only the myristoylated form of the Yes dodecapeptide was an inhibitor of the PAT reaction (Fig. 5B). These results confirm previous observations made in vivo, suggesting that myristoylation of the MetGlyCys sequence of Src-related PTKs and G ␣ subunits was a prerequisite for palmitoylation (9,10). It is not known whether other lipid moieties can substitute for myristate.
Recently, Degtyarev et al. (32) reported that membrane association, rather than myristoylation, was required for palmitoylation of G ␣i1 in vivo. Likewise, it is important to note that G ␣s , which is not myristoylated, is still palmitoylated at Cys-3 (33,34). Although our in vitro assays are performed in cell-free lysates, in the absence of membranes, the assay buffer does contain 0.3% Triton X-100, a detergent concentration above the critical micelle concentration. The presence of a myristate moiety could enhance the ability of the Fyn substrate to productively interact with PAT in a detergent micelle. Alternatively, Fyn and G ␣ proteins may be palmitoylated by different PATs in vivo with different substrate specificities.
Using the cell-free assay and the various FynSH432His 6 substrates, we delineated the substrate requirements of the partially purified PAT (Q-Sepharose pool). The presence of Cys-3 was required for palmitoylation of Fyn in vitro, since the C3S substitution within the FynSH432His 6 substrate completely abolished palmitoylation (Fig. 5). Likewise, the SrcFyn chimera, which lacks Cys-3, was not palmitoylated, and myrSrc peptide failed to inhibit the PAT activity. It is not known whether sequences containing cysteine at positions 4 or 5 can serve as PAT substrates. Mutation of Cys-6 within Fyn did not inhibit palmitoylation in vitro. In contrast, C6S mutants of p59 fyn expressed in vivo exhibit reduced levels of palmitoylation (9). It is possible that additional PATs are responsible for palmitoylating Cys-6 in vivo or that the absence of Cys-6 in the C6S mutant enhances palmitoylation of Cys-3 in vitro. Alternatively, differences observed in palmitoylation of C6S mutants may reflect the use of monkey cells (COS-1) versus bovine brain.
The PAT activity also transferred iodopalmitate onto two other chimeric substrates: YesFynSH432His 6 and a G ␣i0 Fyn-SH432His 6 . 4 Since no homology could be found among the N termini (10-amino acid sequence) of Fyn, Yes, and G ␣i0 other than that encoding for the myristoylation signal and the presence of the cysteine in position three, the myrGlyCys sequence is likely to be necessary and sufficient to direct palmitoylation.
The PAT activity exhibited selectivity for long chain fatty acyl CoAs, as demonstrated by the inhibitory properties of the various acyl CoAs in the assay (Fig. 6). Only fatty acyl CoAs containing 14 carbons or more were significant inhibitors of the PAT, palmitoyl CoA being the best inhibitor and the likely natural substrate of the enzyme. Whether PAT activity can 4  transfer shorter or longer fatty acids onto proteins in vivo remains to be tested. The transfer of [ 125 I]iodopalmitate to FynSH432His 6 occurred via the formation of an alkali and hydroxylamine-sensitive linkage, indicative of a thioester bond. Taken together, these data support the conclusion that PAT catalyzes attachment of palmitate to Cys-3 in the Src family members.
The PAT activity was shown to be membrane bound upon subcellular fractionation of bovine brain (Fig. 3), consistent with the membrane localization of Src-related PTKs and that of G ␣ proteins. Other PAT activities have also been found in membrane fractions, e.g. the PAT that palmitoylates viral glycoproteins and H-Ras (31,35). Whether these PAT activities correspond to identical or different enzymatic activities is not known at the present. The identity of the particular membrane subtype containing the PAT activity is also not known. PAT and its substrates could potentially interact at specific membrane sites. Upon palmitate transfer, the palmitoylated protein would remain stably anchored in the membrane. Identification of membranes enriched for PAT might shed light on how these myrGlyCys-containing proteins are targeted to specific intracellular membranes.
PAT activity could be efficiently solubilized from cellular membranes and fractionated on a Q-Sepharose Fast Flow column (Figs. 3 and 4). However, recovery of the PAT activity from the Q-Sepharose column was low (about 25%) and the increase in specific activity modest (typically 3-4-fold at best). Similar results were obtained with various chromatographic media. These observations, combined with the fact that PAT activity is very labile, have complicated further purification.
In conclusion, dynamic palmitoylation of proteins is a novel and exciting addition to the repertoire of cellular control mechanisms. To further understand this new mechanism, purification, and characterization of the PAT and palmitoyl thioesterase enzymes is imperative. To attain these goals, we designed and utilized a sensitive assay to partially purify and characterize a PAT activity that palmitoylates members of the Src family of PTKs. Once this PAT activity is purified to homogeneity, its encoding cDNA will be cloned, thereby allowing characterization at the molecular level.