Endothelial Nitric Oxide Synthase N-TERMINAL MYRISTOYLATION DETERMINES SUBCELLULAR LOCALIZATION*

2 mM (3-mercaptoethanol, plus 2.5 pg/ml each leupeptin, chymostatin, aprotinin, and soybean trypsin inhibitor). The cellular homogenates were immunoprecipitated with a 1:lOO dilution of immune sera either in the absence or in the presence of an excess (10 pg/ml) of the immunogen peptide. In preliminary experiments, this dilution of the antiserum was found sufficient to precipitate all the NO synthase present in BAEC or transfected cells. The immunoprecipitated samples were incubated for 1 h with protein A-Sepharose (Pharmacia LKB Biotechnology Inc.), washed extensively with Buffer 1, and eluted by heating with Laemmli sample buffer. The samples were analyzed by SDS-poly- acrylamide gel electrophoresis on 7% gels which were then fixed, treated with En3Hance (Du Pont-New England Nuclear), and exposed for 1-3 days to x-ray film at -70 ‘C using an intensifying screen. Preimmune serum precipitated no specifically labeled proteins even after prolonged exposure of the autoradiograms (data not shown). To prepare cytosol and particulate fractions, the cell homogenate was centrifuged at 100,000 X g for 1 h; the particulate fraction was further treated by resuspension in buffer 1 containing 1 M KC1 (plus 10% glycerol) and then centrifuged at 100,000 X g for 30 min. The pellet was then resuspended in Buffer 1 containing 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, and immunoprecipitation of the cy- tosol and particulate fractions was carried out as above. cells transfected with wild-type (WT) and myristoylation-defective (MYR-) cDNA clones for EC-NOS. For each cell type, the proportion of the total NO synthase activity present in the particulate (solid bars) versus soluble (hatched bars) fraction is shown. These data represent means and standard errors of duplicate observations from an experiment that was replicated twice in different preparations with equivalent results. In an independent experiment, also performed in duplicate, the particulate enzyme activity, as a percent of the total, was 86 f 5% for BAEC, 89 f 7% for wild-type, and 5 f 5% for the myr- mutant EC-NOS-transfected cells. In these experiments, no NO synthase activity could be demonstrated in sham- transfected COS-7 cells. In the experiment shown, the NO synthase specific activities obtained in BAEC, wild-type EC-NOS, and myris- toylation- mutant were calculated at 4.3, 2.6, and 3.3 pmol of citrul-line/min/mg of protein, respectively. Following the resolution of the cellular homogenate into particulate and cytosol fractions, the recov-ery of the total enzyme activity initially present in the unfractionated homogenates averaged 70% in three separate experiments.

Nitric oxide synthases in diverse mammalian tissues catalyze the oxidation of L-arginine to L-citrulline plus nitric oxide (NO). In the vascular endothelium, synthesis of NO yields a labile intercellular messenger molecule with potent biological activities, including vascular smooth muscle relaxation. We have recently documented that the endothelial cell NO synthase (EC-NOS) constitutes a genetically distinct tissue-specific enzyme isoform. In further contrast to the soluble NO synthases found in neural tissues and in macrophages, the endothelial enzyme is associated primarily with the particulate fraction. Analysis of molecular clones for the endothelial NO synthase reveals no obvious transmembrane-spanning region, but a consensus motif for Nterminal myristoylation was identified; such a consensus sequence is not evident in the primary sequence of the soluble macrophage and neural NO synthases. We performed oligonucleotide-directed mutagenesis of the myristoylation consensus sequence in the endothelial NO synthase cDNA, and studied the pattern of expression of the wild-type and mutant EC-NOS cDNAs in transient transfection experiments in COS-7 cells. The subcellular localization of heterologous endothelial NO synthase was determined using analyses of enzyme activity as well as immunoprecipitation of biosynthetically labeled NO synthase with a highly specific antipeptide antibody. Expression of the wild-type endothelial NO synthase cDNA in COS-7 cells results in targeting of both enzyme activity and NO synthase immunoreactivity primarily to the particulate fraction. By contrast, transient expression of the myristoylation-mutant cDNA in COS-7 cells yields NO synthase enzyme activity and immunoreactivity associated exclusively with the cytosol fraction. Following biosynthetic labeling with ['Hlmyristate, the NO synthase can be specifically immunoprecipitated from the particulate fraction in endothelial and in COS-7 cells transfected with the wild-type cDNA, but not in cells trans-* This work was supported in part by funds from the American Heart Association and from the National Heart, Lung and Blood Institute (to T. M.) 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 U.S.C. Section 1734 solely to indicate this fact.
$ Recipient of a post-doctoral fellowship from Consejo Nacional de Investigaciones Cientificas y Tbcnicas (Argentina).
Recipient of a Clinician Scientist Award from the American Heart Association. To whom correspondence and reprint requests should be addressed: Cardiovascular Div., Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115. Tel.: 617-732-7376;Fax: 617-732-5132. fected with the myristoylation-mutant EC-NOS cDNA. N-terminal myristoylation of the endothelial NO synthase may provide a potential point of regulation of the biological functions of endothelium-derived NO in situ.
The nitric oxide synthases comprise a family of related genes and are importantly involved in such diverse processes as vascular signaling (endothelium), neurotransmission (brain), and cell-mediated cytotoxicity (macrophage). All tissue-specific NO synthase isoforms thus far characterized appear to share a common overall catalytic scheme involving the oxidation of L-arginine to L-citrulline plus nitric oxide (reviewed in Refs. 1-4). The cofactor requirements for NO synthesis by the various NO synthase isoforms are also strikingly similar in diverse tissues. NO synthase cDNAs from endothelium (5, 6), macrophage (7-9), and brain (lo), show significant sequence similarity, but these isoforms are the products of distinct genes (5). Analyses of the deduced amino acid sequence of the NO synthase cDNAs also reveal shared consensus sequences for the binding of several redox cofactors, as well as a shared homology with cytochrome P450 reductase in the C-terminal half of these molecules. Despite these common enzymological and structural features overall, there is significant sequence divergence at the N termini of the NO synthases. Indeed, the first -236 amino acid residues of the brain isoform (a larger protein) are not present in the endothelial or macrophage NO synthase sequences, and sequence similarities between the latter isoforms are not evident until after -70 residues. The sequence divergence at the N termini of the different NO synthase isoforms may indicate that this region helps to determine tissue-specific functional features of the enzyme family.
The localization of the endothelial NO synthase to the particulate subcellular fraction (11, 12) serves to further distinguish this tissue-specific isoform from the soluble NO synthase enzyme activities found in many other cell types (13-15). Our analysis of molecular clones for the endothelial NO synthase (EC-NOS)' revealed no obvious transmembrane-spanning region, but a consensus motif for N-terminal myristoylation was identified (5); such a consensus sequence is not evident in the primary sequence of the soluble macrophage and neural NO synthases. Myristoylated proteins are often membrane-associated, but myristoylation is neither necessary nor sufficient for membrane attachment (16). Using immunoelectron microscopic techniques, the endothelial NO synthase has been localized to the membrane fraction in bovine aortic endothelial cells.' In the present studies, we performed oligonucleotide-directed mutagenesis to inactivate the myristoylation consensus sequence in the endothelial NO synthase cDNA and then conducted transient transfection experiments to explore the subcellular distribution of the enzyme encoded by wild-type and myristoylation-deficient clones.
PCR-based Mutagenesis of the Myristoylatwn Consensus Sequence-Polymerase chain reaction (PCR)-based mutagenesis (17) of the myristoylation consensus sequence of EC-NOS was used to replace the glycine (GGC) at amino acid position 2 with an alanine residue (GCC) using the wild-type EC-NOS cDNA (5) as template .
The oligonucleotides were a 31-mer sense primer (5"CCATCGA-TATGGCCAACTTGAAGAGTGTGG-3') and a 20-mer antisense primer (5'-AGAGGCGTACAGGATGGTTG-3'). PCR was carried out using the EC-NOS cDNA as template for 30 cycles of denaturation at 94 "C (1 min), annealing at 60 "C (1 min), and elongation at 72 "C (2 min) using standard reaction conditions modified by the addition of 10% Me2S0 to the reaction mixture. The 1.5-kilobase PCR product was treated with the Klenow fragment of Escherichia coli DNA polymerase I and then with the restriction enzyme BglII to yield a 1430-base pair fragment with restriction sites compatible with the EcoRV/BglII-treated wild-type EC-NOS plasmid ( 5 ) , and the mutated fragment was cloned into this plasmid to replace the corresponding fragment of the wild-type cDNA and yield the myristoylation-(myr-) mutant EC-NOS. Nucleotide sequence analysis of the entire 1430-base pair PCR-generated fragment verified the correct wildtype sequence throughout save for the presence of the predicted mutation.
Transient Expression of Wild-type and Myr-Mutant EC-NOS cDNA-Both wild-type and myr-cDNAs were subcloned into the eukaryotic expression vector pcDNANeo (Invitrogen) using compatible XbaI plus Hind111 restriction sites. Subsequently, COS-7 cells were transfected using the DEAE-dextran method (17), as previously applied (5) with the wild-type or the myristoylation-mutant cDNA, or were sham-transfected with pcDNANeo without insert. Typically, the COS-7 cells were studied 72-96 h after transfection; the bovine aortic endothelial cells were isolated and cultured as previously described, and were studied between passage 6 and 10 in culture. Subcellular fractions were prepared (11) from BAEC or from transfected COS-7 cells, and NO synthase activity was assayed by the conversion of ~-[~H]arginine to ~-[~H]citrulline, as described (13). Immunoprecipitation of Endothelial Nitric Oxide Synthase-Polypeptide (PYNSSPRPEQHKSYK-C), deduced from the endothelial clonal antisera were raised directed against a synthetic 15-residue NO synthase cDNA sequence (Ref. 5; residues 599-614, plus Cterminal Cys to facilitate chemical coupling to carrier protein). The peptide was coupled to maleimide-activated keyhole limpet hemocyanin (Pierce Chemical Co.) and used to immunize rabbits, and immune sera were characterized in the immunoprecipitation of endothelial NO synthase from biosynthetically labeled BAEC or from transfected COS-7 cells. Biosynthetic labeling with [35S]methionine (1250 Ci/mmol, Du Pont-New England Nuclear) was for 3 h (100 pCi/ml in methionine-free RPMI medium; GIBCO); for [3H]myristic acid labeling was for 16 h with 400 pCi/ml [3H]myristic acid (49 Ci/ mmol, Amersham Corp.) in RPMI 1640 medium. Labeled cells were harvested by scraping and homogenized by sonication in buffer 1 (50 mM Tris-HC1, pH 7.4 , 0.1 mM EGTA, 2 mM (3-mercaptoethanol, plus 2.5 pg/ml each leupeptin, chymostatin, aprotinin, and soybean trypsin inhibitor). The cellular homogenates were immunoprecipitated with a 1:lOO dilution of immune sera either in the absence or in the presence of an excess (10 pg/ml) of the immunogen peptide. In preliminary experiments, this dilution of the antiserum was found sufficient to precipitate all the NO synthase present in BAEC or transfected cells. The immunoprecipitated samples were incubated for 1 h with protein A-Sepharose (Pharmacia LKB Biotechnology Inc.), washed extensively with Buffer 1, and eluted by heating with Laemmli sample buffer. The samples were analyzed by SDS-polyacrylamide gel electrophoresis on 7% gels which were then fixed, treated with En3Hance (Du Pont-New England Nuclear), and exposed for 1-3 days to x-ray film at -70 'C using an intensifying screen. Preimmune serum precipitated no specifically labeled proteins even after prolonged exposure of the autoradiograms (data not shown). To prepare cytosol and particulate fractions, the cell homogenate was centrifuged at 100,000 X g for 1 h; the particulate fraction was further treated by resuspension in buffer 1 containing 1 M KC1 (plus 10% glycerol) and then centrifuged at 100,000 X g for 30 min. The pellet was then resuspended in Buffer 1 containing 1% Triton X-100, 0.1% SDS, 1% sodium deoxycholate, and immunoprecipitation of the cytosol and particulate fractions was carried out as above.

RESULTS AND DISCUSSION
Using a PCR-based approach, the glycine residue at the +2-position obligatory for N-terminal myristoylation (16) was converted into an alanine residue; this amino acid is unable to serve as a myristate acceptor. We then cloned the wildtype (EC-NOS) and myristoylation-NO synthase cDNAs into a eukaryotic expression vector and transfected these constructs into COS-7 cells. In transient expression experiments, particulate and cytosolic fractions were prepared and assayed for NO synthase enzymatic activity, as shown in Fig.  1, by quantitating the conversion of ~-[~H]arginine to L-[~H] citrulline (13). In BAEC, NO synthase activity is localized primarily to the particulate fraction ( Fig. 1; see also Refs. 11 and 12). Similarly, when the wild-type endothelial NO synthase (EC-NOS) cDNA is expressed in COS-7 cells, the enzymatic activity is principally associated with the particulate fraction. By contrast, the NO synthase enzyme activity expressed by the transfected myristoylation-EC-NOS mutant is associated exclusively with the cytosolic fraction.
We next determined the subcellular localization of the endothelial NO synthase protein, using biosynthetic labeling, and developed methodology for specific immunoprecipitation of the enzyme using antiserum directed against a synthetic oligopeptide deduced from the EC-NOS sequence (Fig. 2). In endothelial cells biosynthetically labeled with [35S]methionine, immunoprecipitation with this antiserum yields a single protein band at the expected molecular mass of 135 kDa. In contrast, immunoprecipitation of EC-NOS from biosynthetically labeled transfected COS-7 cells shows two protein bands,

FIG. 1. Subcellular distribution of the endothelial NO synthase activity in endothelial cells and in transfected COS-7
cells. Shown in this figure is the assay of NO synthase enzyme activity measured in cytosol and particulate fractions prepared from BAEC and from COS-7 cells transfected with wild-type ( W T ) and myristoylation-defective (MYR-) cDNA clones for EC-NOS. For each cell type, the proportion of the total NO synthase activity present in the particulate (solid bars) versus soluble (hatched bars) fraction is shown. These data represent means and standard errors of duplicate observations from an experiment that was replicated twice in different preparations with equivalent results. In an independent experiment, also performed in duplicate, the particulate enzyme activity, as a percent of the total, was 86 f 5% for BAEC, 89 f 7% for wild-type, and 5 f 5% for the myr-mutant EC-NOS-transfected cells. In these experiments, no NO synthase activity could be demonstrated in shamtransfected COS-7 cells. In the experiment shown, the NO synthase specific activities obtained in BAEC, wild-type EC-NOS, and myristoylation-mutant were calculated at 4.3, 2.6, and 3.3 pmol of citrulline/min/mg of protein, respectively. Following the resolution of the cellular homogenate into particulate and cytosol fractions, the recovery of the total enzyme activity initially present in the unfractionated homogenates averaged 70% in three separate experiments.  I, 3 , 4 , and 6) or in the presence (lanes 2,5, and 7) of 10 pg of the immunogen peptide. The immunoprecipitated samples were incubated with protein A-Sepharose, washed extensively, and analyzed by SDS-polyacrylamide gel electrophoresis on 7% gels followed by autofluorography.
one a t 135 kDa and an unexpected band a t 150 kDa. This doublet was not found in untransfected or sham-transfected COS-7 cells but was found reproducibly for both the wildtype EC-NOS and myristoylation-mutant cDNA-transfected cells. The immunoprecipitation of both protein bands is completely blocked in the presence of an excess of the unlabeled immunogen peptide (Fig. 2). It thus appears that both of the protein bands seen in transfected COS-7 cells are the consequence of EC-NOS transfection, and both are specifically immunoprecipitated by the anti-peptide antibody. Among other possibilities, it may be that the higher M , protein reflects a post-translational modification of the transfected endothelial NO synthase, which is, for some reason, more evident when the cDNA is expressed in COS-7 cells, or perhaps there is some anomalous processing of the transfected EC-NOS transcript to yield the upper protein band. In any event, the arguments made herein apply equally well whichever band is being considered. Subcellular fractionations of biosynthetically labeled BAEC and of wild-type and myristoylation-EC-NOS-transfected COS-7 cells are shown in Fig. 3. Following biosynthetic labeling with [35S]methionine (Fig. 3A) or [3H]myristic acid (Fig.  3B), cytosol and particulate fractions were resolved by ultracentrifugation and immunoprecipitated as described above. Biosynthetic labeling with [35S]methionine reveals that the immunoprecipitated NO synthase is localized primarily to the particulate fraction in BAEC and in wild-type EC-NOStransfected COS-7 cells but is exclusively cytosolic in the myristoylation-mutant. The subcellular distribution of NO synthase in the [3H]myristate-labeled cells is consonant with this finding, namely that the [3H]myristate-labeled enzyme is exclusively associated with the particulate fraction in BAEC and in wild-type EC-NOS-transfected cells. In these studies, the chemical identity of the NO synthase acyl adduct as myristate remains presumptive, although exceedingly likely; a definitive determination requires verification by physicochemical analyses of the purified protein. Importantly, there is no incorporation of labeled myristate in the myristoylationmutant-transfected nor in sham-transfected COS-7 cells.
A role for N-terminal myristoylation in localization of the endothelial NO synthase to the particulate subcellular fraction has been proposed on the basis of recent molecular cloning (5) and cell biological considerations (18). The current study establishes, using biochemical and molecular biological approaches, that N-terminal myristoylation of the endothelial NO synthase occurs in uiuo and appears essential for the association of this enzyme with the cellular membrane.
It is plausible that the regulation of endothelial NO synthase myristoylation may affect the biological activity of this enzyme system. Indeed, diverse proteins importantly implicated in intracellular signaling have been shown to undergo N-terminal myristoylation. For example, the a subunits of several of the signal-transducing G proteins undergo N-terminal myristoylation, and this post-translational modification is essential for targeting to the plasma membrane (19, 20). Additionally, the membrane-associated pp60"" oncogene product is myristoylated; myristoylation-defective mutant pp60'" molecules remain cytosolic and are unable to promote cell transformation (21). An additional example is the MARCKS protein (myristoylated _alanine-zich C-kinase substrate), a protein that undergoes translocation from membrane to cytosol upon phosphorylation by protein kinase C (22-25). A common theme in these myristoylated proteins is that the regulation of this post-translational modification can provide a means for the dynamic modulation of subcellular localization and can lead to important changes in cell function. The proportion of EC-NOS present in the cytosol of BAEC or wild-type-transfected COS cells may reflect the dynamic equilibrium between subcellular fractions similar to that seen with other myristoylated proteins. In the present study, the identification of the specific particulate subcellular fraction to which the transfected endothelial NO synthase is targeted remains to be definitively determined. By analogy with the subcellular distribution as determined for bovine aortic endothelial cells, it seems most plausible that the recombinant wild-type enzyme associates with the cellular membrane.
The evident association of the myristoylated endothelial NO synthase with the cell membrane may have important implications for the biochemical transformations of the nitric oxide produced by the enzyme. The lability of free NO and its proclivities to form more stable chemical adducts (such as nitrosothiol compounds; cf Ref. 26) strongly suggest that the intracellular compartment in which NO is synthesized may importantly affect the identity and biological activity of the signaling molecule produced. Furthermore, high concentrations of NO are cytotoxic, and one may postulate that plasma membrane localization of the endothelial NO synthase could serve to facilitate extracellular transport of its product. Thus, the apparent membrane localization of the NO synthase may mitigate cytotoxic effects in the endothelial cell and could lead to enhanced NO signaling in the subjacent vascular smooth muscle cells and the intraluminal platelets. By contrast, the inducible macrophage NO synthase apparently exhibits a robust enzyme activity in the cell cytosol, but this occurs in a terminally activated inflammatory cell in which the autocidal effects of the NO produced may be of relatively little biological consequence. The soluble nature of the neural NO synthase is more challenging to explain in this context, but may indicate that alternative mechanisms for cellular transport and metabolism of NO may exist in different cell types. It also seems plausible that localization of the endothelial NO synthase to the plasma membrane could play a role in signal transduction, perhaps by coupling the enzyme's activation to cell surface receptors or mechanochemical signals. These observations further raise the possibility that the subcellular localization of the endothelial NO synthase may be dynamically regulated and could provide a potential point of control for the biological functions of endothelium-derived NO in situ.