Targeting of Endothelial Nitric-oxide Synthase to the Cytoplasmic Face of the Golgi Complex or Plasma Membrane Regulates Akt- Versus Calcium-dependent Mechanisms for Nitric Oxide Release*

The heterogeneous localization of endothelial nitric-oxide synthase (eNOS) on the Golgi complex versus the plasma membrane has made it difficult to dissect the regulation of each pool of enzyme. Here, we generated fusion proteins that specifically target the plasma membrane or cytoplasmic aspects of the Golgi complex and have assessed eNOS activation. Plasma membrane-tar-geted eNOS constructs were constitutively active, phosphorylated, and responsive to transmembrane calcium fluxes, yet were insensitive to further activation by Akt-mediated phosphorylation. In contrast, cis -Golgi com-plex-targeted eNOS behaved similarly to wild-type eNOS and was less sensitive to calcium-dependent activation and highly responsive to Akt-dependent phosphorylation compared with plasma membrane versions. In plasma membrane- and Golgi complex-targeted constructs, Ser 1179 is critical for NO production. This study provides clear evidence for functional roles of plasma membrane- and Golgi complex-localized eNOS and sup-ports the concept that proteins thought to be regulated and to function exclusively in the plasma membrane of cells can indeed signal and be regulated

Within the vascular endothelium, nitric oxide (NO) is generated via the enzyme endothelial nitric-oxide synthase (eNOS). 1 This highly labile and reactive gas plays a key role in regulating cardiovascular homeostasis, influencing systemic and pulmonary blood pressure, vascular remodeling, and angiogenesis. Given the important and pervasive action of endothelium-derived NO in the cardiovascular system, it is not surprising that the activity of eNOS is highly regulated in a temporal and spatial manner (1). eNOS was originally characterized as a calcium/calmodulin-dependent enzyme (2)(3)(4), and although this is the primary means by which eNOS activity is controlled, discrepancies between the levels of eNOS protein, the amount of NO released, calcium regulation of the enzyme, and the impaired function of mislocalized eNOS have led to the discovery of additional post-translational control mechanisms regulating eNOS activity. It is now generally accepted that eNOS activity is also influenced directly by regulated protein-protein interactions and phosphorylation and indirectly by fatty acylation, which targets the enzyme to specific intracellular domains (1,5,6). Activation of endothelial cells in response to vascular endothelial growth factor (7,8), sphingosine-1-phosphate (9,10), estrogen (11,12), insulin (13), and shear stress (14) results in the Akt-dependent phosphorylation of Ser 1179 on eNOS. Although there is increasing evidence that other phosphorylation sites (Ser 116 , Thr 497 , Ser 617 , and Ser 635 ) (15)(16)(17)(18) contribute to temporal regulation of eNOS function, mutation of Ser 1179 prevents both Akt-mediated phosphorylation and NO release, and dominant-negative Akt reduces endothelium-dependent vascular responses in vitro and in vivo (7,14,19), thus highlighting the importance of this regulatory site. The ability of both Akt-phosphorylated and S1179D eNOS (a constitutively active allele in which Ser 1179 is mutated to an aspartate reside) to release more NO in the absence of a stimulus has led to the proposal that phosphorylation of Ser 1179 mediates the calciumindependent activation of eNOS (7,14,20). eNOS is modified by post-translational myristoylation of Gly 2 and palmitoylation of Cys 15 and Cys 26 (21)(22)(23). These N-terminal fatty acid modifications target eNOS to the Golgi complex and to cholesterol-enriched domains of the plasma membrane, including caveolae and "lipid rafts" (24 -26). Whether in situ within blood vessels or in cultured endothelial cells, the relative proportion of eNOS in the Golgi complex versus the plasma membrane is variable and can depend on the vascular bed, species, and state of confluency (27)(28)(29). The N-myristoylation site mutant of eNOS is catalytically competent, remains cytosolic, and displays reduced NO release. Biochemically, this is due to a reduction in Ser 1179 phosphorylation (30 -32). Phosphorylation of eNOS at Ser 1179 has been observed in the Golgi complex and in the plasma membrane of endothelial cells; however, due to the presence of eNOS in both locations, the individual functions of Golgi versus plasmalemmal eNOS, if any, are not yet known and are matter of mere speculation (30). Therefore, the goal of this study was to determine the individual function of eNOS in the cytoplasmic aspect of the Golgi complex and plasma membrane using molecular targeting with organelle-restricted "zip codes" to deliver functional eNOS to specific intracellular sites and to examine whether eNOS targeted to endomembranes is functionally responsive to extracellular stimuli.

Generation of eNOS Targeting Proteins
The cytosolic myristoylation site mutant (G2A eNOS) is catalytically identical to the wild-type (WT) enzyme (33) and was used as an enzymatic scaffold for attaching the different targeting sequences as outlined below.
Syn17-eNOS-The length of the transmembrane domain of syntaxin-3 determines membrane association and subcellular targeting (35). The truncated transmembrane domain of syntaxin-3 (referred to as Syn17; 18 amino acids, KLIIIIVLVVVLLGILAL) that targets the cis/medial-Golgi complex was fused in-frame to the C terminus of G2A eNOS using primers 5Ј-TGT GCC TCG AGC GGG GCC ACA TG TT-3Ј (sense) and 5Ј-AAA TCT AGA TCT AGA TCA AAT CAA TGC TAA AAT GCC CAG CAA CAC AAC TAC TAG CAC AAT GAT AAT TAT CAA TTT GGG GCC GGG GGT GTC TGG GC-3Ј (antisense).
Syn25-eNOS-A region of syntaxin-3 containing the full-length plasma membrane-targeting domain (27 amino acids, KLIIIIVLVVV-LLGILALIIGLSVGLN) was fused in-frame to the C terminus of G2A eNOS using primer 5Ј-ATC TAG ATC TAG ATC AAT TCA GCC CAA  CGG AAA GTC CAA TAA TCA ATG CTA AAA TGC CCA GCA ACA  CAA CTA CTA GCA CAA TGA TAA TTA TCA ATT TGG GGC CGG  GGG TGT CTG GGC-3Ј (antisense).

Western Blotting
Cells were washed twice with phosphate-buffered saline and lysed on ice in 50 mM Tris-HCl (pH 7.5), 1% (v/v) Nonidet P-40, 10 mM NaF, 1 mM vanadate, 1 mM phenylmethylsulfonyl fluoride, 10 mg/ml aprotinin, and 10 mg/ml leupeptin, and lysates were transferred to an Eppendorf tube and rotated for 45 min at 4°C. Lysates were homogenized in a Dounce homogenizer (50 strokes), and insoluble material was removed by centrifugation at 12,000 ϫ g for 10 min at 4°C, size-fractionated by SDS-PAGE, and Western-blotted as described previously (7).
Immunofluorescence COS-7 cells were transfected as described above and plated onto sterile coverslips. The cells were then fixed in 1:1 acetone/methanol for 3 min at Ϫ20°C and rinsed twice with phosphate-buffered saline plus 0.1% bovine serum albumin for 5 min at room temperature. The cells were incubated with 5% goat serum in phosphate-buffered saline plus 0.1% bovine serum albumin for 30 min at room temperature, followed by incubation for 2 h with primary antibody (polyclonal or monoclonal) at room temperature. Texas Red-labeled anti-rabbit (diluted 1:100) or fluorescein isothiocyanate-labeled anti-mouse (diluted 1:100) secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) was incubated for 1 h at room temperature. Slides were mounted with Slowfade (Molecular Probes, Inc., Eugene, OR), and cells were observed with an inverted Zeiss microscope fitted with a Bio-Rad MRC 600 confocal imaging system. Antibodies for eNOS were obtained from BD Biosciences (anti-eNOS monoclonal antibody), and those for GM130 were from G. Warren (Yale University).

NO Release
Medium (100 l) containing NO (primarily NO 2 Ϫ ) was ethanol-precipitated to remove proteins and refluxed in sodium iodide/glacial acetic acid to convert NO 2 Ϫ to NO. NO was quantitated via specific chemiluminescence following reaction with ozone (Sievers). Net NO 2 Ϫ was calculated after subtracting unstimulated basal release (7).

NOS Activity Measurements
The activity of WT and targeted eNOS was determined in detergentsolubilized lysates of transfected COS-7 cells by measuring the conversion of [ 14 C]arginine to [ 14 C]citrulline as described previously (7).

Statistical Analysis
NO release data are expressed as means Ϯ S.E. Comparisons were made using two-tailed Student's t test or analysis of variance with a post-hoc test where appropriate. Differences were considered as significant at p Ͻ 0.05.

RESULTS
Generation and Characterization of eNOS Targeting Constructs-A cytosolic mutant of eNOS (G2A) that cannot be N-myristoylated was used to generate eNOS fusion proteins that specifically target the Golgi complex or plasma membrane (Fig. 1). Previous work has shown that G2A eNOS retains full catalytic ability in vitro; however, the stimulated release of NO from intact cells expressing this mutant is impaired because the enzyme fails to target the Golgi complex and plasmalemmal rafts/caveolae and is not efficiently phosphorylated by Akt at Ser 1179 (30,33,37). By exploiting these properties of G2A eNOS, we generated Golgi complex-and plasma membranetargeted eNOS to assess the functional importance of subcellular targeting (Fig. 1).
To determine whether the eNOS fusion proteins target their intended destinations, indirect immunofluorescence for eNOS and GM130, a resident peripheral Golgi protein, was performed in transfected COS cells. As shown in Fig. 2, WT eNOS was present in both the plasma membrane (WT-eNOS, left panel, arrows) and the Golgi complex (arrowhead), co-localizing with the Golgi marker GM130 (right panel, arrowhead), whereas G2A eNOS was diffusely distributed throughout the cell (G2A-eNOS, left panel) and did not specifically co-localize with GM130 (right panel). Targeting of eNOS to the Golgi complex with the GRIP domain was confirmed by the presence of eNOS concentrated in a perinuclear pattern (GRIP-eNOS, left panel), but it was also found in a pattern reminiscent of post-Golgi vesicles with the absence of plasma membrane staining. GRIP-eNOS partially co-localized with the Golgi marker GM130 (right panel). Similarly, fusion of eNOS with the syntaxin-3 cis/medial-Golgi complex-targeting motif (Syn17-eNOS) demonstrated Golgi targeting (S17-eNOS, left panel) and partial co-localization with GM130 and no plasma membrane staining (right panel). Fusion of eNOS to the fulllength transmembrane domain of syntaxin-3 (Syn25-eNOS) resulted in dramatic plasma membrane staining, with minor amounts present in the Golgi complex (S25-eNOS, left panel). Membrane targeting of eNOS with the CAAX motif derived from K-Ras (CAAX-eNOS) also showed the majority of eNOS in the plasma membrane (CAAX-eNOS, left panel). Thus, using these targeting motifs, eNOS was targeted to the peripheral aspect of the Golgi network (GRIP-eNOS and Syn17-eNOS) or the plasma membrane (Syn25-eNOS and CAAX-eNOS).
Previous studies have shown that WT and G2A eNOS have similar catalytic abilities and cofactor requirements (38,39) despite the inability of mislocalized G2A eNOS to produce NO after stimulation of cells with calcium-mobilizing agonists (33,37). Therefore, we measured the catalytic competency of the newly generated eNOS fusion proteins in vitro using the conversion of L-[ 3 H]arginine to L-[ 3 H]citrulline as an index of NOS activity in detergent-solubilized COS cell extracts. COS cells do not contain any endogenous NOS isoform, and all activity measurements were eliminated by the NOS inhibitor N -nitro-L-arginine methyl ester hydrochloride. The activity of GRIP-eNOS was slightly greater than that of WT eNOS (89.7 Ϯ 2.7 and 72.8 Ϯ 5.5 pmol of L-citrulline/mg/min, respectively; n ϭ 3). In a separate series of transfectants, Syn17-eNOS, Syn25-eNOS, and CAAX-eNOS exhibited slightly lower catalytic activities compared with WT eNOS (25.4 Ϯ 0.6, 20.8 Ϯ 1.5, 14.3 Ϯ 0.5, and 32.9 Ϯ 1 pmol of L-citrulline/mg/min, respectively; n ϭ 3). These differences in activity were not due to expression differences as assessed by semiquantitative Western blotting of the same extracts.
Plasma Membrane Targeting of eNOS Results in Marked NO Release due to Constitutive Phosphorylation of Ser 1179 -To determine the relevance of intracellular location to eNOS function in intact cells, COS cells were transfected with the various eNOS constructs, and basal NO release was determined by NO-specific chemiluminescence. Cells were subsequently lysed and immunoblotted for total eNOS or phosphorylated eNOS species using phosphorylation state-specific antibodies (Ser(P) 1179 , Thr(P) 497 , Ser(P) 617 , Ser(P) 635 ). As shown in Fig. 3, non-acylated cytosolic G2A eNOS displayed both significantly reduced NO release (A, graph) and hypophosphorylation of Ser 1179 (A) Thr 497 (B), and Ser 617 and Ser 635 (C) compared with WT eNOS (upper gels). On the other hand, the plasma membrane-targeted forms of eNOS fusion proteins (Syn25-eNOS and CAAX-eNOS) produced substantially more basal NO than did WT eNOS, with equivalent levels of eNOS protein expression. As shown in Fig. 3A (upper and lower gels), the level of Ser 1179 phosphorylation was also much greater in the plasma membrane-targeted forms of eNOS relative to the WT enzyme. Localization of eNOS to the peripheral aspects of the Golgi network produced divergent results. Restricted localization of eNOS to the trans-and post-Golgi vesicles via the GRIP targeting motif (GRIP-eNOS) resulted in reduced basal NO release compared with the WT enzyme. In contrast, fusion of eNOS to the syntaxin-3 cis-Golgi complex-targeting motif (Syn17-eNOS) produced equivalent amounts of basal NO compared with the WT enzyme. These changes in NO release did not strictly correlate with basal phosphorylation of Ser 1179 since GRIP-eNOS and Syn17-eNOS phosphorylation was comparable with WT eNOS phosphorylation, but was again greater than G2A eNOS phosphorylation.
We next investigated the relative degree of eNOS phosphorylation of the other regulatory phosphorylation sites. Thr 497 phosphorylation is important for the fidelity of eNOS to produce NO versus superoxide since phosphorylation promotes NO release, whereas dephosphorylation promotes superoxide anion production (18). Phosphorylation of Ser 617 and Ser 635 regulates temporal aspects of calcium/calmodulin regulation of the enzyme (17,40). As shown in Fig. 3B, WT eNOS and plasma membrane-targeted eNOS (Syn25-eNOS and CAAX-eNOS) exhibited enhanced Thr 497 phosphorylation relative to G2A eNOS and the Golgi complex-targeted eNOS constructs. The phosphorylation of Ser 617 was similarly distributed, with the greatest degree of phosphorylation observed with the WT and plasma membrane-targeted eNOS constructs. Low levels of Ser 617 phosphorylation were observed with cytosolic G2A eNOS and eNOS fusion proteins that target the Golgi network (GRIP-eNOS and Syn17-eNOS). GRIP-eNOS displayed the highest level of Ser 635 phosphorylation, followed by WT eNOS and plasma membrane-targeted eNOS (CAAX-eNOS and Syn25-eNOS). Syn17-eNOS (cis-Golgi complex-targeted) was found to have significantly lower phosphorylation of Thr 497 , Ser 635 , and Ser 617 (Fig. 3, B and C). Plasma Membrane-targeted eNOS Is Highly Sensitive to Transmembrane Calcium Fluxes-Thapsigargin, an inhibitor of SERCA, elevates intracellular calcium through store depletion, which in turn triggers calcium influx via capacitive calcium entry (41). The elevation of intracellular calcium is a key determinant of eNOS activation via the increased association of calcium-activated calmodulin. As shown in Fig. 4A, thapsigargin evoked release of substantially more NO from COS-7 cells transfected with plasma membrane-targeted forms of eNOS (Syn25-eNOS and CAAX-eNOS) compared with cells expressing WT eNOS or Golgi complex-restricted eNOS (GRIP-eNOS and Syn17-eNOS). Compared with the WT enzyme, calcium-dependent activation of cytosolic G2A eNOS resulted in the production of significantly less NO, whereas both trans-Golgi-localized GRIP-eNOS and cis/medial-Golgi targeting with Syn17-eNOS were not significantly different from that with WT eNOS.
Golgi Targeting with Syn17-eNOS Is Required for Akt-dependent Activation of eNOS-As stated previously, eNOS can be activated via Akt-dependent phosphorylation of Ser 1179 (7,14). As shown in Fig. 4B, cotransfection of WT eNOS with cDNA for Akt-1 resulted in the increased phosphorylation of Ser 1179 (upper gels) and enhanced production of NO (graph). Previous work has shown that this effect is abolished if Ser 1179 is mutated to alanine (7,14). Fatty acylation and subcellular targeting are required for full activation, as Akt stimulated significantly less NO production (graph) and Ser 1179 phosphorylation (upper gels) with cytosolic G2A eNOS versus the WT enzyme. Although a significant increase in NO release was observed in the presence of Akt, this was not greater than the unstimulated levels for the WT enzyme (Fig. 4B, graph). Transfection with Akt did not stimulate statistically significant NO production from GRIP-eNOS; however, restricted cis/medial-Golgi localization with Syn17-eNOS enabled full Akt-dependent activation of eNOS as determined by NO release and phosphorylation of Ser 1179 to levels seen with WT eNOS. Plasma membrane-targeted eNOS constructs (CAAX-eNOS and Syn25-eNOS) exhibited constitutive NO release and phosphorylation of Ser 1179 . In contrast to WT eNOS and Syn17-eNOS, cotransfection with Akt did not stimulate further production of NO or influence the phosphorylation state of Ser 1179 .

Phosphorylation of Ser 1179 Is Required for the Activation of Golgi Complex-and Plasma
Membrane-restricted eNOS-The ability of Akt to increase eNOS activity is dependent on the specific phosphorylation of Ser 1179 , and mutation of this site to the phospho-mimetic aspartic acid is sufficient to increase NO release from eNOS in the absence of additional stimuli (7,14,20). Therefore, we next investigated whether the phosphorylation state of this site is both sufficient to regulate NO release from cytosolic G2A eNOS and Golgi complex-targeted GRIP-eNOS and necessary for the enhanced activity seen with plasma membranetargeted eNOS. As shown in Fig. 5A, mutation of Ser 1179 to the phospho-mimetic aspartic acid (S1179D) increased NO release from the WT enzyme as described previously (7,14). NO release from cytosolic G2A eNOS and GRIP-eNOS was also increased by the S1179D mutation, restoring the levels of NO release back to the WT enzyme levels. In contrast, the Syn17-S1179D construct was fully activated by the aspartate substitution and produced as much NO as WT eNOS transfected with Akt (compare the second and eighth lanes).
Next, we examined the importance of phosphorylation of Ser 1179 in plasma membrane-targeted constructs that are constitutively active and hyperphosphorylated at Ser 1179 . Mutation of Ser 1179 to alanine (WT S1179A), which prevents phosphorylation, significantly reduced NO production from the WT enzyme as well as from the plasma membrane-targeted forms of eNOS (Syn25-eNOS and CAAX-eNOS), which constitutively produced more NO than the WT enzyme (Fig. 5B). There was no significant difference in the levels of NO produced from WT S1179A, CAAX-S1179A, and Syn25-S1179A. Thus, phosphorylation of eNOS at Ser 1179 restricted to either the cis/medial-Golgi network or the plasma membrane primarily controls NO release. DISCUSSION It is well established that eNOS is localized to both the plasma membrane and the cytoplasmic aspects of the cis/medial-Golgi complex in endothelial cells both in intact blood vessels and in culture (25,27,28,30,33,(42)(43)(44)(45). More recently, transgenesis of the entire human eNOS gene encoding an eNOS-green fluorescent protein recapitulated imaging and functional studies of eNOS-green fluorescent protein in cultured cells (43) and unequivocally documented the bimodal distribution of eNOS into the Golgi complex and plasma membrane of endothelial cells lining intact blood vessels (46). However, the presence of eNOS in both locations has complicated the process of identifying the individual roles played by each pool of eNOS and has made it difficult to determine whether the Golgi pool is merely a biosynthetic precursor pool of enzyme in route to the plasma membrane or an active pool of enzyme.
In this study, we generated fusion proteins that specifically target the plasma membrane or cytoplasmic aspects of the Golgi complex. All constructs were expressed well in cells and exhibited ample NOS activity in vitro. Surprisingly, plasma membrane-targeted eNOS (Syn25-eNOS and CAAX-eNOS) had slightly lower catalytic activity in vitro, but was constitutively active and phosphorylated at Ser 1179 in intact cells and highly responsive to transmembrane calcium fluxes, yet insensitive to further activation by Akt-mediated phosphorylation. Mutation of Ser 1179 to alanine prevented phosphorylation of this site and abolished the "constitutive activation" of eNOS brought about by targeting the plasma membrane. In contrast, cis-Golgi complex-targeted eNOS (Syn17-eNOS) behaved similarly to WT eNOS, was less sensitive to calcium-dependent activation, and was highly responsive to Akt-dependent phosphorylation compared with plasma membrane-targeted versions. We have shown that discrete targeting to the cis/medial-Golgi complex (Syn17-eNOS), but not to the trans-and post-Golgi vesicles (GRIP-eNOS), efficiently regulated eNOS activation by Akt and NO release. The introduction of a negative charge at Ser 1179 (S1179D) was sufficient to fully stimulate NO production from WT eNOS and Syn17-eNOS, consistent with their constitutive phosphorylation brought about by plasma membrane targeting. This study provides the first clear evidence for functional roles of plasma membrane-and Golgi complex-localized eNOS and demonstrates the paramount importance of Ser 1179 phosphorylation in governing eNOS function in these specific locales in intact cells.
The central dogma of eNOS activation describes a model in which eNOS must be localized to plasma membrane microdomains such as caveolae or lipid rafts to produce optimal amounts of NO. This stems from several lines of evidence. Myristoylationand/or palmitoylation-deficient mutants of eNOS fail to target both the plasma membrane and Golgi complex and display reduced activity in vivo (25,26,45). The majority of cellular eNOS and eNOS activity, normalized to caveola protein content, is concentrated in caveolin-1-enriched microdomains (25). Conflu-ent endothelial cells have more plasma membrane and less Golgi eNOS than do subconfluent cells and release more NO in response to agonist challenge (42). The eNOS-binding proteins NOSIP and NOSTRIN displace plasma membrane eNOS to internal membranes (unlikely Golgi) and reduce calcium-dependent NO release (47,48). And finally, depletion of plasma membrane cholesterol with ␤-methyl cyclodextrin and treatment with oxidized low density lipoprotein reduce plasma membrane eNOS to an intracellular compartment (non-Golgi) and attenuate NO release (49,50). The majority of these studies have investigated the impact of subcellular targeting on the calcium-dependent activation of eNOS. Indeed, our data support the concept that targeting eNOS to the plasma membrane with either the fulllength transmembrane domain of syntaxin-3 (Syn25-eNOS) or the prenylated polybasic domain of K-Ras (CAAX-eNOS) results in enzymes that are constitutively active and highly responsive to the calcium-mobilizing agent thapsigargin. However, in addition to calcium, eNOS can be regulated via phosphorylation of Ser 1179 FIG. 4. Calcium-and Akt-dependent activation of plasma membrane-targeted eNOS fusion proteins is greater than that of WT and Golgi complextargeted eNOS constructs. A, transfected COS cells expressing WT eNOS, G2A eNOS, or Golgi complex-targeted (GRIP-eNOS and Syn17-eNOS (S17)) or plasma membrane (PM)-targeted (Syn25-eNOS (S25) and CAAX-eNOS) eNOS constructs were stimulated with thapsigargin (1 M) for 30 min, and stimulated NO release was determined by NO-specific chemiluminescence. Stimulated NO release was determined as the increase in NO release following subtraction of background levels. Data are presented as means Ϯ S.E. (n ϭ 5). *, p Ͻ 0.05 versus the WT enzyme. B, COS cells were transfected with the various eNOS constructs with or without the protein kinase Akt, and NO release was determined over a 24 h-period. Lysates were Western-blotted for the relative degree of Ser 1179 phosphorylation (upper gels) relative to total eNOS (middle gels) and phosphorylated Akt (lower gels). Data are presented as means Ϯ S.E. (n ϭ 5). ϩ, p Ͻ 0.05 versus the WT enzyme; *, p Ͻ 0.05 in the presence of Akt.
by several protein kinases, including Akt. Thus, the relative contribution of subcellular targeting to agonists that utilize calcium-versus Akt-dependent mechanisms to activate eNOS remains to be established despite the evidence provided above.
In the majority of transfected cells, cultured endothelial cells, and intact blood vessels, a significant proportion of eNOS is found localized to the Golgi complex, where it co-localizes with Golgi markers, including mannosidase II and 58-kDa Golgi protein (27,33,51) and GM130 (this study). The function of Golgi complex-localized eNOS is poorly characterized, and given the preeminence placed on plasma membrane targeting, it is often assumed to be an inactive reservoir or immature eNOS in route to the plasmalemma and has more recently been interpreted as eNOS bound to internalized caveolae (51,52). However, there is ample evidence for a functional role of Golgi complex-localized eNOS. As originally described in transfected cells (33,37), eNOS is found in the peripheral aspects of the Golgi complex, and this targeting is required for optimal NO release from cells. More recently, using immunofluorescence microscopy to label eNOS phosphorylated at Ser 1179 and NO imaging with 4,5-diaminofluorescein diacetate, we found that, in response to the angiogenic vascular endothelial growth factor, Golgi complex-localized eNOS is phosphorylated at Ser 1179 , and we identified "bursts" of NO emanating from the perinuclear region of endothelial cells (30). Recently, using a chimeric eNOS fusion protein fused to a calcium-sensing fluorophore, Jobin et al. (52) showed that calcium activation of eNOS imaged in living cells is primarily localized in a perinuclear orientation, identical to the localization of Golgi markers. In addition, acute chemical disruption of the Golgi complex inhibits NO release in some (53), but not all (54), endothelial cells.
In this study, we used a more direct approach to target eNOS specifically to the Golgi complex. Targeting eNOS to the transand post-Golgi vesicles via the GRIP domain resulted in an enzyme that produced very low amounts of basal NO and that was hyporesponsive to both thapsigargin and coexpression of Akt compared with the WT enzyme. The GRIP domain comprises ϳ42 amino acids and is shared by a family of resident peripheral Golgi coiled-coil membrane proteins, including p230, golgin-97, and golgin-245 (34). The GRIP docking site has been identified on a specific subset of Golgi membranes in the trans-Golgi network (55). Targeting eNOS to the cis/medial-Golgi complex (35) with the truncated transmembrane domain of syntaxin-3 (Syn17-eNOS) produced contrasting results. Compared with the WT enzyme, the calcium-dependent activation of Syn17-eNOS was slightly attenuated, but interestingly, this construct was strongly activated by Akt, supporting the idea that even within distinct membrane domains of the Golgi complex, compartmentalization is critical for eNOS activation. In other experiments, we attempted to place eNOS in the lumen of the Golgi complex via fusion with ␤-galactosyltransferase, but this construct did not express well, and the protein was degraded (data not shown). It is possible that targeting to the cytoplasmic aspect of the trans-versus cis-Golgi complex places the enzyme in a more favorable setting to produce NO or to be more accessible to kinases that can influence its activity. To address this question, we constructed phospho-mimetic mutants (S1179D) of GRIP-eNOS and Syn17-eNOS. The introduction of a negative charge at Ser 1179 resulted in increased NO release from GRIP-S1179D; however, its activity was increased only to the level of WT eNOS and was substantially less compared with WT S1179D and Syn17-S1179D. Thus, it is likely that factors other than phosphorylation of Ser 1179 contribute to the reduced activity seen in the GRIP-eNOS construct.
Subcellular targeting is known to influence both eNOS activity and phosphorylation (30,31). In this study, we found that plasma membrane-targeted eNOS (i.e. CAAX-eNOS and Syn25-eNOS) was constitutively phosphorylated at Ser 1179 and that mutation of Ser 1179 to alanine abolished constitutive activation and reduced their activity to that of S1179A eNOS. Golgi complex-targeted Syn17-eNOS exhibited lower basal phosphorylation versus the WT enzyme, but was strongly phosphorylated in the presence of Akt, consistent with the ability of Akt to stimulate NO release. In contrast, trans-Golgi complex-targeted GRIP-eNOS displayed reduced basal and Akt-stimulated FIG. 5. Phosphorylation of eNOS at Ser 1179 is both necessary for the constitutive activation of plasma membrane-targeted eNOS and sufficient to increase NO release from Golgi complextargeted eNOS. COS cells were transfected with WT eNOS, G2A eNOS, Golgi complex-targeted (GRIP-eNOS and Syn17-eNOS) and plasma membrane-targeted (Syn25-eNOS and CAAX-eNOS) eNOS constructs, and site-directed mutants of each construct in which Ser 1179 was mutated to alanine to prevent phosphorylation (WT eNOS, Syn25-eNOS, and CAAX-eNOS) or to the phospho-mimetic aspartic acid (WT eNOS, G2A eNOS, GRIP-eNOS, and Syn17-eNOS). A, mutation of Ser 1179 to aspartic acid increased NO release from WT eNOS, G2A eNOS, and Golgi complex-restricted eNOS constructs (GRIP-eNOS and Syn17-eNOS). B, mutation of Ser 1179 to alanine, which prevents phosphorylation, reduced NO release from WT eNOS and plasma membrane-targeted eNOS constructs (Syn25-eNOS and CAAX-eNOS). Data are presented as means Ϯ S.E. (n ϭ 5). *, p Ͻ 0.05 versus the corresponding unmutated enzyme; ϩ, p Ͻ 0.05 versus WT eNOS. phosphorylation comparable with cytosolic G2A eNOS. In addition to Ser 1179 , eNOS is phosphorylated at Ser 116 , Thr 497 , Ser 616 , and Ser 635 , and all of these sites have been proposed to influence eNOS function. Given the importance of location to the specificity of kinases, it is perhaps not surprising that phosphorylation of eNOS at the other sites varied depending on where the enzyme was targeted. Phosphorylation of Thr 497 , previously characterized as a negative regulatory site (16,56) or more recently as a site important for the fidelity of eNOS coupling (18), was elevated in WT and plasma membranetargeted eNOS. The phosphorylation of Ser 617 followed a similar pattern in that it was elevated in WT and plasma membrane-targeted eNOS, whereas trans-Golgi complex-targeted GRIP-eNOS had the most prominent phosphorylation of Ser 635 . The degree of phosphorylation of Thr 497 , Ser 617 , and Ser 635 was in general lower for cis-Golgi complex-targeted Syn17-eNOS. The functional significance of the differences in relative phosphorylation of these sites is not well understood compared with the better characterized Ser 1179 phosphorylation site.
The heterogeneous localization of eNOS is known to vary depending on the species, degree of confluency, passage, and even across the vascular tree (27,28,42,43). The mechanisms regulating the intracellular distribution of eNOS between the Golgi and plasma membrane pools and the physiological role underlying their distinct regulation remain to be established. However, it is conceivable that by influencing the relative proportion of eNOS in the Golgi complex versus the plasma membrane, the ability of an endothelial cell to respond to calcium-and Akt-dependent agonists can be modified. For example, endothelial cell agonists that utilize primarily Akt-dependent mechanisms, viz. vascular endothelial growth factor, insulin, and estrogen, would preferentially active Golgi eNOS, and conversely, calcium-dependent agonists such as thapsigargin and bradykinin would preferentially activate plasma membrane eNOS. This concept is strengthened by the recent discoveries of two novel eNOS-binding partners, NOSIP and NOSTRIN (47,48), and an indirect regulator, CHIP (57). Overexpression of both NOSIP and NOSTRIN influences eNOS activity only in vivo by redistributing eNOS from the plasma membrane to intracellular sites. CHIP, the C-terminal interacting protein for hsp70, is part of the eNOS-hsp90-hsp70 complex and, when overexpressed in endothelial cells, results in eNOS insolubility in the Golgi complex and loss of NO release from cells. Another consideration is the metabolic fate of locally generated NO. NO is typically thought to react with thiol residues of effector proteins in a diffusion-limited reaction; however, recent data showing that the greatest amount of NO is in the perinuclear region of cells and localized S-nitrosylation reactions influencing vesicle secretion challenge this dogma (58,59). NO generated in the perinuclear/Golgi area would be more highly concentrated in membranes of the Golgi complex and around the nucleus and more apt to produce autocrine actions than NO synthesized in the plasma membrane, which may readily diffuse to extracellular targets. The autocrine role for NO in regulating endothelial cell function is supported by many studies in cultured endothelial cells showing that blockage of NOS reduces growth-stimulated cell migration, proliferation, and organization into tube-like structures (60 -62) and a recent study in which NO was shown to inhibit von Willebrand factor secretion from Weibel-Palade bodies via nitrosylation of the protein NSF (N-ethylmaleimidesensitive factor) (59). Whether the localized release of NO differentially contributes to autocrine (endothelial) or paracrine (vascular smooth muscle, etc.) actions of eNOS-derived NO remains to be established.
In summary, we have shown that plasma membrane-and Golgi complex-localized eNOS can indeed function to produce NO, exhibiting attributes similar to those of the WT enzyme.
We have also shown that phosphorylation of Ser 1179 is of paramount importance for activation of plasma membrane-and Golgi complex-targeted eNOS to produce biologically active NO, regardless of the mechanism of activation of these pools of enzyme (i.e. calcium-dependent agonist versus Akt). Finally, our data strongly support that the concept of endomembranespecific regulation of signaling systems, previously thought to exist only in the plasma membrane, is indeed operational for eNOS as highlighted recently by several classes of signaling molecules, including phosphatidylinositol 3,4,5-trisphosphate (63), Ras (64), and G proteins (65), now appreciated to signal on the Golgi and internalized membranes.