Proteomic Identification of S-Nitrosylated Golgi Proteins: New Insights into Endothelial Cell Regulation by eNOS-Derived NO

Background Endothelial nitric oxide synthase (eNOS) is primarily localized on the Golgi apparatus and plasma membrane caveolae in endothelial cells. Previously, we demonstrated that protein S-nitrosylation occurs preferentially where eNOS is localized. Thus, in endothelial cells, Golgi proteins are likely to be targets for S-nitrosylation. The aim of this study was to identify S-nitrosylated Golgi proteins and attribute their S-nitrosylation to eNOS-derived nitric oxide in endothelial cells. Methods Golgi membranes were isolated from rat livers. S-nitrosylated Golgi proteins were determined by a modified biotin-switch assay coupled with mass spectrometry that allows the identification of the S-nitrosylated cysteine residue. The biotin switch assay followed by Western blot or immunoprecipitation using an S-nitrosocysteine antibody was also employed to validate S-nitrosylated proteins in endothelial cell lysates. Results Seventy-eight potential S-nitrosylated proteins and their target cysteine residues for S-nitrosylation were identified; 9 of them were Golgi-resident or Golgi/endoplasmic reticulum (ER)-associated proteins. Among these 9 proteins, S-nitrosylation of EMMPRIN and Golgi phosphoprotein 3 (GOLPH3) was verified in endothelial cells. Furthermore, S-nitrosylation of these proteins was found at the basal levels and increased in response to eNOS stimulation by the calcium ionophore A23187. Immunofluorescence microscopy and immunoprecipitation showed that EMMPRIN and GOLPH3 are co-localized with eNOS at the Golgi apparatus in endothelial cells. S-nitrosylation of EMMPRIN was notably increased in the aorta of cirrhotic rats. Conclusion Our data suggest that the selective S-nitrosylation of EMMPRIN and GOLPH3 at the Golgi apparatus in endothelial cells results from the physical proximity to eNOS-derived nitric oxide.


Introduction
Nitric oxide (NO) conveys specific cellular signals via Snitrosylation despite its highly reactive and diffusible nature [1,2]. S-nitrosylation is a post-translational modification of cysteine-thiols to form nitroso-thiols [3]. One of the most important factors that specify the targets of S-nitrosylation is the compartmentalization of NO synthase (NOS, a source of NO N ) with its target proteins for S-nitrosylation [1]. This compartmentalization allows for the generation of relatively high local NO concentrations in the vicinity of the target proteins, which enables the formation of S-nitrosocysteine. Endothelial NOS (eNOS) is unique among the NOS family members as it is localized mainly to specific intracellular membrane domains in endothelial cells, including the cytoplasmic side of the Golgi apparatus and the plasma membrane caveolae [4][5][6].
Previously we showed that the localization of eNOS in the cell is an important determinant of protein S-nitrosylation [7]. Using a mutant eNOS that was targeted to the nucleus and wild-type eNOS that is localized on the Golgi apparatus, we demonstrated that protein S-nitrosylation occurs where eNOS is localized. Further, we presented that localization of eNOS at the Golgi apparatus influences Golgi functions such as protein trafficking in endothelial cells. Specifically, eNOS localized at the Golgi reduces the speed of protein transport from the endoplasmic reticulum (ER) to the Golgi apparatus and from the Golgi to the plasma membrane [7].
The Golgi is a membrane organelle that plays essential roles in post-translational modifications such as glycosylation. It is a dynamic structure and its membranes are constantly remodeled during cell growth, migration and division. Therefore, eNOS localization to the Golgi apparatus implies a unique property of this enzyme in regulating Golgi function, likely through S-nitrosylation of Golgi membrane proteins [8]. Thus, identification of S-nitrosylated proteins on the Golgi membranes is required for understanding endothelial cell function regulated by eNOS-derived NO.
In this study, we performed a proteomic analysis [9] of Golgi membranes isolated from rat livers and identified 78 putative Snitrosylated proteins and the target cysteine residues. Nine of them were Golgi resident and Golgi/endoplasmic reticulum (ER)associated proteins that are largely representative of critical Golgi-related cellular functions such as protein trafficking and glycosylation. Among them, we confirmed that at least two proteins, extracellular matrix metalloproteinase inducer (EMM-PRIN) and Golgi phosphoprotein 3 (GOLPH3), are endogenously S-nitrosylated and co-localized with eNOS at the Golgi apparatus in endothelial cells, providing strong evidence for compartmentalization-induced selective S-nitrosylation.

Materials and Methods
Plasmids A yellow fluorescent protein (YFP)-tagged rat EMMPRIN cDNA construct was a kind gift from Drs. Andrew P. Halestrap and Marieangela C. Wilson (University of Bristol, UK) [10]. A red fluorescent protein (RFP)-tagged wild-type eNOS (WTeNOS-RFP) was generated as described [7]. An HA-tagged human GOLPH3 in pBABE-Puro vector was obtained from Addgene (Cambridge, MA).

Isolation of an enriched Golgi membrane fraction
Rat liver Golgi membranes were isolated from adult male Sprague-Dawley rats as described [11]. Briefly, male rat livers were homogenized using a 150-mm mesh sieve in a buffer containing 100 mM potassium phosphate buffer, pH6.7, 5 mM magnesium chloride, 0.5 M sucrose and protease inhibitors. The homogenates were layered over 0.86 M sucrose buffer, followed by a layer of 0.25 M sucrose buffer, and then centrifuged at 100,0006g for 1 hour at 4uC. After the centrifugation, the Golgi membrane fraction was accumulated at the interface between 0.5 M and 0.86 M sucrose buffers (Fraction II in Figure 1A). This Golgi membrane fraction was then adjusted to 0.25 M sucrose, placed on 0.5 M (middle layer) and 1.3 M sucrose (bottom layer) buffers, and then centrifuged at 8,0006g for 30 min at 4uC. The Golgi membrane was collected at the 0.5 M/1.3 M sucrose interface (Fraction III in Figure 1A).

Quantification of protein S-nitrosylation
S-nitrosocysteine (Cys-NO) formation was determined by triiodide reduction coupled to chemiluminescence detection using a Sievers 280 NO analyzer as described [9]. Approximately a total of 400 mg Golgi membrane proteins was used. Golgi membrane proteins were incubated with different concentrations of NO donors, Cys-NO and diethylamine NONOate (DEANO), for 10 min at 37uC, and followed by acetone precipitation for 1 hour. Protein pellets were resuspended in an isolation buffer and adjusted to a concentration of 1 mg/ml. The isolation buffer contained 20 mM HEPES, pH7.7, 1 mM diethylene triamine pentaacetic acid (DTPA), 0.1 mM neocuproine, 1% Triton X-100 and 2 mM methyl methanethiosulfonate (MMTS). Protein samples were then split into two equal volumes; one sample was added 1/10 volume of 1% sulfanilamide (SNA)/glacial acid and incubated on ice for at least 15 min to remove nitrite contamination. The other sample was added 1/10 volume of 1% SNA/glacial acid and 30 mM HgCl 2 , incubated at room temperature (RT) for 10 min, and then placed on ice. Snitrosoglutathione (GSNO) was used as a standard. Approximately 15 to 50 mg of Golgi membrane proteins were injected into a reaction vessel containing 5 ml of 60 mM potassium iodide and 10 mM iodine in glacial acetic acid at 37uC.

Biotin-switch assay
Approximately a total of 2.0 mg Golgi membrane proteins was used for the biotin-switch assay [12] necessary to perform a proteomic analysis described below, while an approximately 200 mg of Golgi membrane proteins was used per treatment group for dose-response and time-course experiments with an NO donor, S-nitroso-N-acetyl-D,L-penicillamine (SNAP). Golgi membrane proteins were also treated with or without another NO donor, nitrosocysteine (Cys-NO, 10 mM) for 15 min at 37uC. After cold acetone precipitation, pellets were resuspended in a lysis buffer (HEN buffer containing 1%Triton X-100 and protease inhibitors) to obtain 1 mg/ml. Samples were adjusted to 0.5 mg/ ml containing 2.5% SDS and 200 mM MMTS and incubated at 50uC for 20 min with every 5 min of vortexing to block free thiols. After blocking, cell extracts were precipitated with two volumes of 220uC acetone, incubated at 220uC for 20 min, centrifuged at 12,0006g for 10 min at 4uC, washed four times with acetone, and resuspended in 0.2 ml of HENS buffer (25 mM HEPES, pH7.7/ 0.1 mM EDTA/0.01 mM neocuproine/1% SDS). Then, 0.4 mM biotin-HPDP and 5 mM ascorbate were added and incubated at 25uC for 1 hour while rotating. As a control for non-specific HPDP labeling, ascorbate was omitted. Proteins were precipitated with acetone. Samples in which protein digestion was performed for further proteomic analysis were resuspended in 0.45 ml of 0.1 M ammonium bicarbonate and 0.5% SDS [9]. Otherwise, samples were processed for detection of biotinylated proteins as previously described [12,13].

Protein digestion and affinity peptide capture
Biotinylated protein samples were incubated with trypsin (1:30 enzyme/protein ratio) at 37uC for 18-24 hours in the dark as described [9]. The samples were passed through Ultra-free-MC 10-kDa cutoff filters that had previously been rinsed with methanol and washed with H 2 O. The filtrate containing the peptides was recovered and incubated with approximately 50 ml of dry, washed streptavidin-agarose beads per mg of initial protein for 30 min with gentle mixing. The samples were centrifuged at 5,0006g for 5 min, and the supernatants were discarded. The beads were washed five times with 10 volumes of 1 M ammonium bicarbonate, followed by five washes with 10 volumes of deionized water. Between washes, the samples were centrifuged at 1,0006g for 1 min. An elution buffer containing 70% formic acid was incubated with the beads for 30 min with gentle mixing. The captured peptides were recovered by centrifuging the beads at 5,0006g for 4 min and collecting the supernatants. The captured peptides were evaporated to approximately 5 ml in vacuum, resuspended in 20 ml of 0.1% formic acid, and desalted using Zip-Tips.

Analysis by LC-MS/MS
Desalted samples were analyzed on a Thermo LTQ linear trap mass spectrometer equipped with a Thermo micro electrospray source, a Thermo Surveyor pump and autosampler (Thermofisher Scientific, San Jose, CA) as described [9]. MS/MS spectra were searched with SEQUEST (Bioworks Browser 3.1 SR1) against the rat NCBI RefSeq database concatenated with corresponding reverse sequences. Cysteine modification by MMTS (+46 atomic mass units) and by biotin-HPDP (+428 atomic mass units) was specified as variable modifications. Peptide spectrum matches (PSMs) were empirically filtered by SEQUEST cross correlation scores (X c ) using PSMs assigned to the reverse database to achieve an estimated false discovery rate of ,5%. RefSeq accessions were mapped on to UniProt UniRef accessions.
Immuno-labeling of eNOS (rabbit, 1:10; Novus Biologicals, Littleton, CO, Cat#:NB120-15280), EMMPRIN (goat, 1:100; Santa Cruz, CA, Cat#:sc-9757), GM130 (mouse, 1:500; BD Biosciences, CA, Cat#:610822) and GOLPH 3 (rabbit, 1:100; Abcam, MA, Cat#:AB91492) was performed in bovine aortic endothelial cells (BAECs) grown in 8-well slides. The cells were Asterisks indicate endogenous biotin-containing proteins, thus considered as non-specific bands. Dithiothreitol (DTT), which cleaves nitroso-cysteine bonds, serves as a negative control. Shown are representative blots from 3 independent experiments. (C) S-nitrosylated Golgi membrane proteins were increased time-dependently in response to 100 mM SNAP in vitro for indicated incubation times (0, 30 and 60 min). Arrows indicate those proteins increased in response to SNAP. Asterisks indicate endogenous biotin-containing proteins, thus considered as non-specific bands. Shown are representative images from 3 independent experiments. (D) The sample quality was verified before performing a proteomic analysis. Successful biotinylation (i.e., S-nitrosylation) by the biotin-switch assay was determined by Western blot. Golgi membranes were incubated with or without 10 mM S-nitrosocysteine (Cys-NO) for 15 min at 37uC. Then, the biotin-switch assay was performed in the presence or absence of 2 mM ascorbic acid (AA) (right panel). Proteins were separated by SDS-PAGE and an equal protein loading to each lane was confirmed by Ponceau staining (left panel). Subsequent Western blotting using an anti-biotin antibody detected biotinylated proteins (right panel). Arrows indicate unique bands of biotinylated proteins that appeared in the presence of AA. Those biotinylated proteins increased by Cys-NO treatment (the 4 th lane in the right panel) were identified by mass spectrometry. doi:10.1371/journal.pone.0031564.g001 fixed with 4% paraformaldehyde for 10 min at 4uC, permeabilized with 0.1% Triton X-100/PBS for 15 min at RT, blocked with 5% donkey serum/0.3% BSA/PBS for 30 min at RT, and incubated with primary antibodies overnight at 4uC. Then, the cells were incubated with secondary antibodies conjugated to 488 or 568 (1:500 in PBS) for 1 hour at RT. After wash, the cells were mounted with DAPI media (Invitrogen, CA) for nuclear staining according to the manufacturer's instruction. Images were taken using a Nikon E800 Microscope with a Plan-Fluorchromat 406/ 0.75 objective (Nikon, Melville, NY).

Generation of cirrhotic rats and isolation of the aorta from cirrhotic rats
Male Sprague-Dawley rats (Harlan Sprague-Dawley Laboratories, Indianapolis, IN), weighing 100-125 g, were exposed to carbon tetrachloride (CCl 4 ) by inhalation for 12 weeks as described [15]. The aorta from three cirrhotic and three agematched control rats were used for analysis. Vessels were lysed in the lysis buffer mentioned above and the biotin switch assay was performed using an equal amount of proteins (500 mg) per group. Biotinylated proteins (i.e., S-nitrosylated proteins) were purified using streptavidin agarose beads and detected by Western blot. All procedures were performed in accordance with the ''Principles of Laboratory Animal Care'' and were approved by the Animal Care and Use Committee at the Veterans Affairs Healthcare System of Connecticut (Protocol#: YI0002).

Golgi membrane proteins were successfully enriched
We isolated Golgi membranes from rat livers. We chose this tissue for several reasons. First, rat liver is the most commonly used organ for studies of Golgi function and the protocol for isolation of Golgi membranes is well established [11]. Second, in order to perform the biotin-switch assay and the subsequent proteomic analysis, a large quantity of Golgi membrane proteins (at least 1.0-2.0 mg per treatment) was required.
The yield of Golgi membrane proteins was approximately 1.0-2.0 mg per liver. An examination of Golgi membrane proteins by Western blotting showed a successful enrichment of the final Golgi membrane fraction, which was used for further analysis of Snitrosylated Golgi membrane proteins ( Figure 1A). There was slight contamination of endoplasmic reticulum (ER), while contamination of nuclear membranes was very minimal ( Figure 1A).

Golgi membrane proteins are S-nitrosylated in response to NO donors
To investigate S-nitrosylation of Golgi membrane proteins, we first determined S-nitrosylated protein levels by the biotin-switch assay (Figures 1B and C) and the tri-iodide chemiluminescent method ( Figure S1). A dose and time-dependent increase in Snitrosylated proteins in response to SNAP treatment of Golgi membranes was determined ( Figures 1B and C). Similarly, Snitrosylation of Golgi membrane proteins was increased in response to treatment with S-nitrosocysteine (Cys-NO) and DEANO ( Figure S1). As reported previously [9], Cys-NO was a more effective S-nitrosylating agent than DEANO.

Proteomic analysis and identification of S-nitrosylated Golgi and Golgi/ER-associated proteins
Based on the results in Figure S1 and our previous experience in exposing cells to different concentrations of Cys-NO [9], we decided to carry out S-nitrosylation of Golgi membrane proteins for a proteomic analysis using 10 mM Cys-NO, a concentration that results in the formation of S-nitrosylated proteins without compromising cellular functionality [8]. Before conducting a proteomic analysis, we ran a Western blot after the biotin switch assay for Golgi membrane proteins to verify whether Snitrosylated proteins were biotinylated in the presence or absence of ascorbic acid (AA) during the biotin-switch assay. In the presence of AA, nitroso-thiol residues (i.e., S-nitrosylated cysteine residues) are converted to thiol residues, which then react with biotin-HPDP, a thiol-specific biotinylating reagent. Therefore, biotinylation that occurs in the absence of AA can be considered non-specific. As indicated by arrows in Figure 1D (the 4th lane on the right membrane), the Cys-NO treatment in the presence of AA resulted in unique bands of biotinylated (i.e., previously Snitrosylated) proteins. Biotinylated proteins were subjected to tryptic digestion followed by streptavidin-based enrichment of biotinylated peptides, and subsequent tandem mass spectrometry analysis. Biotinylated peptides that were identified from samples treated with biotin-HPDP alone (in the absence of AA) were considered non-specific and were excluded.
We identified 78 putative S-nitrosylated proteins from our Golgi membrane samples (Table S1), nine of which were Golgi resident and Golgi/ER-associated proteins (Table 1), while the rests were considered proteins in transit through the Golgi apparatus. This proteomic analysis also allowed us to identify the site where Snitrosylation occurs on the protein.

EMMPRIN and GOLPH3 are S-nitrosylated in endothelial cells
Based on availability and specificity of antibodies, we chose two proteins from Table 1, EMMPRIN (also known as CD147 or basigin), an inducer of matrix metalloproteinases (MMPs) that are involved in vascular remodeling [16][17][18], and Golgi phosphoprotein 3 (GOLPH3), a Golgi-localized oncoprotein implicated in protein trafficking, receptor recycling, and glycosylation [19], and confirmed that these proteins are S-nitrosylated in endothelial cells. We performed the biotin-switch assay coupled with Western blot for lysates collected from bovine aortic endothelial cells (BAECs) that were treated with a calcium ionophore, A23187, a well-known agonist of eNOS activation. Figure 2A shows that EMMPRIN is Snitrosylated. We then tested whether EMMPRIN is endogenously Snitrosylated in unstimulated BAECs. For this experiment, we immunoprecipitated endothelial cell lysates with an S-nitrosocysteine antibody or nonspecific rabbit IgG followed by Western blotting for EMMPRIN. As shown in Figure 2B (left panel), EMMPRIN was Snitrosylated at a detectable level without eNOS stimulation (i.e., at the basal level) and was increased in response to eNOS stimulation ( Figure 2B, middle and right panels, also see Figure S3).
S-nitrosylation of GOLPH3 was also confirmed in endothelial cell lysates by the biotin-switch assay ( Figure 2C). Like EMM-PRIN, GOLPH3 was S-nitrosylated at the basal level (without A23187) and S-nitrosylation of GOLPH3 increased about 3-fold in response to A23187 treatment in endothelial cells ( Figure 2D). Collectively, these observations suggest that EMMPRIN and GOLPH3 are S-nitrosylated in endothelial cells at their basal levels and that S-nitrosylation of these proteins increases upon eNOS stimulation.

EMMPRIN and GOLPH3 are co-localized with eNOS at the Golgi apparatus in endothelial cells
First, we determined the cellular localization of EMMPRIN and GOLPH3 in endothelial cells ( Figure 3A). We found that EMMPRIN was co-localized with eNOS at the Golgi apparatus ( Figure 3A, upper panel). Further, we observed that GOLPH3 was highly concentrated on the Golgi apparatus ( Figure 3A, lower panel) as indicated by its co-localization with GM130, a Golgi marker. These results indicate that EMMPRIN and GOLPH3 colocalize with eNOS at the Golgi apparatus in endothelial cells. In addition to BAECs, we investigated their cellular localization in COS cells. We co-expressed eNOS with YFP-tagged EMMPRIN and HA-tagged GOLPH3 in COS cells. Consistent with our observation in endothelial cells, our results showed that these three proteins were co-localized on the Golgi membranes ( Figure 3B).
We next examined if these proteins might be physically associated with each other in endothelial cells. Interestingly, EMMPRIN was co-immunoprecipitated with both eNOS and GOLPH3 in endothelial cells ( Figure 3C, Figure S2). These results suggest that eNOS interacts with its substrates on the Golgi. EMMPRIN may act as a scaffold for eNOS and GOLPH3, forming a complex at the Golgi.

EMMPRIN S-nitrosylation is increased in the aorta isolated from cirrhotic rats
We also examined whether S-nitrosylation of EMMPRIN increases in pathological conditions where excessive eNOSderived NO production is involved. We chose the aorta of cirrhotic rats with portal hypertension for this purpose, because in the arteries such as the aorta NO production is increased and extracellular matrix remodeling occurs [20], a potential role for EMMPRIN. Furthermore, the aorta provides a relatively large amount of proteins that are required for the biotin-switch assay to detect S-nitrosylation, compared to other arteries. EMMPRIN exists in glycosylated (an active form, 43-66 kDa) and nonglycosylated forms (,28 kDa). The total EMMPRIN level in the aorta of cirrhotic rats was approximately 2-fold higher than that of the normal aorta (Figure 4, left panel). However, as indicated by Hsp90 content (loading control), the lysates prepared from the aorta of normal and cirrhotic rats had a similar protein content. Most interestingly, we found that S-nitrosylated EMMPRIN was also roughly 2-fold higher in the aorta of cirrhotic rats than that of normal rats and that glycosylated EMMPRIN was only Snitrosylated (Figure 4, right panel).

Discussion
Using a mass spectrometry-based approach, we identified for the first time nine Golgi resident and Golgi/ER-associated proteins from Golgi membranes isolated from rat livers that are potential targets for S-nitrosylation. The purity of the isolated Golgi membranes was reasonable. Although their enriched fractions showed a low level of ER contamination, it was not problematic since we identified Golgi/ER-associate proteins as well as Golgi proteins. Actually, ER contamination is almost inevitable since many ER resident proteins recycle through the Golgi where they are recognized by the KDEL receptor and retrieved back to the ER by COPI vesicles [21]. In fact, some researchers state that no organelle fractionation technique produces a ''pure'' preparation [22].
Importantly, among the nine Golgi resident and Golgi/ER proteins identified, we validated S-nitrosylation of EMMPRIN and GOLPH3 in endothelial cells by the biotin-switch assay coupled with Western blot. In addition, we found that EMMPRIN and GOLPH3 are basally S-nitrosylated and that S-nitrosylation of these proteins increases in response to eNOS stimulation. The immunofluorescence results show that EMMPRIN, GOLPH3 and eNOS are co-localized at the Golgi apparatus in endothelial cells. Furthermore, EMMPRIN was co-immunoprecipitated with both eNOS and GOLPH3, suggesting that these three proteins may form a complex at the Golgi apparatus. Collectively, these findings indicate that eNOS-derived NO modulates S-nitrosylation of EMMPRIN and GOLPH3 in endothelial cells.
eNOS is localized on the cytoplasmic side of the Golgi apparatus [4][5][6]. To our knowledge, the membrane topology of EMMPRIN at the Golgi apparatus is not known. However, given that glycosylation sites of EMMPRIN are located only on the Nterminus [16,23], it is likely that the N-terminus is on the luminal side of the Golgi apparatus, while the C-terminus is on the cytoplasmic side. It is possible that eNOS-derived NO could Snitrosylate the N-terminus of EMMPRIN by diffusing across the Golgi membrane. GOLPH3 is located on the cytoplasmic surface of the Golgi apparatus by binding to a membrane lipid, phosphatidylinositol 4-phosphate [PtdIns(4)P] [19,24,25]. It is possible that a complex of eNOS, EMMPRIN and GOLPH3 is formed on the cytoplasmic side of the Golgi apparatus through the C-terminus of EMMPRIN.
The mass spectrometry-based analysis also identified putative cysteine residues for S-nitrosylation on these Golgi resident and Golgi/ER-associated proteins. For example, cysteine 87 of EMMPRIN is an S-nitrosylated site. This residue is also known to form a disulfide bond and is located within the extracellular domain that is also known to induce matrix metalloproteinase 2 (MMP-2) expression/activation [16][17][18]. Thus, it is possible that S-nitrosylation of this cysteine residue may influence MMP-2 induction. The functional importance of Cys 288, a site for GOLPH3 S-nitrosylation identified in this study, has not been reported. Investigations on the effect of S-nitrosylation on protein function in relation to its cysteine residue will be highly appreciated.
EMMPRIN S-nitrosylation may have implication for pathological vasculature. It is known that the aorta and other arteries of cirrhotic rats exhibit excessive vasodilation due to elevated eNOSderived NO production [26][27][28]. Further, increased NO level causes thinning of the aorta in cirrhotic rats, as an NOS inhibitor ameliorates aortic wall thickness [20]. These observations suggest that active vascular remodeling takes place in these arteries of cirrhotic animals and that increased eNOS-derived NO facilitates such arterial thinning in those animals. Since EMMPRIN is an inducer of MMPs and could be involved in vascular remodeling, we examined whether EMMPRIN is S-nitrosylated in the aorta isolated from cirrhotic rats. Interestingly, EMMPRIN S-nitrosylation was notably higher in the aorta of cirrhotic rats than that of normal rats. Importantly, only the glycosylated form of EMM-PRIN is S-nitrosylated. Given that the Golgi is the major site of glycosylation and that eNOS is localized on the Golgi, this finding also supports that the primary site of EMMPRIN S-nitrosylation is the Golgi apparatus. Although the effect of S-nitrosylation on EMMPRIN functions is still not clear, it is possible that eNOSderived NO enhances S-nitrosylation of mature EMMPRIN, induces MMP expression/activation and facilitates active arterial remodeling in cirrhotic animals.
Presently, the selectivity of S-nitrosylation of proteins in vivo is not completely understood. A recent study by Doulias et al. [29] has suggested that the selectivity of some proteins requires secondary structure elements and correct folding. As such glycosylation of EMMPRIN may be required for correct protein folding that will facilitate S-nitrosylation.
GOLPH3, also called GPP34, GMx33, MIDAS or yeast Vps74p, is known to be necessary for protein transport from the Golgi to the plasma membrane [19]. Using a temperature sensitive mutant protein, ts045-VSVG-EGFP, we previously demonstrated that eNOS localized to the Golgi apparatus delays protein transport from the Golgi apparatus to the plasma membrane, while a nuclear targeted form of eNOS does not affect the protein transport [7]. Thus, it is tempting to speculate that S-nitrosylation of GOLPH3 at the Golgi apparatus by eNOS-derived NO may  . S-nitrosylation of EMMPRIN is increased in the aorta isolated from cirrhotic rats. Aorta samples were lysed in a lysis buffer. Three aorta samples were combined per group to obtain a sufficient amount of proteins for the biotin-switch assay. Lysates before the biotin-switch assay were blotted for EMMPRIN and a loading control, heat shock protein 90 (Hsp90) (input, left panel). Equal amounts of proteins (500 mg) in the lysates were used for the biotin-switch assay. Biotinylated proteins were captured by streptavidin agarose beads and blotted with an EMMPRIN antibody (right panel). The aorta from cirrhotic rats showed a higher level of S-nitrosylated EMMPRIN than that of normal rats. Interestingly, only those glycosylated EMMPRIN were Snitrosylated in the aorta. doi:10.1371/journal.pone.0031564.g004 play a role in the delayed protein trafficking observed in cells expressing wild-type eNOS.
The other seven Golgi resident and Golgi/ER-associated proteins that we identified are also known for their importance in cellular regulation. ERGIC-53 (also known as p58) and transmembrane emp24 domain-containing protein 9 (TMED9) are involved in protein trafficking between the ER and the Golgi apparatus [30,31]. ERGIC-53 is an ER-Golgi intermediate compartment protein and serves as a molecular chaperone for the ER to the Golgi transport of a specific subset of secreted proteins such as coagulation factors V and VIII. Its defect causes a combined deficiency of coagulation factors V and VIII [31]. TMED9 is a member of p24 family [32]. It has been shown that p24 family members play an important role in retrograde transport from the Golgi to the ER by facilitating the formation of COPI-coated vesicles [30].
Giantin, a tail-anchored transmembrane protein with a large (350-380 kDa) N-terminal cytoplasmic region, is known to regulate Golgi structure. It is localized mostly at the rims of Golgi cisternae and influences diverse morphological changes of the Golgi by facilitating COPI vesicle formation and fusion [33,34]. Further, four additional proteins, namely alpha-6-fucosyltransferase, fukutin related protein, polypeptide N-acetylgalactosaminyltransferase II and D-glucuronic acid C5-epimerase, are known to have implication in protein glycosylation.
As evidence is reported that NO impairs ERGIC-53 transport from the ER to the Golgi in macrophages [35], it is possible that these and other functions of these proteins are regulated by Snitrosylation. Furthermore, besides the nine Golgi resident and Golgi/ER-associated proteins, we identified 69 potential S-nitrosylated proteins. They are considered cargo proteins and listed in Table S1. It is also possible that these cargo proteins are Snitrosylated at the Golgi apparatus and distributed to their designated cellular location, which may subsequently impact the functions of these proteins. Given that the Golgi is an important site for post-translational modifications such as phosphorylation, glycosylation and sulfation, it may also be important for S-nitrosylation.
In conclusion, we have identified potential S-nitrosylated Golgi proteins, which are important regulators of extracellular matrix remodeling, signal transduction, protein trafficking, glycosylation and maintenance of the Golgi structure. Importantly, we have verified S-nitrosylation of EMMPRIN and GOLPH3 in endothelial cells, two example proteins among those identified. We have also showed a high prevalence of S-nitrosylated EMMPRIN in the aorta of cirrhotic rats. These findings provide new insights into organelle specific S-nitrosylation of proteins regulated by eNOS-derived NO. Figure S1 S-nitrosylation of Golgi membrane proteins is increased in a dose-dependent manner. Tri-iodide chemiluminescence was used for the measurement of S-nitrosylated proteins formed in response to graded concentrations of NO donors, S-nitrosocysteine (Cys-NO) and diethylamine nitric oxide (DEANO). (TIF) Figure S2 EMMPRIN forms a complex with eNOS and GOLPH3 in endothelial cells. Bovine aortic endothelial cell (BAEC) lysates were immunoprecipitated using an anti-EMM-PRIN (EMP) antibody or control (Goat) IgG followed by western blot for indicated proteins. This is a repeat of the experiment shown in Figure 3C but using a salt-containing wash buffer (50 mM Tris-HCl, pH7.5, 0.1 mM EDTA/EGTA, and.150 mM NaCl). EMMPRIN was co-immunoprecipitated with both eNOS and GOLPH3, although the efficiency was slightly reduced for eNOS by the addition of 150 mM salt in the wash buffer. This result confirms our finding in Figure 3C that EMMPRIN forms a complex with eNOS and GOLPH3 in endothelial cells. (TIF) Figure S3 Inhibition of eNOS activity by N(G)-nitro-Larginine methyl ester (L-NAME) decreases the formation of S-nitrosocysteines in endothelial cells. Bovine aortic endothelial cells (BAECs) were incubated in cell culture media (Dulbecco's Modified Eagle's Media; DMEM) in the absence of fetal bovine serum (FBS) for 24 hours. Then 100 mM ATP was added to activate eNOS for 30 min (middle lane). Some cells were pretreated with 100 mM L-NAME in the absence of FBS for 24 hours before the addition of ATP (right lane). Cells were lysed, and 150 mg protein from each sample was analyzed by SDS-PAGE and western blotting using S-nitrosocysteine (rabbit, 1:500, Sigma, St. Louis, MO, Cat#: N5411) and eNOS (mouse, 1:1000, BD Biosciences, San Jose, CA, Cat#: 610296) antibodies. Proteins are S-nitrosylated at the base level without L-NAME and ATP treatment (left lane). The level of S-nitrosylated proteins is increased by ATP treatment (middle lane) but not when the cells are pretreated with L-NAME (right lane). Collectively, these results indicate that the increase in S-nitrosocysteine levels in BAECs is NOS activity dependent. (TIF)

Supporting Information
Table S1 List of all the putative S-nitrosylated proteins from Golgi membrane samples. We identified 78 putative Snitrosylated proteins from our Golgi membrane samples, nine of which were Golgi resident and Golgi/ER-associated proteins (Table 1), while the rests were considered proteins in transit through the Golgi apparatus. This proteomic analysis also allowed us to identify the site where S-nitrosylation occurs on the protein. (XLS)