Nitric oxide promotes endothelial cell survival signaling through S-nitrosylation and activation of dynamin-2

Tumor necrosis factor-alpha (TNF-α) constitutes a prominent apoptotic stimulus for endothelial cells (EC) in pathobiologic conditions (Zhang et al., 2005). Conversely, nitric oxide (NO) signals are known to mediate cell survival and resistance to apoptosis, a key component of EC-based angiogenesis (Zeng et al., 2001). The mechanisms that regulate this counterbalance of external apoptotic stimuli and survival signaling are not fully defined.

Pathways by which extracellular ligand binding to membrane receptors regulates downstream effectors of EC survival are susceptible to endocytosis-dependent regulation (Bhattacharya et al., 2005; Gliki et al., 2001; Pierce et al., 2000). In this regard, the dynamin family of large GTPase are recognized as drivers of both clathrin and caveolar vesicle internalization (Henley et al., 1998; Oh et al., 1998; Sontag et al., 1994). Thus, the relationship between dynamin and ensuing downstream signal transduction, which regulates the counterbalances of EC survival and apoptosis, may be significant. This is especially true in EC in which the dynamin-2 isoform is closely associated with key EC survival proteins, including the NO generating enzyme, endothelial NO synthase (eNOS), and the VEGF receptor, KDR (Bhattacharya et al., 2005; Chatterjee et al., 2003; Predescu et al., 2001).

Dynamin endocytic function is regulated through post-translational modifications including tyrosine phosphorylation (Ahn et al., 2002; Shajahan et al., 2004). NO also confers post-translational modifications on specific target proteins, especially eNOS binding partners, through S-nitrosylation of specific cysteine residues (Stamler et al., 2001). Since eNOS and dynamin interact closely within cells (Chatterjee et al., 2003; Icking et al., 2005; Wang et al., 2006), in the present study we explored the role of NO regulation of dynamin function as a membrane trafficking switch that regulates EC survival signaling.

Endocytic internalization of caveolar receptors and proteins is important for the propagation of several survival signaling pathways (Lua and Low, 2005; Pierce et al., 2000). Since the large GTPase dynamin-2 is essential for endocytosis and downstream signal transduction (Praefcke and McMahon, 2004; Shajahan et al., 2004), and TNF-α is a prototypical apoptotic ligand in EC (Zhang et al., 2005), we examined the effect of dynamin inhibition on TNF-α-induced apoptosis. Transfection of EC with the well-characterized GTPase mutant, K44A, in the form of an adenoviral construct (AdK44A), increased EC apoptosis in response to TNF-α as measured by levels of cleaved caspase-3 from cell lysates (Fig. 1A). As a complementary approach to ascertain the influence of dynamin inhibition on EC survival signaling, we also depleted dynamin from EC using a siRNA approach. Initial experiments evidenced the high transfection efficiency of siRNA in bovine aortic endothelial cells (BAEC) and specificity of the dynamin-2 siRNA (Fig. S1A,B in supplementary material). As shown in Fig. 1B, even in the absence of an apoptotic stimulus, dynamin-2 knockdown from EC resulted in prominent changes in EC morphology and increased EC death as assessed by caspase-3 activation and morphologic detection of apoptotic EC. These data indicate that dynamin promotes survival signaling in EC.

NO has been implicated in EC survival signaling through multiple mechanisms (Papapetropoulos et al., 2001; Rossig et al., 2000). Since dynamin and eNOS interact within EC (Chatterjee et al., 2003; Wang et al., 2006), we next explored the hypothesis that NO may reciprocally regulate dynamin function and thereby promote EC survival signals. First, to examine the effects of endogenous eNOS-derived NO on EC survival, we transfected EC with eNOS-siRNA or scrambled RNA and examined activated caspase-3 levels as a measure of apoptosis in the presence of TNF-α and/or the eNOS agonist, VEGF. Initial experiments evidenced the specificity and effectiveness of the eNOS-siRNA to reduce eNOS protein levels and activity in EC (see Fig. S1C in supplementary material). Although VEGF-stimulated NO production protected EC from TNF-α-induced apoptosis, this was not evident in eNOS-siRNA-transfected cells [Fig. 2A; representative western blot with triplicate samples (upper panel), densitometric quantification of pooled experiments (lower panel)]. Transfection of EC with eNOS-siRNA did not result in large levels of caspase-3 cleavage in the absence of an apoptotic stimlus such as TNF-α (data not shown). In a complementary approach, EC were incubated with TNF-α alone or in combination with VEGF and/or the NOS inhibitor L-NAME for 24 hours and analyzed for apoptosis by Hoechst 33342 stain. VEGF reduced TNF-α-induced apoptosis in an L-NAME-inhibitable manner (Fig. S2A in supplementary material). Next, as a complementary approach to assess the effects of NO, we used the NO donor sodium nitroprosside (SNP; 100 μM). SNP protected EC from TNF-α-induced apoptosis as evidenced by diminished levels of TNF-α-induced activated caspase-3 (Fig. 2B; representative western blot with triplicate samples (upper panel) and densitometric analysis (lower panel)]. This protective effect of SNP was observed at concentrations as low as 0.01 μM and was observed in response to an alternative NO donor, S-nitrosoglutathione (GSNO), as well (Fig. S2B in supplementary material). These studies confirm an important role for NO in EC survival signaling.

Next, we directly examined whether NO survival signals could be transduced through dynamin. EC were transfected with AdGFP or AdK44A and then incubated with TNF-α and/or SNP. Although SNP protected EC from TNF-α-induced apoptosis in cells transfected with AdGFP, this protection was not evident in EC transfected with AdK44A [Fig. 3A; representative western blot for cleaved caspase-3 levels (left panel) and densitometric analysis (right panel)]. AdK44A also blocked EC tube formation in response to endogenous NO generated by VEGF (Fig. 3B), demonstrating a physiologic relevance of dynamin-dependent EC survival signals in the context of angiogenesis.

To examine whether dynamin-dependent NO survival signaling occurs through the prototypical downstream cGMP-protein kinase G (PKG) pathway, cells were pretreated with equimolar concentrations of SNP (100 μM) or the cGMP analogue and PKG agonist 8-Br-cGMP, prior to incubation with TNF-α. Although SNP protected EC from TNF-α-induced apoptosis, no protection was afforded by 8-Br-cGMP even up to a concentration of 100 μM and in the presence of adenoviral overexpression of Flag-tagged PKG (AdPKG; Fig. 3C). As a positive control, 8-Br-cGMP did increase EC migration at this concentration, as previously noted by others (Kawasaki et al., 2003) (vehicle: 100% and 8-Br-cGMP: 283%; n=3), and the PKG construct increased PKG activity from cell lysates of transfected cells by 2.5-fold (relative activity/mg protein; AdGFP: 27.7 versus AdPKG: 69.3). Thus, NO, independent of the prototypical cGMP-PKG pathway, counterbalances TNF-α-induced apoptosis in EC in a dynamin-dependent manner.

Since inhibition of dynamin GTPase activity prevented NO-dependent EC survival signaling, we next directly examined whether NO promotes dynamin-dependent endocytosis. EC were preincubated with the NO donor GSNO (100 μM) or vehicle for 10 minutes and assayed for uptake of fluorescence-conjugated transferrin (Tfn), a marker of clathrin-dependent endocytosis, or lactosylceramide (LacCer), a marker of caveolae-dependent endocytosis (Henley et al., 1998; Oh et al., 1998). Fluorescent ligand uptake was analyzed quantitatively using a highly sensitive laser scanning cytometry (LSC) method and also measured qualitatively by confocal microscopy. Incubation of EC with GSNO increased Tfn (Fig. 4A) and LacCer (Fig. 4B) uptake as compared with vehicle (Tfn: control R1=6%, GSNO R1=15%. LacCer: control R1=13%, GSNO R1=55%). To directly test whether these effects of NO were mediated through activation of dynamin GTPase, EC were transfected with mock virus or AdK44A prior to GSNO incubation as the mutant lacks GTPase domain function, rendering it resistant to GTPase domain activation. AdK44A blocked Tfn uptake as previously noted (Altschuler et al., 1998) and more importantly, GSNO did not increase Tfn uptake in EC transfected with AdK44A (control R1=27%, K44A R1=17%, K44A+GSNO R1=18%) (Fig. 4C). These data demonstrate that NO promotes dynamin GTPase activity and endocytosis.

Dynamin endocytic function is regulated in part by its GTPase activity (Altschuler et al., 1998). To directly assess whether NO activates dynamin GTPase function, we pursued in vitro GTPase activity studies to measure the catalytic activity of purified recombinant dynamin in the presence and absence of GSNO. A representative experiment is shown in Fig. 4D, demonstrating that GSNO increases dynamin GTPase activity by 300% compared with dynamin alone (lane 4 versus lane 3). Similar results were obtained with an alternative NO donor, diethylamine NONOate (DEA-NO) (data not shown). These studies indicate that NO donors can enhance dynamin GTPase activity and promote endocytosis.

As NO-based nitrosylation influences vesicle exocytosis (Matsushita et al., 2003) and NO can act independently of the canonical cGMP-PKG survival pathway in EC, we next considered whether NO promotes dynamin function through nitrosylation. Indeed, while these studies were near completion, a paper was published demonstrating that the dynamin-1 isoform of the dynamin GTPase family is a target for nitrosylation by NO in a process that promotes internalization of pathogens into cells (Wang et al., 2006). To directly examine whether dynamin-2 is a nitrosylation target of NO, we used the biotin-switch method and UV-visible spectroscopy (Jaffrey et al., 2001). Nitrosylation of immunoprecipitated EC dynamin was detected in response to GSNO (100 μM), whereas nitrosylation was not detected in the absence of GSNO (Fig. 5A, left). The NO donor, DEA-NO (100 μM), also nitrosylated purified recombinant GST-dynamin as well as its GTPase domain (Fig. 5A, right). By contrast, biotin switch did not reveal prominent nitrosylation of the GST-PRD domain (see GST-PRD, bottom-right in Fig. 5A), providing evidence that GST nitrosylation was not confounding the biotin-switch assay. Dynamin nitrosylation was also increased in aortic tissues of mice genetically deficient in caveolin-1, which show a phenotype of increased basal NO generation (Razani et al., 2002), supporting the concept of NO-dependent dynamin nitrosylation in situ (Fig. 5B). Although cells from these mice are devoid of caveolae (Razani et al., 2002), prior work has demonstrated that caveolar proteins such as eNOS and dynamin may reside in non-caveolar compartments such as Golgi and lipid rafts (Cao et al., 2001; Shah et al., 1997; Sowa et al., 2001). Thus, dynamin may be exposed to eNOS-derived NO gradients within one of these subcellular compartments in Cav^-/- EC. We confirmed dynamin nitrosylation by UV-visible spectrophotometric analysis, an alternative method to detect nitrosothiol modification of proteins (Jaffrey et al., 2001). As seen in Fig. 5C, incubation of dynamin or the GTPase subdomain with DEA-NO revealed an absorbance peak at ∼320 nm, reflecting the light absorbance changes indicative of a nitrosothiol bond. However, the recombinant protein alone, in the absence of incubation with a NO donor, did not reveal an absorbance peak at 320 nm. Furthermore, no absorbance peak was detected with the NO donor alone, in the absence of recombinant protein (data not shown).

Finally, to directly demonstrate that NO survival signals could indeed promote EC survival through a dynamin-dependent pathway, we next sought to identify the specific cysteine residue that was responsible for the functional effects observed in response to dynamin nitrosylation. There are two cysteine residues in the dynamin-2 GTPase domain. We pursued a site-directed mutagenesis strategy in which each of these cysteines (C24 and C86) was changed to an alanine, a residue that is not susceptible to nitrosylation. Constructs were subcloned into a retroviral expression vector to allow for high-level overexpression in BAEC. Since NO promoted endocytosis in EC through a dynamin-dependent pathway, we first examined NO-induced endocytosis in BAEC transfected with the various mutant constructs. As seen in Fig. 6A, BAEC transfected with C86A had no significant increase in endocytosis in response to GSNO (control R1=10%, GSNO R1=12%). Conversely, NO-induced endocytosis was maintained in BAEC transfected with C24A (control R1=12%, GSNO R1=65%). We also examined the effects of the mutants in our model of TNF-α-induced apoptosis. Overexpression of C86A, but not C24A, abolished the protection conferred by NO donors in the face of TNF-α-induced apoptosis (Fig. 6B). A recent publication that emanated during the conduct of our studies revealed that nitrosylation of dynamin-1 at amino acid 607 in the PH domain regulated microbe internalization (Wang et al., 2006). Since this cysteine residue at amino acid 607 is conserved in dynamin-2, we mutated this residue to alanine as well to examine its effect on NO survival signaling. Interestingly, the C607A construct showed a phenotype similar to the C86A construct. For example, cells transfected with C607A showed neither GSNO-induced endocytosis (Fig. 6A; R1=11%, GSNO R1=12%) nor SNP-induced protection from TNF-α-induced apoptosis (Fig. 6B). To confirm that C86 and C607 of dynamin-2 could indeed both be nitrosylated by NO, we performed biotin-switch assays using purified recombinant protein constructs. Although dynamin nitrosylation was detected in both the C86A and C607A mutant proteins, nitrosylation was not detected in a mutant protein in which both C86 and C607 were changed to alanine (Fig. 6C). In summary, these studies indicate that both C86 and C607 are crucial for NO-mediated protection against TNF-α-induced apoptosis.

EC-based angiogenesis and vascular homeostasis are governed by a counterbalance of opposing EC survival signals and apoptosis signals. In this regard, the present studies provide novel observations that enhance our understanding of this process. We find that inhibition of dynamin GTPase function markedly perturbs EC survival signaling and that activation of dynamin promotes EC survival signals. Notably, we identify a mechanism by which NO activates dynamin through post-translational S-nitrosylation. Thus, these studies delineate a novel cross-talk between TNF-α-driven apoptosis signals and NO survival signaling that is regulated by dynamin nitrosylation, thereby uncovering a surprising and novel angiogenic capacity for dynamin GTPase in EC.

The role of NO in EC survival is controversial as NO promotes EC survival through numerous potential mechanisms but may also promote apoptosis through a mitochondrial pathway (Rossig et al., 2000). Adding to this complexity is the diversity in signaling pathways by which NO effects on EC survival are conferred (Kawasaki et al., 2003; Rossig et al., 2000). In the present study, NO survival signals did not transduce through the canonical cGMP-PKG pathway, and thus prompted us to examine alternative mechanisms of NO signaling. Increasing evidence indicates a key role for S-nitrosylation of NO target proteins, whereby NO reacts with specific cysteine thiols to form an S-nitrosylated derivative that regulates the biological activity of the target protein (Iwakiri et al., 2006; Stamler et al., 2001; Williams et al., 2003). Thus, NO signaling in EC appears to occur through complementary cGMP-dependent and -independent pathways. Indeed, in neuronal cells, recent studies have delineated a complex interplay of both cGMP-dependent and nitrosylation-dependent pathways by which NO promotes neuronal survival, thereby providing interesting parallels with NO signaling in EC (Ciani et al., 2002; Riccio et al., 2006). Interestingly, Ras GTPase, one of the most established nitrosylation targets of NO, demonstrates GTPase activation in response to nitrosylation, and maintains a GTPase domain architecture that is comparable to the dynamin GTPase domain (Williams et al., 2003). This prompted us to query and then establish, through the use of several complementary approaches, that dynamin-2 is indeed a NO nitrosylation target that can promote NO-dependent EC survival.

Since NO increased dynamin GTPase activity and endocytosis, we anticipated that cysteine nitrosylation may occur in the GTPase subdomain. Indeed, sequence analysis of the GTPase domain of dynamin-2 revealed two cysteines within the GTPase domain: C24 and C86. Although mutation of C24 did not influence NO activation of endocytosis and survival signaling, mutation of C86 prevented NO-dependent endocytosis and survival signaling. In fact, amino acids 85-87 of dynamin-2 (His85-Cys86-Lys87) constitutes a putative `nitrosylation motif' whereby a hydrophobic cysteine is juxtaposed with hydrophilic residues (histine and lysine). Extrapolation of this sequence to the crystal structure of the GTPase domain of dynamin (Protein Data Bank ID code 2AKA) (Reubold et al., 2005) reveals that this cysteine is exposed to the surface of the molecule, in close apposition to the GTP-binding pocket, and also within a proximity of 6 Å to aromatic amino acids (phenylalanines 83 and 91). This proximity to aromatic amino acids has been postulated to enhance cysteine-nitrosylation susceptibility based on computer modeling studies (Hao et al., 2006). Interestingly, in the course of our work, a paper was published by Wang et al. that demonstrated that NO nitrosylates the dynamin-1 isoform of the dynamin GTPase family (Wang et al., 2006). In that study, nitrosylation was demonstrated to occur at cysteine 607 within the PH domain of dynamin-1. Since this cysteine is conserved within the dynamin-2 isoform as well, we examined this residue as a possible NO nitrosylation target in dynamin-2. Similar to that observed with the C86 mutant, the C607 mutant also prevented NO endocytosis and survival signaling in EC. Thus, the current studies corroborate the recent work by Wang et al. (Wang et al., 2006) and mechanistically extend that work, by demonstrating an essential role for nitrosylation of both C86 and C607 of dynamin-2. These studies also extend the work of Wang et al. by demonstrating an important role for dynamin nitrosylation in the process of EC survival signaling and angiogenesis.

The present studies are of particular interest in relation to the role of dynamin in the prototypical VEGF pathway of signaling in EC (Papapetropoulos et al., 2001). Prior studies from our group and others have demonstrated that perturbation of dynamin function alters the localization and function of both eNOS and the VEGF receptor, KDR, with ensuing impairment of VEGF signaling and eNOS activation (Bhattacharya et al., 2005; Cao et al., 2001; Maniatis et al., 2006). The present studies extend these concepts by demonstrating that inhibition of dynamin function in EC potentiates EC apoptosis and antagonizes EC survival signaling. Interestingly, in epithelial-type cells which, unlike EC, do not utilize a VEGF-NO survival pathway, dynamin overexpression has been shown to inhibit cell survival signals (Fish et al., 2000), indicating a cell-type specificity of dynamin membrane trafficking functions.

Thus, the present studies highlight a novel mechanism by which NO survival signaling in EC is conferred. Furthermore, they identify dynamin, through its regulation of membrane traffic, as an EC-specific survival factor that regulates the coordinated balances of apoptosis, survival and ensuing EC-based angiogenesis.

BAEC (P2-6; Clonetics, San Diego, CA) were used in these studies. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), penicillin (100 IU/ml) and streptomycin (100 IU/ml). Pharmacologic reagents used in these studies included L-NAME (1 mM), 8-Br-cGMP (100 μM), VEGF-165 (10 ng/ml for migration and 50 ng/ml for survival studies) and TNF-α (30 ng/ml). Complementary NO donor compounds were used for various assays described below, at doses that did not exceed 100 μM, a concentration threshold commonly used for in vitro studies (Nisoli et al., 2005). Transfection of BAEC was performed with siRNA, adenoviral vectors or retroviral vectors. siRNA transfection was performed using Oligofectamine reagent (Invitrogen) with a final concentration of 100 nM. Functional and biochemical assays were performed between 48 and 72 hours after transfection. Replication-incompetent serotype 5 adenoviral vectors included a well-characterized GTPase domain point mutant, K44A, that functions as a dominant negative for GTP-dependent dynamin functions (Chatterjee et al., 2003) (AdK44A; University of Iowa Vector Core), wild-type PKG (AdPKG; K. Bloch, Massachusetts General Hospital, Boston, MA) and green fluorescent protein (GFP) (AdGFP; Bill Sessa, Yale University). For adenoviral transduction, BAEC were transfected with PBS/0.1% albumin containing 50 MOI (multiplicity of infection) of adenovirus for 1 hour. The vector solution was then aspirated and, after washing with PBS, complete culture medium was replenished. Transfection efficiency using this approach was at almost 100%. The retroviruses were produced from a 293T packaging cell line (Zeng et al., 2002). In brief, 293T cells were grown to 30-40% confluency and cotransfected with pMD.MLV gag.pol, pMD.G and pMMP vectors containing the cDNA sequence of interest. Cell culture supernatant containing high-titer viruses were collected after 48 hours. For transduction of EC, cells were grown to 40-50% confluency and retroviral stock was added to the culture medium at a ratio of 1:4 with polybrene 4-8 μg/ml. Cells were utilized for functional and biochemical assays after 24 hours.

Dynamin-2 constructs (provided courtesy of Mark McNiven, Mayo Clinic, Rochester, MN) were subcloned into the GST fusion protein vector pGEX-1. GST constructs were transformed into BL21 (DE3) and purified as we described previously (Cao et al., 2001). Specificity and quality of GST-dynamin-2 constructs were assessed by Coomassie Blue staining. Point mutations of cysteine 24, 86 and 607 to alanine were generated by PCR (QuikChange; Stratagene) using three pairs of primers; 5′-CGGTCAGAGCGCCCACCTGGACC-3′, 5′-GTTTTTGCACGCCAAGTCCAAAA-3′ and 5′-CGAACTGGCTGCCGACTCCCAGG-3′, respectively. Constructs were subcloned into the pMMP retroviral vector or pGEX-1 GST vectors using standard molecular biology approaches. All constructs were verified by sequencing.

Recombinant dynamin protein and subdomains were purified from Escherichia coli as previously described (Cao et al., 2001). In brief, dynamin, its subdomains or mutants in the GST plasmid were transformed into BL21 (DE3), induced with isopropyl β-D-thiogalactopyranosid (IPTG) (1 mM) overnight, and lysed by sonication with lysozyme. Samples were resonicated after addition of Triton X-100 to a final concentration of 1%. Cell debris was removed by centrifugation and supernatant was mixed with glutathione sepharose beads and agitated for 2 hours at 4°C. Samples were centrifugated and pellets washed three times in PBS with 1% Triton X-100. Specificity and quality of GST fusion proteins was assessed by Coomassie Blue staining of SDS-PAGE gels.

Dynamin (0.5 μM) was incubated with or without the NO donor GSNO (100 μM) for 30-60 minutes at room temperature in a buffer (135 mM NaCl, 5 mM KCl, 20 mM HEPES, pH 7.4, 1 mM MgCl2) as previously described (Ahn et al., 2002). Reactions (20 μl) were initiated by the addition of 10 μCi/ml [α-³²P] GTP. GTP hydrolysis was analyzed by thin-layer chromatography on PEI cellulose (Sigma), and results were quantified using phosphorImager and ImageQuant software (Molecular Dynamics, Sunnyvale, CA). Quantitation of cold GTP and GDP at each time point was performed on a PhosphorImager (Molecular Dynamics).

Recombinant proteins were incubated with or without the NO donor DEA NONOate (DEA-NO; 50 μM) for 20-60 minutes and passed through a Sephadex G-60 column to remove the NO donor. The S-nitrosylation characteristic absorbance peak at 320-340 was measured by 200-450 nm UV-visible wavelength scanning (Jaffrey et al., 2001).

Recombinant proteins were diluted or mouse aorta/treated cells were lysed in HEN buffer [250 mM HEPES, pH 7.7, 1 mM EDTA, 0.1 mM neocuproine, 1% Nonidet P-40, 150 mM NaCl, 1 mM PMSF, protease inhibitor mixture (Sigma)] as we and others have previously done (Jaffrey et al., 2001; Perri et al., 2006). GSNO (100 μM) or DEA-NO (50 μM) was added to the protein solution or cell lysates for 20 minutes in the dark. The solution or lysates were then mixed with an equal volume of methyl methanethiosulfonate (MMTS) buffer (25 mM HEPES, pH 7.7, 0.1 mM EDTA, 10 μM neocuproine, 5% SDS, 20 mM MMTS) and incubated at 50°C for 20 minutes with frequent vortexing. After free MMTS was removed by acetone precipitation, the precipitates were resuspended in HENS buffer (25 mM HEPES, pH 7.7, 0.1 mM EDTA, 10 μM neocuproine, 1% SDS). After adding 50 mM ascorbate in the mixture, to reduce non-modified SNO, the S-nitrosylated proteins were then modified with biotin in HPDP buffer (25 mM HEPES, pH 7.7, 0.1 mM EDTA, 1% SDS, 10 μM neocuproine, 10 mM ascorbic acid sodium salt and 0.2 mM N-[6-(biotinamido)hexyl]-3′-(2′-pyridyldithio) propionamide (biotin-HPDP; Pierce). Mixture was vortexed and allowed to react at room temperature for 1 hour in the dark. The samples were then added to Laemmli buffer without a reducing reagent and separated by SDS-PAGE followed by immunoblotting with anti-biotin antibody (Cell Signaling). In some experiments, cells were subjected to NO donor and, after cell lysates were immunoprecipitated with anti-dynamin antibody (Transduction Laboratories), the immunoprecipitates were blocked with MMTS and subsequent modification with biotin-HPDP. In some experiments, to establish specificity of nitrosylation, cells were incubated with ascorbic acid (2.5 mM) in conjunction with the NO donor.

Tube formation assays were performed as we have previously described (Gulati et al., 2003). In brief, four slide chambers (Lab-Tek Chamber Slides, Nunc) were coated with 150 μl Matrigel per chamber (10 ng/ml, BD Biosciences) and incubated at 37°C for 30 minutes to promote gelling. Twenty-four hours after transduction with AdGFP or AdK44A, BAEC [1×10⁵ cells in 1000 μl of DMEM medium containing 1% of basal endothelial growth supplement (ECGS, ScienCell)] were seeded in each of the Matrigel-coated wells. VEGF or vehicle was added to the Matrigel-coated wells and, after 12 hours of incubation at 37°C in a 5% CO₂ humidified atmosphere, the center of each well was photographed (three consecutive pictures) with a light microscope. The length of the tubes was measured at 100× magnification with Image-Pro Plus (Media Cybernetics, version 5.1). All assays were done in three independent experiments.

After transfection with the appropriate adenovirus and/or siRNA and incubation with test compounds, BAEC were collected and lysed in a buffer (50 mM Tris, 01 mM EDTA, 0.1 mM EGTA, 0.1%SDS, 1% NP-40, 0.1% deoxycholic acid, pH 7.5) (Cao et al., 2001). Lysates were prepared for measurement of caspase-3 after 24 hours of incubation with test compounds. Protein lysates were prepared and quantified using Bio-Rad protein assay. Equal amounts of protein (30 μg/well) were loaded and separated by SDS-PAGE and transferred to PVDF membranes. After blocking, blots were incubated with primary antibody (GAPDH mAb, Ambion; eNOS mAb, Transduction Laboratories; caspase-3 Ab, Cell Signaling; β-actin Ab, Sigma; Flag M2 Ab, Sigma; dynamin pAb, Transduction Laboratories) for 2 hours at room temperature or overnight at 4°C, incubated with appropriate secondary antibody, and protein-band signals detected with ECL (Amersham Biosciences). Quantification was performed by densitometric analysis of band density using the software ImageJ (NIH).

For imaging of fluorescent-labeled molecule uptake, cells were preincubated with the NO donor GSNO (100 μM) or vehicle for 10 minutes and then incubated with 5 μg/ml Tfn Alexa-Fluor 488 or 0.5-2.5 μM BODIPY C5-lactosylceramide (LacCer). After a series of washing steps to remove residual fluorescent lipid remaining at the plasma membrane, cells were fixed with 1% paraformaldehyde, stained with Hoechst and mounted for analysis. For apoptosis detection, 0.3×10⁶ BAEC were seeded into each well of six-well plates. Cells were transfected with appropriate adenovirus or siRNA and/or treated with test compounds at concentrations as indicated. Floating cells together with the remaining attached cells were collected and fixed with 4% formaldehyde. Cell pellets were suspended in Hoechst 33342 (10 μg/ml) and apoptotic cells were identified by conventional fluorescence microscopy (5100TV; Zeiss, Germany) based on the presence of chromatin condensation and nuclear fragmentation. Each experimental condition was examined for apoptosis in triplicate with 1000 cells from four different high-power fields. Each experiment was performed on three independent occasions.

Quantification of endocytosis was performed using an inverted-format laser scanning cytometer (LSC; iCyte-Compucyte, Cambridge, MA). This system has a large field of depth, so cells are illuminated uniformly throughout their volume, which allows accurate whole-cell immunofluorescence intensity measurements with objective and quantitative analysis (Dmitrieva et al., 2004). Fluorescent-labeled Tfn uptake and BODIPY-LacCer internalization were measured to assess clathrin- and caveolae-mediated endocytosis, respectively (Henley et al., 1998; Oh et al., 1998). In brief, after incubation of cells with NO donor and fluorescent compounds as described above, two to three areas (5×5 mm²) were chosen in each chamber for fluorescence intensity measurement. Cells were recognized by LSC by nuclear Hoechst stain and cytoplasmic green fluorescence was counted per cell and plotted as a histogram as previously described (Ozawa et al., 2005). Green fluorescence intensity cut lines were calculated between the groups to quantitatively analyze endocytosis of the fluorescent compounds. Histograms were quantitated as the percentage of cells with fluorescent intensity above the specified cut-point (R1), with the same cutpoint applied to all samples within a given experiment.

Statistical analyses were performed using Student's t-test. P<0.05 was considered as statistically significant.

The authors thank Martin Fernando-Zapico and Raul Urrutia for critical review of the manuscript, Mark McNiven for providing dynamin subdomain constructs and Resham Bhattacharya and Yehia Daaka for helpful discussions. These studies were supported by R01 DK59615, R01 DK59388 and P50CA102701 Pancreatic Cancer SPORE Development Grant (V.S.) and an American Heart Association Scientist Development Grant (AHA0435063N to S.C.).