Endothelial nitric oxide synthase overexpression by neuronal cells in neurodegeneration: a link between inflammation and neuroprotection


Address correspondence and reprint requests to Emilio Clementi, Department of Preclinical Sciences, University of Milano, via GB Grassi 74, 20157 Milano, Italy. E-mail: emilio.clementi@unimi.it; and Maria Teresa Bassi, E. Medea Scientific Institute, via Don Luigi Monza 20, 23842 Bosisio Parini (Lc), Italy. E-mail: mariateresa.bassi@bp.inf.it


The roles of neuronal and inducible nitric oxide synthases in neurones have been extensively investigated; by contrast, the biological significance of endothelial nitric oxide synthase (eNOS) overexpression that occurs in several pathological conditions has not yet been studied. We have started addressing this issue in a cell model of neurodegeneration, i.e. human SKNBE neuroblastoma cells transfected with a mutant form of alsin, a protein causing an early-onset type of amyotrophic lateral sclerosis, ALS2. We found that eNOS, which is endogenously expressed by these cells, was activated by tumour necrosis factor-α (TNF-α), a proinflammatory cytokine that plays important roles in ALS2 and several neurodegenerative diseases. The TNF-α-dependent eNOS activation occurred through generation, by sphingosine-kinase-1, of sphingosine-1-phosphate, stimulation of its membrane receptors and activation of Akt, as determined using small interference RNA and dominant negative constructs specific for the enzymes and receptors. eNOS activation by TNF-α conferred cytoprotection from excitotoxicity and neurotoxic cues such as reactive oxygen species, endoplasmic reticulum stress, DNA damage, and mutated alsin itself. Our results suggest that overexpression of eNOS by neurones is a broad-range protective mechanism activated during damage and establish a link of pathophysiological relevance between this enzyme and inflammation accompanying neurodegenerative diseases. These findings also question the concept that high NO output in the presence of oxidative stress leads always to peroxynitrite formation contributing to neurodegeneration.

Abbreviations used



amyotrophic lateral sclerosis

alsinWT and alsinMT

alsin wild-type or mutated in p.G540E


guanosine 3′5′-cyclic monophosphate


endothelial NO synthase


inducible NO synthase


Nω-nitro-L-arginine methyl ester




neuronal NO synthase


nitric oxide




phosphate buffered saline


pertussis toxin


sphingosine 1 phosphate


small interfering RNA


sphingosine kinase 1


SK1 G82D mutant


tumour necrosis factor-α

Nitric oxide (NO), a diffusible messenger of many forms of intercellular communication and intracellular signalling, regulates crucial physiological processes in the nervous system such as learning and memory, neuronal survival and differentiation (Calabrese et al. 2007). In addition, it regulates synapse formation and patterning, thus playing a role in embryonal and adult neurogenesis and development (Gibbs 2003; Estrada and Murillo-Carretero 2005). NO may contribute also to neuronal damage when generated at high concentrations (Bo et al. 1994; Vodovotz et al. 1996; Iravani et al. 2002; Ebadi and Sharma 2003; Moncada and Bolanos 2006).

Nitric oxide is generated following activation of three isoforms of NO synthases (NOS), the endothelial (eNOS) neuronal (nNOS) and inducible (iNOS) isoforms, that differ from each other in terms of intracellular localization, activation properties and sensitivity to regulation by protein interactions and second messenger molecules (Alderton et al. 2001); increasing evidence indicates that NO plays different biological functions depending on its generating enzyme (Kone et al. 2003).

Neuronal NO synthase is constitutively expressed in neurones and astrocytes and associated more often with physiological actions of NO, while iNOS is induced by pro-inflammatory conditions in microglia, astrocytes and neurones, and mediates generation of NO in pathophysiological conditions (Jaffrey and Snyder 1995; Heneka and Feinstein 2001).

The role of eNOS in the nervous system has been studied only to a limited extent and investigations have concentrated mainly on its role as a regulator of cerebral blood flow in synaptic plasticity, memory, neurogenesis and behaviour (Wilson et al. 1999; Reif et al. 2004; Chen et al. 2005; Garthwaite et al. 2006; Pereira de Vasconcelos et al. 2006). Yet, eNOS may be expressed also by neuronal cells: high levels of the enzyme have been consistently reported in specific neuronal populations under pathological conditions as diverse as peripheral nerve injury, brain ischemia and several neurodegenerative diseases, including amyotrophic lateral sclerosis (ALS) Alzheimer’s and Parkinson’s diseases (Sohn et al. 1999; Levy et al. 2000; de la Monte et al. 2000; Leker et al. 2001; Kashiwado et al. 2002). The biological significance of the enhanced expression of eNOS in pathology has not been studied.

We have addressed this issue in a cell model that reproduces cellular events occurring during neurodegeneration, i.e. human SKNBE neuroblastoma cells transfected with a mutant form (p.G540E) of alsin (alsinMT), a protein involved in early onset ALS, juvenile primary lateral sclerosis and in infantile forms of ascending spastic paralysis. Alsin is implicated in endosome–membrane trafficking and fusion, with roles in promoting neurite outgrowth in neuronal cultures (Hadano et al. 2007) and in neuroprotection against cytotoxicity (Hadano et al. 2001; Panzeri et al. 2006). We found that eNOS, which is expressed endogenously by these cells, was activated by tumour necrosis factor-α (TNF-α), a key proinflammatory cytokine in ALS and other neurodegenerative diseases (Moreau et al. 2005), in a pathway involving generation of sphingosine-1-phosphate (S1P) and activation of Akt. eNOS activation by TNF-α protected neuronal cells against a variety of neurotoxic stimuli. These results start unravelling the role of neuronally-expressed eNOS, and establish for the first time a link of pathophysiological relevance between expression of this enzyme and the inflammatory response accompanying neuronal damage in neurodegenerative diseases.

Materials and methods

Cell culture and transfection

SKNBE human neuroblastoma cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% foetal bovine serum (Hyclone Celbio, Milan Italy), L-glutamine (2 mM), penicillin (50 UI/mL), streptomycin (50 μg/mL) at 37°C under a humidified 95% to 5% (vol/vol) mixture of air and CO2. Expression of wild-type alsin (alsinWT) and the p.G540E alsinMT was obtained by transfection with pcDNA3MyEGFP-ALS2wt and pcDNA3MyEGFP-ALS2mutG540E vectors, whose preparation has been described in detail elsewhere (Panzeri et al. 2006). To silence eNOS and nNOS we took advantage of the small interfering RNA (siRNA) technique using specific siRNA duplexes targetting the transcripts for these two enzymes, using the following sequences 5′-CCC CCA AGA CCU ACG UGC Att-3′ for eNOS and 5′-UGG GCG GCC CUU GGU GGA Ctt-3′ for nNOS. The scrambled siRNA selected showed no homology to any human DNA sequence based on a BLAST search. To inhibit human sphingosine kinase 1 (SK1) activity we used the well characterized dominant-negative SKG82D mutant (SK1-DN) carrying a FLAG-epitope (DYKDDDDK), subcloned into the pcDNA3 vector (De Palma et al. 2006). Cell transfections with the various constructs, siRNAs, SK1-DN, and the respective controls (i.e. scrambled siRNA, empty pcDNA3 vector) were carried out in cells plated in six well dishes at 90% confluency 24 h prior to their use in the various experimental settings. Transfections were carried out using Lipofectamine 2000 (Invitrogen-Life Science, Carlsbad, CA, USA) according to manufacturer’s instructions. Transfection efficiency was more than 87% and did not vary significantly among the various preparations. The scrambled siRNA showed no homology to any human DNA sequence based on a BLAST search.

Cell death induction and analysis

Cell death was induced by administration of NMDA (N-Methyl-D-Aspartic-Acid), thaspigargin, tunicamycin (Xu et al. 2005; Deniaud et al. 2007), etoposide (Kruman et al. 2004), As2O3 or H2O2 (Sciorati et al. 2006) in the presence or absence of pertussis toxin (PTx), 8 Bromo guanosine 3′5′-cyclic monophosphate (8 Br-cGMP) H-[1,2,4]oxadiazolo[4,3-α]quinoxalin-1-one (ODQ) (all from Sigma), TNF-α or KT5823 (Alexis Italia, Florence, Italy). Cells were detached and stained with fluorescein isothiocyanate-annexin V and 7 amino actinomycin D (BD Bioscience, San Jose, CA, USA) according to the kit’s manufacturer’s instructions and analysed by flow cytometry in a FACStar Plus (Becton Dickinson, Sunnyvale, CA, USA) as described (Steensma et al. 2003; Sciorati et al. 2006). Cells showing staining for annexin V and excluding actinomycin D were considered apoptotic. Subcellular compartmentalisation of Bax was assessed by cell fractionation experiments as described (Panzeri et al. 2006). In brief, cells were scraped in 0.3 M sucrose, 10 mM Mes K+, 1 mM K2EGTA, 1 mM MgSO4, 1 mM dithiothreitol, pH 6.5, supplemented with a protease inhibitors cocktail (Complete Roche Diagnostics, Rotkreuz, Switzerland) and homogenised in the same buffer with 10 strokes using a syringe with a 27-gauge needle. Homogenates were centrifuged at 1200 g for 5 min at 4°C, and the post-nuclear supernatant was centrifuged at 10 000 g for 30 min at 4°C to separate the cytosolic (supernatant) and mitochondrial (pellet) fractions. Protein content in the fractions was assessed by the bicinchoninic acid procedure (Perbio, Bezons, France). Equal amounts of proteins for each fraction were then analysed by Western blotting as described in the section below.

Protein expression by western blot analysis

For western blot analyses cells were washed in phosphate buffered saline (PBS), lysed in a buffer containing 150 mM NaCl, 15 mM MgCl2, 1 mM EGTA, 50 mM HEPES-KOH, 10% glycerol, 1% Triton X-100, pH 7.5, supplemented with a protease inhibitors cocktail (Complete Roche Diagnostics). Fifty μg of lysates were then subjected to 10% sodium-sodecyl-sulfate polyacrilamide gel electrophoresis as described (Bulotta et al. 2001). The separated proteins were electroforetically trasferred to 0.2 μm-pore nitrocellulose membranes (Schleicher & Schuell, Dassel, Germany) at constant 250 mA for 2 h and the relevant antigens revealed using polyclonal antibodies (Abs) anti phosphorylated eNOS or phosphorylated Akt (Transduction Laboratories, Lexington, KY), Bax or Bcl-xL (Cell Signalling Technology, MA, USA) or monoclonal Abs against Akt, eNOS, iNOS nNOS, (Transduction Laboratories) or glyceraldehyde phosphate dehydrogenase (GAPDH; Biogenesis, Poole, UK). After incubation with appropriate secondary Abs, blots were developed with the enhanced chemiluminescence procedure (ECL-Amersham Bioscience, Little Chalfont, UK) (Bulotta et al. 2001).

Assay of NOS activity

Nitric oxide synthase activity was assayed in intact cells by measuring the conversion of L-[3H] arginine (Amersham Bioscience) into L-[3H] citrulline. SKNBE cell monolayers transfected with the various constructs as described in the Results were washed and then incubated for 20 min at 37°C in a reaction buffer containing: 145 mM NaCl, 5 mM KCl, 1 mM MgSO4, 10 mM glucose, 1 mM CaCl2 and 10 mM HEPES, pH 7.4, with or without Nω-nitro-L-arginine methyl ester (L-NAME) or PTx. At the end of the pre-incubation, 2.5 μCi/mL of L-[3H] arginine was added 5 min before cell stimulation with TNF-α. Non-stimulated cells were run in parallel. At the time point indicated in the results the monolayers were washed with 2 mL of ice-cold PBS, pH 7.4, supplemented with 5 mM L-arginine and 4 mM EDTA. 0.5 mL of 100% cold ethanol was added to the dishes and left to evaporate before a final addition of 2 mL of 20 mM HEPES, pH 6.0. Separation of L-[3H] citrulline from L-[3H] arginine was obtained by DOWEX 50X8-400 chromatography (Sigma) as described (Bulotta et al. 2001).

Measurement of cGMP formation

SKNBE cells expressing the various constructs were detached by trypsinisation and samples of suspended 2 × 106 cells were treated for 15 min at 37°C in the presence or absence of TNF-α (100 ng/mL) in 1 mL of a solution containing: 125 mM NaCl, 5 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 2 mM CaCl2, 6 mM glucose, 1 mM L-arginine, 0.6 mM isobuthylmethylxanthine and 25 mM HEPES, pH 7.4, with or without L-NAME (1 mM). The reaction was terminated by addition of ice-cold trichloroacetic acid (final concentration 6%). After ether extraction cGMP levels were measured using a radioimmunoassay kit (NENTM, Boston, MA) and normalised to cellular proteins as described above.

Measurement of SK activity

Cell monolayers transfected with the SK1-DN or the mock vector were challenged at 37°C with or without TNF-α (100 ng/mL), washed twice in cold PBS, scraped and lysed with a Dounce homogenizer (40 strokes) in a buffer containing 20 mM Tris-HCl, 1 mM β-mercaptoethanol, 1 mM EDTA, 1 mM sodium orthovanadate, 40 mM β-glycerophosphate, 15 mM NaF, 10 μg/mL leupeptin, aprotinin and soybean trypsin inhibitor, 1 mM phenylmethylsulfonyl fluoride, and 0.5 mM 4-deoxypyridoxine, pH 7.4. Lysates were centrifuged at 10 000 g for 10 min at 4°C. To determine SK activity equal amounts of protein (60 μg) from cell extracts were incubated in an assay buffer containing 20 mM Tris–HCl (pH 7.4), 40 mM β-glycerophosphate, 15 mM NaF, 5% glycerol, 0.5 mM 4-deoxypyridoxine and protease inhibitors, in the presence of 20 μM [3H] sphingosine (23 Ci/mmol), dissolved in 5% Triton X-100 (final concentration 0.25%), and 1 mM ATP containing 5 mM MgCl2. Reactions were initiated by addition of ATP and incubated for 4 h at 37°C. The formed [3H] S1P in the organic phase was separated by thin layer chromatography on silica gel using CH3(CH2)3OH/CH3COOH/H2O (3 : 1 : 1, vol : vol : vol) as previously described (De Palma et al. 2006). The radioactive spots corresponding to authentic S1P were identified using internal standard and quantified by scraping from the plates and counting the radioactivity in a Beckman LS1000 β-counter.

Statistical analysis

The results are expressed as means ± standard error of the mean (SEM); n represents the number of individual experiments. Statistical analysis was carried out using the Student’s t test for unpaired variables (two-tailed). Statistical probability values (p) of less than 0.05 were considered significant. Single, double or triple crosses and asterisks in the figures panels refer to p of < 0.05, < 0.01 and < 0.001, respectively, measured in the various experimental conditions as detailed in each figure legend.


eNOS is activated by TNF-α in SKNBE neuroblastoma cells

Wild-type SKNBE cells express nNOS and eNOS but not iNOS, as detected using Abs recognising each specific NOS isoform by Western blotting (Fig. 1a). Cell transfection with alsinWT or alsinMT did not change the expression levels of nNOS or eNOS, nor it induced expression of iNOS. eNOS activation by TNF-α in endothelial cells mediates relevant actions of this cytokine (De Palma et al. 2006). In addition, TNF-α plays key roles in ALS and other degenerative diseases. We thus investigated whether TNF-α triggered eNOS activation in neuronal cells and the biological role of such an activation. To investigate effects on eNOS effectively and specifically we took advantage of a well-characterised technique, i.e. silencing of enzyme expression using specific siRNAs. This technique was optimised for the SKNBE cell system so that siRNAs specific for either eNOS or nNOS reduced the expression of each target enzyme selectively without affecting the expression of the other isoform (Fig. 1b). Transfection with a scrambled siRNA, used as control of specificity, did not affect expression of either NOS isoform (Fig. 1b).

Figure 1.

 TNF-α activates eNOS in neuronal cells. (a, b) Expression of eNOS, iNOS and nNOS in or SKNBE, wild-type (WT SKNBE) or transfected with alsinWT or alsinMT (a), or after transient transfection with eNOS, nNOS or scrambled siRNAs (b). Shown are the results of one out of three reproducible experiments; GAPDH was revealed in parallel as a loading control. The graph in (b) shows the densitometric values ± SEM (= 3). (c) NOS activity stimulated by TNF-α (100 ng/mL) in SKNBE treated with L-NAME (1 mM) or transfected with the eNOS siRNA, the nNOS siRNAs or the scramble, measured as L-citrulline formation over time (= 4). Asterisks in (b) and (c) indicate statistical significance versus scramble, calculated as described in the Materials and Methods section.

We examined the time-course of NOS activation induced by TNF-α (100 ng/mL) in cells transfected with the eNOS specific siRNA, the nNOS specific siRNA, the scrambled control, results were compared to those obtained using the broad NOS inhibitor L-NAME (1 mM). TNF-α triggered NOS activation, which was time-dependent with a peak of activity after 30 min (Fig. 1c). This pattern of activation was consistent with the time-course of activation of the enzyme by TNF-α in endothelial and epithelial cells (Bulotta et al. 2001; De Palma et al. 2006). NOS activation was significantly inhibited in SKNBE cells expressing the eNOS specific siRNA, in a way similar to that observed with L-NAME, but not in cells transfected with the scramble or transfected with the nNOS-specific siRNA. These results indicate that TNF-α activates eNOS in SKNBE cells.

To investigate whether eNOS activation led to NO generation we measured formation of cGMP, a good proxy for NO since soluble guanylate cyclase is activated by nanomolar concentrations of the gas (Garthwaite and Boulton 1995). Exposure of scramble-transfected SKNBE cells to TNF-α for 30 min induced increases in cGMP levels over non-stimulated controls. cGMP values before and after stimulation with TNF-α were 0.36 ± 0.02 and 2.45 ± 0.2 pmol/mg/min (= 4). Such cGMP generation was prevented by transfection with the eNOS-specific siRNA (values before and after stimulation with TNF-α were 0.34 ± 0.02 and 0.68 ± 0.2 pmol/mg/min, = 4) in a way similar to that obtained using L-NAME (not shown). Transfection with the nNOS-specific siRNA, used as an internal control, yielded cGMP values of 0.33 ± 0.01 and 2.28 ± 0.1 pmol/mg/min before and after stimulation with TNF-α (= 4). Thus the nNOS-specific siRNA did not inhibit the effect of TNF-α on cGMP formation. These results indicate that TNF-α activates eNOS and generates bioactive NO in SKNBE cells.

Activation of eNOS by TNF-α confers protection against neuronal damage induced by alsinMT and excitotoxicity

A prominent feature of ALS is death of motor neurones, essentially via apoptosis (Nixon 2005). In a previous study (Panzeri et al. 2006) we demonstrated that alsinMT is cytotoxic for neuronal cells because it induces apoptosis per se. In addition, alsinMT increases apoptosis induction by NMDA that activates specific glutamate receptors involved in excitotoxicity (Hardingham and Bading 2003; Nixon 2005). We also found that apoptosis induction by alsinMT was accompanied by specific changes in the expression and localization of members of the Bcl-2 family of proteins. In particular, alsinMT increased the expression and translocation from the cytosol to the mitochondria of proapoptotic Bax and decreased the expression of antiapoptotic Bcl-xL (Panzeri et al. 2006).

We decided to investigate the role of eNOS activation and NO generation by TNF-α on the pathway of alsinMT-induced apoptosis. Both basal and NMDA-stimulated apoptosis increased in SKNBE cells transfected with alsinMT, as shown by the increased phosphatidylserine exposure on the outer leaflet of the plasma membrane, detected using Annexin V, in cells excluding 7 amino actinomycin D (Steensma et al. 2003); by contrast, alsinWT conferred cytoprotection (Fig. 2a; see also Panzeri et al. 2006). Treatment with TNF-α inhibited significantly apoptosis triggered by NMDA in both alsinWT- and alsinMT-transfected cells (Fig. 2a). In addition, TNF-α inhibited significantly basal apoptosis of alsinMT-expressing cells. When cells were co-transfected also with the eNOS specific siRNA the antiapoptotic effects of TNF-α was no longer observed (Fig. 2a).

Figure 2.

 eNOS-activation by TNF-α in neuronal cells is cytoprotective. SK-N-BE cells were transiently transfected with alsinWT (AlsinWT-GFP), or alsinMT (AlsinMT-GFP), both GFP-tagged at their N-terminus, or the vector carrying only GFP (pcDNA3-GFP), used as control. When indicated eNOS siRNA was also co-transfected. Apoptosis was triggered by an 8 h administration of NMDA (1 mM) in the presence/absence of TNF-α (100 ng/mL) and compared to that of the control treated with the scrambled siRNA (sc). (a) Apoptosis, assessed by measuring in flow cytometry the % of cells staining positive for annexin V and excluding actinomycin D (7AAD). (b and c) Western blot analysis of Bax and Bcl-xL expression (b) and of the NMDA-induced Bax translocation from the cytosol (C) to the mitochondria (M) (c). On the same nitrocellulose filters were routinely carried out the evaluation of GAPDH and cytochrome c oxidase subunit IV (COX-IV), used as controls of purity of the cytosolic and mitochondrial fractions, respectively. Shown are both representative images and graphs, the latter reporting the ratio of densitometric values ± SEM of Bax in the cytosol and mitochondria versus those of GAPDH (= 4), expressed as %. In all experiments, the expression level of the alsinWT and alsinMT in each sample was routinely assessed by staining with an anti GFP Ab, and found not to differ significantly among the various preparations. Asterisks indicate statistical significance versus sc; crosses indicate statistical probability versus sc + NMDA, calculated as described in the Materials and Methods section.

The expression of Bax and its translocation to the mitochondria were increased in cells transfected with alsinMT, and both parameters were further enhanced by NMDA treatment (Fig. 2b and c). TNF-α inhibited both these effects of alsinMT expression. Consistently, TNF-α prevented the reduction in Bcl-xL levels induced by alsinMT both in the presence and absence of NMDA (Fig. 2c). Of importance, TNF-α was ineffective in cells also expressing the eNOS specific siRNA (Fig. 2b and c). When TNF-α was administered alone to SKNBE cells it did not have significant cytotoxic effects even at very high (1 mM) concentrations (not shown). Taken together these results indicate that TNF-α acts as a neuroprotective agent against alsinMT- and NMDA-induced cytotoxicity through activation of eNOS and generation of NO. The antiapoptotic effect of the TNF-α/NO pathway is mediated through an action that counteracts the specific apoptogenic signalling events induced by alsinMT and NMDA.

TNF-α induced eNOS activation and cytoprotection occur through stimulation of SK1 and activation of S1P receptors

Activation of eNOS by TNFα in endothelial cells depends on its phosphorylation by Akt in a pathway involving activation of SK1, with generation of S1P and stimulation of specific, PTx-sensitive, S1P membrane receptors (De Palma et al. 2006).

We investigated whether this pathway operates also in neuronal cells and the functional relevance of it. To inhibit S1P membrane receptors we used PTx. To inhibit the activity of SK1 effectively and specifically we used a well characterised dominant-negative SK1 (SK1-DN) (Pitson et al. 2000; De Palma et al. 2006). Expression of SK1-DN was checked by detecting its FLAG epitope by western blotting and FACS analyses (Fig. 3a). In cells transfected with SK1-DN activation of SK was inhibited by 78 ± 7.2% with respect to that of cells transfected with the empty vector (not shown, = 6). Transfection with SK1-DN or treatment with PTx inhibited activation by TNF-α of both Akt and eNOS, measured using specific Abs recognising the active, phosphorylated forms of the enzymes, with inhibition of NOS activity (Fig. 3b and c). Thus, generation of S1P and activation of S1P receptors are necessary for Akt and eNOS activation by TNF-α in neuronal cells.

Figure 3.

 TNF-α activates eNOS through sequential activation of and SK1 and S1P membrane receptors. SKNBE cells were transfetced with SK1-DN, the pcDNA3 empty vector (e.v.) and incubated for 8 h with or without PTx (100 ng/mL). (a) expression of SK1-DN, detected using an anti-FLAG Ab, both by western blot and by FACS analyses (= 6). RFI (relative fluorescence intensity) values represent the efficiency of transfection ± SEM. (b, c) Cells were incubated with or without TNF-α for 30 min, after which were assessed: (b) active, phosphorylated Akt (P-Akt) and eNOS (P-eNOS) were detected by western blotting. GAPDH was revealed in parallel as a loading control. Shown are both representative images and quantitative values, expressed described in Fig. 1 (= 4); (c) NOS activity, measured as L-citrulline formation (= 4). Asterisks indicate statistical significance versus e.v., crosses indicate statistical probability versus e.v. + TNF-α, calculated as described in the Materials and Methods section.

Cell transfection with either SK1-DN or treatment with PTx reversed the protective action of TNF-α fully in both alsinWT- and alsinMT-transfected cells, without affecting the basal rate of apoptosis (Fig. 4a and b). These results confirm unambiguously that TNF-α activates eNOS through generation of S1P and activation of S1P receptors and that such pathway of activation has functional relevance in neuroprotection.

Figure 4.

 TNF-α cytoprotective action in neuronal cells requires SK1 and S1P membrane receptors. SKNBE cells expressing AlsinWT or AlsinMT were transfetced with SK1-DN or empty pcDNA3 vector (e.v.) (a) or incubated for 8 h with or without (NT) PTx (100 ng/mL) (b). Cells were treated for a further 8 h in the presence/absence of NMDA (1 mM) and TNF-α (100 ng/mL) in various combinations as specified in the keys to the graphs and apopotosis measured as described in Fig. 2 (= 4). Asterisks indicate statistical significance versus e.v. or NT; crosses versus indicate statistical probability e.v.+NMDA or NMDA, calculated as described in the Materials and Methods section.

Activation of eNOS by TNF-α protects against several neurotoxic cues

The process of neurodegeneration involves a multiplicity of pathogenic mechanisms other than excitotoxicity, including generation of reactive oxygen species, endoplasmic reticulum stress and DNA damage (Xu et al. 2005; Potashkin and Meredith 2006). We investigated whether the NO-dependent protective action of TNF-α was limited to NMDA-induced death or counteracted also these other pathogenic mechanisms. To mimic endoplasmic reticulum stress cells were treated with thaspigargin, a blocker of the SERCA pumps, or tunicamycin, an inhibitor of N-glycosylation (Xu et al. 2005; Deniaud et al. 2007); DNA damage was investigated using etoposide (Kruman et al. 2004); generation of reactive oxygen species was achieved using As2O3 and H2O2 (Sciorati et al. 2006). In addition, we used the protein kinase C inhibitor staurosporine (Panzeri et al. 2006).

As reported in Table 1, all of these stimuli induced apoptosis of SKNBE cells, and in all cases co-stimulation with TNF-α yielded significant protection from damage. Such protection was no longer observed if cells were incubated also with L-NAME or PTx, or transfected with SK1-DN. These results indicate that inhibition of apoptosis by TNF-α via NO is a general phenomenon that may protect neuronal cells from a variety of toxic stimuli involved in neurodegeneration.

Table 1.   Activation of eNOS by TNF-α protects from several cytotoxic stimuli
Percentage (%) apoptotic cells (annexin V+/7AAD−)
Cytotoxic stimulusTreatment
e.v.e.v./TNF-αe.v./TNF-α + L-NAMEe.v./TNF-α  +  PTxSK1DN + TNF-α
  1. SKNBE cells were transfected with SK1-DN or empty vector (e.v.), incubated for 8 h in the presence or absence of PTx (100 ng/mL) and exposed for 12 h to staurosporine (1 μM), thaspigargin (1 μM) tunicamycin (7 μg/mL), etoposide (1 μM), As2O3 (20 μM) or H2O2 (100 μM) in the presence/absence of TNF-α (100 ng/mL) and L-NAME (1 mM) as indicated. Apoptosis was detected as described in Fig. 2a (= 3). Asterisks indicate statistical significance versus e.v.; crosses indicate statistical probability versus e.v./TNF-α, respectively, calculated as described in the Materials and Methods section.

No stimulus5.45 ± 0.316.19 ± 0.426.93 ± 0.465.89 ± 0.288.11 ± 1.00
Staurosporine32.0 ± 1.4111.4 ± 1.47***44.1 ± 2.17+++34.0 ± 2.25+++34.0 ± 1.51+++
Thapsigargin67.0 ± 3.4423.2 ± 1.14***72.0 ± 3.19+++65.0 ± 2.67+++71.4 ± 3.38+++
Tunicamycin48.0 ± 2.2310.9 ± 0.41***56.2 ± 3.03+++43.0 ± 2.80+++59.1 ± 3.00+++
Etoposide55.0 ± 2.3521.1 ± 1.31***59.0 ± 3.27+++51.0 ± 2.55+++63.2 ± 3.61+++
As2O361.0 ± 4.1116.0 ± 0.98***66.8 ± 2.41+++54.0 ± 2.74+++69.5 ± 3.46+++
H2O273.0 ± 5.0127.1 ± 0.76***69.3 ± 4.90+++60.0 ± 3.04+++81.3 ± 4.27+++

The protective action of TNF-α via NO depends on cGMP generation

We analysed the role of cGMP signalling in mediating the effect of NO. To this end cells were exposed to TNF-α in the presence of guanylate cyclase inhibitor ODQ (1 μM) (Boulton et al. 1995). ODQ reversed the protective effect of TNF-α on NMDA-induced apoptosis (Fig. 5a). The protective effect of TNF-α was likewise inhibited by its co-incubation with KT5823 (1 μM), an inhibitor of protein kinase G (Clementi et al. 1995). Similar results were obtained using H2O2 as the apoptogen (not shown). Neither ODQ, nor KT5823 had any effect when administered without TNF-α. To assess the effect of cGMP further SKNBE cells were exposed to NMDA and H2O2 in the presence of increasing concentrations of the membrane-permeant cGMP mimic 8 Br-cGMP. As shown in Fig. 5b, low concentrations of 8 Br-cGMP did not protect from apoptosis induced by NMDA or H2O2, consistent with the absence of protection observed when TNF-α was administered to cells expressing the eNOS-specific siRNA (see Fig. 2), i.e. conditions under which the generation of cGMP was only two fold over basal. At higher concentration however, 8 Br-cGMP was protective and such protection was concentration-dependent. In no case a direct cytotoxic effect of 8 Br-cGMP was observed (Fig. 5b). Taken together, these results indicate that the NO-dependent protective effect of TNF-α on NMDA and H2O2 takes place via a mechanism involving activation of guanylate cyclase, formation of cGMP and activation of protein kinase G.

Figure 5.

 The cytoprotective action of the NO generated by TNF-α is mediated through generation of cGMP and activation of protein kinase G. SK-N-BE cells were transiently transfected with alsinWT GFP-tagged at its N-terminus. Apoptosis was triggered by an 8 h administration of NMDA (1 mM) in the presence/absence (NT) of TNF-α (100 ng/mL), ODQ (1 μM), KT 5823 (1 μM) (a) or increasing concentrations of 8 Br-cGMP (b). Apoptosis was assessed by measuring in flow cytometry the % of cells staining positive for annexin V and excluding actinomycin D (7AAD). Values in (b) are reported as % of those observed in cells exposed to NMDA in the absence of 8 Br-cGMP. Asterisks indicate statistical significance versus NT; crosses indicate statistical probability versus NMDA, calculated as described in the Materials and Methods section (= 6).


Amyotrophic lateral sclerosis is a quickly progressive, lethal, degenerative disorder of motor neurones for which only palliative therapy is available (McGuire et al. 1996; Pasinelli and Brown 2006). Genetic studies have increased our understanding of the causes of this disease with identificaton of defects in specific genes, namely SOD1, alsin, senataxin, dynactin and synaptobrevin-VAMP (vescicle-associated membrane protein) associated protein B (Pasinelli and Brown 2006). The pathogenesis of neuronal death in ALS is however still undefined.

Multiple perturbations of cellular function in ALS motor neurones leading to cell damage and apoptosis have been identified. Among these particularly relevant are excessive excitatory tone, protein misfolding, impaired energy production, abnormal calcium metabolism, altered mitochondrial function and axonal transport, generation of reactive oxygen species and activation of proteases and nucleases leading to apoptosis (Bruijn et al. 2004; Bacman et al. 2006). Several factors may instigate these phenomena such as latent infections by viral and non-viral agents, environmental toxins and autoimmune reactions (MacGowan et al. 2001; Bruijn et al. 2004; Pasinelli and Brown 2006). Among these instigating events, activation and proliferation of astrocytes and microglia, accompanied by increased levels of several proinflammatory cytokines are believed to play relevant roles (Bruijn et al. 2004; Pasinelli and Brown 2006).

Particular attention in this respect has been devoted to TNF-α, whose increased levels have often been associated with detrimental consequences to neurones and myelin (Robertson et al. 2001; He et al. 2002). The actual role of this cytokine in ALS pathogenesis, however, is still debated and studies in several models of neurodegeneration, including studies in TNF-α knockout mice, support also a neuroprotective role (Carlson et al. 1999; Acarin et al. 2002; Marchetti et al. 2004; Turrin and Rivest 2006). Thus, it has been suggested that TNF-α generation may be part of a neuroprotective mechanism, which, however, deranges and eventually contributes to neuronal toxicity (Ghezzi and Mennini 2001; Moreau et al. 2005).

In this study we provide clear evidence that TNF-α exerts a basal protective effect in neuronal cells expressing a cytotoxic mutant form of alsin involved in the early onset form of ALS, ALS2 (Panzeri et al. 2006). Cytoprotection was preserved also in the presence of NMDA, a classical excitotoxic agent whose release by activated microglial cells, also in response to TNF-α, has been reported to play a concurrent key role in ALS pathogenesis (Pardo et al. 2006; Takeuchi et al. 2006; Van Den Bosch et al. 2006).

The protective action of TNF-α was accompanied by inhibition of changes in the ratio of Bax and Bcl-xL and of translocation of Bax to mitochondria. An altered ratio and localisation of these proteins contributes to apoptosis induction in a variety of physiopathological conditions. In particular, decreases in Bcl-xL and increases in Bax have been found consistently in mouse models of ALS (Mu et al. 1996; Kostic et al. 1997; Gonzalez de Aguilar et al. 2000). Thus, TNF-α inhibits a specific pathway of apoptosis known to be relevant in ALS. Interestingly, Bax plays a role also in stimulating neuromuscular denervation in ALS, a key pathogenic event that occurs earlier than apoptosis and through different pathways (Gould et al. 2006). TNF-α might therefore play a protective role also in early events of motorneuronal damage preceding apoptotic degeneration.

We have investigated the signalling pathway initiated by TNF-α to exert its protective effect and found that it involves two relevant events, the first of which is generation of S1P. S1P, in turn, activates eNOS via stimulation of Akt, a well known pathway of activation of eNOS. (Fulton et al. 2001). The involvement of S1P and eNOS in cytoprotection shed further light on the mechanism of action and role of TNF-α in ALS pathogenesis. S1P contributes to differentiation and survival of neurones, and plays a role in their plasticity by favouring modifications of shape through changes in actin and other cytoskeletal proteins (Buccoliero and Futerman 2003). In addition, Akt has been reported to be cytoprotective in ALS (Li et al. 2003). Thus, the involvement of S1P generation in eNOS activation suggests that TNF-α activates a complex signalling pathway in which protective and regenerative neuronal events are integrated.

The involvement of eNOS is even more intriguing. eNOS is highly expressed in the motor neurones in the anterior horns of spinal cord in ALS patients, while absent in healthy controls (Kashiwado et al. 2002). Moreover, eNOS expression is increased especially in motor neurones of cervical, lumbar, and sacral cord, which seldom degenerate until the late stages of the disease (Kashiwado et al. 2002). Our observation that eNOS activation by TNF-α is cytoprotective against a variety of toxic stimuli involved in the pathogenesis of ALS, including NMDA, DNA damage, endoplasmic reticulum stress, and generation of reactive oxygen species, suggests for this enzyme a novel broad-range protective function in damaged neurones.

The protective action mediated by eNOS-generated NO depended on generation of cGMP. This observation is important because cGMP mediates physiological actions of NO, including NO-dependent protection from apoptosis induced by several apoptogenic stimuli in a variety of cell systems (Sciorati et al. 1997; Kim et al. 1998; Shen et al. 1998; Fiorucci et al. 1999; Liu and Stamler 1999; De Nadai et al. 2000; Bulotta et al. 2001; Barsacchi et al. 2002). Interestingly, we did not observe protection at low concentrations of cGMP, suggesting that a critical output of NO is necessary to reach the threshold of cGMP concentration needed to activate the cGMP-dependent protection mechanism.

Endothelial NO synthase activity is sensitive to many feedback events including protein/protein interactions, changes in plasma membrane lipid concentrations and regulatory multi-site phosphorylation of specific serine threonine and tyrosine residues (Fulton et al. 2001; Goligorsky et al. 2002). In addition, eNOS mRNA expression is controlled both at the transcriptional and post-transcriptional phases, and epigenetic mechanisms appear to modulate tissue-specific eNOS expression (Dudzinski and Michel 2007). Furthermore, Akt and NO themselves may regulate eNOS expression (Sud et al. 2008). While our model system is testing somehow acute phenomena, profound changes in protein and cellular lipid expression and structure occur in neurones in the course of chronic degenerative processes such as ALS (Bruijn et al. 2004). Such changes cannot be predicted by our studies in vitro, may affect the various components of the complex eNOS regulatory machinery and lead to a reduced or enhanced ability of TNF-α to trigger eNOS activation and NO output.

Some studies have suggested that the ability of TNF-α to switch from cytoprotective to cytotoxic may involve NO generation. Indeed TNF-α affects eNOS expression in a time-dependent way, by inducing first an increase and then a downregulation (Yoshizumi et al. 1993; Bove et al. 2001). In addition, high levels of TNF-α may trigger expression of iNOS, leading to generation of micromolar concentrations of NO. Such high concentrations of NO in the presence of oxidative stress conditions may then react with superoxide anions at near diffusion-limited rates, even faster than dismutation of superoxide anions by superoxide dismutase, to generate the highly oxidant and nitrosant species peroxynitrite (Ischiropoulos and Beckman 2003). Formation of peroxynitrite has been suggested by some studies to contribute to the pathogenesis of ALS (Almer et al. 1999; Ischiropoulos and Beckman 2003; Martin et al. 2005; Raoul et al. 2006). In our experiments we did not detect cytotoxic effects of TNF-α even when administered at very high concentration. In addition we generated pro-oxidant conditions (e.g. with H2O2) and found that NO generated endogenously by TNF-α was of benefit. Likewise in preliminary experiments we did not observe apoptosis when SKNBE cells were challenged with H2O2 in the presence of 50 μM of the NO donor DETA-NO that generates micromolar concentrations of NO (Clementi et al. 1998). Thus the concept that high NO output in the presence of oxidative stress leads to generation of peroxynitrite and that this contributes to ALS pathogenesis has to be taken with caution. Likewise, further investigations are necessary to establish how and whether the TNF-α-to-NO protective signalling turns into a cytotoxic one, and the relevance of the protective versus toxic actions of these messengers in the context of ALS disease progression.

A possible caveat stems from the in vitro model that we used. At variance with primary cells, they can be transfected with high efficiency, thus offering a clear advantage for molecular dissection of signal transduction pathways (Panzeri et al. 2006). However, SKNBE are of neoplastic origin and may thus only partially recapitulate the sequence of events taking place in primary neurones in vivo (Opel et al. 2007).

Increases in TNF-α accompanied by high levels of eNOS expression in neurones are not an exclusive feature of ALS but occur also in Alzheimer’s and Parkinson’s diseases, diffuse Lewy body disease, Pick’s disease and progressive supranuclear palsy (Sohn et al. 1999; Levy et al. 2000; de la Monte et al. 2000; Leker et al. 2001). It is therefore conceivable that the TNF-α/S1P/NO signalling pathway and its protective role in neurones we describe here has biological relevance broader than ALS and plays a relevant role also in other neurodegenerative disorders.


We thank Dr Andrea Martinuzzi, E. Medea Scientific Institute, Conegliano, for fruitful discussions and criticisms. This work was supported by grants from: Italian Association for Cancer Research (E.C.); Telethon GGP07006 (E.C.), the Association Française contre les Myopathies projects n. 12769 and 12780 (M.T.B. and E.C.), Ministero della Salute (Prog PS-Neuro-ex.Art 56/05/7, RF2006-70, RF2007-75) (E.C. and M.T.B.), Fondazione Romeo ed Enrica Invernizzi (E.C.).