Different pathways for iNOS-mediated toxicity in vitro dependent on neuronal maturation and NMDA receptor expression


Address correspondence and reprint requests to S. Golde, Cambridge Centre for Brain Repair, University of Cambridge, Forvie Site, Robinson Way, Cambridge, CB2 2PY, UK. E-mail: sg274@cus.cam.ac.uk


Co-localization of activated microglia and damaged neurones seen in brain injury suggests microglia-induced neurodegeneration. Activated microglia release two potential neurotoxins, excitatory amino acids and nitric oxide (NO), but their contribution to mechanisms of injury is poorly understood. Using co-cultures of rat microglia and embryonic cortical neurones, we show that inducible NO synthase (iNOS)-derived NO aloneis responsible for neuronal death from interferon γ(IFNγ) +lipopolysaccharide (LPS)-activated microglia. Neurones remain sensitive to NO irrespective of maturation state but, whereas blocking NMDA receptor activation with MK801 has no effect on NO-mediated toxicity to immature neurones, MK801 rescues 60–70% of neurones matured in culture for 12 days. Neuronal expression of NMDA receptors increases with maturation in culture, accounting for increased susceptibility to excitotoxins seen in more mature cultures. We show that MK801 delays the death of more mature neurones caused by the NO-donor DETA/NO indicating that NO elicits an excitotoxic mechanism, most likely through neuronal glutamate release. Thus, similar concentrations of nitric oxide cause neuronal death by two distinct mechanisms: NO acts directly upon immature neurones but indirectly, via NMDA receptors, on more mature neurones. Our results therefore extend existing evidence for NO-mediated toxicity and show a complex interaction between inflammatory and excitotoxic mechanisms of injury in mature neurones.

Abbreviations used

cyclosporin A








glial fibrillary acidic protein


inducible nitric oxide synthase


interferon gamma




lactate dehydrogenase




341495 (2S)-2-Amino-2-[(1S,2S)-2-carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid




mitochondrial permeability transition


[manganese(III)tetrakis(4-benzoic acid)porphyrin]


3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide


2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide disodium


nitric oxide


neuronal nitric oxide synthase


poly ADP ribose polymerase, SOD, superoxide dismutase


[hydroxydiazenesulfonic acid 1-oxide.disodium salt]


tumour necrosis factor alpha




W, (N-(3-(aminomethyl)benzyl)acetamidine).

Microglia are resident cells of monocytic lineage in the CNS. They exist in a resting state as ramified cells but can be activated by a wide range of stimuli including pro-inflammatory cytokines and factors released during neuronal injury. Activation is characterized by changes in morphology, phenotype and behaviour including increased proliferation and recruitment to the site of injury. The functions of activated microglia include phagocytosis, antigen presentation and toxic effects on invading pathogens (Banati et al. 1993; Aschner et al. 1999; Raivich et al. 1999). However, excess activation may be deleterious as activated microglia can cause neuronal death in culture (Piani et al. 1991; Chao et al. 1992; Kingham et al. 1999; Flavin et al. 2000). They may contribute to neuronal injury resulting from trauma, ischaemia or inflammation and are associated with demyelination and neurodegeneration (Banati and Graeber 1994; Diemel et al. 1998; Giulian 1999; Kaul et al. 2001). Activated microglia are found around the β-amyloid-plaques in Alzheimer's disease (AD; Itagaki et al. 1989; Miyazono et al. 1991; McGeer and Rogers 1992; Wa et al. 1996) and in the lesions of multiple sclerosis (Trapp et al. 1998; Luchinetti et al. 2000) in close association with damaged axons (Trapp et al. 1998). Anti-inflammatory treatments influence the progression of AD (McGeer and McGeer 1995; McGeer et al. 1996; Stewart et al. 1997; Lim et al. 2000), and modulation of the inflammatory response in multiple sclerosis may slow the rate of brain atrophy (Rudick et al. 1999; Zivadinov et al. 2001), supporting the idea that activated microglia contribute actively to neurodegeneration through inflammatory mechanisms.

Although substantial evidence for microglia-induced neurodegeneration has been provided both by in vitro and in vivo studies, the molecules mediating neurodegeneration are still not fully characterized, perhaps because these studies have been performed in different species, using various neuronal cell types, of different ages. Several mediators have been suggested including tumour necrosis factor α (TNFα; Meda et al. 1995; de Bock et al. 1998), Interleukin-1 (IL-1; Chao et al. 1995), proteases (Rogove and Tsirka 1998; Flavinet al. 2000) and reactive oxygen intermediates (Beckman et al. 1994), but glutamate receptors and nitric oxide (NO) have been most commonly implicated (Piani et al. 1991, 1992; Boje and Arora 1992; Chao et al. 1992; Giulian et al. 1993; Dawson et al. 1994; McMillian et al. 1995; Jeohn et al. 2000; Bal-Price and Brown 2001).

NO is released from activated microglia and astrocytes following induction of nitric oxide synthase (iNOS; Chao et al. 1992; Brown et al. 1995; Murphy 2000; Bal-Price and Brown 2001), and neurones are particularly sensitive to NO-mediated toxicity (Leist et al. 1997; Wei et al. 2000). However, the precise contribution of NO to microglia-induced neurodegeneration is unclear. Several studies using mixed neuronal–glial cultures together with non-isoform specific NOS inhibitors suggest the involvement of NO (Boje and Arora 1992; Chao et al. 1992; Dawson et al. 1994; McMillian et al. 1995; Jeohn et al. 2000) but do not identify the cellular source, the sequence of molecular events, or the potential role of co-factors. The mechanism by which NO kills neurones and its relationship to excitotoxicity is also unclear, as NO and glutamate could act independently or consecutively to cause neurotoxicity. Nitric oxide may, for example, induce glutamate release by neurones (Meffert et al. 1994; Montague et al. 1994; McNaught and Brown 1998; Bal-Price and Brown 2001), which then stimulates NMDA receptors, triggering excitoxicity. However, when NMDA receptors are activated, Ca2+ influx stimulates NO production through the calcium- and calmodulin-dependent neuronal nitric oxide synthase (Garthwaite and Boulton 1995), potentially leading to neuronal death (Dawson et al. 1991; Strijbos et al. 1996). Bal-Price and Brown (2001) recently showed that activated glia kill co-cultured cerebellar granule cells due to glial-derived NO inducing neuronal glutamate release followed by excitotoxicity.

In this study, we sought to define the involvement of NO and excitotoxins in microglia-induced death of cortical neurones. We studied neurones of differing age to assess whether maturation, which can be associated with differential expression of cell surface receptors such as the NMDA receptor (Piani et al. 1992), can alter the mechanism of microglial neurotoxicity. We show that microglia-derived NO alone is responsible for neuronal toxicity of co-cultured immature as well as more mature neurones. This effect is mediated by the NO molecule itself, rather than its peroxynitrite metabolite. Furthermore, we demonstrate that NO injures neurones by two distinct mechanisms: NO is directly toxic to immature cortical neurones but acts on mature neurones primarily through activation of NMDA receptors. Mechanisms of nitric oxide toxicity therefore differ depending on the target cell.

Materials and methods


Stock solutions of the following substances were stored in working aliquots at − 20°C: 1400 W [N-(3-(aminomethyl)benzyl)acetamidine; 0.5 mm in H2O; Calbiochem, La Jolla, CA, USA], aminoguanidine [40 mm in Dulbecco's modified Eagle medium (DMEM), Sigma, St Louis, MO, USA], l-NAME (1 m in DMEM, Sigma), catalase (20 000 µ/mL in DMEM, Sigma), MnTBAP [manganese(III) tetrakis(4-benzoic acid)porphyrin] (10 mm in PBS, Calbiochem), MK801 [(5R,1OS)-(+)-5-methyl-10,11-dihydro-5H-dibenzo[a,d]-cyclohepten-5,10-imine; Dizocilpine] (1 mm in H2O, Tocris Cookson, Bristol, UK), NBQX (10 mm in DMEM, Tocris), LY 341495 (10 mm in DMEM, Tocris), cyclosporin A (400 µm in DMSO, Calbiochem), 1,5-Isoquinolinendiol (10 mm in DMSO, Alexis Biochemicals, San Diego, CA, USA), soluble TNF-receptor (1 mg/mL in PBS, Therapeutic Antibody Centre, Oxford, UK) and recombinant TNFα (10 mg/mL in PBS, R&D Systems, Mineapolis, MN, USA). Vinyl-L-NIO [N5-(1-Imino-3-butenyl)-L-ornithine] (0.5 mm in H2O, Calbiochem) was stored at −70°C for a maximum of 3 weeks. FK506 (1 mm in DMSO, Calbiochem) and guvacine (1 m in PBS, Tocris) were prepared immediately before use. Stock solutions (50 mm) of DETA-NONOate [(Z)-1-[2-(2-Aminoethyl)-N-(2-ammonioethyl)amino]diazen-1-ium-1,2-diolate]; DETA/NO] and Sulfo-NONOate [(Hydroxydiazenesulfonic acid 1-oxide.disodium salt); Sulfo/NO; both Alexis Biochemicals] were prepared in 10 mm NaOH immediately before use. Uric acid (Sigma) was dissolved in 1 N NaOH at a concentration of 25 mg/mL and diluted in H2O to a 100-mm stock solution that was used on the same day. Oxyhaemoglobin (Sigma) was dissolved at 10 mg/mL in PBS and stored at 4°C for a maximum of 3 weeks.

Neuronal cultures

Whole brains from embryonic day 16 Sprague–Dawley rat embryos (Charles River, Margate, UK) were kept in Hank's balanced salt solution without calcium and magnesium (HBSS, Gibco, Paisley, UK) and cleaned of meninges. Cerebral cortices were dissected free and dissociated in HBSS. Tissue pieces were incubated in trypsin (0.1%, Sigma, in HBSS without calcium and magnesium) for 20 min at 37°C, washed in DNAse (0.001%, Sigma, in HBSS), and manipulated in triturating solution (1 g Albumax, Gibco; 50 mg trypsin inhibitor, Sigma; and 1 mg DNAse, Sigma, per 100 mL HBSS) using flame polished Pasteur pipettes. Cells were resuspended in serum-free defined culture medium [DMEM, Gibco; 1% penicilline-streptomycine-fungizone (PSF), Gibco; and 2% B27 supplement, Gibco]. Viable cells, demonstrated by Trypan blue exclusion, were plated on poly-l-lysine (0.01% in distilled water, Sigma) coated 24-multiwell plates (Nunclon, Life Technologies, Paisley, UK) at a density of 2.26 × 105 cells/cm2 or onto poly-l-lysine-coated 13 mm-diameter glass coverslips at a density of 5.6 × 104 cells/cm2 or 1.13 × 105/cm2 cells for immunocytochemistry of neurones grown 1 or 12 days in vitro, respectively. Cultures were incubated in culture medium at 37°C in a 5% CO2 humidified atmosphere.

Microglial cultures

Microglia were isolated from mixed glial cell cultures as previously described (Giulian and Baker 1986). Briefly, cells dissociated from neonatal rat (Sprague–Dawley) cerebral hemispheres were plated in 75-cm2 poly-l-lysine-coated tissue culture flasks (Orange Scientific, Triple Red, Thame, UK) at a density of two brains per flask in culture medium consisting of DMEM supplemented with 10% fetal calf serum (FCS). Culture medium was changed after 24 h and then twice per week. After 12 days, cultures contained a confluent glial cell layer with top-dwelling amoeboid microglia. The loosely adherent microglia were harvested using a rotary shaker (Lukham R300) at 70% for 20 min. After centrifugation, (180 g for 5 min), cells were re-suspended in serum free defined culture medium (DMEM with 1% PSF and 2% B27 supplements). Cell viability was determined by Trypan blue exclusion, and viable cells were plated at a final density of 1.12 × 105 cells/cm2 onto 24-multiwell plates (Nunclon); 2.8 × 104 cells/cm2or 5.6 × 105 cells/cm2 onto 13 mm-diameter glass coverslips for staining; or 5.6 × 106 cells/cm2 onto 6-multiwell plates (Orange Scientific) for conditioning medium. After 20 min, cultures were washed once with HBSS to remove non-adherent contaminating glia and incubated in serum-free defined medium for 3 days with or without microglial activators – LPS (1 µg/mL, Sigma) and IFNγ (100 µ/mL, Serotec Ltd, Oxford, UK). Conditioned media were prepared by incubating microglia for 24 h in serum-free defined medium (1 mL/4 × 106 microglia) with or without microglial activators. Each medium was subsequently collected and cleared of cells and debris by centrifugation (2880 g for 20 min). The supernatant was then diluted 1 : 1 with fresh medium and applied to neuronal cultures.

Co-cultures of microglia and cortical neurones

Co-cultures were prepared by plating a suspension of isolated microglia on neurones maintained for 1 or 12 days in vitro in a 1 : 2 ratio. Cells were incubated in serum-free defined medium for 30 min and then washed once with HBSS to remove non-adherent contaminating macroglia before incubation in serum-free defined medium for 3 days with or without LPS (1 µg/mL) or IFNγ (100 µ/mL). For experiments using Transwell culture inserts (Millipore, Bedford, MA, USA), microglia were plated into the well insert and transferred onto neuronal cultures after the washing step.

NO2 determination

NO2 levels, measured with the Griess reagent, were taken as an estimate of NO generation. Griess reagent (equal volumes of 0.1% N-1-naphthylethylenediamine dichloride, Sigma, in water and 1% sulfanilamide, Sigma, plus 5% H3PO4 in water) was added to an equal volume of cell culture supernatant (100 µL) and incubated for 20 min at room temperature. The optical density was measured at 570 nm and a standard curve established using NO2 (Sigma).

Lactate dehydrogenase assay

Lactate dehydrogenase (LDH) release into the culture medium was used as a measure of cell death. LDH is only released when membrane integrity breaks down. The Sigma in vitro-toxicology assay ‘tox-7’ was used to determine LDH activity in the cell culture supernatants according to the manufacturers instructions. For determination of cytotoxicity in neuronal cultures from DETA/NO, exposure cell culture medium was discarded, cells were washed once with PBS, then lysed and assayed for LDH activity according to the manufacturers instructions. Cytotoxicity = 100 − (100 × OD490nm of treated cultures/OD490nm of untreated cultures).


The viability of cell cultures was estimated using 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma). This tetrazolium dye is oxidized by mitochondrial dehydrogenases and is a sensitive marker for mitochondrial function. Cells were incubated with MTT (0.5 mg/mL in culture medium) for 1 h at 37°C and then washed in PBS before adding 10% Triton X-100 in 0.1 m HCl and shaking for at least 20 min at room temperature to dissolve the formazane crystals before optical density of the mixture was measured at 570 nm.

Uptake of [3H]GABA

Cell cultures were washed with 500 µL DMEM and incubated for 30 min at 37°C with 1 mmβ-alanine (Sigma) to inhibit astrocytic uptake of [3H]GABA (Chao et al. 1992). The incubation medium was then discarded, and the cultures further incubated with 200 nm[3H]GABA (92 Ci/mmol; Amersham Biosciences UK Ltd, Little Chalfont, UK) in the presence of 1 mmβ-alanine (Sigma). GABA-uptake inhibitor guvacine (1 mm, Tocris) was added to neurones and microglia, alone and in co-culture, to determine non-specific uptake (Johnston et al. 1975). Since there was no increase above background in uptake by microglia exposed to guvacine, experimental differences in specific uptake reflect neuronal function only. Following the incubation with [3H]GABA, cells were washed with PBS and then extracted with NaOH (0.4 N, 500 µL/well). Radioactivity present in the extract was estimated by scintillation counting after neutralization with HCL (1 N, 50 µL/well). Values are presented after correction for non-specific uptake.

L929 TNFα-bioassay

TNFα concentrations were calculated using the TNFα-sensitive mouse fibroblast cell line L929 as described (Stangel et al. 1997). Briefly, L929 cells were plated at a density of 7500 cells/well on a 96-well plate and incubated overnight in RPMI-1640 medium (Gibco) with 10% FCS, 2 mm glutamine and 1% PSF. This was then discarded and the cells incubated in medium containing 2.66 µg/mL actinomycine D (Sigma), 1.33 mm homocysteine (Sigma) and 1.33 mm adenosine (Sigma). Cell culture supernatants were added at 3 concentrations (1 : 1, 1 : 5, 1 : 25) in aliquots of 50 µL. A standard curve was obtained using recombinant TNFα in a concentration range of 100 ng to 10 fg/mL. To test specificity of the cytotoxic effect of supernatants, soluble TNFα receptor (10 µg/mL) was added to wells containing recombinant TNFα and supernatant from activated microglial-neuronal co-cultures in a concentration of1 : 1. Soluble TNFα-receptor neutralized recombinant TNFα (10 ng/mL) and fully prevented toxicity of supernatant from activated microglial-neuronal co-cultures. This medium was discarded after incubation for 24 h at 37°C and an MTT assay performed as described above.


Cells were fixed in 4% paraformaldehyde for 10 min at room temperature followed by permeabilization with methanol for 10 min at − 20°C. The coverslips were then incubated with DMEM/10% FCS to block non-specific antibody binding. All further wash and incubation steps were also done in DMEM/FCS unless otherwise stated. Incubation with primary antibodies was carried out overnight at 4°C or for 1 h at 37°C. Monoclonal antibodies against β-Tubulin, isotype III (Chemicon, International, Temecula, CA, USA and Sigma) were used at 1 : 200 dilution; hybridoma supernatants against A2B5 (clone 105, ECACC) and GalC (clone IC-07, ECACC) were used at 1 : 5 and 1 : 10, respectively. Rabbit-Immunserum against GFAP (Chemicon) was used at 1 : 500. After washing three times, the coverslips were incubated with secondary antibodies for one hour at 37°C. Goat-anti-Mouse-IgG1 and -IgG2b, anti-IgM-TRITC, anti-IgG3-FITC (all Harlan-Sera-Laboratory) andanti-rabbit-AMCA (Vector Laboratories, Peterborough, UK) were all used at 1 : 100. Biotinylated goat anti-mouse Ig (Chemicon) was used at 1 : 200 followed by incubation with streptavidin-coupled rhodamine (1 : 200, Serotec) for 30 min at 37°C. For microglial staining, fluorescein isothiocyanate (FITC)-labelled B4-Isolectin (2 mg/mL, 1 : 50, Sigma) was used before permeabilization with methanol, with an additional blocking step after paraformaldehyde fixation. To label dead cells, coverslips were incubated with propidium iodide (0.5 mg/mL in PBS) before fixation. DNA damage was detected using an in situ immunocytochemical assay based on TdT-mediated dUTP nick end labelling (TUNEL) according to the manufacturers' recommendations (Roche/Boehringer Mannheim, Lewes, UK). All coverslips were incubated with Hoechst 33342 (1 : 5000 in PBS) to visualize nuclei before mounting with vectashield solution. The number of β-Tubulin and TUNEL-positive cells was determined under a fluorescent microscope at × 40 magnification by counting the number of immunoreactive cells in five random fields on three to four coverslips per condition for three independent experiments. For staining against NMDA receptor subunit NR1, cells were fixed in 4% paraformaldehyde, permeabilized with 0.25% Triton (Sigma) in PBS for 5 min, blocked with 10% normal goat serum (NGS) in PBS and then incubated with the monoclonal anti-NMDA-NR1-antibody (OMA-04010, Cambridge Biosciences, Cambridge, UK) at a dilution of 1 : 100 in 5% NGS in PBS at 4°C overnight. Secondary reagents (biotinylated anti-mouse Ig and TRITC–streptavidin) were used as above.

Preparation of cell extracts, SDS–PAGE and western blotting

Neurones were grown in culture for 1, 4 or 12 days as indicated, then washed with PBS and lysed for 15 min in ice-cold lysis buffer (50 mm Tris, pH 7.5, 150 mm NaCl, 1 mm MgCl2, 1 mm CaCl2, 1 mg/mL pepstatin, 1 µg/mL leupeptin, 1 µg/mL aprotinin, 1 µg/mL pepstatin) containing 1% (v/v) NP40. Lysates were cleared of debris by centrifugation (15 500 g for 15 min) and supernatants stored at − 20°C before processing. Protein concentration of lysates was determined using the BCA Protein Assay (Pierce, Rockford, IL, USA). Cell lysates were resolved on 8% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) gels and electroblotted onto polyvinylidene difluoride (PVDF) membranes (Immobilon-P, Millipore, Bedford, MA, USA). Blots were incubated for 1h at room temperature in blocking buffer (5% milk, 0.05% Tween-20 in PBS), followed by incubation with primary antibodies (mouse monoclonal anti-NMDA-NR1, Cambridge Biosciences, 1 : 500, goat polyclonal anti-NMDA-NR2A and anti-NMDA-NR2B, Santa Cruz Biotechnology, Inc., Heidelberg, Germany, both 1 : 100, and mouse monoclonal antiβ-Actin, clone AC-15, Sigma, 1 : 10 000) in blocking buffer over night at 4°C. Blots were then washed three times for 5 min with blocking buffer and incubated with horseradish-peroxidase-coupled anti-mouse or anti-goat Ig 1 : 2000 (Dako and Santa Cruz Biotechnology) in blocking buffer for 1 h at room temperature. The immunoblots were processed using Renaissance® Chemiluminescence detection system (NEN, Boston, MA, USA) according to manufacturer's instructions and immunoreactive bands were visualized after exposure to Hyperfilm ECL (Amersham Biosciences UK). Blots were then stripped using the Re-Blot-Kit (Chemicon International) according to the manufacturers' instructions.

Statistical analysis

All quantitative data are shown as means from at least three experiments each with three repeats ± SEM. Data were analysed using Student's t-test or one-way anova, followed by Newman–Keuls post-hoc test as stated in the figure legends.


Characterization of primary cell cultures

Neurones derived from rat embryonic cortex were cultured in defined serum-free conditions optimized for neuronal survival resulting in an almost pure culture of β-tubulin-positive neurones (98.5%). The remaining cells consisted of GFAP-positive astrocytes (0.2%), GalC-positive oligodendrocytes (0.4%), pre-oligodendrocytes staining for the early oligodendroglial marker O4 (0.9%) and no microglia. Co-culture with microglia in transwell inserts for 3 days changed the astrocyte content to 2.3%, but had no effect on the number of oligodendrocytes. Microglial cultures contained ≧ 95% B4-Isolectin-positive cells with ≦ 5% contaminating GFAP-positive astrocytes and A2B5-positive oligodendrocyte precursors. The effects of microglia were studied on immature and more mature neurones (1 and 12 days in vitro, respectively).

IFNγ + LPS-activated microglia are neurotoxic in co-culture but not via conditioned medium

We co-cultured cortical neurones (1 day in vitro) and microglia to determine whether microglia activated with IFNγ + LPS directly cause neuronal death. Significant loss of neurones was seen at 3 days. Hoechst staining showed condensation and fragmentation of nuclei and nick end labelling with the TUNEL method confirmed DNA strand breakage (Figs 1a and b), suggesting an apoptotic mechanism of cell death. Counting β-tubulin- and TUNEL-positive cells clearly revealed selective neuronal loss in activated but not non-activated co-cultures (Fig. 1c). We further quantified the effect of IFNγ + LPS on microglia-neuronal co-cultures as well as individual neuronal and microglial cultures, as controls, by determining LDH release into the culture medium and metabolism of MTT as markers of cellular viability. Activated microglial cultures released slightly more LDH than under non-activated conditions, emphasizing that activation is associated with some microglial death, as previously shown using chromogranin A as the microglial activator (Kingham et al. 1999). Against this baseline, LDH release increased significantly in the activated co-cultures, indicating enhanced cell death. Co-culture with non-activated microglia showed no increase in LDH-release above that attributable to loss of neurones and microglia, demonstrating that cell death requires microglial activation (Fig. 2a). MTT metabolism was reduced in activated compared to non-activated co-cultures, confirming the specificity for toxicity of activated microglia (Fig. 2b). IFNγ and LPS had no direct effect on LDH release or MTT metabolism of primary neuronal cultures. To establish that decreased MTT metabolism in co-cultures mainly reflects neuronal effects, the functional activity of cortical neurones was estimated by [3H]GABA-uptake (Fig. 2c). The addition of IFNγ and LPS to co-cultures reduced neuronal uptake of GABA but had very little effect on neurones in isolation. The demonstration that LDH-release and MTT metabolism correspond well both with number and function of neurones validates the use of LDH and MTT assays in subsequent experiments. Some neuronal loss was revealed after 3 days when microglia were plated into transwell culture inserts, allowing diffusion of soluble substances but no direct cell contact, and co-cultured with 1-day-old neurones. Neuronal death was observed only when microglia were activated and even then was confined toan area immediately underneath the transwell insert. Conditioned medium from activated or non-activated microglia had no effect on neuronal cultures when added at 1 day in vitro for 3 days (data not shown). Treatment of co-cultures with either IFNγ or LPS alone did not increase neurotoxicity (data not shown).

Figure 1.

Toxic effect of IFNγ- and LPS-activated microglia on neurones. (a) Cortical neurones (1 day in vitro, visualized with anti-β-tubulin – red) survived well and formed long processes after 3 days in co-culture with microglia. (b) Addition of IFNγ and LPS to co-cultures of neurones and microglia induced neuronal death after 3 days. Note the reduction of β-tubulin-positive neurones and the presence of condensed, TUNEL-positive nuclei (green) in most of the remaining cells, quantified in (c). *p < 0.05, **p < 0.01 (Student's t-test).

Figure 2.

Characterization of the effect of IFNγ and LPS on viability and function of microglia, neurones and co-cultures. (a) IFNγ and LPS induced significant increase of LDH release when added to microglia-neuronal co-cultures for 3 days, but not in neuronal or microglial control cultures. (b) MTT metabolism showed significantly reduced viability only in IFNγ- and LPS-treated co-cultures, but not in controls, although microglia showed a trend to reduced viability when activated that does not reach significance. (c) Reduced 3[H]GABA-uptake confirmed that cell death in activated co-cultures was due to loss of neurones, whereas neuronal control cultures were not affected by the activators. **p < 0.01 (Student's t-test).

Microglial secretion of nitric oxide is essential for neurotoxicity

Activation of microglia with LPS and IFNγ in combination triggered the release of NO (c. 50 µm nitrite/3 days; see Fig. 3a) due to induction of the microglial isoenzyme of nitric oxide synthase, iNOS. As expected, non-activated microglia and neuronal control cultures secreted very small amounts (< 2 µm/3 days) of NO (Fig. 3a). The observation that microglia are toxic to neurones in direct co-culture or in close transwell proximity is consistent with a predominant neurotoxic role in vitro for unstable microglial molecules such as NO, peroxynitrite or oxygen radicals. In order to determine whether iNOS-derived NO contributes to neurotoxicity, and to distinguish its role from nNOS, cultures were treated with 1400 W, an irreversible inhibitor of nitric oxide synthase 200 times more specific for iNOS than nNOS (Garvey et al. 1997). It was noted that 1400 W (50 µm) completely inhibited microglial NO but was moderately neurotoxic at this dose, whereas 1400 W (10 µm) selectively inhibited microglial NO-production by 80% (Fig. 3a). Activated co-cultures treated with 1400 W (10 µm) showed a reduced LDH release compared to those without the iNOS-inhibitor (Fig. 3b). MTT levels in activated co-cultures with iNOS inhibition were comparable to those seen in non-activated co-cultures (Fig. 3c). Together with the immunocytochemical observations (Figs 3d–f), these quantitative data demonstrate that the enhanced neuronal loss can be abolished by inhibition of microglial NO production. The partially selective iNOS inhibitor aminoguanidine (AG, 400 µm), and the non-selective NOS-inhibitor l-NAME (10 mm), were also neuroprotective in activated co-cultures (Table 1). This confirms the importance of iNOS-derived NO for neurotoxicity. Protection was incomplete both with aminoguanidine and l-NAME, possibly due to less effective inhibition of NO-production (54% and 43%, respectively) or due to simultaneous inhibition of nNOS. Conversely, treatment of the activated co-cultures with Vinyl-l-NIO, a mechanism-based potent inhibitor of neuronal nitric oxide synthase (nNOS) that does not inhibit iNOS (Babu and Griffith 1998) at doses up to 50 µm, did not protect neurones (Table 2) strongly suggesting that nNOS-derived NO does not contribute to the observed microglial toxicity.

Figure 3.

Inhibition of iNOS with 1400 W completely abolishes microglial neurotoxicity. (a) Treatment with IFNγ and LPS triggered substantial production of nitric oxide (determined as accumulation of nitrite after 3 days with the Griess reaction) of microglia cultured on their own and in co-cultures. iNOS inhibitor 1400 W abolished 80% of this NO production. Neuronal controls and non-activated microglia, alone or in co-culture, did not show increased NO production. (b) 1400 W significantly decreased cell death (as demonstrated by LDH release) in activated co-cultures and had no effect on controls. (c) Viability (MTT metabolism) of activated co-cultures was increased tolevels in non-activated co-cultures through inhibition of iNOS with 1400 W, *p < 0.05, **p < 0.01, ***p < 0.001. (One-way anova and Newman–Keuls post-hoc test.) Immunocytochemistry demonstrated intact neurones (β-tubulin – red) and microglia (B4-Isolectin – green) in non-activated co-cultures after 3 days (d), loss of neurones and altered morphology of microglia in IFNγ- and LPS-activated co-cultures (e) and complete rescue of neurones in activated co-cultures by inhibition of iNOS with 1400 W (f).

Table 1.  Effects of NOS-inhibition, catalase, AMPA/Kainate receptor- and mGluR-blockage on microglial toxicity to immature and more mature neurones
TreatmentCo-cultures with neurones (1 day in vitro)Co-cultures with neurones (12 days in vitro)
NO2 in µmMTT as ratio to controlNO2 in µmMTT as ratio to control
  1. Co-cultures of microglia and neurones (1 day in vitro and 12 days in vitro, respectively), and microglia and neurones alone were incubated for 3 days with or without IFNγand LPS, and with or without pharmacological agents in the concentrations indicated. MTT metabolism was determined at 3 days. None of these agents had an effect on MTT metabolism of neurones or microglia alone at the indicated concentration (data not shown). Changes in MTT metabolism compared to IFNγ+ LPS-treated co-cultures are indicated if significant: ap < 0.05, bp < 0.01, cp < 0.001 (one-way anova and Newman-Keuls post-hoc test).

None1.83 ± 0.5418.4 ± 3.81
IFNγ + LPS64.62 ± 10.160.29 ± 0.0278.89 ± 1.810.17 ± 0.07
IFNγ + LPS + aminoguanidine29.8 ± 9.230.64 ± 0.05c29.31 ± 3.891.03 ± 0.17b
(400 µm)
IFNγ + LPS + l-NAME36.65 ± 7.960.57 ± 0.03c55.61 ± 3.460.94 ± 0.12a
(10 mm)
IFNγ + LPS + catalase11.84 ± 4.180.58 ± 0.05c49.00 ± 1.870.67 ± 0.24a
(200 u/mL)
IFNγ + LPS + LY 34149550.96 ± 13.110.26 ± 0.0678.17 ± 3.250.18 ± 0.06
(100 µm)
IFNγ + LPS + NBQX56.86 ± 16.790.32 ± 0.0979.15 ± 1.810.15 ± 0.05
(100 µm)
Table 2.  Effects of soluble TNFα-receptor, NMDA receptor blockage and peroxynitrite scavengers on microglial neurotoxicity
TreatmentCo-cultures with neurones (1 day in vitro) MTT as ratio to control
  1. Co-cultures of microglia and neurones (1 day in vitro) as well as microglia and neurones alone were incubated for 3 days with or without IFNγand LPS, and with or without TNFαreceptor or pharmacological agents in the concentrations indicated. MTT metabolism was determined at 3 days and is shown here for the co-cultures. None of these agents had an effect on MTT metabolism of neurones or microglia alone at the indicated concentration (data not shown). None of the agents reversed the MTT decrease in co-cultures induced by microglial activation (one-way anova and Newman–Keuls post-hoc test).

IFNγ + LPS0.13 ± 0.02
Vinyl-LNIO (50 µm)0.10 ± 0.02
MnTBAP (100 µm)0.07 ± 0.01
Uric acid (1 mm)0.08 ± 0.02
Soluble TNFα receptor (10 µg/mL)0.14 ± 0.08
MK801 (10 µm)0.10 ± 0.02

Having established the essential role of microglia-derived NO for neurotoxicity in activated co-cultures, the effect of other microglia-derived products – TNFα and glutamate-receptor agonists – was then assessed. Human recombinant soluble TNFα receptor, at a concentration (10 µg/mL) that effectively blocked toxicity of supernatants from LPS- and IFNγ-activated rat microglia-neuronal co-cultures (containing 1–10 ng/mL TNFα) on the TNFα-sensitive cell line L929, had no effect when added to the co-cultures (Table 2). Likewise, blockade of NMDA receptors with MK801 (10 µm), AMPA/kainate receptors with NBQX (100 µm) and metabotropic glutamate receptors with LY 341495 (100 µm), respectively, did not influence these immature neurones (Tables 1 and 2).

To examine whether the combination of NO with reactive oxygen species formed by microglia upon activation is responsible for neurotoxicity, the superoxide dismutase (SOD) mimetic MnTBAP and the peroxynitrite scavenger uric acid were added to co-cultures. No decrease in microglial neurotoxicity could be observed at concentrations of 100 µm and 1 mm, respectively (Table 2). Catalase (200 and 500 u/mL), added to test whether hydrogen peroxide participates in microglial neurotoxicity in co-cultures, unexpectedly was found substantially to inhibit microglial NO production and protect neurones (Table 1). Oxyhaemoglobin (10 µm), which inactivates NO through formation of nitrate and methaemoglobin, decreased neurotoxicity (85.6% ± 4.2% of β-tubulin-positive neurones in activated co-cultures with Hb compared to 9.6 ± 3.7% without Hb, p < 0.001), supporting the direct neurotoxic role of NO.

The nitric oxide donor DETA-NONOate is sufficient to cause neuronal death

To examine whether nitric oxide is not only necessary but also sufficient to cause microglia-induced neuronal cell death, the slow releasing NO-donor DETA-NONOate (DETA/NO) was added to neuronal cultures. Sulfo-NONOate (SULFO/NO), a nitrous oxide-releasing compound, was used in parallel cultures as control. Cell death was apparent within 24 h following DETA/NO (0.5 mm) exposure (not shown). Specific dose-dependent loss of viable cells following exposure to the NO-donor DETA/NO for 3 days was demonstrated by reduction in MTT metabolism (Fig. 4a) and a parallel increase in cytotoxicity as determined by LDH activity in remaining cells (Fig. 4b). The accumulation of nitrites was measured in order to compare effects of this nitric oxide donor and microglial nitric oxide in co-culture. The nitrate value resulting from addition of DETA/NO for 3 days was plotted against MTT metabolized in the same cultures, and the relationship between nitrite accumulation and MTT-metabolism analysed (Fig. 4c). Total loss of viable cells was observed for nitrite values ≧ 60 µm, whereas those <20 µm had almost no effect on neuronal viability. Nitrite values ≧ 20 µm and < 60 µm were associated with a decrease in neuronal viability, although there was wide variability in this effect. Nitrite production was around 50–60 µm after 3 days in co-cultures activated with LPS and IFNγ, indicating that activated microglia release amounts of NO that mark the threshold for 100% neuronal death from DETA/NO. Concentration of nitrite after 3 days was 6.7 ± 1.9 µm and 19.1 ± 6.1 µm for independent activation of microglia with IFNγ and LPS, respectively. No neurotoxicity was observed under either condition.

Figure 4.

DETA/NO, a slowly releasing NO donor, causes dose-dependent neuronal death. Neuronal cultures (1 day in vitro) were exposed to increasing concentrations of the NO donor DETA/NO (▮) for 3 days or to SULFO/NO (▵) as control. (a) DETA/NO caused a dose-dependent decrease in viability (MTT) and (b) an increase in cytotoxicity. Cytotoxicity was calculated from measurement of LDHactivity after lysis of cells remaining intact in the culture wells. Cytotoxicity = 100 – (100 × OD490nm of treated cultures/OD490nm of untreated cultures). (c) NO production was quantified in each experiment with NO donors after 3 days via accumulation of nitrates using the Griess reaction. Mean values from three determinations ± SEM (n = 3 per data point) obtained in three different experiments were plotted against the MTT-value reached in these experiments after 3 days. A dose-dependent decrease of MTT could be observed. At nitrate values similar to values in the activated co-cultures (> 50 µm), viability was close to zero.

Potential intracellular mechanisms of NO-dependent death of immature neurones

Having established that microglia-derived nitric oxide directly kills immature cortical neurones, we further investigated which intraneuronal mechanisms could mediate NO-dependent cell death. NO can act directly on mitochondria-inhibiting respiration, stimulating the production of superoxide, which is then converted to hydrogen peroxide, and inducing mitochondrial permeability transition (MPT; Brown and Borutaite 2001). Furthermore, NO can also cause death in some cell types through DNA damage and overactivation of poly ADP ribose polymerase (PARP; Zhang et al. 1994). We tested whether inhibition of MPT or PARP, and scavenging hydrogen peroxide would each influence NO-donor-induced death of immature neurones. In pure cultures of immature (1 day in vitro) neurones, cytotoxicity from 0.5 mm DETA/NO was unchanged at 24 h and 3 days by cyclosporin A which potently inhibits MPT but also acts on calcineurin (1 µm; however, a dose of 2 µm wascytotoxic at 3 days), FK506 (a calcineurin inhibitor) andcatalase (200 µ/mL; a dose of 500 u/mL was sligthlycytotoxic at 3 days; Table 3). The PARP-inhibitor 1,5-Isoquinolinendiol (100 µm) was also unable to rescue neurones from toxicity of DETA/NO. 1,5-Isoquinolinendiol (10 µm) showed a significant, but very small, effect (< 10% reduction in cytotoxicity) at 3 days, indicating that PARP-overactivation plays no major role in NO-mediated death of immature neurones (Table 3).

Table 3.  Intracellular mechanisms of NO dependent death in immature neurones
 Neuronal cultures (1 day in vitro) + 24hNeuronal cultures (1 day in vitro) + 72 h
TreatmentCytotoxicity in percentage of controlCytotoxicity in percentage of control
  1. Neuronal cultures (1 day in vitro) were treated with or without 0.5 mm DETA/NO and with the indicated drugs. After 24 or 72 h, neuronal cultures were lysed and LDH activity in cell lysate was used to determine cytotoxicity as described in Materials and methods. Changes in cytotoxicity compared to untreated controls are indicated if significant: ap < 0.05, bp < 0.01, cp < 0.001 (one-way anova and Newman–Keuls post-hoc test).

No treatment37.01 ± 11.860.00 ± 0.0386.26 ± 1.400.02 ± 0.01
CsA 2 µm49.88 ± 14.468.07 ± 5.5585.66 ± 1.8055.82 ± 11.72c
CsA 1 µm48.11 ± 15.213.78 ± 0.6685.77 ± 2.0713.29 ± 4.88
FK 506 1 µm35.66 ± 7.51− 2.06 ± 3.1984.33 ± 1.76− 2.15 ± 1.89
1,5-Iso. 100 µm34.65 ± 7.816.78 ± 3.3380.26 ± 3.044.71 ± 6.22
1,5-Iso. 10 µm31.57 ± 6.872.33 ± 1.7576.82 ± 3.60b− 4.06 ± 1.81
Catalase 200 u/mL26.96 ± 0.226.51 ± 3.1780.18 ± 0.699.26 ± 2.02
Catalase 500 u/mL26.06 ± 0.687.35 ± 5.2179.48 ± 0.2420.50 ± 6.87a

More mature cortical neurones are also susceptible to iNOS-derived NO but can be partially rescued by NMDA receptor blockage

In order to determine whether maturation alters susceptibility to microglial factors, experiments were repeated using neurones that had first been cultured for 12 days. At this timepoint, expression of NR1 (an essential subunit of NMDA receptors) was marked not only around cell bodies but also along axons (Figs 5e and f) whereas, at 4 days in vitro, NMDA-NR1-expression was less pronounced and confined to the cell bodies (Figs 5c and d). No staining could be observed at 1 day invitro (Figs 5a and b). The change in NMDA-NR1 expression over time in culture was confirmed by western blot (Fig. 5g) showing no signal from lysates of neurones cultured for 1 day compared to a strong signal from 12-day cultures. Western blot analysis of NR2 subunits (Fig. 5h) demonstrates that both NR2A and NR2B are expressed in 12-day-old cultures. NR2A expression shows only little change from 1- to 12-day-old neurones; NR2B expression, however, markedly increases between day 1 and day 4 in vitro and plateaus thereafter. Therefore, neurones cultured for 1 day lack functional NMDA receptors, as they do not express NR1, whereas both NR1 and NR2 subunits are expressed at 12 days in culture. When (12-day-old) NMDA receptor-expressing neurones were co-cultured with LPS- and IFNγ-activated microglia, decreases in the number of β-tubulin-positive neurones and MTT metabolism were observed at 3 days (Fig. 6a,b), demonstrating that immature and mature neurones are equally susceptible to microglia. MTT metabolism of neurones grown for either 1 or 12 days in vitro did not decrease after 3 days culture in medium conditioned by either activated or non-activated microglia (data not shown) indicating that, for each, stable soluble molecules such as glutamatergic amino acids derived from activated microglia play no role in mediating toxicity. MTT metabolism and the number of β-tubulin-positive neurones of activated co-cultures increased to control levels in the presence of 1400 W, showing that more mature neurones can also be rescued by iNOS inhibition. Aminoguanidine and l-NAME were also neuroprotective (Table 2). However, in contrast to immature neurones, treatment with the NMDA receptor blocker MK801 significantly increased neuronal cell number and MTT metabolism compared to activated co-cultures not exposed to MK801, indicating that activation of the NMDA receptor is a co-factor for NO-mediated toxicity in the more mature neurones; nevertheless, MK801 only achieved 60% (MTT) to 70% (neuron number) protection compared with 1400 W. These two neuroprotective agents did not have synergistic effects (Figs 6a and b). Blocking AMPA/Kainate receptors with NBQX (100 µm) and mGlu receptors with LY 341495 (100 µm) had no effect (Table 1). Treatment of neuronal cultures at 12 days in vitro with the NO donor DETA/NO (0.5 mm) led to 54% cell death at 12 h. Addition of MK801 reduced the DETA/NO-induced cell death at 12 h to 15% (Fig. 6c). However, comparing the number of live cells (Fig. 6c) with MTT metabolism (Fig. 6d) showed an effect on reduction in MTT resulting from exposure to DETA/NO but less so on cell death, indicating that not all metabolically compromised cells necessarily die. Conversely, MK801 did not protect neurones from prolonged (3 days) exposure to high levels of DETA/NO (0.5 mm; Figs 6c and d).

Figure 5.

Expression of NMDA receptor subunits in cultured cortical neurones. Cortical neurones were stained with a monoclonal antibody against the NR1-subunit of NMDA receptors (red) after 1 day (a and b), 4 days (c and d) or 12 days (e and f) in culture. Phase-micrographs of fixed cells were overlaid (using Openlab 2.2.5) with immunocytochemical stainings (a, c and e) to result in the images (b, d and f). An increase in NMDA–NR1-expression was evident with maturation of neurones in culture. (g) For western blots, 3.75, 7.5 and 15 µg total cell lysate from neurones after 1, 4 and 12 days in vitro were blotted and immunostained with anti-NMDA–NR1 antibody. Representative signals for NMDA–NR1 and β-Actin (control) from 7.5 µg total cell lysate are shown. A very strong signal for NMDA–NR1 was also obtained from the lysate of 12-day-old cultures when only 3.75 µg total protein were loaded, whereas lysates of 1-day-old cultures showed no NMDA–NR1-signal in any of the amounts tested, clearly demonstrating the up-regulation of NMDA-NR1-expression in neurones with maturation. (h) For analysis of NMDA–NR2A and NMDA–NR2B expression 7.5, 15 and 30 µg total cell lysate from neurones after 1, 4 and 12 days in vitro were blotted and immunostained with antibodies against NR2A, NR2B and β-Actin (control). Representative signals from 15 µg total cell lysate are shown.

Figure 6.

Microglial nitric oxide mediates its effect on more mature cortical neurones partially via NMDA receptor-activation. After maturation in culture for 12 days, cortical neurones were co-cultured with microglia for 3 days. The number of β-tubulin-positive (B4-Isolectin-negative) neurones (a) and MTT metabolism (b) were greatly reduced in co-cultures activated with IFNγ + LPS. Treatment with 1400 W abolished this effect completely. NMDA receptor-blocker MK801, which had no effect on younger neurones, significantly increased neurone number to 70% (a) and MTT metabolism to 60% (b) of the control value. The combination of 1400 W and MK801 had no effect above the effect of 1400 W alone. 1400 W and MK801 either alone or in combination were not toxic to microglial or neuronal control cultures, as established by MTT assay in parallel control cultures (not shown). (c) After 12 h of incubation with the NO donor DETA/NO (0.5 mm), 50% of the neurones (12 days in vitro) had died (propidium iodide-positive); however, in the presence of MK801 (10 µm), this early cell death was significantly attenuated (15% propidium iodide-positive neurones). After prolonged incubation with the NO donor for 72 h, most neurones had died and MK801 did not afford significant protection. (d) MTT metabolism was reduced compared to cell number (see c) at 12 h in DETA/NO-treated cultures both in the presence and absence of MK801, demonstrating a metabolic compromise following exposure to NO, whereas after 72 h MTT metabolism paralleled the cell death in (c). *p < 0.05, **p < 0.01, ***p < 0.001, n.s. = not significant (one-way anova and Newman–Keuls post-hoc test).


In this study, we demonstrate that microglial nitric oxide is essential and sufficient to kill co-cultured cortical neurones. Whereas NO is directly toxic to immature cortical neurones, it acts via activation of NMDA receptors in more mature cortical neurones.

INOS-derived NO is necessary and sufficient to mediate neurotoxicity from activated microglia

Microglia-induced neuronal death has previously been described but the mechanisms are not fully characterized. (Piani et al. 1991, 1992; Boje and Arora 1992; Chao et al. 1992; Giulian et al. 1993; Dawson et al. 1994; McMillian et al. 1995; Matsuoka et al. 1999; Jeohn et al. 2000; Bal-Priceand Brown 2001). Potent excitotoxins recovered from medium conditioned by microglia, either untreated or treated with LPS or IFNγ (Piani et al. 1991), PMA (Piani et al. 1992) or Zymosan A (Giulian et al. 1993), have been implicated as the principal mediators of microglial toxicity to cerebellar neurones (Piani et al. 1991, 1992) and ciliary ganglion cells (Giulian et al. 1993), acting via NMDA receptors. However, none of these studies tested for the involvement of NO, shown in separate studies to be critical for microglial toxicity towards cortical neurones (Chao et al. 1992; Jeohn et al. 2000), cerebellar granule cells (Boje and Arora 1992; Bal-Price and Brown 2001) or cholinergic neurones (McMillian et al. 1995) using microglia activated by IFNγ and LPS (Chao et al. 1992; Dawson et al. 1994; Jeohn et al. 2000; Bal-Price and Brown 2001), LPS and several cytokines (Boje and Arora 1992) or LPS alone (McMillian et al. 1995).

These studies, using mixed neuronal–glial cultures and mostly employing isoform non-specific inhibitors of NOS, leave unresolved the issues of whether microglia act directly on neurones or via a bystander cell, for example by secreting IL-1 to induce astrocytic release of NO (Chao et al. 1995); whether the toxic NO is produced by the inducible isoform (iNOS) in microglia or astrocytes or by nNOS, in neurones, possibly downstream of NMDA receptor-activation (Dawson et al. 1991); and whether, if microglial NO is causative, it acts directly or by influencing the secretion of other toxic mediators.

In the present work, using a highly specific inhibitor of iNOS (1400 W) in a co-culture system virtually free from astrocytes, we demonstrate that NO is necessary for neurotoxicity and that the source of neurotoxic NO is microglial iNOS. Unlike selective inhibition of neuronal NOS, inhibition of iNOS with aminoguanidine or the non-isoform selective NOS inhibitor l-NAME protect against microglial neurotoxicity. This underlines the importance of microglia-derived nitric oxide for neuronal death in the activated co-cultures.

NO might not be directly neurotoxic but might act by triggering the secretion of other potentially toxic molecules, such as TNFα or excitotoxins. NO itself influences microglial activation, as inhibition of NO-synthesis reduces secretion of TNFα (Deakin et al. 1995; de Bock et al. 1998) and glutamate (Kingham et al. 1999). Therefore, iNOS inhibition could be indirectly neuroprotective. However, we find that microglia-conditioned medium is not neurotoxic and consequently, NO cannot act indirectly by increasing secretion of stable toxins such as glutamate or TNFα from microglia. Furthermore, we show that soluble TNFα receptor, which efficiently blocked toxicity of microglia-derived TNFα to L929 cells, does not rescue immature rat cortical neurones, excluding TNFα as mediator of microglial toxicity to these neurones.

The NO donor DETA/NO is directly neurotoxic in doses that produce the same cumulative amount of NO as measured in our co-cultures – demonstrating that NO alone is sufficient to cause neuronal death. The comparison of NO production and viability of immature neurones indicates that a relatively high threshold level of NO (> 20 µm nitrite/3 days) is necessary to cause neuronal death, and this is not reached after activation with IFNγ or LPS alone. LPS becomes neurotoxic for rat cortical neurone–glia cultures when the number of microglia and therefore the total NO production are greatly increased (Kim et al. 2000), supporting our conclusion that the amount of NO production rather than a different repertoire of microglial factors determines LPS toxicity to neurones.

Peroxynitrite is not involved in microglial neurotoxicity

When NO is secreted it quickly reacts with O2 to produce N2O3 and NO2, or with O2 to produce ONOO (Murphy 2000). It is not clear when toxic effects are due to NO or its reaction products. We found that whereas oxyhaemoglobin, a nitric oxide scavenger, decreased the neurotoxicity of microglia, there was no effect on cultures treated with the cell-permeating SOD-mimetic and ONOO scavenger MnTBAP or the ONOO scavenger uric acid. B27 supplement, containing various antioxidants, may have attenuated the effects of oxygen radicals. It follows that the substantial neuronal loss we observed is most likely due to actions of the NO molecule itself, rather than generalized ONOO production, although a local increase in superoxide is not excluded. Nor can a contribution of microglia-derived hydrogen peroxide be excluded as treatment of microglia-neuronal co-cultures with catalase strongly reduced NO production from microglia and protected neurones. Although unexpected, a catalase-mediated change in intracellular oxidative status might influence the likelihood for NFκB activation, a pathway that is also involved in the transcriptional up-regulation of iNOS (Baeuerle et al. 1996). However, catalase did not prevent neuronal death induced by DETA/NO. Therefore, our results could be consistent with part of the neurotoxic effect of microglia being mediated by H2O2 but we exclude a role for H2O2 in the toxicity of NO donor for cortical neurones.

Intraneuronal mechanisms of NO-induced cell death in immature neurones

The cellular targets of NO itself or its reaction products responsible for neuronal death are not fully defined. Direct chemical modification of DNA has been linked to neurotoxicity (Murphy 2000). DNA damage may lead to prolonged activation of the DNA repair enzyme Poly (ADP-ribose)-polymerase (PARP), which then depletes the cell of NAD+ and ATP, followed by necrosis (Zhang et al. 1994). However, this mechanism does not seem to be a major mechanism of NO-dependent cell death of immature cortical neurones since inhibition of PARP does not substantially reduce neurotoxicity of DETA/NO.

NO exposure can furthermore increase cytosolic calcium levels, for instance by activation of calcium release from mitochondria (Schweizer and Richter 1994) or indirect activation of NMDA receptors, and stimulate calcineurin, which in turn can activate key regulators of apoptosis, such as Bad, leading to cell death (Wang et al. 1999). We tested for involvement of calcineurin in DETA/NO-induced death of immature neurones, using a specific inhibitor (FK 506) but found no effect.

NO and its derivatives can influence mitochondrial function at multiple levels. Mitochondrial permeability transition (MPT) can be induced by NO resulting in neuronal cell death (Bosca and Hortelano 1999; Borutaite et al. 2000). However, blocking MPT with cyclosporin A did not attenuate NO-dependent death of immature neurones.

NO produced from cytokine-stimulated cells can inhibit respiration of donor and target cells in co-culture (Brown et al. 1995, 1998) and this has been implicated as a major mechanism of NO-induced cell death (Heales et al. 1999). NO itself has been shown not only to cause a reversible inhibition of complex IV of the respiratory chain but also irreversibly to inhibit complex I (Clementi et al. 1998), and NO donors cause death of PC12 cells by blocking respiration (Bal-Price and Brown 2000). Direct and potent inhibition of mitochondrial function by NO could therefore be a possible mechanism for the direct toxicity of NO to immature neurones.

NO induces cell death in immature and mature cortical neurones by different mechanisms

In more mature cultures, microglial NO still causes neuronal death. However, in contrast to immature cortical neurones, neurones cultured for 12 days express both NR1 and NR2 subunits of the NMDA receptor and microglial toxicity is now partially dependent on its activation. The observation that 1400 W alone completely rescues more mature neurones from microglial toxicity shows that NMDA receptor-activation cannot kill independent of NO. Excitotoxins are probably not derived from activated microglia, as conditioned medium has no effect and therefore the most likely source is neuronal. In isolated brain synaptosomes, NO donors lead to Ca2+-independent glutamate release via inhibition of cytochrome oxidase, ATP depletion and reversal of the Na+/glutamate co-transporter (McNaught and Brown 1998). We show here that treatment with the nitric oxide donor DETA/NO for 12 h has a greater effect on MTT metabolism than cell count, demonstrating that mitochondrial activity is indeed inhibited in viable cells. Addition of the NMDA receptor blocker MK801 leads to a marked survival of neurones exposed to DETA/NO at this early time point, indicating that NO elicits an excitotoxic mechanism. Bal-Price and Brown (2001) suggest a mechanism for NO-induced excitotoxicity showing that inhibition of respiration by DETA/NO or specific respiratory chain inhibitors causes rapid ATP decrease and release of glutamate from cerebellar granule neurones. This glutamate release is required for NO-mediated death of cerebellar granule neurones as inhibition of NMDA receptor-activation prevents iNOS-dependent death in glial–neuronal co-cultures and delays neuronal death from exposure to DETA/NO. Glutamate released from neurones is furthermore implicated in irreversible inhibition of respiration in neurones following NO exposure (Stewart et al. 2002), pointing to a vicious circle of mitochondrial damage and NMDA receptor-activation culminating in neuronal death. The present study and the work of Bal-Price and Brown (2001) each show that nitric oxide can kill different NMDA receptor-expressing neurones, cerebellar and cortical, by eliciting excitotoxicity. However, the results of this study indicate that this is not exclusively the pathway leading to neuronal death, as immature cortical neurones are equally sensitive to NO yet do not die through an excitotoxic mechanism.

Our findings demonstrate that NO produced by activated microglia is necessary and sufficient to mediate neurotoxicity but also reveal that similar concentrations of microglial NO act through at least two distinct mechanisms depending on maturation of the target neurones: NO is directly toxic to immature neurones but acts primarily through an excitotoxic mechanism in more mature cells that express NMDA receptors.


We thank Dr Jonathan Lindquist and Dr Martin Stangel for helpful discussions. This work was supported by a Marie-Curie Individual Fellowship of the European Commission (SG) the Raymond and Beverley Sackler Research Centre (SG), and the Medical Research Council (SG and SC).