Tumor necrosis factor alpha stimulates NMDA receptor activity in mouse cortical neurons resulting in ERK-dependent death

Abstract

Multiple cytokines are secreted in the brain during pro-inflammatory conditions and likely affect neuron survival. Previously, we demonstrated that glutamate and tumor necrosis factor alpha (TNFα) kill neurons via activation of the N-methyl-d-aspartate (NMDA) and TNFα receptors, respectively. This report continues characterizing the signaling cross-talk pathway initiated during this inflammation-related mechanism of death. Stimulation of mouse cortical neuron cultures with TNFα results in a transient increase in NMDA receptor-dependent calcium influx that is additive with NMDA stimulation and inhibited by pre-treatment with the NMDA receptor antagonist, dl-2-amino-5-phosphonovaleric acid, or the α-amino-3-hydroxy-5-methylisoxazole-4-pro-pionate/kainate receptor antagonist, 6,7-dinitroquinoxaline-2,3-dione. Pre-treatment with N-type calcium channel antagonist, ω-conotoxin, or the voltage-gated sodium channel antagonist, tetrodotoxin, also prevents the TNFα-stimulated calcium influx. Combined TNFα and NMDA stimulation results in a transient increase in activity of extracellular signal-regulated kinases (ERKs) and c-Jun N-terminal kinases (JNKs). Specific inhibition of ERKs but not JNKs is protective against TNFα and NMDA-dependent death. Death is mediated via the low-affinity TNFα receptor, TNFRII, as agonist antibodies for TNFRII but not TNFRI stimulate NMDA receptor-dependent calcium influx and death. These data demonstrate how microglial pro-inflammatory secretions including TNFα can acutely facilitate glutamate-dependent neuron death.

Tumor necrosis factor alpha (TNFα) is a pleotropic cytokine capable of stimulating a variety of cell types with a diverse set of consequences (Locksley et al. 2001; Watts 2005). Its effects on target cells occur via binding to two receptors, CD120a (p55, TNFRI) and CD120b (p75, TNFRII) (Hehlgans and Mannel 2002). Both neurons and glia have been reported to express these receptors (Kinouchi et al. 1991; Dopp et al. 1997). The cytoplasmic domain of TNFRI contains a ‘death domain’ (DD) sequence of 60–80 amino acids responsible for recruiting the adaptor protein TRADD, capable of transducing diverse signaling pathways including activation of nuclear factor κB (NFκB) or caspases (Bradley and Pober 2001; MacEwan 2002). Although TNFRII lacks the DD sequence, it contains a region of 78 amino acids responsible for recruiting the adaptor protein, TRAF2, allowing subsequent activation of NFκB as well (Rothe et al. 1995). A conflicting body of literature has demonstrated that neuronal stimulation via either receptor mediates protection or death (Barger et al. 1995; Li et al. 2004).

In several experimental systems, TNFα stimulation mediates toxicity in the context of a pro-inflammatory milieu, suggesting that its effects are superimposed upon responses initiated by additional stimuli (Gelbard et al. 1993; Chao and Hu 1994; Floden et al. 2005). Expression and secretion of TNFα, particularly by microglia, has been observed in neurotoxic paradigms and implicated in mechanisms of neuron loss (Chao et al. 1995; Combs et al. 2001). For example, elevated central nervous system TNFα levels have been reported from multiple sclerosis, Alzheimer's disease (AD), stroke/ischemic, traumatically injured, and epileptic brains (Tchelingerian et al. 1993; Akiyama et al. 2000; Gimsa et al. 2000; Yin et al. 2003; Ravizza et al. 2005).

N-methyl-d-aspartate (NMDA) receptor activity-dependent death is one such response likely modulated by TNFα stimulation. NMDA receptor activity is regulated by a variety of post-translational modifications, including nitrosylation and phosphorylation (Wang and Salter 1994; Choi et al. 2000), which allows varying neuronal stimuli to modulate NMDA receptor-dependent calcium influx (Rostas et al. 1996; Yu et al. 1997). Increased NMDA receptor activity can facilitate the well-characterized excitotoxic death mechanism within neurons. Although this response is typically dependent upon elevated intracellular calcium levels, excitotoxic death can also require activation of members of the mitogen-activated protein (MAP) kinase family (Satoh et al. 2000; Hughes et al. 2003).

Our previous work demonstrated that microglia stimulated with β-amyloid peptide secretes TNFα and glutamate to kill mouse cortical neuron cultures over a 72-h time course. Cell death was dependent upon coincident stimulation of TNFα and NMDA receptors and subsequent activation of neuronal inducible nitric oxide synthase (iNOS) (Floden et al. 2005). Based upon prior studies, we hypothesized that this death mechanism involved a specific cross-talk response allowing TNFα receptor stimulation to modulate NMDA receptor-dependent calcium influx and activation of MAP kinases. To determine whether a TNFα and NMDA receptor-dependent signaling cross-talk event was responsible for the death of our cultures, we have employed our same primary mouse neuron culture system treated with recombinant TNFα and NMDA and quantitated effects on NMDA receptor-dependent calcium influx, intracellular signaling responses and resultant death.

Our findings demonstrate a mechanism by which NMDA receptor activation leads to extracellular signal-regulated kinase (ERK)-dependent neuronal death in the presence of the appropriate cytokine containing environment and provide insight into a neuron loss mechanism relevant to AD and other inflammatory neurodegenerative conditions.

Materials and methods

Materials

The anti-phosphoERK antibody (pTyr-204), anti-ERK2 antibody, and affinity-purified horseradish peroxidase-conjugated secondary antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). For immunostaining, polyclonal anti-phosphoERK was acquired from Cell Signaling Technology (Beverly, MA, USA). Anti-phospho-c-N-terminal kinase (JNK) and anti-JNK antibodies were obtained from cell signaling. The specific JNK inhibitor (Bonny et al. 2001), JNK peptide inhibitor 1,d-stereoisomer (d-JNKI1) and the specific mitogen-activated protein kinase kinase (MEK) inhibitor (Favata et al. 1998), 1,4-diamino-2,3-dicyano-1,4-bis (2-aminophenylthio) butadiene (UO126) were both purchased from Alexis Biochemicals (Carlsbad, CA, USA). TNFα (cat. # 410-MT), TNFRI agonist antibody (Pollock et al. 2002; Soond et al. 2003) (cat. # AF-425-PB), brain-derived neurotrophic factor (BDNF), and interleukin-1β (IL-1β) were purchased from R& D Systems (Minneapolis, MN, USA). TNFRII agonist antibody (rat monoclonal HM102) (cat. # ab7369) was purchased from Novus Biologicals Inc. (Littleton, CO, USA). According to manufacturer specifications, the agonistic properties of the antibody were tested in a proliferation assay with mouse thymocytes (thymidine uptake was measured), which showed that 3 μg/mL (HM102) leads to cell activation (0.3 and 1 μg/mL did not lead to cell activation). Normal rat IgG (negative control for TNFRII agonist antibody) (cat. # sc-2026) was purchased from Santa Cruz Biotechnology. TNFα was resuspended in sterile phosphate-buffered saline (PBS) containing 1% bovine serum albumin (BSA) for use. In some cases, TNFα and TNFRI and TNFRII agonist antibodies were dialyzed to remove any manufacturer contaminants before stimulation. To dialyze TNFRI agonist antibody, TNFRII agonist antibody, and TNFα, we used disposable Slide-A-Lyzer MINI Dialysis Units (3.5 K MWCO) from Pierce Biotechnology Inc. (Rockford, IL, USA) according to the manufacturer instructions in PBS containing 1% BSA at 4°C. The 6,7-dinitroquinoxaline-2,3-dione (DNQX) (cat. # 0189) was purchased from TOCRIS Bioscience (Ellisville, MO, USA). The ω-conotoxin GVIA (cat. # CA-235) was purchased from BIOMOL Research Laboratories Inc. (Plymouth Meeting, PA, USA). The dl-2-amino-5-phosphonovaleric acid (APV), tetrodotoxin, NMDA, glutamate, and all other materials not specified were purchased from Sigma (St Louis, MO, USA).

Tissue culture

Neurons were cultured from cortices of E16 mice (C57B1/6J). Briefly, meninges-free cortices were isolated, trypsinized and plated onto poly-l-lysine-coated (0.05 mg/mL) tissue culture wells (260 cell/mm²) for 7 days in vitro before use. Neurons were grown in glutamine containing Neurobasal media with B27 supplements (Life Technologies, Rockville, MD, USA) to consistently provide neuronal cultures greater than 95% pure and able to survive for at least 1 month in vitro. Culture purity is routinely evaluated by cell counting after immunostaining, to identify the neuronal cytoskeletal protein, microtubule-associated protein 2 (MAP2).

Cell stimulation

Neurons were stimulated for varying times with or without kinase inhibitors (UO126, d-JNKI1), tetrodotoxin, ω-conotoxin, APV, DNQX, IL-1β, TNFα, BDNF, NMDA, TNFRI agonist antibody, TNFRII agonist antibody, or glutamate. For western blot analysis, cells were lyzed as below after specified times. To perform immunohistochemical analysis, cells were fixed as described below at defined times.

Neuron viability

To assess neuronal survival, neurons were fixed following 72 h stimulation, stained using a neuron-specific anti-MAP2 antibody, and a counting grid was placed under the wells to count neuron numbers from four identical fields for each condition. MAP2-positive cells with visible nuclei and immunostained processes, at least one cell diameter in length, are counted as neurons. The average number of neurons (±SEM) was calculated for each condition. Each experiment was performed in quadruplicate three to four times. In addition, neuronal viability was determined by measuring the cellular release of lactate dehydrogenase (LDH) from culture media of neurons (eight wells/condition) stimulated for 72 h using a commercial non-radioactive assay (Promega, Madison, WI, USA). Absorbance measurements were taken at 490 nm. Neuron viability was also determined by repetitive counting of phase bright neurons. Cells from the same field (20× magnification), each from six sister wells per condition were repetitively counted at 24, 48, and 72 h and averaged (±SD).

Western blotting

Neurons were stimulated as described above, then lyzed in ice-cold RIPA buffer [1% Triton, 0.1% sodium dodecyl sulfate (SDS), 0.5% deoxycholate, 20 mmol/L Tris (pH 7.4), 150 mmol/L NaCl, 10 mmol/L NaF, 1 mmol/L Na₃VO₄, 1 mmol/L EDTA, 1 mmol/L EGTA, and 0.2 mol/L phenylmethylsulfonyl fluoride] and insoluble material was removed by centrifugation (14 000 g, 10 min, 4°C). Protein concentrations were quantitated by Bradford method (Bradford 1976). Proteins were resolved by 7.5% or 10% SDS – PAGE and incubated in primary antibody overnight, 4°C, followed by incubation in 2° horseradish peroxidase-conjugated antibody. Antibody binding was detected via enhanced chemiluminescence (Pierce, Rockford, IL, USA).

Immunohistochemistry

To perform immunocytochemistry, neurons were stimulated as described above and fixed in 4% paraformaldehyde, 37°C, 30 min at selected times. Fluorescent immunodetection of active ERK was performed using anti-phosphoERK antibody. Antibody binding was visualized using Texas Red conjugated anti-rabbit secondary antibody (Santa Cruz). To visualize the nucleus, 4′,6′-diamidino-2-phenylindole (Invitrogen, Carlsbad, CA, USA) was used. Images were captured using a Zeiss (Zeiss, Thornwood, NY, USA) LSM 510 Meta confocal microscope.

Measurement of intracellular Ca²⁺

Primary cortical neurons were grown on cover slips (MatTek Cultureware, Ashland, MA, USA) and incubated with 2 μmol/L Fura-2 (Molecular Probes, Eugene, OR, USA) for 45 min at 37°C (5% CO₂–95% air). The cells were washed twice with calcium-containing buffer (10 mmol/L HEPES, 120 mmol/L NaCl, 5.4 mmol/L KCl, 1 mmol/L MgCl₂, 1 mmol/L CaCl₂, 10 mmol/L glucose, pH 7.4). Cells were then pre-treated in the calcium-containing buffer with and without drugs (30 min) and bath stimulated in the same buffer, once per culture, to assess changes in intracellular calcium levels. To measure stimulated fluorescence changes, fluorescence intensity of Fura-2-loaded cells was monitored with a CCD camera-based imaging system (Compix Inc., Cranbery, PA, USA) mounted on an Olympus XL70 inverted microscope equipped with an Olympus 40× (1.3 NA) fluor objective. A monochrometer dual wavelength enabled alternative excitation at 340 and 380 nm, whereas the emission fluorescence was monitored at 510 nm with an Okra Imaging camera (Hamamatsu Photonics, Hamamatsu, Japan). The images of multiple cells collected at each excitation wavelength were processed using the C imaging, PCI software (Compix) to provide ratios of Fura-2 fluorescence from excitation at 340 nm to that from excitation at 380 (F₃₄₀/F₃₈₀). Analog plots of the fluorescence ratio (340/380) in single cells are shown.

Statistical analysis

Experiments were performed in quadruplicate at least three times unless noted. Mean values (±SEM) for each experiment were determined and values statistically different from controls were calculated using one-way anova. The Tukey–Kramer multiple comparisons post-test was used to determine p-values.

Animal care and use

All procedures were reviewed and approved by the University of North Dakota Institutional Animal Care and Use Committee. Mice were housed at the Center for Biomedical Research on a 12 h light/dark cycle and were allowed food and water ad libitum.

Results

TNFα stimulation potentiates NMDA receptor activity and ERK-dependent neuron death

Our prior studies demonstrated that TNFα stimulation of mouse cortical neuron cultures was sufficient to induce death in the presence of non-toxic concentrations of NMDA or glutamate over a delayed time course of 72 h (Floden et al. 2005). To determine the nature of the signaling cross-talk response, we first assessed whether TNFα stimulation led to changes in NMDA receptor-mediated calcium influx (Fig. 1). Stimulation with TNFα alone resulted in a rapid increase in calcium influx that was inhibited via pre-treatment with the NMDA receptor antagonist, APV (Figs 1b and c). We next pre-treated neurons with an α-amino-3-hydroxy-5-methylis-oxazole-4-propionate (AMPA)/kainate receptor antagonist, DNQX (Kendrick et al. 1996), to determine whether the TNFα-dependent calcium influx occurred via specific modulation of only NMDA receptor activity. DNQX pre-treatment also inhibited the ability of TNFα to stimulate changes in calcium influx, suggesting that TNFα was not affected via direct modulation of NMDA receptor activity (Fig. 1c). To assess whether basal electrical activity was required for the TNFα-mediated change in NMDA receptor activity, we first pre-treated neurons with the voltage-gated sodium channel antagonist, tetrodotoxin, prior to stimulation with TNFα. Tetrodotoxin pre-treatment abolished the ability of TNFα to stimulate changes in NMDA receptor-mediated calcium influx, suggesting that neuronal electrical activity is required for TNFα effects (Fig. 1d). To further establish whether synaptic transmitter release was needed to evoke the change in calcium influx, we pre-treated the neuronal cultures with the N-type calcium channel blocker, ω-conotoxin GVIA (Horne and Kemp 1991). As predicted, ω-conotoxin prevented the TNFα-mediated calcium influx, suggesting that synaptic glutamate release was required (Fig. 1d). Interestingly, combined TNFα and NMDA stimulation did not lead to a significant increase in peak calcium influx values but rather a sustained increase in calcium influx compared with NMDA stimulation alone (Fig. 1e). Collectively, our results suggest that TNFα modulates synaptic release of glutamate to mediate an NMDA receptor-dependent calcium influx that is dependent upon initial activity of the AMPA/kainate receptors.

Fig. 1.

TNFα stimulates increased intracellular Ca²⁺ levels in primary cortical neurons. Neurons were loaded with Fura-2 then stimulated to measure changes in Ca²⁺ influx in NMDA and TNFα-stimulated primary cortical neurons (E16, 7 days in vitro), in the presence or absence of APV, DNQX, ω-conotoxin, and tetrodotoxin. (a) Demonstrates representative fluorescence traces from NMDA (100 μmol/L) stimulated cells in the absence or presence of APV following pre-treatment with the drug (25 μmol/L, 30 min). (b–d) Contain representative traces from TNFα (50 ng/mL) and dialyzed TNFα stimulated cells in the absence or presence of APV (25 μmol/L), DNQX (50 μmol/L), tetrodotoxin (1 μmol/L), and ω-conotoxin (9 μmol/L), 30 min pre-treatment. (e) Shows a representative trace following dual stimulation with TNFα and NMDA. (f) Demonstrates a representative trace from cells stimulated with PBS only. (g) Represents average peak calcium influx values (±SEM) in each set of treatments from three independent experiments. The number of cells imaged is indicated in brackets (*p < 0.05 from NMDA).

Based upon our observations of TNFα-mediated changes in calcium influx, we hypothesized that the cytokine would play a role in NMDA receptor-dependent toxicity. Although our cultured mouse cortical neurons at 7 days in vitro are relatively resistant to NMDA-dependent toxicity, our prior work has demonstrated that TNFα induces robust death of these cells in combination with NMDA receptor stimulation (Floden et al. 2005). To assess the role of TNFα stimulation in NMDA receptor-dependent excitotoxic death, we performed a time course survival analysis of neurons stimulated with or without TNFα and NMDA as previously described (Floden et al. 2005). Cells were stimulated for increasing times (0, 24, 48, and 72 h) and evaluated for cell survival. Counting live cells with morphologic characteristics of neurons from the same field showed that neurons displayed a progressive decrease in viability over 72 h in the presence of combined TNFα and NMDA stimulation (Figs 2a and b). Importantly, this is consistent with our results derived from cell counting based upon immunocytochemical detection of neurons (Floden et al. 2005). As a further confirmation that neurons die during combined TNFα and NMDA treatment, media aliquots were collected from cells stimulated for 72 h for quantifying cellular release of LDH (Fig. 2c). Our results confirmed that combined treatment with NMDA and TNFα is required to reduce neuronal viability in our culture system. For consistency, we relied upon immunocytochemical detection of histologically characteristic neurons and cell counting to assess neuron survival in the subsequent studies.

Fig. 2.

TNFα and NMDA stimulation leads to neuron death after 72 h. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L), and mouse TNFα (50 ng/mL). Stimuli were added to neurons for 0, 24, 48, and 72 h and neuron viability was determined. (a) Representative fields of neurons were photographed during stimulation with and without TNFα and NMDA at each time point to demonstrate morphologic changes. Sister cultures were fixed at each age and immunostained with anti-MAP2 antibody and nuclei visualized by DAPI to demonstrate representative changes in MAP2 immunoreactivity at each time point (arrowheads indicate degenerated neurons). (b) The same field of phase bright, morphologically defined neurons from six sister cultures was counted at each time point and the average number of cells (±SD) was graphed for each condition. Data are representative of three independent experiments. (c) Media aliquots were taken from cells stimulated 72 h with or without TNFα and NMDA for 72 h and LDH concentrations were quantitated using a commercial assay. Data are representative of two independent experiments. (*p < 0.05 from control, ***p < 0.001 from control).

Although our data demonstrate a role for TNFα stimulation in NMDA receptor-dependent excitotoxic death, several studies have reported that TNFα stimulation can contribute to either trophic or toxic support of neurons (de Bock et al. 1998; Carlson et al. 1998; Marchetti et al. 2004; Floden et al. 2005; Zou and Crews 2005). Moreover, others have demonstrated that TNFα pre-treatment of cultured neurons will condition them to resist a subsequent, toxic stressor (Houzen et al. 1997; Furukawa and Mattson 1998; Marchetti et al. 2004). To determine whether acute or prolonged TNFα stimulation was required for potentiation of the NMDA-dependent death, we pre-treated our cultures with TNFα for 24 h prior to stimulation with NMDA or glutamate (Fig. 3a). TNFα pre-treatment had no ability to condition neurons to die via subsequent stimulation with NMDA or glutamate alone, demonstrating that the neurotoxic action of TNFα involved an acute effect coincident with NMDA stimulation.

Fig. 3.

TNFα and glutamate/NMDA stimulated neuron death is acutely stimulated and involves MAP kinase activation. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L), glutamate (25 μmol/L), mouse TNFα (1– 100 ng/mL), mouse IL-1β (1–100 ng/mL), and APV (5 and 25 μmol/L). (a) Stimuli were added to neurons for 72 h with and without a 24-h pre-treatment of 50 ng/mL TNFα. (b) Neurons were stimulated with NMDA and increasing concentrations of TNFα or IL-1β for 72 h. (a, b) After treatment, cells were fixed, stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.05 from control, **p < 0.001 from control). (c, d) For western blots, stimuli were added to neurons for 5 min with and without 30 min pre-treatment with APV. Cells were lyzed and proteins resolved by 10% SDS-PAGE and western blotted using anti-pERK, anti-ERK2 (loading control), anti-pJNK, and anti-JNK (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence.

Because a variety of factors can be secreted from reactive glia, it is possible that multiple cytokines beyond TNFα will utilize a similar signaling cross-talk response to modulate NMDA receptor-dependent events. For example, it has been reported that IL-1β stimulation also potentiates NMDA receptor-dependent death similar to our observations made using TNFα (Chao et al. 1995; Viviani et al. 2003). To determine whether the effects we observed were cytokine specific, we compared the abilities of both TNFα and IL-1β to potentiate NMDA receptor-mediated death. Although IL-1β was significantly toxic alone, it was unable to potentiate NMDA-dependent death at any concentration employed (Fig. 3b). On the contrary, TNFα stimulation resulted in a significant loss of neurons in a dose-dependent fashion in the presence of NMDA (Fig. 3b).

Next, we began identifying the mechanism of cell death that occurred following combined stimulation with TNFα and NMDA. As both in vivo and in vitro evidences demonstrate that ischemic and excitotoxic death can require activation of members of the MAP kinase family of serine/threonine kinases (Satoh et al. 2000; Sugino et al. 2000; Ferrer et al. 2001; Kuroki et al. 2001; Jin et al. 2002; Li et al. 2002; Noshita et al. 2002; Stanciu and DeFranco 2002; Hughes et al. 2003; Zablocka et al. 2003), we examined the neuron cultures for changes in activity of MAP kinases. All members of the MAP kinase family, JNK, p38 MAP kinase, and ERK have the capacity to mediate death in different paradigms (Morishima et al. 2001; Harper and Wilkie 2003; Hetman and Gozdz 2004). We first determined whether changes in active MAP kinase levels occurred during our toxic stimulation. Five minutes of combined TNFα and NMDA stimulation led to a rapid increase in levels of the active forms of ERKs and JNKs (Fig. 3c) but not p38 MAP kinase (data not shown). This increase was dependent upon NMDA receptor activity, as pre-treatment with 25 μmol/L APV prevented the increase in the active form of ERKs (Fig. 3d). More importantly, inhibition of ERK activity via maintained treatment with the MEK inhibitor, UO126, provided protection from TNFα and NMDA stimulated death in a dose-dependent fashion by 72 h (Figs 4a and b). Conversely, maintained treatment of the cultures with the JNK inhibitory peptide, d-JNKI1, had no ability to rescue neurons from TNFα and NMDA-mediated death (Figs 4c and d). Interestingly, JNK inhibition alone using the inhibitory peptide was significantly toxic to our cultures at 72 h, suggesting that a basal level of JNK activity is required for neuron viability in our system.

Fig. 4.

TNFα and NMDA-stimulated neuron death requires ERK activity. Mouse cortical neurons (E16, 7 days in vitro) were stimulated with NMDA (100 μmol/L) and/or mouse TNFα (50 ng/mL), in the absence or presence of pre-treatment (30 min) with (a) UO126 (10, 50, and 100 nmol/L), or (c) D-JNKI1 (100, 500, and 1000 nmol/L). Stimuli were added to neurons with and without drugs for 72 h then cells were fixed, stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.01 from control, **p < 0.001 from control, ***p < 0.05 from TNFα+NMDA, ****p < 0.001 from TNFα+NMDA). For western blots, stimuli were added to neurons for 5 min following pre-treatment (30 min) with (b) UO126 (100 nmol/L) or (d) D-JNKI1 (100, 500, and 1000 nmol/L). Cells were lyzed and proteins resolved by 10% SDS-PAGE and western blotted using anti-pERK, anti-ERK2 (loading control), anti-pJNK, and anti-JNK (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence.

ERK-dependent neuron death is unique to stimulation with TNFα and NMDA

However, alteration of ERK activity is a common event in many signal transduction pathways. To assess whether increased ERK activity itself was sufficient to initiate a toxic signaling response, we examined other ERK activating stimuli for their effects on neuronal viability. Although toxic IL-1β stimulation resulted in elevated levels of active ERKs, there was no potentiation of cell loss when the stimulation was combined with NMDA (Fig. 5). Neurons stimulated with BDNF also demonstrated elevated levels of active ERKs. In this case, ERK activity was potentiated when combined with TNFα stimulation (Fig. 6a). However, neither BDNF alone nor combined BDNF and TNFα stimulation produced a decrease in cell viability. Instead, both treatments resulted in a significant increase in cell survival compared with unstimulated controls (Fig. 6b) consistent with the neurotrophic action of BDNF. These results collectively demonstrate that increased ERK activity itself is not toxic to our cultures but rather is a required component of the signaling response initiated by TNFα and NMDA.

Fig. 5.

IL-1β stimulation does not potentiate NMDA-dependent death and ERK activity. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L) and mouse IL-1β (50 ng/mL). (a) For western blots, stimuli were added to neurons for 5 min then cells were lyzed and proteins resolved by 10% SDS-PAGE and western blotted using anti-pERK, anti-ERK2 (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence. (b) Stimuli were added to neurons for 72 h then cells were fixed, stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.001 from control).

Fig. 6.

TNFα and BDNF stimulation does not result in neuron death. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L), mouse TNFα (50 ng/mL) and human BDNF (50 ng/mL). (a) For western blots, stimuli were added to neurons for 5 min then cells were lyzed and proteins resolved by 10% SDS-PAGE and western blotted using anti-pERK, anti-ERK2 (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence. (b) Stimuli were added to neurons for 72 h then cells were fixed, stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.01 from control, **p < 0.001 from control, ***p < 0.001 from TNFα + NMDA).

To better characterize the increase in ERK activity required for TNFα and NMDA-mediated death, we next determined the temporal and spatial profile of activity following stimulation. Although non-toxic NMDA stimulation alone resulted in a prolonged increase in levels of active ERKs, the toxic TNFα and NMDA combined stimulation produced an abbreviated time course of activation (less than 30 min) (Fig. 7a). After observing this attenuation and considering the fact that ERK activity was required for death, we determined whether transient stimulation was sufficient for the delayed death we observed. As expected, transient 15 min stimulation with TNFα and NMDA was sufficient to induce neuron death by 72 h (Fig. 7b). Finally, we assessed the spatial profile of ERK activity in our stimulated neurons to determine whether kinase activity was localized to specific subcellular compartmentation during toxic stimulation (Fig. 8). Active ERK immunoreactivity was visualized following different stimulation times to establish its temporal and spatial profile. Combined TNFα and NMDA stimulated active ERK immunoreactivity within the somatodendritic region of the neurons at 5 min, which decreased after 15 min of combined stimulation. Nevertheless, the spatial staining pattern was not different from that induced by NMDA stimulation alone. Taken together, these data demonstrated that neurotoxic ERK activity appeared dependent upon a temporal rather than spatial activation pattern. Time course differences in ERK activity reportedly produce different substrate specificities and, more importantly, differences in stimulated transcriptional alteration (Stanciu and DeFranco 2002; Chu et al. 2004; Levinthal and DeFranco 2004; Rusanescu et al. 2005). Indeed, our prior work demonstrated that increased protein levels and activity of iNOS are required for combined TNFα and NMDA-dependent death (Floden et al. 2005).

Fig. 7.

TNFα and NMDA stimulate a transient increase in ERK activity. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L) and mouse TNFα (50 ng/mL). (a) For western blots, stimuli were added to neurons for increasing time then cells were lyzed and proteins resolved by 10% SDS-PAGE and western blotted using anti-pERK, anti-ERK2 (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence. (b) Stimuli were added to neurons for increasing times then removed and replaced with fresh Neurobasal media. Cells were fixed after 72 h then stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.001 TNFα+NMDA treatment compared with paired time control).

Fig. 8.

Active ERK immunoreactivity increases after TNFα and NMDA treatment. Mouse cortical neurons were cultured for 7 days then stimulated, 5, 15, and 30 min in the absence or presence of TNFα (50 ng/mL) and NMDA (100 μmol/L). Cells were fixed in 4% paraformaldehyde, 37°C, and then immunostained using anti-pERK antibody. Antibody binding was visualized using Texas Red conjugated secondary antibody.

Stimulation via TNFRII and not TNFRI mediates NMDA-dependent neuron death

As mentioned above, although the biological consequences of TNFα stimulation are typically characterized as components of TNFRI-mediated signaling events, neurons express both low- and high-affinity receptors. As both receptors have been implicated in TNFα-dependent effects on neurons, we determined which receptor subtype was required for mediating the toxicity with NMDA. Activation of the specific TNFα receptors was achieved via stimulation with subtype-specific agonist antibodies in the absence or presence of NMDA (Figs 9a and b). Agonist antibodies for TNFRII but not TNFRI resulted in increased active ERK levels and subsequent neuron death when combined with NMDA, demonstrating that the TNFRII-dependent signaling events mediated the cross-talk with the NMDA receptor (Fig. 9c). Finally, to verify that stimulation of TNFRII mediated an increase in NMDA-dependent calcium influx similar to that observed using TNFα (Fig. 1), we stimulated cells as before using the agonist antibodies for TNFRI and TNFRII (Fig. 10). As expected, TNFRII agonist antibodies stimulated a transient increase in intracellular calcium levels, which was inhibited by pre-treatment with APV, DNQX, and ω-conotoxin (Figs 10b and c). On the contrary, TNFRI agonist antibodies had no effect on altering intracellular calcium levels (Fig. 10a).

Fig. 9.

Agonist antibodies for TNFRII potentiate NMDA-dependent neuronal death. Mouse cortical neurons (E16, 7 days in vitro) were cultured in the absence or presence of NMDA (100 μmol/L), or 0.01, 0.1, 1 μg/mL TNFRI or TNFRII dialyzed agonist antibodies. (a) Stimuli were added to neurons for 72 h then cells were fixed, stained using anti-MAP2 antibody, and counted. Neurons from four fields/condition were counted in quadruplicate wells and averaged ± SEM. Graphs are representative of three independent experiments. (*p < 0.001 from control). (b) For western blots, cells were stimulated for 5 min then lyzed, quantitated and proteins resolved via 10% SDS-PAGE and western blotted using anti-pERK and anti-ERK2 (loading control) antibodies. Antibody binding was visualized via enhanced chemiluminescence.

Fig. 10.

TNFRII agonist antibodies stimulate Ca²⁺ influx in primary cortical neurons. Primary cortical neurons (E16, 7 days in vitro) were loaded with Fura-2 and then stimulated to measure changes in Ca²⁺ influx in TNFα (50 ng/mL), TNFRI (1 μg/mL) agonist antibody, and TNFRII (1 lg/mL) agonist antibody stimulated cells with and without 30 min pre-treatment with APV (25 μmol/L), DNQX (50 μmol/L), or ω-conotoxin (9 μmol/L). (a) Demonstrates a representative trace from cells stimulated with TNFRI agonist antibody, dialyzed TNFRI agonist antibody, with and without APV pre-treatment. (b, c) Show representative traces from TNFRII agonist antibody, dialyzed TNFRII agonist antibody, rat IgG (isotype negative control) with and without pre-treatment with DNQX, ω-conotoxin, APV. (d) Represents average peak calcium influx values (±SEM) in each set of treatments from three independent experiments. The number of cells imaged is indicated in brackets. (*p < 0.05 from TNFRII).

Discussion

In this study we have demonstrated that TNFα stimulation of mouse cortical neurons acts via the low-affinity TNFα receptor, TNFRII, to modulate NMDA receptor-mediated calcium influx. Stimulation with TNFα or TNFRII agonist antibodies was not toxic to the neurons. However, when TNFα or TNFRII agonist antibody stimulation was combined with NMDA, a potent neurotoxicity was induced. More interestingly, combined TNFα and NMDA stimulation resulted in transient activation of ERKs compared with a prolonged activation observed following stimulation with NMDA alone. It is important to point out that although our data demonstrate that TNFα stimulates an NMDA receptor-dependent increase in calcium concentrations, this is likely a complex process involving several steps before the activation of the NMDA receptors. For example, we have demonstrated that AMPA/kainate receptor activity is required for the calcium concentration change as well as activity of N-type calcium channels and voltage-gated sodium channels. Collectively, these data suggest the hypothesis that TNFα stimulates synaptic glutamate release resulting in subsequent activation of AMPA receptors and membrane depolarization, leading to final NMDA receptor-dependent calcium influx. In support of this notion, prior studies have demonstrated that TNFα stimulation of dissociated neuron cultures stimulates AMPA receptor localization to the plasmalemma over a time scale of minutes in the range of the effects we observe (Ogoshi et al. 2005; Stellwagen and Malenka 2006). Future work is required to determine whether this AMPA receptor insertion and activation described by others is involved in the combined treatment of TNFα and NMDA, and ERK-dependent death we have observed. It is also possible that TNFα stimulation inhibits glutamate uptake by the small amount of contaminating glia in the cultures as a source of additional excitatory stimulation. Although we have not tested this hypothesis, this likely has an important role during in vivo, particularly, because TNFα stimulation has been reported to both increase and decrease glutamate transporter expression depending upon the paradigm (Romera et al. 2004; Korn et al. 2005).

Because of the changes we observed in active MAP kinase levels, we decided to assess the contribution of these enzymes to the resultant neuron loss. Cell death was dependent upon activation of ERKs and not JNKs, as treatment with the specific MEK inhibitor, UO126, and not the JNK inhibitory peptide, d-JNKI1, was able to prevent death stimulated by combined TNFα and NMDA stimulation. Complimentary results from experiments performed using similar systems have demonstrated a crucial role for robust ERK activation during NMDA-dependent death in both neuronal cell lines and cortical neuron cultures (Stanciu and DeFranco 2002; Levinthal and DeFranco 2004). Although we observed that ERK activation was required for cell death in our system, others have reported in different culture paradigms that the MAP kinase family members, p38 and JNK, can also propagate a neurotoxic signaling response (Cao et al. 2004). For example, JNK activity has been implicated in MDA receptor-dependent death in rat cortical neuron cultures (Centeno et al. 2006), phosphatase inhibition-associated death of rat cortical neuron cultures (Yoon et al. 2006), and trophic factor-dependent apoptosis in mouse dorsal root ganglia cultures (Besirli et al. 2005; Ben-Zvi et al. 2006). It is possible that the relative contribution of a particular MAP kinase to neuron death is largely dependent upon the stressor stimulus. In addition, the age of the cultures used may also represent a significant factor in kinase-dependent death. Indeed, the cortical neuron studies cited above used older culture ages than our current work. However, it has been our experience that older cultures contain a greater degree of contaminating cell types confounding possible interpretation of the data, and we have elected in the current study to use cultures at 7 days in vitro where our neuronal purity remains high. Collectively, these results demonstrate the broad contribution of the MAP kinase family to transducing diverse cytotoxic neuronal stimuli.

One caveat of this culture age, however, is their relative resistance to NMDA receptor-stimulated toxicity in spite of the presence of functional NMDA receptors (Fogal et al. 2005). Although our data demonstrate that neuron death in this paradigm is dependent upon combined TNFα and NMDA stimulation, we cannot rule out the possibility that older culture ages or neurons in vivo may respond differently to TNFα stimulation. Study of age-dependent effects of TNFα modulation of NMDA receptor-dependent neuronal death represents an intriguing possibility for future effort. Indeed, TNFα expression is developmentally regulated in the brain of both rodents and humans with maximal levels appearing at approximately 10 weeks gestation or embryonic day 12 (E12), for humans and mice, respectively (Gendron et al. 1991; Mousa et al. 1999).

Extracellular signal-regulated kinase activation also has a demonstrated role in mediating a neuroprotective response. Importantly, prolonged or delayed activation of ERK appears to be critical for neuroprotection in some paradigms (Levinthal and DeFranco 2004; Levinthal and Defranco 2005; Zhu et al. 2005). As suggested by these results, one explanation for the varying ability of ERK activity to mediate a toxic or protective response depends upon the temporal pattern of enzyme activation. In agreement with this hypothesis, we observed that the combined toxic TNFα and NMDA stimulation involved a transient activation of ERK when compared with that induced by non-toxic NMDA stimulation alone. Additionally, prior studies have also suggested that the spatial pattern of ERK activation can determine survival or death in neurons (Levinthal and DeFranco 2004) However, in our culture paradigm, we observed no nuclear translocation of active ERK. We did observe that active ERK inmmunoreactivity increased within the somatodendritic region during TNFα and NMDA combined stimulation compared with unstimulated cells. Therefore, some spatial specificity of ERK activity occurred but the substrate significance of this remains unclear.

Although we have not yet identified ERK substrate-specific phosphorylation changes, it is likely that varying kinetics of ERK activation mediate alternate transcriptional events in our cells. In fact, it is known that differing lengths and amplitudes of ERK activation regulate the phosphorylation state and stability of several immediate early genes to regulate subsequent second generation gene expression in mammalian cells (Murphy et al. 2002). Indeed, our prior study demonstrated that toxic TNFα and NMDA stimulation required increased protein levels and activity of iNOS (Floden et al. 2005). A requirement for cytotoxic protein synthesis during neuronal death processes is certainly an established mechanism of death in specific toxicity assays (Martin et al. 1988).

Prior studies from other groups have also documented this ability of TNFα to modulate neuron viability. An extensive body of literature exists, characterizing the pleiotropic nature of this cytokine to mediate protective (Carlson et al. 1998) or toxic (Nicholas et al. 2001; Zassler et al. 2003) neuron stimulations. One reason for the varying ability of TNFα to provide a neurotoxic or neuroprotective stimulus likely depends upon the relative contribution of TNFRI or TNFRII-dependent signaling to the stimulation. Although our data demonstrated that TNFRII mediates the NMDA receptor-dependent death, these observations contrast with other reports, suggesting a neurotoxic and neurotrophic role of TNFRI and TNFRII, respectively (Natoli et al. 1997; Yang et al. 2002; Li et al. 2004; Marchetti et al. 2004). One explanation for the observed differences is likely because of differing culture purities and conditions as well as the use of cells derived from receptor knockout animals in other studies. In addition, we have not yet defined the precise signaling pathway employed by TNFRII, leading to the changes in ERK activity and NMDA receptor-dependent calcium influx. However, prior studies in other cell types have implicated various MAP kinase activities in the signaling response employed by both the low- and high-affinity receptors (Natoli et al. 1997; Yang et al. 2002; Marchetti et al. 2004) consistent with our observations.

Although we have focused our efforts on understanding the ability of TNFα to modulate neurotransmitter receptor signaling, we are aware that the inflammatory milieu will involve multiple mediators besides glutamate and TNFα. Certainly, IL-1β also has a demonstrated ability to increase NMDA receptor-mediated calcium influx via modulating tyrosine phosphorylation of the receptor (Viviani et al. 2003). In addition, IL-6 stimulation can also increase NMDA receptor-mediated calcium influx and resultant neurotoxicity in some neuron culture paradigms (Qiu et al. 1995). Therefore, our data demonstrating a specific mechanism of TNFα and NMDA receptor-dependent death are only one of several toxicity pathways possibly occurring concurrently in vivo during inflammatory states.

Finally, it is known that NMDA receptor-expressing neurons are a population vulnerable to AD-associated loss possibly correlating with the degree of dementia (Greenamyre and Young 1989; Francis et al. 1993). These observations support a hypothesis of NMDA receptor-dependent excitotoxicity explaining neuron loss in AD. However, the excitotoxic hypothesis has been confounded by a lack of mechanism for excitotoxic stimulation of the NMDA receptor-expressing neurons in AD (Albin and Greenamyre 1992). Our data suggest that excitotoxic levels of glutamate are not necessary in AD brains or other pathophysiology, if physiologic levels of TNFα, such as that employed in our studies, are present (Furukawa and Mattson 1998). More importantly, brain TNFα levels are reportedly increased in several conditions, besides AD including multiple sclerosis, epilepsy and ischemia/injury (Tchelingerian et al. 1993; Akiyama et al. 2000; Gimsa et al. 2000; Yin et al. 2003; Ravizza et al. 2005). Therefore, our results suggest a possible mechanism by which this cytokine may potentiate loss of NMDA receptor-expressing neurons in multiple pro-inflammatory states.

Abbreviations

AD

Alzheimer's disease

AMPA

α-amino-3-hydroxy-5-methylisoxazole-4-propionate

APV

2-amino-5-phosphopet-anoic acid

BDNF

brain-derived neurotrophic factor

BSA

bovine serum albumin

DD

death domain

DNQX

6,7-dinitroquinoxaline-2,3-dione

ERK

extracellular signal-regulated kinase

IL-1β

interleukin-1β

iNOS

inducible nitric oxide synthase

JNK

c-Jun N-terminal kinase

LDH

lactate dehydrogenase

MAP

mitogen-activated protein

MAP2

microtubule-associated protein 2

MEK

mitogen-activated protein kinase kinase

NFκB

nuclear factor κB

NMDA

N-methyl-d-aspartate

PBS

phosphate-buffered saline

TNFα

tumor necrosis factor α