Transforming Growth Factor-α AttenuatesN-Methyl-D-aspartic Acid Toxicity in Cortical Cultures by Preventing Protein Synthesis Inhibition through an Erk1/2-dependent Mechanism

Transforming growth factor-α (TGF-α), a ligand of the epidermal growth factor receptor, reduces the infarct size after focal cerebral ischemia in rat, but the molecular basis underlying the protection is unknown. Excitotoxicity and global inhibition of translation are acknowledged to contribute significantly to the ischemic damage. Here we studied whether TGF-α can rescue neurons from excitotoxicity in vitro and how it affects calcium homeostasis, protein synthesis, and the associated Akt and extracellular signal-regulated kinase 1/2 (Erk1/2) intracellular signaling pathways in mixed neuron-glia cortical cultures. We found that 100 ng/ml TGF-α attenuated neuronal cell death induced by a 30-min exposure to 35 μm N-methyl-d-aspartic acid (NMDA) (as it reduced lactate dehydrogenase release, propidium iodide staining, and caspase-3 activation) and decreased the elevation of intracellular Ca2+ elicited by NMDA. TGF-α induced a prompt and sustained phosphorylation of Erk1/2 and prevented the loss of Akt-P induced by NMDA 3 h after exposure. The protective effect of TGF-α was completely prevented by PD 98059, an inhibitor of the Erk1/2 pathway. Studies of incorporation of [3H]leucine into proteins showed that NMDA decreased the rate of protein synthesis, and TGF-α attenuated this effect. TGF-α stimulated the phosphorylation of the eukaryotic initiation factor 4E (eIF4E) but did not affect eIF2α, two proteins involved in translation regulation. PD 98059 abrogated the TGF-α effect on eIF4E. Our data demonstrate that TGF-α exerts a neuroprotective action against NMDA toxicity, in which Erk1/2 activation plays a key role, and suggest that the underlying mechanisms involve recovery of translation inhibition, mediated at least in part by eIF4E phosphorylation.

TGF-␣ 1 is protective in models of permanent and transient focal ischemia induced by occlusion of the middle cerebral artery in the rat (1,2). Although TGF-␣ is known to promote neuronal survival (3) and to induce proliferation and differentiation of astrocytes in vitro (4), little is known about its effects on neural cells. TGF-␣ binds to the epidermal growth factor receptor (EGFR), which stimulation induces its dimerization, autophosphorylation on tyrosine residues, and triggers a cascade of reactions that requires the contribution of adapter proteins and kinases. EGFR activates several signal transduction pathways, among others are phosphatidylinositol 3-kinase/protein kinase B (PI 3-kinase/Akt) and the p44/p42 mitogen-activated protein kinase, also referred to as extracellular signal-regulated kinases 1/2 (Erk1/2) (5).
Among the very early consequences of energy depletion after an ischemic insult to the brain are membrane depolarization that induces excitatory amino acid release and inhibition of global protein synthesis, and both significantly contribute to the development of brain infarct (6 -8). Indeed, glutamate receptor overactivation results in an increase in intracellular calcium and extensive neuronal death by excitotoxicity (9). Likewise, persistent blockade of protein synthesis is associated with brain damage (10), and neuronal survival after focal ischemia may depend on the recovery of protein synthesis (11). Intracellular Ca 2ϩ overload by activation of glutamate receptors (12) or disturbances of endoplasmic reticulum Ca 2ϩ homeostasis decrease protein synthesis (13)(14)(15)(16). Protein synthesis is mainly regulated at the initiation step. Initiation of translation is a complex process that requires an initiator methionyl-tRNA and the eukaryotic initiation factors (eIF) for the assembly of mRNA and the ribosome subunits (17). Availability and activity of some of these factors, which depends on their phosphorylation state, will determine the efficiency of translation (18). Phosphorylation of eIF2␣ is well known to lead to a decrease in protein synthesis, and increasing evidence accumulates to show the critical involvement of phospho-eIF4E in translation (18). Mnk1, a kinase activated by Erk1/2 and p38 (19), phosphorylates eIF4E. Cellular stresses such as excitotoxicity or hydrogen peroxide in neurons (12,20,21), oxygen and glucose deprivation (22), and serum deprivation (23) in PC12 cells down-regulate protein synthesis. On the other hand, growth factors and neurotrophins enhance translation in cultured neurons, but different agents regulate specific translation factors through activation of Akt and/or Erk1/2 signaling pathways (24 -26).
In order to better understand how TGF-␣ exerts its neuroprotective action in vivo, we examined the effect of TGF-␣ in a model of excitotoxicity in cortical cultures. Thereafter, we stud-ied some of the transduction signaling pathways activated by TGF-␣ and its effect on NMDA-induced increase in cytosolic Ca 2ϩ , the rate of amino acid incorporation into proteins, and the phosphorylation of eIF2␣ and eIF4E.
Cell Cultures-Mixed primary cortical cultures of neurons and glia were prepared from 18-day-old Sprague-Dawley rat embryos (IFA-CREDO, Lyon, France) as described previously (27). Cells were resuspended in MEM supplemented with 10% fetal calf serum and 100 g/ml gentamycin and seeded onto poly-L-lysine (5 g/ml)-precoated 24-well plates (Nunc, Roskilde, Denmark) at a density of 3680 cells/mm 2 . Medium was partly changed on 4, 7, and 10 days in vitro (DIV) with MEM supplemented with B27. For immunocytochemistry, cells were plated on poly-L-lysine-coated glass coverslips. For the determination of intracellular Ca 2ϩ , cells were plated at a density of 3180 cells/mm 2 on 48-well plates. For the determination of leucine-specific activity (SA) in [ 3 H]leucyl-tRNA, cells were plated at a density of 3470 cells/mm 2 on 6-well plates.
Treatments-Excitotoxic lesion was performed on 12 DIV by treating cultures for 30 min with NMDA. Medium was thereafter replaced with MEM supplemented with B27 and gentamycin. Unless otherwise stated, recombinant human TGF-␣ was added twice to the culture medium: on 7 DIV and immediately after the 30-min NMDA incubation on 12 DIV. PD 98059 at 40 M was added 30 min before NMDA and was also present in the medium after NMDA washout. Tyrphostin AG 1478 at 10 M was added after NMDA removal 10 min before TGF-␣ on 12 DIV.
Measurement of LDH Activity-Cell death was estimated 24 h after the lesion by measuring the activity of lactate dehydrogenase (LDH) released in the medium according to a modification of the method of Wroblewski and LaDue (28). Briefly, the decrease in 0.75 mM NADH absorbance at 340 nm was followed in a phosphate buffer (50 mM, pH 7.4) in the presence of 4.2 mM pyruvic acid as substrate. Serial dilutions of medium from 0.2% Triton X-100-lysed cells were used to construct a standard curve, and cell death was expressed as the percentage of maximal death in the dose-response study.
In all other experiments, the value of LDH activity in control cultures was subtracted from values of treated cultures, and 35 M NMDA values were normalized to 100% (maximum cell death) as described in Bruer et al. (29). Results of LDH release were expressed in percent of NMDA and presented as means Ϯ S.E.
Propidium Iodide Nuclear Staining-Cultures pre-treated or not with TGF-␣ were lesioned with NMDA. Sixteen hours after NMDA lesion, cultures were incubated with 7.5 M of propidium iodide (PI) for 30 min at 37°C and protected from light. After two washes with PBS, cells were fixed for 30 min with 4% paraformaldehyde at 4°C and washed with PBS. PI-positive cells were counted in 8 fields per well and the sum of the 8, corresponding to a total area of 1.1776 mm 2 , was calculated. Results were expressed as percent of control.
Measurement of Caspase-3 Activity-The caspase-3 activity enzymatic assay was performed according to Valencia and Moran (30) using 100 g of proteins and 25 M Ac-DEVD-AMC as the substrate. Fluorescence of AMC, generated by cleavage of Ac-DEVD-AMC (excitation/ emission 380/460 nm), was monitored every 5 min for 30 min in a CytoFluor 2350 Millipore scanner. Enzymatic activity was calculated as ⌬ fluorescence/mg protein/min and expressed as percent of control. We checked that 50 M Ac-DEVD-CHO, the caspase-3 selective inhibitor, completely blocked NMDA-induced caspase-3 activity.
NeuN Immunocytochemistry-After exposure for 5 days to 100 ng/ml TGF-␣, mixed cultures were washed with 10 mM cold phosphate-buffered saline (PBS) and fixed for 30 min with 4% paraformaldehyde at 4°C. Cells were then washed, and endogenous peroxidase activity was inhibited by a 2-min incubation in 1% H 2 O 2 diluted in methanol/PBS (30:70). Blocking was then performed for 30 min at room temperature in PBS containing Triton X-100 and 7% horse serum. Cells were then incubated for 1 h with the primary antibody (anti-NeuN 1:100), washed, and incubated with a mouse biotinylated horseradish peroxidase secondary antibody. After a wash, cells were incubated with the ABC reagent according to the instructions of the manufacturer. The reaction was developed in 0.5 mg/ml 3,3Ј-diaminobenzidine tetrahydrochloride. Three control wells and 3 TGF-␣-treated wells within a same 24-well plate were analyzed. NeuN-immunoreactive cells were counted in 3 fields per well, and the sum of the 3, corresponding to a total area of 0.44 mm 2 , was calculated. The result was expressed in NeuN-positive cells per mm 2 . A t test was performed for statistical analysis. Western Blot-Cultures were washed with cold PBS and harvested at several time points in RIPA lysis buffer (10 mM PBS, 1% Igepal AC-630, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with a protease inhibitor mixture (Complete) and 1 mM sodium orthovanadate. Protein content was determined by the Bradford assay (Bio-Rad). Proteins were separated by electrophoresis on 10 or 12% polyacrylamide gels in denaturing conditions and transferred to polyvinylidene difluoride Immobilon-P membranes (Millipore). After 1 h blocking in 20 mM Tris-HCl, 150 mM NaCl, 0.1% Tween 20 (T-TBS) containing 5% albumin bovine and 5% non-fat dry milk, membranes were incubated overnight at 4°C with the following primary antibodies: mouse anti-phospho-p44/p42 MAPK (Thr-202/Tyr-204), rabbit anti-phospho-Akt (Ser-473), rabbit anti-phospho-eIF4E (Ser-209), rabbit anti-phospho-eIF2␣ (Ser-51), rabbit anti-p44/p42, mouse anti-eIF4E diluted 1:1000, and goat anti-eIF2␣ diluted 1:100. After two washes in T-TBS, membranes were then incubated for 1 h at room temperature with either anti-rabbit, anti-mouse, or anti-goat horseradish peroxidase-conjugated antibody at 1:2000. The reaction was visualized using a chemiluminescence detection system based on the luminol reaction. Reprobing the membranes with an antibody against ␤-tubulin diluted 1:15,000 was carried out to check equal loading in the lanes.

Incorporation of [ 3 H]Leucine into Proteins-
A time course of leucine uptake was performed in order to determine the time at which [ 3 H]leucine steady state between the medium and the cells was reached. Culture medium was withdrawn, and cells were incubated in 300 l of MEM/B27 (commercial MEM contains 0.052 g/liter of leucine) in the presence of 4 Ci/ml [ 3 H]leucine for different times ranging from 1 to 30 min at 37°C. Medium was then removed, and cells were washed once with 0.5 mg/ml non-radioactive leucine in PBS, harvested in lysis buffer (20 mM Tris-HCl, pH 7.6, 10 mM potassium acetate, 1 mM dithiothreitol, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, 0.25% Igepal), and processed as described in Alcazar et al. (13). Briefly, lysates were spun for 30 min at 12,000 ϫ g at 4°C. Supernatants were collected, and an aliquot was taken for protein determination before addition of 10% trichloroacetic acid. Proteins were precipitated by a 30-min centrifugation at 12,000 ϫ g at 4°C, and the supernatants (trichloroacetic acid soluble fraction) were separated from the pellet (trichloroacetic acid precipitable fraction). The trichloroacetic acid-precipitable fraction was resuspended in 0.2 N NaOH, and radioactivity was measured to determine [ 3 H]leucine incorporation into proteins. Radioactivity was also measured in the trichloroacetic acid-soluble fraction to calculate intracellular soluble radioactivity (dpm/mg of protein) at each time point. According to the uptake results showing that steady state was already reached by 20 min, in further experiments we chose to study the effect of treatments on the incorporation of [ 3 H]leucine into proteins (dpm/mg protein/min) after 30 min of incubation with the radiotracer.
The content of intracellular free leucine (nanomole) was measured by high pressure liquid chromatography in a separate experiment as follows. After the treatments, medium was replaced with MEM/B27 without [ 3 H]leucine, and 30 min later the cells were washed with leucine-free PBS and processed as above. Five l of the trichloroacetic acid-soluble fraction were brought to pH 10 with 1 M NaOH and mixed with 15 l of the derivatizing reagent o-phthaldialdehyde at 1 mg/ml. Separation of endogenous leucine was carried out in a reverse-phase C 18 column (Tracer Nucleosil C 18 , 5-m particle size, 10 ϫ 0.4 cm; Teknokroma, Spain) using the following mobile phase: solution A (100 mM sodium acetate, 5.5 mM triethylamine, 11% acetonitrile, pH 5.5) and solution B (acetonitrile) mixed with a gradient program from 100 to 30% of solution A within 60 min at 0.8 ml/min. Leucine content was calculated after fluorimetric detection (excitation/emission 360:450 nm) using an external standard method according to Suñ ol et al. (31).
Leucine-specific Activity in [ 3 H]Leucyl-tRNA-Under steady-state conditions, the rate of protein synthesis is a function of the SA of the FIG. 1. TGF-␣, through activation of EGFR, reduces NMDA-induced excitotoxicity in cortical cultures. A, NMDA dose-response in cortical cultures. Cultures were exposed on 12 DIV to increasing concentrations of NMDA (5-300 M) for 30 min. NMDA was then washed out, and death was studied 24 h later by LDH activity. Data are means Ϯ S.E. of results expressed as percent of maximum cell death, a value obtained from Triton X-100 lysed cultures (n ϭ 6 cultures). B, cultures treated for 5 days with increasing concentrations of TGF-␣ were lesioned with 35 M NMDA on 12 DIV. Thirty min later, the medium was removed and replaced with fresh medium supplemented with TGF-␣ (1-100 ng/ml). LDH activity was determined at 24 h post-lesion. LDH value in control and NMDA-treated cultures were set to 0 and 100%, respectively. Data were normalized (expressed as % of NMDA) and presented as means Ϯ S.E. (n ϭ 3-4 cultures). Open bars, control; filled bars, NMDA. C, control and TGF-␣-pre-treated (100 ng/ml) cultures were lesioned with 35 M NMDA. Sixteen hours later, cells were incubated with 7.5 M propidium iodide (PI), and PI-stained nuclei were counted under the fluorescence microscope (see "Experimental Procedures"). The number of PI-positive nuclei per mm 2 was 349 Ϯ 27.6 in control cultures. NMDA increased the number of PI-stained nuclei, an effect that was attenuated by TGF-␣. Data are expressed as percent of control (means Ϯ S.E. of n ϭ 4 cultures). D, caspase-3 activity was calculated as ⌬ fluorescence/mg protein/min. Data are mean Ϯ S.E. of n ϭ 8 cultures from three independent experiments of results expressed as percent of control. Caspase-3 activity increased at 16 h after exposure to NMDA, and this was prevented by TGF-␣. E, TGF-␣ reduces the NMDA-evoked increase in intracellular Ca 2ϩ . Control and TGF-␣-pre-treated cultures were loaded with 15 M Fluo-3/AM. Cells were then exposed to NMDA or NMDA-free HEPES buffer, and the fluorescence emitted at 530 nm was monitored from 0.5 to 10 min. Data are mean Ϯ S.E. of results expressed as percent of control (n ϭ 16 cultures from three independent experiments). Open circle, TGF-␣; open triangle, NMDA ϩ TGF-␣; filled square, NMDA. F, EGFR inhibitor AG 1478 completely prevents TGF-␣ neuroprotection. Cultures treated with TGF-␣ were exposed to NMDA; AG 1478 was added after the lesion as indicated under "Experimental Procedures." LDH activity was determined at 24 h post-lesion. Data are means Ϯ S.E. of results expressed as percent of NMDA (n ϭ 10 -16 cultures from two independent experiments). C, control cultures; N, NMDA; AG, AG 1478; NϩT, NMDA ϩ TGF-␣; NϩTϩAG, NMDA ϩ TGF-␣ ϩ AG 1478. ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05 indicate a significant difference versus control; $$$, p Ͻ 0.001; $$, p Ͻ 0.01; $, p Ͻ 0.05 versus NMDA; &&&, p Ͻ 0.001 versus NMDA ϩ TGF-␣.
precursor pool for the incorporation of amino acids into proteins (32)(33)(34). The immediate precursor pool is not the free amino acid in the intracellular space but the corresponding aminoacyl-tRNA. For this reason we determined in a separate group of experiments the SA of leucyl-tRNA for certain conditions. Isolation of aminoacyl-tRNA and deacylation was performed according to a modification of the protocol described by Keen et al. (32). Medium from control, TGF-␣-, or NMDAtreated cultures was removed, and cells were incubated in 1 ml of MEM/B27 containing 10 Ci/ml [ 3 H]leucine for 30 min at 37°C. Medium was then removed, and the cells were washed once with 3 ml of leucine-free PBS and kept at Ϫ80°C. Cells from 6 wells per condition were resuspended in a 3.5-ml final volume of buffer (250 mM NaCl, 10 mM MgCl 2 , 1 mM Na 2 EDTA, 0.4 mg/ml bentonite, 10 mM sodium acetate, pH 4.5) and vigorously mixed with 3.5 ml of saturated phenol at pH 5 for 1 h at 4°C. The mixture was centrifuged for 5 min at 1500 ϫ g at room temperature in Phase Lock Gel heavy tubes, and the aqueous phase containing total RNAs was collected and chromatographed in a 0.7 ϫ 10 cm DE52 DEAE-cellulose column with buffer (250 mM NaCl, 10 mM MgCl 2 , 1 mM Na 2 EDTA, 10 mM sodium acetate, pH 4.5) at a flow rate of 4 ml/min. One-min fractions were collected for a total of 60 ml. Aminoacyl-tRNAs were eluted with a higher salt concentration (700 mM NaCl, 10 mM MgCl 2 , 1 mM Na 2 EDTA, 10 mM sodium acetate, pH 4.5) in 0.7-min fractions for a total of 30 ml. Radioactivity was measured in 0.5-ml aliquots, and the peak fraction containing the [ 3 H]leucyl-tRNA was further processed to deacylate the aminoacyl-tRNA. pH was adjusted to 9 by addition of 400 mM sodium borate, pH 9.5, and the samples were incubated at 37°C for 90 min. Free leucine was separated from tRNA in Microcon 30 ultrafiltration systems by spinning at 14,000 ϫ g for 25 min at 15°C. The ultrafiltrate was concentrated in a speed-vac system and analyzed by high pressure liquid chromatography as described above. The content of free leucine (nmol) was measured, and the associated radioactivity (dpm) was determined to calculate [ 3 H]leucine SA (dpm/nmol) derived from the corresponding leucyl-tRNA.
Statistical Analysis-Results are expressed as mean Ϯ S.E., and the number of replicates (n) is indicated in the legend of each figure and table. Unless indicated, one-way analysis of variance (ANOVA) with the Bonferroni post hoc test was performed to determine which groups were significantly different. The Kruskall Wallis analysis followed by the Dunn's test for multiple comparisons was used to compare groups with non-homogenous variance. Two-way ANOVA was performed to compare changes in intracellular calcium (Fig. 1E) and the effect of NMDA and TGF-␣ on neuronal death (Fig. 1B). One symbol indicates p Ͻ 0.05, two p Ͻ 0.01, and three p Ͻ 0.001.

RESULTS
TGF-␣ Reduces NMDA-induced Excitotoxicity-A NMDA dose-response study was undertaken to estimate the concentration that kills 50% of NMDA-sensitive neurons, as assessed by LDH release. LDH release assay, measuring cell permeability, was previously shown to correlate with cell death (35). EC 50 was 37.9 Ϯ 7.2 M (Fig. 1A), so we chose to perform further experiments with 35 M NMDA. TGF-␣ at 1, 30, or 100 ng/ml was added 5 days before and after the NMDA lesion, as indicated under "Experimental Procedures." According to LDH release, TGF-␣ reduced NMDA-induced neuronal death by 40% (p Ͻ 0.05) at the highest concentration (Fig. 1B). This result was further validated with propidium iodide staining that showed attenuation of NMDA-induced neuronal loss by TGF-␣ (Fig. 1C). Furthermore, in agreement with previous reports (36 -38), NMDA induced activation of caspase-3, which again was prevented by TGF-␣ (Fig. 1D).
Additionally, we checked that addition of 100 ng/ml TGF-␣ on 7 DIV did not alter the number of neurons. After exposure for 5 days to TGF-␣, mixed cultures were fixed on 12 DIV and stained with the antibody against NeuN, a marker of neuronal nuclei. Control and TGF-␣-treated cultures showed 892 Ϯ 132.2 and 936 Ϯ 130.5 NeuN-positive cells per mm 2 , respectively, and no statistically significant difference was found between the two groups.
Because excitotoxicity induces disturbances in calcium homeostasis, we studied changes in cytosolic calcium in our model. Free intracellular calcium in cultures exposed to NMDA was raised to 2-3-fold over control values (Fig. 1E). The effect FIG. 2. Activation of Erk1/2 by TGF-␣ is involved in neuroprotection against NMDA toxicity. NMDA induces loss of Akt-P at 3 h, and this effect is prevented by TGF-␣. A, cultures were treated for 5 min with NMDA and/or increasing concentrations of TGF-␣. Proteins were then resolved on 10% SDS-PAGE and processed for Western blot. Immunoblotting was performed with an Erk1/2-P antibody (p42/p44-P), and the membrane was then reprobed for Akt-P and then ␤-tubulin. B, cells cultured for 5 days with TGF-␣ were treated with PD 98059 and exposed to NMDA for 30 min as indicated under "Experimental Procedures." Erk1/2-P, Akt-P, and ␤-tubulin Western blot were performed with cell extracts harvested 3 h after NMDA lesion. PD 98059 inhibits the raise of Erk1/2 phosphorylation in the presence of TGF-␣ and NMDA. Phosphorylation of Akt is reduced by NMDA at 3 h after exposure (B), but not at earlier time points (A). C, cultures were treated as in B, and LDH activity was measured at 24 h post-lesion. Data are mean Ϯ S.E. (n ϭ 9 -18 cultures from two independent experiments) of results expressed as % of NMDA. ***, p Ͻ 0.001 versus control; $$$, p Ͻ 0.001 versus NMDA; &&, p Ͻ 0.01 versus NMDA ϩ TGF-␣. D, EGFR blockade with AG 1478 prevents TGF-␣-induced Erk1/2 phosphorylation. Cultures treated with TGF-␣ were exposed to NMDA; AG 1478 was added after the lesion as indicated under "Experimental Procedures." Cells were harvested 1 h after the lesion and processed for Erk1/2-P, Erk1/2 total, and Akt-P Western blot. The amount of total protein was not affected by any treatment. PD, PD 98059; N ϩ PD, NMDA ϩ PD 98059; N ϩ T ϩ PD, NMDA ϩ TGF-␣ ϩ PD 98059, for other abbreviations see Fig. 1 legend.
was observed as early as 30 s and was maintained for at least 10 min. TGF-␣ reduced the NMDA-induced increase in intracellular calcium by about 25% at all time points (p Ͻ 0.05).

TGF-␣ Neuroprotection Is Mediated through EGFR by Activation of Erk1/2-TGF-␣ is an endogenous ligand of EGFR.
Here we found that the protective effect of TGF-␣ was mediated through EGFR as AG 1478, a specific inhibitor of the tyrosine kinase activity of EGFR prevented the protective effect of TGF-␣ (Fig. 1F). EGFR is coupled to several intracellular signaling pathways. Here we studied Akt and Erk1/2 phosphorylation after TGF-␣ and NMDA treatments. TGF-␣ induced a strong phosphorylation of Erk1/2 in a concentration-dependent manner at 5 min ( Fig. 2A). NMDA caused a comparatively moderate Erk1/2 phosphorylation (Fig. 2A). In the presence of both drugs Erk1/2 activation was higher than after single treatment ( Fig. 2A). However, the effect of NMDA was transient as it was detected at 5 min during NMDA exposure, but it was no longer found at 1 (Fig. 2D) or 3 h after NMDA removal (Fig. 2B), whereas Erk1/2 phosphorylation by TGF-␣ was maintained at these times. We then wanted to figure out whether Erk1/2 activation was responsible for TGF-␣ protection. Preincubation with PD 98059, an inhibitor of MAPK kinase (MEK1/ 2), totally blocked Erk1/2 phosphorylation (Fig. 2B). PD 98059, which did not compromise neuronal survival on its own, fully prevented the TGF-␣ neuroprotective effect, but it did not alter NMDA-induced toxicity (Fig. 2C). These data demonstrate that the MEK/Erk1/2 pathway is involved in a TGF-␣ neuroprotective process. In addition, we observed that AG 1478 blocked TGF-␣-induced phosphorylation of Erk1/2, showing that this effect is mediated through activation of EGFR (Fig. 2D).
In contrast to the TGF-␣-induced Erk1/2 activation, we did not observe any activation of Akt after treatments (Fig. 2, A  and D). However, 3 h after NMDA lesion the level of Akt-P was reduced in relation to control (Fig. 2B). This effect was not seen at earlier time points, i.e. during 5 min of NMDA exposure ( Fig.  2A) and at 1 h after NMDA (Fig. 2D), likely reflecting an early step in neuronal degeneration. Because TGF-␣ reduced neuronal death, maintenance of Akt-P in the NMDA ϩ TGF-␣ condition is in accordance with the protection. No major effect of PD 98059 on Akt phosphorylation was seen (Fig. 2B).

H]Leucine Incorporation into Proteins Is Reduced by NMDA, and This Effect Is Prevented by TGF-␣-Because early
and persistent inhibition of protein synthesis has been reported in vivo and in vitro after ischemia and excitotoxicity, we investigated whether NMDA altered this parameter in cortical cultures. We first performed a kinetic study of cellular [ 3 H]leucine uptake from the medium to determine the time at which [ 3 H]leucine reached steady state. [ 3 H]Leucine uptake was linear during the first minutes and reached a plateau from 20 min (Fig. 3A). Therefore, we decided to incubate cells for 30 min with [ 3 H]leucine for protein synthesis studies. NMDA inhibited the incorporation of [ 3 H]leucine into proteins in a dose-dependent manner (Fig. 3B) in parallel to its neuronal toxicity (Fig.  1A), indicating that the degree of inhibition of [ 3 H]leucine incorporation correlates with the severity of the lesion. TGF-␣ prevented the reduction of [ 3 H]leucine protein incorporation induced by NMDA (Fig. 3C). This effect of TGF-␣ was not observed in the presence of the Erk1/2 inhibitor PD 98059 (Fig.  3C). Yet PD 98059 affected [ 3 H]leucine protein incorporation by itself (Fig. 3C) suggesting that Erk activity is involved in maintaining protein synthesis under control conditions. Therefore, we cannot conclude that PD 98059 reverses the effect of TGF-␣ on protein synthesis recovery after NMDA.
Effect of Treatments on Intracellular Free Leucine and Leucyl-tRNA-The amount of trichloroacetic acid-soluble radioactivity was reduced by NMDA in relation to controls (Table   FIG. 3. (Table I), although differences did not reach statistical significance. Therefore, the reduced amount of trichloroacetic acid-soluble intracellular radioactivity under steady-state conditions was associated with a faint reduction in the amount of intracellular free leucine. Yet this small change in intracellular free leucine by NMDA was not limiting for protein synthesis, as the SA of leucyl-tRNA (the actual precursor pool for protein synthesis) was not modified by NMDA (Table I). This indicated that amino acids were available for protein synthesis, and thus the deep inhibition of amino acid incorporation into proteins caused by NMDA was due to alterations in the process of translation. This view was supported by the observation that treatment with TGF-␣ alone, despite not decreasing the incorporation of radioactivity into proteins in relation to controls (Fig. 3C), reduced, as NMDA did, trichloroacetic acid-soluble intracellular radioactivity, and it slightly reduced (non significantly) the content of intracellular leucine (Table I). But again, the SA of leucyl-tRNA was unchanged after TGF-␣ (Table I), thus implying that this treatment affected the molecular mechanism underlying the process of translation rather than the availability of amino acids for protein synthesis.

NMDA inhibits [ 3 H]leucine incorporation into proteins, and TGF-␣ prevents it.
NMDA Reduces and TGF-␣ Activates the Phosphorylation of eIF4E; eIF2␣ Phosphorylation Is Not Affected by NMDA or by TGF-␣-Because translation is predominantly regulated at the initiation level and some of the eIFs can be limiting factors, we examined the phosphorylation state of eIF4E and eIF2␣. The former is more efficient after phosphorylation, whereas the phosphorylated eIF2␣ binds to the GDP/GTP exchanger eIF2B and inhibits it so that a new round of translation initiation is blocked. NMDA did not induce a phosphorylation of eIF2␣ (Fig.  4A). This indicates that eIF2␣ does not mediate NMDA downregulation of protein synthesis. Acute addition of TGF-␣ did not modify eIF2␣ phosphorylation either (Fig. 4A).
We then considered eIF4E because it can be phosphorylated after Erk1/2 activation (19). We observed a slight dephosphorylation of eIF4E after NMDA exposure (Fig. 4, B and C), which might contribute to the inhibition of protein synthesis after NMDA removal. TGF-␣ induced a strong phosphorylation of eIF4E, which was detected even after NMDA treatment (Fig. 4,  B and C). The level of total eIF4E protein was not affected by TGF-␣ (Fig. 4, B and C). Therefore, eIF4E phosphorylation might contribute to the maintenance of protein synthesis by TGF-␣ after NMDA (Fig. 3C). PD 98059 prevented eIF4E phosphorylation by TGF-␣ (Fig. 4C), consistently with its inhibitory effect on TGF-␣-induced Erk1/2 activation (see above). Thus, this result shows that Erk1/2 activation mediates phosphorylation of eIF4E by TGF-␣. DISCUSSION In the present paper, we show that TGF-␣ exerts a neuroprotective effect in cortical cell cultures subjected to an excitotoxic lesion. Through binding to EGFR, TGF-␣ reduced neuronal death by attenuating the calcium influx triggered by NMDA and by persistently activating the MEK/Erk1/2 pathway. Subsequently, eIF4E, a downstream target of Erk1/2, was phosphorylated and facilitated protein synthesis. Transient exposure to NMDA caused persistent inhibition of the rate of amino acid incorporation into protein without altering the FIG. 4. NMDA decreases eIF4E phosphorylation and TGF-␣ stimulates it through activation of Erk1/2. eIF2␣ phosphorylation is not affected by any treatment. A, cells were treated with NMDA for 30 min, changed to a fresh medium, and harvested 30 min later. TGF-␣ was added simultaneously to NMDA (60) or after the lesion (30). Proteins from duplicate samples were resolved on 12% SDS-PAGE. Immunoblots were incubated first with anti-eIF2␣-P, then with anti-eIF2-␣, and finally with anti-␤-tubulin. B, blot A was incubated with anti-eIF4E-P and then with anti-eIF4E. C, cultures treated for 5 days with TGF-␣ were exposed to NMDA. PD 98059 was added as described under "Experimental Procedures." Cells were harvested 30 min after medium removal and processed for eIF4E-P, eIF4E, and ␤-tubulin Western blot. availability of precursor amino acid. Inactivation of eIF4E, but not eIF2␣, was related to this process. TGF-␣ prevented the collapse of translation elicited by NMDA and therefore largely contributed to decrease neuronal death. In addition, our results suggest that phospho-Erk1/2 and phospho-eIF4E play a critical role in the maintenance of protein synthesis in cortical neurons. In this study, a direct correlation is found between the recovery of protein synthesis after treatment with a growth factor and its neuroprotective effect against an excitotoxic injury. TGF-␣ Increases Neuronal Survival after an Excitotoxic Lesion-TGF-␣ exerts beneficial effects against NMDA-induced toxicity in mixed neuron-glia cortical cultures. In this experimental condition toxicity is directed toward neurons, as astrocytes in culture do not express NMDA receptors (39). Treatment with TGF-␣ attenuated neuronal calcium overload triggered by NMDA receptor overactivation and reduced neuronal cell death. Likewise, glial cell line-derived neurotrophic factor was shown to improve neuronal survival after NMDA by reducing Ca 2ϩ influx in pure cortical neuron cultures (40). A significant neuroprotective effect of TGF-␣ against NMDA toxicity was observed at rather high doses (100 ng/ml). Likewise, protection of neurons against oxygen deprivation was found with 100 ng/ml epidermal growth factor (5). We reported previously that 50 ng of TGF-␣ is protective in vivo in the rat brain against focal cerebral ischemia (1, 2). Yet glutamate-mediated excitotoxicity is not the only neurotoxic mechanisms accounting for ischemic neuronal damage, as other disturbances (such as the glial reaction, inflammation, and oxidative stress) are also involved in the generation of infarction. It is unknown whether TGF-␣ might exert beneficial effects against the pathogenesis of ischemic damage, other than reducing NMDAmediated excitotoxicity.
The Effect of TGF-␣ Is Mediated by Activation of the EGFR/ Erk1/2 Signaling Pathway-We then aimed to identify the signal transduction pathways stimulated by TGF-␣. We first confirmed that TGF-␣ conferred neuroprotection acting on the EGFR, as inhibition of the tyrosine kinase activity coupled to EGFR with AG 1478 (41) suppressed TGF-␣-induced neuroprotection. Following EGFR activation, the phosphorylated tyrosines act as binding sites for molecules containing the Src homology 2 domain, such as phospholipase C␥ and adapter proteins such as Grb2 and Shc (reviewed in Ref. 5). The adapter proteins function as docking sites for other signaling molecules such as Cbl that associates to PI 3-kinase, activating in its turn Akt, and recruitment of Grb2 activates the MAPK pathway through Ras-GTP (5). Thus, we studied the phosphorylation of Akt and Erk1/2 and observed that Akt was not activated, whereas Erk1/2 was highly and persistently phosphorylated after exposure to TGF-␣. In accordance with this result, many effects of TGF-␣ in the central nervous system depend on the stimulation of the Erk1/2 pathway (42). Evidence for the critical involvement of the Erk1/2 pathway in TGF-␣ function was supported by the observation that AG 1478 and PD 98059 both abolished Erk1/2 phosphorylation and TGF-␣ neuroprotection in our model. The finding that Erk1/2 phosphorylation was important for neuronal survival is in agreement with previous studies. For instance, estrogen rescues cortical neurons from glutamate toxicity through an Erk1/2-dependent mechanism (43). Interestingly, Akt and Erk1/2 are activated after an ischemic episode (44 -47), but the underlying mechanism remains unknown. Whereas Akt is clearly involved in anti-apoptotic processes (48,49), the beneficial effect of activated Erk1/2 in neurons is still a matter of controversy. Indeed, although Erk1/2 signaling regulates synaptic plasticity, long term potentiation, and survival (50), its inhibition prevents neuronal damage resulting from focal cerebral ischemia (44) and excitotoxicity in vitro (51). Glutamate receptor stimulation in cortical cultures triggers Erk1/2 phosphorylation (52). In agreement with the latter report, NMDA activated Erk1/2 in our cultures in a transient manner. However, this activation was not responsible for NMDA toxicity because PD 98059 did not impede NMDA-induced toxicity. In contrast, NMDA caused dephosphorylation of Akt (in relation to controls), although this was not seen until 3 h after the lesion. TGF-␣ prevented this effect of NMDA even though treatment with TGF-␣ alone did not activate Akt. Because it is likely that dephosphorylation of Akt by NMDA at this time point would impair neuronal survival, we cannot exclude the possibility that preventing Akt dephosphorylation by TGF-␣ was involved in neuroprotection. Yet the facts that Akt dephosphorylation by NMDA was not an early event and that TGF-␣ by itself did not activate Akt suggest that the maintenance of Akt-P after NMDA and TGF-␣ was an indicator of neuronal survival.
Transient Exposure to NMDA Causes Persistent Inhibition of Protein Synthesis and eIF4E Dephosphorylation-We then addressed suppression of translation, another important aspect involved in neuronal death (6,8,10). We found a persistent inhibition of amino acid incorporation into proteins following transient exposure to NMDA, which was prevented by TGF-␣. The effect of NMDA was not a consequence of reduced amino acid availability as the leucyl-tRNA precursor pool for protein synthesis after NMDA was not different from control. Yet a small reduction in the free leucine intracellular pool was detected after NMDA, but this was not limiting for the formation of leucyl-tRNA. Treatment with TGF-␣ alone did not alter leucine incorporation into proteins in relation to control. However, unexpectedly, TGF-␣ by itself decreased, as NMDA did, the free leucine pool, but again this had no effect on the pool of leucyl-tRNA. Therefore, we conclude that NMDA inhibited protein synthesis by a mechanism other than reducing amino acid availability, and that TGF-␣ prevented the down-regulation of protein synthesis caused by NMDA.
In order to find out the molecular mechanism underlying protein synthesis inhibition and recovery, we focused our study on two translation factors, namely eIF2␣ and eIF4E. eIF2␣ is a critical regulator of protein synthesis as it mediates the binding of the Met-tRNA to the ribosome. eIF2␣ has been extensively studied because it is phosphorylated after ischemia and cellular stress (21,53,54). Depletion in endoplasmic reticulum Ca 2ϩ , which is suspected to occur in various pathologies including cerebral ischemia, leads to phosphorylation of eIF2␣ and subsequent inhibition of protein synthesis (13,15). However, eIF2␣ was not phosphorylated after NMDA exposure compared with the control condition. Therefore, the inhibition of protein synthesis observed after treatment with NMDA is not caused by inactivation of this factor. We also examined the phosphorylation state of eIF4E after NMDA. eIF4E is the m 7 G cap mRNA-binding protein that associates to eIF4G and eIF4A to form the eIF4F complex. This complex interacts with cap mRNA and facilitates its binding to the ribosome. Affinity of eIF4E for the cap is increased after phosphorylation (18). Here we report that eIF4E suffered dephosphorylation (i.e. inactivation) after transient exposure to NMDA, which likely contributed to persistent inhibition of protein synthesis when cells were no longer exposed to the toxin.
TGF-␣ Promotes Protein Synthesis Recovery and eIF4E Phosphorylation-Regarding the mechanism involved in the effect of TGF-␣ preventing NMDA-induced protein synthesis inhibition, we observed that TGF-␣ induced eIF4E phosphorylation after NMDA exposure and that this was mediated through activation of Erk1/2. Several growth and trophic factors cause eIF4E phosphorylation after Erk1/2 activation in various cells under different experimental conditions (19,23). eIF4E was shown to play a pivotal role in the activation of translation by epidermal growth factor and nerve growth factor in PC12 cells after nutrient deprivation (23). In cortical cultures, brain-derived neurotrophic factor is a better stimulator of protein synthesis than insulin because it phosphorylates eIF4E, whereas insulin does not (25). In a similar manner, eIF4E phosphorylation by TGF-␣ might be involved in the recovery of protein synthesis after NMDA exposure. Besides Erk1/2 activation, regulation of eIF4E may occur through another pathway. Indeed, some proteins, eIF4E-binding proteins (eIF4E-BP), associate with eIF4E and impede the formation of the eIF4F complex. Under the phosphorylated state, they dissociate from eIF4E and no longer block the complex formation. The mammalian target of rapamycin, a downstream kinase of Akt (18), is the enzyme responsible for eIF4E-BP phosphorylation in cortical neurons (25). Many growth factors, such as IGF-1 (26), brain-derived neurotrophic factor (25), and epidermal growth factor (55) activate the mammalian target of rapamycin and phosphorylate eIF4E-BP. However, because TGF-␣ did not activate Akt signaling pathway in our cultures, we assume that eIF4E-BP did not play a major role in the regulation of eIF4E. Beyond its effect on eIF4E, TGF-␣ did not affect eIF2␣ phosphorylation in the absence or presence of NMDA.
Altogether our data show that TGF-␣ unlocked the blockade of translation by activating the EGFR/Erk1/2 route and reduced the Ca 2ϩ entry elicited by NMDA, thereby promoting neuronal survival. Although TGF-␣ has pleiotropic effects, the present results indicate that the recovery of NMDA-induced persistent protein synthesis inhibition is critical for the neuroprotective effect of TGF-␣ against excitotoxicity. Because disturbances in calcium homeostasis and protein synthesis occur after an ischemic insult, a partial restoration of these parameters after in vivo TGF-␣ treatment may contribute to its protective effect.