Erythropoietin Receptor-mediated Inhibition of Exocytotic Glutamate Release Confers Neuroprotection during Chemical Ischemia*

Erythropoietin (EPO) reduced Ca2+-induced glutamate (Glu) release from cultured cerebellar granule neurons. Inhibition was also produced by EPO mimetic peptide 1 (EMP1), a small synthetic peptide agonist of EPO receptor (EPO-R), but not by iEMP1, an inactive analogue of EMP1. EPO and EMP1 induced autophosphorylation of Janus kinase 2 (JAK2), a tyrosine kinase that associates with EPO-R. Furthermore, genistein, but not genistin, antagonized both the phosphorylation of JAK2 and the suppression of Glu release induced by EPO and EMP1. During chemical ischemia, substantial amounts of Glu were released from cultured cerebellar and hippocampal neurons by at least two distinct mechanisms. In the early phase, Glu release occurred by exocytosis of synaptic vesicle contents, because it was abolished by botulinum type B neurotoxin (BoNT/B). In contrast, the later phase of Glu release mainly involved a BoNT/B-insensitive non-exocytotic pathway. EMP1 inhibited Glu release only during the early exocytotic phase. A 20-min exposure of hippocampal slices to chemical ischemia induced neuronal cell death, especially in the CA1 region and the dentate gyrus, which was suppressed by EMP1 but not iEMP1. However, EMP1 did not attenuate neuronal cell death induced by exogenously applied Glu. These results suggest that activation of EPO-R suppresses ischemic cell death by inhibiting the exocytosis of Glu.

Brain ischemia induces delayed neuronal cell death, especially in the CA1 region of the hippocampus (1). Glutamate (Glu) 1 and related excitatory amino acids are well-known neu-rotoxins, and their extracellular levels increase dramatically in the course of brain ischemia. The excitotoxicity hypothesis suggests that, in neuronal hypoxia/ischemia, neurodegeneration can be triggered by cytoplasmic Ca 2ϩ overload, which occurs when N-methyl-D-aspartate receptors are overstimulated by excessive Glu (2). Numerous pharmacological approaches have been explored to prevent or attenuate neuronal cell death in ischemia, however, no satisfactory methods have been developed (3)(4)(5).
The hematopoietic growth factor erythropoietin (EPO) is a primary regulator of mammalian erythropoiesis and is produced by kidney and liver in an oxygen-dependent manner (6,7). The EPO receptor (EPO-R) is a member of the type 1 superfamily of single-transmembrane cytokine receptors. EPO binding to EPO-R induces receptor oligomerization and subsequent activation by autophosphorylation of Janus kinase 2 (JAK2), a protein-tyrosine kinase that associates with EPO-R. JAK2 phosphorylates a cytoplasmic transcription factor called signal transducer and activator of transcription (STAT), leading to translocation of STAT into the nucleus and regulation of transcription (8). EPO and EPO-R are also expressed in mammalian brain, where hypoxia strongly stimulates EPO production (9,10). Although activation of EPO-R has been suggested to play a neuroprotective role (10,11), little is known about the action of EPO in the nervous system. Recently, we showed that EPO inhibits Ca 2ϩ -induced dopamine release from clonal rat pheochromocytoma PC12 cells (12); however, it has not yet been deduced whether EPO also suppresses neurotransmitter release from central neurons. In the present study, we investigated the effect of EPO on Glu release from cultured cerebellar granule cells and hippocampal neurons. We found that the activation of EPO-R attenuates Ca 2ϩ -induced Glu release from these neurons possibly through activation of JAK2 tyrosine kinase. We also found that EPO-R activation protects hippocampal neurons from ischemic neuronal damage through the inhibition of Glu release via exocytosis.
Antibody-Anti-EPO-R and anti-phospho-JAK2 antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) and BIO-SOURCE International (Camarillo, CA), respectively.
Glu Release Assay-Cells were washed once with the low K ϩ solution (140 mM NaCl, 4.7 mM KCl, 1.2 mM KH 2 PO 4 , 2.5 mM CaCl 2 , 1.2 mM MgSO 4 , 11 mM glucose, 15 mM HEPES-NaOH, and pH 7.4). Following pretreatment in various conditions as indicated in the figure legends, the cells were incubated for 2 min in the low K ϩ solution at 37°C. The release of Glu during this period represented the basal release. Then they were incubated for 2 min with the low K ϩ solution containing 5 M ionomycin to evaluate ionomycin-stimulated release of Glu. For the assay of Glu release during chemical ischemia, cells were incubated in glucose-free low K ϩ solution supplemented with 1 mM KCN (chemical ischemia solution). The extracellular solutions were centrifuged at 15,000 rpm for 5 min at 4°C and supernatants were stored at Ϫ80°C until use. Glu content was determined by reverse-phase HPLC on a Crestpak C18S column (4.6 ϫ 150 mm) (JASCO, Tokyo), using precolumn derivatization with o-phthalaldehyde and fluorescence detection (FP-920-S, JASCO, Tokyo) (17).
Treatment of Neuronal Cells with Botulinum Type B Neurotoxin (BoNT/B)-BoNT/B was prepared as described (18) and used to treat neuronal cells as described (19). In brief, cells were incubated in DMEM containing B27 supplement (Life Technologies, Inc.) in the presence or absence of 10 nM BoNT/B at 37°C for 18 h, and Glu release was measured as described above.
Organotypic Cultures of Rat Hippocampal Slices-Organotypic slice cultures of rat hippocampus were prepared as described (20). The hippocampus was removed from Wistar rats of either sex 7-9 days after birth, and 400-m slices were prepared using a McIlwain tissue chopper (Mickle Laboratory Engineering Co.). Each slice was placed on a filter (Millicell-CM, 30 mm, Millipore), then transferred to 6-well microplates (3810-006, Asahi Techno Glass, Tokyo, Japan). Four slices were placed in each well, containing 1.0 ml of DMEM supplemented with 25% heat-inactivated horse serum, and cultured for 1 week.
Assay of Neuronal Cell Death in Organotypic Slice Cultures-After 1 week of culture, 2 M propidium iodide (PI), a fluorescent marker of degenerating cells, was added to the culture medium (21). After 4 h, PI incorporated into the slice was visualized by fluorescence microscopy (5ϫ objective lens, 520-to 555-nm bandpass filter). Subsequently, the slices were incubated in chemical ischemia solution for 20 min in the presence or absence of either EMP1 or iEMP1. After washing, the slices were further cultured for an additional 24 h in normal culture medium. 2 M PI was added to the culture medium, and the uptake was again visualized as described above. Fluorescence was detected using a high sensitivity video camera (SIT-camera C2400-8, Hamamatsu Photonics, Hamamatsu, Japan) at constant sensitivity, and quantified as the average intensity using an image analyzer (Argus 50, Hamamatsu Photonics). The total fluorescence intensity for a given area (about 180 ϫ 180 for whole-slice images), clipped from a total of 256 ϫ 241 pixels, was calculated by digitizing the fluorescence intensity by frame memory, and the average fluorescence intensity was obtained by dividing the intensity by the total pixels.
Immunoblotting-Proteins were denatured at 100°C for 5 min in SDS sample buffer (2% SDS, 10% (v/v) glycerol, 50 mM Tris-HCl, and pH 6.8) in the presence of 10% (v/v) 2-mercaptoethanol. SDS-polyacrylamide gel electrophoresis (PAGE) was performed on linear 2-15% acrylamide gradient gels (Daiichi Pure Chemicals, Tokyo). After separation by SDS-PAGE, the proteins were transferred to polyvinylidene difluoride membranes following standard procedures with a semi-dry transblotting apparatus. The membranes were blocked in 10% (w/v) nonfat milk in Tris-buffered saline (TBS) and incubated with antibodies overnight at 4°C. After washing in TBS containing 0.05% Tween 20, membranes were incubated for 1 h at room temperature with peroxidase-labeled anti-rabbit IgG antibody in TBS containing 10% nonfat milk. After washing, the immunoreactive bands were visualized by enhanced chemiluminescence (Amersham Pharmacia Biotech, Uppsala, Sweden) and a luminescence image analyzer with an electronically cooled charge-coupled device camera system (LAS-1000, Fuji Photo Film Co., Tokyo, Japan).

EPO Inhibits Ca 2ϩ -induced Glu Release from Cerebellar
Granule Cells-Cerebellar granule cells release substantial amounts of Glu in response to 5 M ionomycin treatment in the presence of extracellular Ca 2ϩ . As shown in Fig. 1A, EPO reduced Ca 2ϩ -dependent Glu release in a concentration-dependent manner. The inhibition appeared at concentrations above 10 ng/ml (0.28 nM) and about 30% inhibition was achieved at 50 ng/ml (Fig. 1A). The inhibition of Glu release by EPO was induced rapidly and attained a maximal level within 20 min after at 50 ng/ml (Fig. 1B). No significant change was observed in basal Glu release in low K ϩ solution.
EMP1, a Synthetic EPO-R Agonist, Inhibits Ca 2ϩ -induced Glu Release in an Activity-dependent Manner-The inhibition of Ca 2ϩ -induced Glu release was also induced by another EPO-R agonist, EMP1. EMP1 is a synthetic 20-mer peptide having no sequence similarity to EPO (22). EMP1 binds to an extracellular region of EPO-R distinct from the EPO binding site and induces receptor dimerization and activation (23). As shown in Fig. 2, EMP1 also inhibited Ca 2ϩ -induced Glu release from cerebellar granule cells in a concentration-dependent manner. Inhibition by EMP1 was specific, because no significant decrease in Glu release was observed with iEMP1, an inactive analogue of EMP1 with two Cys residues mutated to Ser (22). These results indicate that Ca 2ϩ -induced Glu release from cerebellar granule cells was reduced by the activation of EPO-R.
Activation of JAK2 by EPO and EMP1-JAK2 is a tyrosine kinase associated with EPO-R, and the activation of JAK2 is essential for signal transduction via the EPO-R. As shown in Fig. 3, genistein, a tyrosine kinase inhibitor, but not genistin, an inactive analogue of genistein, abolished EPO and EMP1induced inhibition of Glu release from cultured cerebellar granule cells. Activation of JAK2 was further investigated by immunoblotting of cellular homogenates using antibodies specific for tyrosine-phosphorylated JAK2 and for EPO-R. As shown in Fig. 4, the intensity of the immunoreactive band increased markedly after a treatment with EPO. Pretreating the cells with genistein but not with genistin blocked the tyrosine phosphorylation of JAK2. Tyrosine phosphorylation of JAK2 was also induced by EMP1 but not by iEPM1. The amounts of EPO-R estimated by immunoblotting did not change significantly in these conditions. These results suggest that the activation of JAK2 is involved in EPO-R-mediated suppression of Glu release from cerebellar granule cells.
Activation of EPO-R Suppresses Exocytotic Glu Release from Cerebellar Granule Cells Induced by Chemical Ischemia-Next we examined in cultured cerebellar granule cells whether Glu release induced by ischemia could be suppressed by activation of EPO-R (24). The extracellular solution was replaced with low K ϩ solution without glucose and supplemented with 1 mM KCN to induce chemical ischemia. The extracellular solution was changed every 5 min, and the amount of Glu released into the solution was determined. As shown in Fig. 5A, chemical ische-mia induced substantial Glu release from cultured cerebellar granule cells. After a lag of several minutes, Glu release gradually increased with time and attained a maximal level about 30 min after treatment. To determine whether chemical ischemia-induced Glu release occurred via exocytosis, or another mechanism, we examined the effect of BoNT/B. BoNT/B blocks neurotransmitter release by specifically cleaving vesicle-associated membrane protein-2, a synaptic vesicle protein essential for exocytosis (25). We found that there was a difference in the BoNT/B sensitivity of Glu release between early and late phase. As shown in Fig. 5B, Glu release in the early phase (15-20 min after the ischemic treatment) was almost abolished by BoNT/B, indicating that exocytosis was the predominant pathway involved. On the other hand, Glu release in the late phase (45-50 min after the ischemic treatment) was inhibited only partially by BoNT/B, consistent with an alternative secretory mechanism. EMP1 effectively inhibited Glu release induced by chemical ischemia in the early phase but not in the late phase. These results indicate that Glu was released by at least two different mechanisms during chemical ischemia and that EMP1 only suppressed Glu release by exocytosis from synaptic vesicles.

Activation of EPO-R Suppresses Exocytotic Glu Release from Hippocampal Neurons Induced by Chemical Ischemia-Glu
was also released when cultured rat hippocampal neurons were exposed to chemical ischemia. As shown in Fig. 6A, after a lag period, Glu release gradually increased with time and attained a maximal level about 45 min after treatment. As shown in Fig.   6B, Glu release was markedly suppressed by EMP1 and BoNT/B in the early phase (15-20 min after the ischemic treatment). Genistein, but not genistin, abolished EMP1-mediated inhibition of Glu release, indicating that tyrosine kinase was involved in the inhibition (Fig. 6B). In contrast, release was inhibited only slightly by either BoNT/B or EMP1 in the late phase (45-50 min after the ischemic treatment), suggesting that an EMP1-insensitive non-exocytotic mechanism also occurs in hippocampal neurons.

Activation of EPO-R Protects Hippocampal Neurons from Ischemic Neuronal Damage through the Inhibition of Glu
Release by Exocytosis-Because an excessive release of Glu in ischemia induces neuronal cell death in the hippocampus, we next examined whether neuronal cell death could be attenuated by activation of EPO-R. Hippocampal slices in organotypic culture were maintained in chemical ischemic conditions for 20 min to induce Glu release by exocytosis. After washing with normal low K ϩ solution, slices were cultured for an additional 24 h in normal culture conditions. Neuronal cell death was evaluated before and 24 h after ischemic treatment by incorporation of PI, a fluorescent marker. PI incorporation was dramatically increased 24 h after chemical ischemia, especially in the CA1 region and the dentate gyrus, suggesting that exocytotic Glu release induced neuronal cell death in these regions (Fig. 7, A and B). Neuronal cell death was markedly suppressed by treating the slice with EMP1 during chemical ischemia, whereas its inactive analogue iEMP1 showed no suppressive effect. These results indicate that EPO-R activation could protect neurons from ischemic damage.
Because it was not clear whether the neuronal cell death was prevented by EPO-R activation via a presynaptic or postsynaptic mechanism, we next examined the effects of EPO-R activation on neuronal cell death induced by exogenously applied Glu (Fig. 8). If EMP1 suppressed cell death through a postsynaptic mechanism, protection against exogenous Glu would be expected. As shown Fig. 8, exogenously applied Glu induced neuronal cell death not only in the CA1 region and the dentate gyrus but also in the CA2 and CA3 regions. EMP1 did not provide significant protection from Glu-induced cell death in any of these regions even reducing Glu concentration down to 0.5 mM (PI fluorescent ratios were 1.01 Ϯ 0.22 and 1.04 Ϯ 0.10 in control slices; 3.19 Ϯ 0.52 and 2.84 Ϯ 0.60 in slices treated with 0.5 mM Glu; 3.05 Ϯ 1.52 and 1.52 Ϯ 0.39 in slices treated with the chemical ischemia, in the absence or presence of 10 mM EMP1, respectively). These results suggest that EPO-R activation attenuates neuronal cell death by inhibiting Glu release via exocytosis. DISCUSSION We showed in the present study that EPO and EMP1 inhibit Ca 2ϩ -induced Glu release from cultured cerebellar granule cells and hippocampal neurons. EPO-R activation also protected hippocampal neurons from ischemic damage by reducing the exocytosis of Glu. This is the first demonstration that EPO-R activation could exert its neuroprotective effects by presynaptic mechanisms.
The expression of EPO-R in cultured cerebellar granule cells was confirmed by immunoblotting. The inhibitory effect of EPO-on Ca 2ϩ -induced Glu release was observed in a nanomolar concentration range, very close to the reported K d for EPO binding to the EPO-R (26). Inhibition was also induced by EMP1, another EPO-R agonist with a completely different structure and binding site on the EPO-R (22,23). EPO and EMP1 induced tyrosine phosphorylation of JAK2, a tyrosine kinase that associates with EOP-R, in cultured cerebellar granule cells. Pretreating the cells with genistein but not with genistin abolished the EPO-and EMP1-induced suppression of Glu release and the phosphorylation of JAK2. All of these results clearly indicate that in cultured cerebellar granule cells the inhibitory effect of EPO and EMP1 on Ca 2ϩ -induced Glu release is mediated through activation of the EPO-R.
EPO and EMP1 inhibited Glu release induced by ionomycin. Because voltage-dependent Ca 2ϩ channels are not involved in Glu release in these conditions, inhibition must occur by mod- ulating a step subsequent to Ca 2ϩ entry. The onset of inhibition by EPO/EMP occurred very rapidly after application, suggesting that regulation is likely to involve post-translational modulation of a pre-existing protein. Previously, we have shown that phosphorylation of GAP-43 at Ser 41 was decreased after EPO/EMP1 treatment in PC12 cells (12). However, we have not detected any dephosphorylation of GAP-43 in cultured cerebellar granule cells, suggesting that this process is not necessary for EPO/EMP1-induced inhibition of neurotransmitter release (data not shown). The identity of the protein substrate(s) of tyrosine kinase essential for this modulation has yet to be elucidated. Various protein kinases, including protein kinase C, cAMP-dependent protein kinase, mitogen-activated protein kinase, and TrkA receptor protein-tyrosine kinase, are known to stimulate neurotransmitter release from a variety of neuronal preparations (27,28). However, reports on protein kinases involved in negative regulation of neurotransmitter release are extremely scarce to date. We have shown for the first time that EPO-R activation inhibits Ca 2ϩ -induced catecholamine release from PC12 cells (12) and Ca 2ϩ -induced Glu release from neurons of hippocampus and cerebellum. Thus, EPO-R-mediated suppression of release seems not to be restricted to any particular type of neurotransmitter, and it will be important to dissect the underlying mechanisms to improve our understanding of the molecular basis of synaptic plasticity.
There are a number of reports showing that there are two phases of Glu release during ischemia (29). We found in the present study that Glu release in chemical ischemia was induced by at least two different mechanisms in cultured cerebellar granule cells and hippocampal neurons. In the early phase (up to 20-min treatment), Glu release was almost abolished by BoNT/B, indicating that it occurs via exocytosis of the contents of synaptic vesicles. In contrast, Glu release in the later phase was only partially inhibited by BoNT/B, suggesting that a non-exocytotic mechanism is involved, possibly by reversal of neuronal Glu transporters (30). EMP1 almost abolished Glu release in the early phase, but scarcely affected the release in the late phase of ischemia. These results indicate that during chemical ischemia the activation of EPO-R only suppressed exocytotic Glu release. The excessive release of Glu in brain ischemia is known to trigger neuronal cell death. In the present study, we found that a 20-min exposure of cultured hippocampal slices to chemical ischemia induced neuronal cell death selectively in the CA1 region and the dentate gyrus, which suggests that Glu released by exocytosis was enough to induce the neuronal cell death in ischemia. EMP1 suppressed cell death induced by the chemical ischemia for 20 min. Activation of EPO-R was involved in this process, because iEMP1, the inactive analogue of EMP1, did not produce any protective effect. Neuronal cell death induced by exogenously applied Glu was not suppressed by EMP1. Thus, it is reasonable to conclude that EMP1 attenuated neuronal cell death presynaptically by inhibiting exocytotic Glu release, rather than through modification of a post-synaptic step downstream of Glu receptor activation.
Previously, EPO has been reported to protect neurons against ischemia-induced cell death in living gerbils. Infusion of EPO into the lateral ventricles of gerbils prevented ischemia-induced learning disability and rescued hippocampal CA1 neurons from lethal ischemic damage (11,31). Because EPO rescued cultured neurons from NO-induced death, it was concluded that EPO exerts its neuroprotective effect by postsynaptic mechanisms, possibly by reducing the NO-mediated formation of free radicals or antagonizing their toxicity (11,32). EPO also protected cultured hippocampal and cerebral cortical neurons from glutamate neurotoxicity (10).
However, treatment of neurons with EPO ϳ8 h prior to exposure to glutamate was required, and RNA and protein synthesis were also necessary, to demonstrate a neuroprotective effect. The action of the EPO-R revealed in the present study is clearly different, because EPO and EMP1 confer protection very quickly by inhibiting presynaptic functions. Thus, the activation of EPO-R seems to protect neurons from ischemic damage through multiple mechanisms, and presynaptic and postsynaptic mechanisms are likely to be involved in inducing protection in the early and the late phases of ischemia, respectively.
Adenosine A 1 receptor agonists had been used to prevent neuronal cell death in ischemia, because they depress Glu release from cortical slice preparations, however, cardiovascular side effects have constituted major obstacles to clinical implementation (3). EPO has proven to be a safe therapeutic agent with minimal adverse effects (33,34). However, it is thought that direct delivery of EPO into the brain is not a practical approach in clinical contexts, because the blood-brain barrier effectively excludes large glycosylated molecules (35,36). Recently, EPO was found to cross the blood-brain barrier and intraperitoneal administration of recombinant EPO before or up to 6 h after focal brain ischemia reduced injury (37). These observations raise the possibility that EPO and EMP1 could be used therapeutically to attenuate neuronal damage in the early phase of ischemia. Further studies, including elucidation of the molecular mechanisms of JAK2-mediated inhibition of neurotransmitter release, are required with a view to developing novel strategies for neuroprotection in ischemia.