Ischemia-like Oxygen and Glucose Deprivation Mediates Down-regulation of Cell Surface γ-Aminobutyric AcidB Receptors via the Endoplasmic Reticulum (ER) Stress-induced Transcription Factor CCAAT/Enhancer-binding Protein (C/EBP)-homologous Protein (CHOP)*

Background: ER stress associated with cerebral ischemia induces the expression of the transcription factor CHOP. Results: Interaction with CHOP down-regulates cell surface GABAB receptors and, thus, GABAB receptor-mediated neuronal inhibition. Conclusion: Interaction of CHOP with GABAB receptors in the ER prevents forward trafficking of the receptors. Significance: This mechanism is expected to contribute to excitotoxicity in cerebral ischemia. Cerebral ischemia frequently leads to long-term disability and death. Excitotoxicity is believed to be the main cause for ischemia-induced neuronal death. Although a role of glutamate receptors in this process has been firmly established, the contribution of metabotropic GABAB receptors, which control excitatory neurotransmission, is less clear. A prominent characteristic of ischemic insults is endoplasmic reticulum (ER) stress associated with the up-regulation of the transcription factor CCAAT/enhancer-binding protein-homologous protein (CHOP). After inducing ER stress in cultured cortical neurons by sustained Ca2+ release from intracellular stores or by a brief episode of oxygen and glucose deprivation (in vitro model of cerebral ischemia), we observed an increased expression of CHOP accompanied by a strong reduction of cell surface GABAB receptors. Our results indicate that down-regulation of cell surface GABAB receptors is caused by the interaction of the receptors with CHOP in the ER. Binding of CHOP prevented heterodimerization of the receptor subunits GABAB1 and GABAB2 and subsequent forward trafficking of the receptors to the cell surface. The reduced level of cell surface receptors diminished GABAB receptor signaling and, thus, neuronal inhibition. These findings indicate that ischemia-mediated up-regulation of CHOP down-regulates cell surface GABAB receptors by preventing their trafficking from the ER to the plasma membrane. This mechanism leads to diminished neuronal inhibition and may contribute to excitotoxicity in cerebral ischemia.

GABA B receptors are G i/o protein-coupled receptors composed of the two obligatory and functionally distinct subunits GABA B1 and GABA B2 . GABA B1 harbors the binding site for orthosteric ligands, whereas GABA B2 contains a binding site for allosteric modulators, recruits the G protein, and is required for trafficking of the receptors from the endoplasmic reticulum (ER) 2 to the plasma membrane (reviewed in Ref. 1). GABA B1 contains an ER retention signal in the C-terminal domain that retains unassembled GABA B1 in the ER. Heterodimerization with GABA B2 masks the ER retention signal and permits ER export of the receptor heterodimers (2)(3)(4). GABA B receptors are abundantly expressed throughout the mammalian central nervous system, where they mediate slow and persistent inhibition. According to their prominent role in regulating neuronal excitability, GABA B receptors have been implicated in a variety of neurological disorders, including cerebral ischemia.
In cerebral ischemia, excessive glutamatergic neurotransmission eventually leads to neuronal death (5). Decreased GABAergic activity appears to contribute to neuronal overexcitation (6), and there are indications that GABA B receptors are down-regulated under ischemic conditions (7)(8)(9)(10). This suggests that impaired GABA B receptor signaling contributes to excitotoxicity. In line with these findings, enhancing GABA B receptor activity during ischemic insults by application of the GABA B receptor agonist baclofen has been reported to be neuroprotective in vitro and in vivo (10 -19).
Cerebral ischemia induces ER stress, which is characterized by the accumulation of proteins in the ER, leading to the acti-
Plasmids-Plasmids containing full-length cDNA of human CHOP and deletion mutants of CHOP (C-terminal deletion (⌬LZ) and N-terminal deletion (⌬N)) have been described previously (22). Plasmids containing wild-type and mutant rat GABA B1 and rat GABA B2 were provided by Bernhard Bettler (University of Basel, Basel, Switzerland) and are described in Ref. 3.
Cell Culture-Primary cortical neurons were prepared from E18 embryos of time-pregnant Wistar rats as described previously (26). Briefly, minced E8 cortex was incubated for 15 min with papain solution (0.5 mg/ml PBS, 1 mg/ml BSA, 10 mM glucose, and 10 g/ml DNase I), washed with Dulbecco's modified Eagle's medium containing 10% fetal calf serum and titrated using a Pasteur pipette. Neurons were plated at a density of 120,000 cells onto poly-L-lysine coverslips (12 mm) in 24-well plates and kept in culture at 37°C and 5% CO 2 for 11-15 days.
Neurons in neuron-glia cocultures were transfected with plasmids using magnetofection, exactly as described in Ref. 27. 60,000 cells were plated on 18-mm coverslips and kept in culture for 11-15 days at 37°C and 5% CO 2 . Magnetofection was performed for 30 min on a prewarmed magnetic plate.
Immunocytochemistry and Confocal Laser-scanning Microscopy-Multiplex-labeling immunocytochemistry was performed as described previously (26). For the visualization of cell surface GABA B receptors, living neurons were incubated with primary antibodies for 2 h at 4°C in ACSF (2 mM CaCl 2 , 2 mM MgCl 2 , 30 mM L-glucose, 5 mM KCl, 119 mM NaCl, and 25 mM HEPES (pH 7.4)) containing 10% normal goat serum. For staining of intracellularly localized proteins, neurons were subsequently fixed with 4% paraformaldehyde for 15 min at room temperature and permeabilized for 6 min with 0.2% Triton X-100. Neurons were then incubated with primary antibodies for 1 h (in PBS/10% normal goat serum) at room temperature, washed four times for 5 min with PBS, and incubated with secondary antibodies for 1 h. After four washes with PBS, neurons were mounted in fluorescence mounting medium and analyzed by confocal laser-scanning microscopy (LSM510 Meta, Zeiss). Images were acquired using a Zeiss ϫ100 plan apochromat oil differential interference contrast objective (1.4 numerical aperture) at 512 ϫ 512 pixel resolution for fluorescence intensity measurements or 1024 ϫ 1024 pixel resolution for colocalization studies. For each neuron, five optical sections spaced by 0.4 m were taken.
Fluorescence intensity measurements were performed using the Mac Biophotonics ImageJ software (version 1.41n). For analysis of cell surface protein expression, cells were outlined carefully, and the mean fluorescence intensity of the soma was subtracted. For total protein expression analysis, somata of neurons were outlined carefully, and the mean intensity of the fluorescence signals was measured. An area of each image con-taining no specific signals was selected for determining background staining and was subtracted from the image.
Colocalization studies were performed using Imaris (version 7.1.1, Bitplane, Zurich, Switzerland). Images were smoothed using the median filter tool (filter size, 3 ϫ 3 ϫ 1) and processed further by setting threshold cutoffs for each channel to exclude background staining. Colocalization channels were built (colocalization intensity 255, constant value), and protein clusters (Ͼ15 pixels) as well as colocalized clusters were counted within a randomly selected, 30-m 2 area of the somata.
Gene Expression Assays-GABA B1 , GABA B2 , and CHOP RNA levels were determined in cortical primary neurons using real-time PCR. Total RNA was extracted from neurons using the GenElute mammalian total RNA miniprep kit (Sigma-Aldrich) according to the recommendations of the manufacturer. Reverse transcription was performed with the QuantiTec reverse transcription kit (Qiagen). Quantitative real-time PCR (7900HT fast real-time PCR system, Applied Biosciences) was done using the prepared cDNA and TaqMan gene expression assays (Applied Biosciences) for GABA B1 (Gabbr1, assay ID Rn00578911_m1), GABA B2 (Gabbr2, assay ID Rn00582550_ m1), CHOP (Ddit3, assay ID Rn00492098_g1), and ␤-actin (Actb, assay ID Mm00607939_s1) as a control. Quantification of RNA levels was done using the ⌬⌬Ct method.
Baclofen-induced ERK1/2 Phosphorylation-GABA B receptor activity was indirectly determined by measuring the levels of baclofen-induced ERK1/2 phosphorylation (28 -31). Cortical neurons were incubated with the GABA B receptor agonist baclofen for 10 min at 37°C, 5% CO 2 or were left untreated for controls. Subsequently, the cultures were placed on ice, fixed with 4% paraformaldehyde for 15 min at 4°C, and permeabilized with 0.2% Triton X-100 for 6 min. For determination of ERK1/2 phosphorylation, cultures were incubated overnight at 4°C with antibodies directed against total ERK1/2 as well as with antibodies against diphosphorylated ERK1/2. After washing with PBS, secondary antibodies were added for 1 h at room temperature, and neurons were analyzed by measuring fluorescence intensities of total and diphosphorylated ERK1/2 levels using confocal laser-scanning microscopy. Levels of phosphorylated ERK1/2 were normalized to total ERK1/2 levels. The specificity of the baclofen-induced ERK1/2 phosphorylation was determined using the GABA B receptor antagonist CPG 56999A.
In Situ PLA-The in situ PLA is a highly sensitive antibodybased method for the visualization of protein-protein interactions and posttranslational modification in cultured cells and tissue sections (32,33). This method employs two primary antibodies detecting the proteins of interest raised in different species and corresponding secondary antibodies (PLA probes) tagged with oligonucleotides. Only when the proteins of interest are in close proximity (Ͻ30 nm), specific connector oligonucleotides can be hybridized and ligated to the oligonucleotides attached to the secondary antibodies, forming a circular oligonucleotide. Rolling circle amplification then creates a large DNA strand to which numerous fluorophore-labeled oligonucleotides (detection probes) are hybridized. This generates a bright fluorescent spot that can be easily detected by microscopy. Quantification is done by counting the number of spots.
Here we used in situ PLA for analyzing the interaction of CHOP with GABA B receptors and the heterodimerization of GABA B1 and GABA B2 . In situ PLA was performed using the Duolink kit (Olink Bioscience) according to the protocol of the manufacturer, as described previously (34). The specificity of the PLA signal was verified for both pairs of antibodies in HEK 293 cells expressing or not expressing one of the interaction partners. Furthermore, leaving out one of the primary antibodies completely prevented PLA signals. For in situ PLA, neurons were fixed and permeabilized as described above and incubated with primary antibodies (in PBS/10% normal goat serum) overnight at 4°C. After in situ PLA, neurons were analyzed by confocal laser-scanning microscopy as described above. Proteinprotein interactions were quantified by counting signal dots using the ImageJ software. Image stacks (five optical sections spaced by 0.4 m) of individual neurons were merged, visualizing maximum intensities, and the number of maxima per area was determined.
Forward Trafficking Assay-For visualization of the amount of GABA B receptors inserted into the plasma membrane within a time period of 16 h, neurons were incubated in culture medium with antibodies directed against GABA B2 for 2 h at 37°C. Following washes with ACSF to remove unbound primary antibody, cells were incubated for 2 h with a large excess of Alexa Fluor 488-conjugated secondary antibody to label (i.e. mask) the existing pool of cell surface receptors. After washing, the neurons were further incubated in culture medium for 16 h at 37°C to allow neosynthesis of GABA B receptors and forward trafficking of receptors to the plasma membrane. Neurons were then placed on ice and stained at 4°C for receptors newly inserted into the plasma membrane using antibodies directed against GABA B2 and Cy3-conjugated secondary antibodies. Neurons were then processed for confocal laser-scanning microscopy. Controls for judging the efficiency of labeling (i.e. masking) the pool of cell surface receptors were treated in exactly the same manner but kept at 4°C for the 16-h incubation period.
Electrophysiology-Neurons of thapsigargin-treated or untreated control cultures were recorded in the whole-cell voltage clamp configuration at room temperature. Spontaneously occurring postsynaptic currents (sPSCs) were recorded before, during, and after the application of 50 M baclofen at a holding potential of Ϫ60 mV. Patch electrodes were pulled from borosilicate glass and filled with 120 mM CsCl, 10 mM EGTA, 10 mM HEPES (pH 7.4), 4 mM MgCl 2 , 0.5 mM GTP, and 2 mM ATP. The external solution contained 140 mM NaCl, 10 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES (pH 7.4), and 10 mM glucose. Recordings were performed with a HEKA EPC-7 amplifier and Patch Master v2.11 software (HEKA Elektronik, Germany). Baclofen (50 M) was applied locally using an outlet tube (inner diameter, 200 m) of a custom-designed, gravity-fed microperfusion system positioned 50 -100 m of the recorded neuron. All synaptic events displaying amplitudes above the background noise (5-12 pA) were identified and analyzed offline using the MiniAnalysis 6.0.7 software (Synaptosoft). Mean amplitudes and frequency values were obtained from 1-minlong recordings for each experimental condition and normalized.
Oxygen and Glucose Deprivation (OGD) Model of Ischemia-Primary cortical neurons were washed twice with DMEM without glucose or with DMEM containing glucose for controls. For OGD, neurons were incubated for 10 min at 37°C in 95% N 2 /5% CO 2 prebubbled, glucose-free DMEM in an airtight box filled with 95% N 2 /5% CO 2 . Control cultures were incubated in DMEM containing glucose at 37°C in a physiological oxygencontaining environment. Subsequently, cells were washed twice with DMEM containing glucose and incubated for 24 h in their original culture medium.

CHOP Interacts with GABA B Receptors and Down-regulates Cell Surface Receptors in Neurons-
We have previously shown that the stress-induced transcription factor CHOP, which is only marginally expressed under normal physiological conditions, interacts with GABA B receptors by binding to the coiled coil motif in the C-terminal domain of GABA B2 and with a so far unidentified site in GABA B1 (22). Upon overexpression in HEK 293 cells, the interaction of CHOP with GABA B receptors resulted in the intracellular accumulation and a reduced cell surface expression of the receptors by an as yet unknown mechanism (22). To confirm an interaction of GABA B receptors with CHOP in neurons, we overexpressed CHOP in cultured cortical neurons and tested for an interaction with GABA B receptors using the in situ PLA (32,33). In non-transfected neurons, only very few PLA dots were detected, in line with a low expression level of CHOP under normal physiological conditions. However, overexpression of CHOP dramatically increased PLA signals, indicating numerous GABA B receptor-CHOP interactions (Fig. 1A).
Because ER stress-induced up-regulation of CHOP has been shown to play an important role in ischemia-induced neuronal death (35)(36)(37)(38)(39)(40)(41)(42), we then analyzed the mechanism of CHOPinduced down-regulation of cell surface GABA B receptors by exposing cultured cortical neurons to ER stress. ER stress was induced by treating neurons for 2 h with the sarco/endoplasmic reticulum Ca 2ϩ -ATPase blocker thapsigargin. After a 16-h recovery period, neurons were analyzed for CHOP and GABA B receptor interaction using an in situ PLA. Under control conditions, i.e. in untreated neurons, only few interactions were observed (Fig. 1B). However, upon up-regulation of CHOP with thapsigargin, numerous interactions with GABA B receptors were detected (454 Ϯ 53% of control, n ϭ 30, p Ͻ 0.001, Fig. 1B). Thus, up-regulation of CHOP by ER stress resulted in an interaction with GABA B receptors. Under these conditions, no in situ PLA signals were generated using CHOP and GABA A receptor antibodies, documenting the specificity of the CHOP-GABA B receptor interaction (Fig. 1C).
These findings suggest that the interaction of CHOP with GABA B receptors caused a down-regulation of the receptors from the cell surface and their intracellular accumulation.
To verify that the loss of cell surface receptors was caused by the interaction of CHOP with GABA B receptors, neurons were transfected with mutant forms of CHOP that are unable to interact with GABA B1 (CHOP⌬N, N-terminal deletion (22)) or GABA B2 (CHOP⌬LZ, deletion of C-terminal leucine zipper motif (22)). Neurons overexpressing wild-type CHOP displayed a significantly reduced level of cell surface GABA B2 compared with control neurons transfected with EGFP (63 Ϯ 4%, n ϭ 82, p Ͻ 0.001, Fig. 2C). However, overexpressing either of the CHOP mutants did not affect cell surface expression of GABA B receptors (CHOP⌬LZ, 92 Ϯ 5%, p Ͼ 0.05, n ϭ 79; CHOP⌬N, 103 Ϯ 5%, n ϭ 101, p Ͼ 0.05), suggesting that downregulation of cell surface receptors was mediated by its interaction with CHOP.
To test whether CHOP-induced down-regulation of GABA B receptors affects GABA B receptor-mediated neuronal inhibition, spontaneous synaptic activity was measured in electrophysiological experiments using the whole-cell voltage clamp configuration. In untreated control neurons, application of baclofen considerably decreased the amplitude (66 Ϯ 6% of control, n ϭ 9) and the frequency (38 Ϯ 8% of control, n ϭ 5) of sPSCs (Fig. 3B). However, after up-regulation of CHOP by treating cultures with thapsigargin, baclofen-induced inhibition of sPSCs was strongly reduced (amplitude, 87 Ϯ 6% of control, n ϭ 9, p Ͻ 0.05; frequency, 65 Ϯ 9% of control, n ϭ 5, p Ͻ 0.05; Fig. 3B). This finding is consistent with an up-regula-tion of CHOP and, subsequently, reduced levels of functional GABA B receptors available for neuronal inhibition.
CHOP Interacts with GABA B Receptors in the ER and Interferes with Receptor Heterodimerization-Next, we investigated the mechanism of CHOP-mediated down-regulation of cell surface GABA B receptors. We envisioned two scenarios. 1) CHOP disrupts functional GABA B receptor heterodimers on the cell surface, which would lead to their internalization and degradation, or 2) CHOP interacts with GABA B receptors in the ER and inhibits receptor heterodimerization and, consequently, forward transport to the cell surface. To test these two scenarios, we analyzed the colocalization of CHOP and GABA B receptor subunits in different cellular compartments in untreated control neurons and neurons treated with thapsigargin to up-regulate CHOP. In thapsigargin-treated neurons, GABA B receptors accumulated in the ER, as indicated by an increased colocalization of GABA B1 (control, 10 Ϯ 0.7 clusters, n ϭ 20; thapsigargin, 15 Ϯ 0.8 clusters, n ϭ 21, p Ͻ 0.001) and GABA B2 (control, 7 Ϯ 0.7 clusters, n ϭ 32; thapsigargin, 13 Ϯ 0.7 clusters, n ϭ 32, p Ͻ 0.001) with the ER marker PDI (Fig. 4,  A and B). Hardly any CHOP was observed in the ER under control conditions, whereas CHOP accumulated in the ER after thapsigargin treatment (GABA B1 -stained neurons, control, 2 Ϯ 0.4 clusters, n ϭ 20; thapsigargin, 9 Ϯ 0.9 clusters, n ϭ 21, p Ͻ 0.001; Fig. 4A; GABA B2 -stained neurons, control, 1.5 Ϯ 0.4 clusters, n ϭ 32; thapsigargin, 9 Ϯ 0.6 clusters, n ϭ 32, p Ͻ 0.001; Fig. 4B). Triple colocalization of GABA B -CHOP-PDI was basically absent under control conditions but increased significantly after up-regulation of CHOP by thapsigargin (Fig.  4, A and B; GABA B1 -CHOP-PDI, control, 1.4 Ϯ 0.3 clusters, n ϭ 20; thapsigargin, 6 Ϯ 0.7 clusters, n ϭ 21, p Ͻ 0.01; GABA B2 -CHOP-PDI, control, 0.7 Ϯ 0.3 clusters, n ϭ 32; thapsigargin, 6 Ϯ 0.5 clusters, n ϭ 32, p Ͻ 0.001). These findings indicate that GABA B receptors interact with CHOP in the ER, which leads to the accumulation of GABA B1 and GABA B2 in the ER.
The identical colocalization patterns of GABA B1 ASA/GABA B2 receptors and endogenous GABA B receptors indicate normal trafficking capabilities of the mutant receptors.
Because GABA B1 and GABA B2 interact via the leucine zippers in their C-terminal domains, binding of CHOP to GABA B2 may prevent GABA B receptor heterodimerization and, thus, inhibit forward trafficking of the receptors. Therefore, we tested whether up-regulation of CHOP interferes with heterodimerization of GABA B1 and GABA B2 using in situ PLA. In thapsigargin-treated neurons, we observed a significant reduction of in situ PLA signals (64 Ϯ 5% of control, n ϭ 32, p Ͻ 0.001, Fig. 5C), indicating the presence of considerably less GABA B1 /GABA B2 heterodimers compared with untreated control neurons. This result suggests that the interaction of CHOP with GABA B receptors disrupts or prevents the heterodimerization of GABA B1 and GABA B2 .
These findings suggest a mechanism in which CHOP interacts with GABA B receptors in the ER after cellular stress to prevent heterodimerization of GABA B1 and GABA B2 . Because only heterodimerized GABA B receptors can leave the ER (2-4), this mechanism is expected to impede forward trafficking of newly synthesized receptors to the plasma membrane.
Up-regulation of CHOP Prevents Forward Trafficking of GABA B Receptors to the Cell Surface-To test whether up-regulation of CHOP impairs forward trafficking of GABA B receptors to the cell surface, we transfected neurons with HA-tagged GABA B2 and CHOP. Twenty-four hours after transfection, newly synthetized HA-tagged receptors inserted into the plasma membrane were detected using an anti-HA antibody. In control neurons only transfected with HA-GABA B2 , strong cell surface expression of HA-tagged GABA B2 was detected (Fig.  6A). However, in neurons transfected with CHOP, staining for cell surface HA-GABA B2 was strongly reduced (36 Ϯ 4% of control, n ϭ 25, p Ͻ 0.001, Fig. 6A). This reduction of cell surface HA-GABA B2 was not observed in neurons expressing a mutant of CHOP (CHOP⌬LZ) that is unable to bind to GABA B2 (80 Ϯ 7% of control, n ϭ 25, p Ͼ 0.05, Fig. 6A). These results suggest that the interaction of CHOP with GABA B2 impairs forward trafficking of GABA B receptors to the cell surface.
This result was confirmed for native GABA B receptors with an immunofluorescence-based forward trafficking assay using untreated control neurons and thapsigargin-treated neurons. After masking the existing receptor pool on the cell surface with primary and secondary antibodies, the insertion of new receptors into the plasma membrane was tested after 16 h. Membrane insertion of GABA B receptors was reduced drastically in neurons expressing high levels of CHOP (16 Ϯ 6% of control, n ϭ 33, p Ͻ 0.001, Fig. 6B).
Down-regulation of Cell Surface GABA B Receptors by CHOP in the OGD Model of Ischemia-Next, we tested whether upregulation of CHOP under ischemic conditions mediates down-regulation of cell surface GABA B receptors. We used the OGD in vitro model of ischemia, in which cultured cortical neurons were deprived of oxygen and glucose for 10 min, followed by a recovery period of 24 h. 24 h after OGD, CHOP expression in neurons was increased significantly (157 Ϯ 5% of control, n ϭ 311, p Ͻ 0.001, Fig. 7A). As expected, cell surface GABA B receptors decreased considerably 24 h after OGD in CHOP-expressing neurons (44 Ϯ 3% of control, n ϭ 49, p Ͻ 0.001, Fig. 7B).
To test whether the interaction of CHOP with GABA B receptors is responsible for the down-regulation of cell surface receptors after OGD, we overexpressed CHOP⌬LZ in cortical neurons. CHOP⌬LZ is a mutant form of CHOP that is not able to interact with GABA B2 and does not affect cell surface numbers of GABA B receptors (Fig. 1C). Endogenous CHOP was significantly up-regulated after OGD in transfected (154 Ϯ 8% of control, n ϭ 20, p Ͻ 0.001) and in non-transfected neurons (167 Ϯ 12% of control, n ϭ 22, p Ͻ 0.001, Fig. 8). In non-transfected neurons, GABA B2 receptors were down-regulated after OGD (60 Ϯ 7% of control, n ϭ 27, p Ͻ 0.001, Fig. 8), whereas in neurons overexpressing CHOP⌬LZ, despite similar up-regulation of endogenous CHOP, the down-regulation of GABA B receptors was completely prevented (113 Ϯ 7% of control, n ϭ 45, p Ͼ 0.05, ; Fig. 8). These results verify that down-regulation of cell surface GABA B receptors after ischemic conditions is mediated by their interaction with CHOP.

DISCUSSION
GABA B receptor-mediated neuronal inhibition critically depends on the availability of receptors in the plasma membrane. Receptor numbers might be altered under pathological conditions and a loss of receptors resulting in diminished neuronal inhibition is expected to contribute to the disease state. We showed previously that the ER stress-induced transcription factor CHOP interacts with GABA B receptors, causing their down-regulation from the cell surface upon coexpression in HEK 293 cells (22). This finding suggests that CHOP, besides its function as transcription factor, may regulate GABA B receptormediated neuronal inhibition by affecting the availability of cell surface receptors. Because CHOP is expressed at very low levels under normal physiological conditions but is highly up-regulated after induction of ER stress (21,(43)(44)(45), CHOP-induced down-regulation of cell surface GABA B receptors may be a contributing factor to neurological disease states associated with ER stress, such as stroke, Alzheimer and Parkinson disease, or bipolar disorders (20).
In this study, we verified that ER stress-induced CHOP expression also down-regulates cell surface GABA B receptors in neurons and disclosed the mechanism underlying this effect. An in vitro model of ischemia suggests that this mechanism is operative in cerebral ischemia.
We induced up-regulation of CHOP in cultured neurons by inhibition of the sarco/endoplasmic reticulum Ca 2ϩ -ATPase with thapsigargin, which leads to the depletion of Ca 2ϩ from the ER, thereby causing ER stress (46). Thapsigargin up-regulated CHOP in neurons and considerably down-regulated GABA B receptors from the cell surface. The reduced level of cell surface receptors consequently affected downstream signaling of GABA B receptors, as shown by impaired ERK1/2 phosphorylation and reduced baclofen-induced inhibition of spontaneous neuronal activity.
Down-regulation of cell surface GABA B receptors was mediated by the interaction of the receptors with CHOP because overexpression in neurons of mutant forms either lacking the site binding to GABA B1 or the site binding to GABA B2 prevented down-regulation of the receptors. Most importantly, up-regulation of CHOP did not affect the mRNA levels of GABA B1 and GABA B2 , ruling out a contribution of impaired subunit transcription (Fig. 2D).
Up-regulated CHOP selectively accumulates together with GABA B receptors in the ER, suggesting that interaction with CHOP retains GABA B receptors in the ER and prevents their forward trafficking to the cell surface. This conclusion is supported by the finding that mutant GABA B receptors not containing the CHOP binding site in GABA B2 do not accumulate in the ER and that the insertion of new receptors into the plasma  membrane is strongly reduced upon up-regulation of CHOP. The mechanism that interferes with forward trafficking of the receptors appears to involve prevention or disruption of receptor heterodimerization, as indicated by our in situ PLA experiments. It is currently not clear whether CHOP prevents heterodimerization of newly synthesized GABA B1 and GABA B2 by binding to GABA B1 and GABA B2 or directly disrupts already existing heterodimers. In either case, preventing heterodimerization exposes the ER retention signal in the C-terminal domain of GABA B1 , which prohibits ER exit of GABA B1 (2)(3)(4). In addition, GABA B2 contains a C-terminal sequence (amino acids 841-862) that is important for forward trafficking (47). Binding of CHOP to the leucine zipper domain of GABA B2 , upstream of this motif, might sterically interfere with the function of this motif and prevent ER export of GABA B2 .
The mechanism of CHOP-induced down-regulation of cell surface receptors appears to be operative under ischemic conditions. This is not surprising because cerebral ischemia has been shown to be associated with ER stress and the profound up-regulation of CHOP (35)(36)(37)(38)(39)(40)(41)(42). Using the OGD in vitro model of ischemia, we found that up-regulated CHOP mediates downregulation cell surface GABA B receptors, which depended on the interaction of CHOP with the receptors.
So far, the effect of cerebral ischemia on GABA B receptor expression levels has been rarely investigated, and the results are difficult to correlate because different animal species and experimental conditions were used. In vivo models of ischemia demonstrated a loss of GABA B receptors 1-4 days after the insult (7-9). Because considerable ischemia-induced neuronal death occurs during this time period, the loss of GABA B receptors might be due, at least in part, to a loss of GABA B receptorexpressing neurons. A recent in vitro study on cultured hippocampal slices using a similar OGD paradigm as in this study detected down-regulation of total GABA B2 but no effect on the expression of total GABA B1 (10). In this study, we detected down-regulation of both GABA B1 and GABA B2 selectively from the cell surface but did not observe changes in total receptor expression. The reason for this discrepancy is not clear yet, but it may be caused by different experimental conditions used (organotypic hippocampal slices, 45-min OGD versus cultured cortical neurons, 10-min OGD).
Cerebral ischemia is associated with excessive glutamate release, and overstimulation of glutamate receptors triggers a signaling cascade leading to neuronal death (48). The enhanced neuronal activity also increases extracellular levels of GABA (6), which, in turn, should activate GABA B receptors located at glutamatergic synapses to counteract, at least partially, the increased neuronal excitation to reduce excitotoxicity. The down-regulation of functional GABA B receptors from the cell surface might, therefore, be one factor that fosters excitotoxicity. It is well established that sustained activation of GABA B receptors (by application of baclofen) under ischemic conditions reduces neuronal cell death (10 -19). In this regard, it is conceivable that restoring normal cell surface expression of GABA B receptors under ischemic conditions would also reduce excitotoxicity. This view is supported by a recent study showing that mild hypothermia reverses down-regulation of total GABA B1 after cerebral ischemia by an unknown mechanism and reduces neuronal death (8).
One opportunity to restore normal cell surface GABA B receptor levels would be the application of small synthetic peptides interfering with the CHOP-GABA B receptor interaction. Further experiments are required to identify small peptide sequences that inhibit the interaction of CHOP with GABA B receptors but do not interfere with receptor heterodimerization. This approach provides the opportunity to unambiguously test the hypothesis whether CHOP-induced down-regulation of cell surface GABA B receptors promotes excitotoxicity. If this is the case, small interfering peptides may be of potential therapeutic use to limit neuronal death under ischemic conditions. Such an intervention would be highly specific because CHOP is significantly expressed only upon ER stress, and it targets a specific protein-protein interaction in response to the ischemic insult and is, thus, expected to be associated with a low risk of side effects.