The excessive neuronal excitation underlying several clinically important diseases is often treated with GABA allosteric modulators in an attempt to enhance inhibition. An alternative strategy would be to enhance directly the sensitivity of postsynaptic neurons to GABA. The GABAC receptor, normally found only in the retina, is more sensitive to GABA and demonstrates little desensitization compared with the GABAA receptor. We constructed an adenovirus vector that expressed cDNA for both the GABAC receptor ρ1 subunit and a green fluorescent protein (GFP) reporter and used it to transduce cultured hippocampal neurons. Transduced neurons were identified by fluorescence, double immunocytochemistry proved colocalization of the ρ1protein and the reporter, Western blot verified the expected molecular masses, and electrophysiological and pharmacological properties confirmed the presence of functional GABAC receptors. ρ1-GFP transduction resulted in an increased density of GABAA receptors as well as expression of novel GABAC receptors. This effect was not reproduced by addition of TTX or Mg2+ to the culture medium to reduce action potentials or synaptic activity. In a model of neuronal hyperexcitability induced by chronic blockade of glutamate receptors, expression of GABAC receptors abolished the hyperactivity and the consequent delayed neuronal death. Adenovirus-mediated neuronal GABAC receptor engineering, via its dual mechanism of inhibition, may offer a way of inhibiting only those hyperexcitable neurons responsible for clinical problems, avoiding the generalized nervous system depression associated with pharmacological therapy.
Enhancement of inhibitory neuronal activity may be beneficial for treating CNS diseases characterized by excessive excitation. Experimental models of epilepsy, for example, demonstrate a decrease in hippocampal GABAA receptor expression (Friedman et al., 1994;Rice et al., 1996) and function (Gibbs et al., 1997), suggesting that loss of inhibitory control plays a pathogenic role (Mody, 1998). Postischemic and traumatic neuronal death is reduced by the GABA potentiator diazepam (Schwartz et al., 1995; O'Dell et al., 2000). Enhancement of the GABA effect at the receptor is a reasonable strategy for increasing inhibition; this is the mechanism of many clinically useful drugs including anticonvulsants, anxiolytics, and general anesthetics (Hevers and Luddens, 1998). An alternative to receptor modulation is changing the number or type of GABA receptors, with consequent suppression of hyperexcitability. In epilepsy, in which relatively few hyperexcitable neurons are responsible for the initiation of seizures, overexpression of GABA receptors restricted to those few neurons might enhance inhibition with fewer side effects than conventional systemic pharmacotherapy (Jallon, 1997; Loscher, 1998).
The common ionotropic GABAA receptor is a pentamer formed of a combination of α, β, and γ subunits. Binding of the ligand GABA opens the integral chloride ion channel, driving the membrane potential toward the chloride equilibrium potential and thus reducing sensitivity to excitatory neurotransmitters. The rarer GABAC receptor, another member of the inotropic GABA receptor family, is a homomeric assembly of ρ1 subunits (Shimada et al., 1992;Amin and Weiss, 1996); a search for endogenous ρ mRNA using reverse transcription-PCR found ρ1 only in bipolar cells of the retina (Boue-Grabot et al., 1998), although other types of ρ subunits are found elsewhere in the CNS (Wang et al., 1994; Enz and Cutting, 1999). Recent evidence that ρ1 subunits do not coimmunoprecipitate in vitro with α1, α5, or β1 subunits of GABAAsuggests that ρ1 subunits do not associate with GABAA subunits to form receptors (Hackam et al., 1998). Pharmacologically, the GABAC receptor is characterized by insensitivity to the GABAAantagonist bicuculline and to the GABAB agonist baclofen (Bormann and Feigenspan, 1995; Bormann, 2000). Compared with the GABAA receptor, the GABAC receptor is 40 times as sensitive to GABA, is much slower, requiring eight times as long to activate and deactivate, and shows little or no desensitization (Amin and Weiss, 1994). Because of these favorable properties, forced expression of GABAC receptors should significantly enhance the effects of GABA in hyperexcitable neurons.
Adenovirus can transduce hippocampal neurons both in slice and in culture with high efficiency (Wilkemeyer et al., 1996;Griesbeck et al., 1997) and, at least in the short term, without affecting electrical excitability or synaptic transmission (Smith et al., 1997; Lissen et al., 1998; Johns et al., 1999). We report the use, in cultured hippocampal neurons, of a replication-deficient adenovirus designed to express cDNA encoding the GABACreceptor ρ1 subunit. Overexpression of the ρ1 subunit should result in formation of GABAC receptors with predictable pharmacological properties and without disruption of the native GABAA receptors.
MATERIALS AND METHODS
Rat primary hippocampal culture. Sprague Dawley rat pups (1- to 2-d-old) were decapitated, and the hippocampi were dissected out in ice-cold HBSS (Life Technologies, Gaithersburg, MD). The tissue was enzymatically digested with papain and bovine serum albumin (each at 1 mg/ml; Sigma, St. Louis, MO) for 20 min at 37°C. Cells were disaggregated by trituration and plated on Matrigel-coated 35 mm tissue culture plates (Becton Dickinson, Bedford, MA) in Neurobasal medium (Life Technologies) supplemented with 2 mm l-glutamine, 10% fetal calf serum (HyClone, Logan, UT), 5% horse serum, and B-27 supplement (Life Technologies). After 2–3 d of growth in a 95% O2–5% CO2 humidified incubator at 37°C, the dishes were treated with 10 μm cytosine arabinoside for 24 hr to suppress the growth of glia. Thereafter, the medium was switched to a Neurobasal medium containing 5% horse serum and changed every 2–4 d until the cultures were used for experiments.
Induction of neuronal hyperexcitability. Dissociated hippocampal neurons grown with glutamate receptors and synapses chronically blocked (>14 d) by kynurenate (1 mm) and magnesium (10 mm MgCl2) became hyperexcitable; when the kynurenate–Mg2+medium was replaced by a control salt solution without the blockers, neurons displayed seizure-like activity, consisting of bursts of action potentials riding on paroxysmal depolarization shifts (Furshpan and Potter, 1989).
Recombinant adenovirus. The recombinant E1- and E3-deleted replication-deficient human adenovirus type-5 was created via homologous recombination between the pXCR shuttle vector and the pBHG10 parent vector (Bett et al., 1994). So that the ρ1 subunit and green fluorescent protein (GFP) proteins would be expressed independently, the shuttle vector was modified to contain two expression cassettes, both driven by the Rous sarcoma virus promoter followed by a multiple cloning site and a polyadenylation sequence. GFP cDNA was subcloned into the first cassette, and human ρ1 subunit cDNA (a gift of Dr. Gary Cutting, Johns Hopkins University, Baltimore, MD) was subcloned into the second. The pBHG10 plasmids and the shuttle vector containing the transgene were cotransfected into human embryonic kidney 293 (HEK293) cells using the Ca-phosphate method (Life Technologies). Lytic plaques were isolated and expanded, and the presence of the transgene and the absence of the E1 gene were confirmed by PCR. High-titer adenovirus, twice purified by CsCl gradient centrifugation, was stored as a 10% glycerol suspension at −80Co. The titer of each adenovirus preparation was determined by counting GFP-positive plaques formed in a virus-transduced confluent HEK293 monolayer overlaid with low-melt agarose. Because the number of viable cells in a culture dish was unknown, we report the concentration of virus used for transduction of cultured cells as plaque-forming units (pfu) per milliliter rather than as pfu per cell. pBHG10 and pXCR plasmids were purchased from Microbix (Toronto, Ontario, Canada).
Whole-cell recording and data collection. Patch electrodes were pulled from 1.2-mm-outer-diameter borosilicate capillary glass (WPI, Sarasota, FL) and fire polished. The electrodes had a typical resistance of 5–10 MΩ when filled with intracellular solution. For voltage-clamp experiments the solution was composed of (in mm): 140 CsCl, 4 NaCl, 2 MgCl2, 10 K-EGTA, and 10 HEPES. For current-clamp recordings of action potentials, 140 mm K-gluconate replaced CsCl. Solutions were titrated to pH 7.3 with CsOH or KOH and supplemented with 2 mm Mg-ATP. The external solution contained (in mm): 140 NaCl, 2.8 KCl, 1 MgCl2, 3 CaCl2, 10 HEPES, and 10 glucose; the solution was titrated to pH 7.4 with NaOH. Recordings were made using an AxoPatch 200A amplifier (Axon Instruments, Foster City, CA). A typical access resistance of ∼15 MΩ in the whole-cell mode of patch clamp was compensated by 75%. Cell input capacitance was approximated by reading the capacitance compensation dial of the amplifier. Recorded membrane currents were filtered at 5 kHz, digitized using Clampex v8.0, and analyzed with Clampfit v6.0 (Axon Instruments). Kinetic parameters were determined by simultaneously fitting a monoexponential rising phase (activation) and biexponential decay phase (desensitization) to the evoked current from the beginning to the end of the duration of drug application. A separate monoexponential decay function was fit to the deactivation phase of the current after washout of GABA. A syringe pump delivered external solutions at 15 ml/hr through orifices of a θ tube mounted on a piezoelectric transducer (Burleigh Instruments, Fishers, NY). Command steps at 120 sec intervals rapidly moved the perfusion ports, exposing the cell to either the control or the drug solution. The perfusion device allowed exchange of solution in ∼15 msec (10–90% rise time) for the whole-cell recording configuration. Because of the limited exchange rate, the kinetic parameters determined should be considered approximate and to represent only the upper limit of the true process. For antagonist applications, GABA and the antagonist were applied simultaneously. Because GABA binding to its receptor is rather slow (Jones et al., 1998), rapid equilibrium should be established well within the time scale of tens of milliseconds relevant to our study even without antagonist preexposure. All experiments were performed at room temperature (20–25Co); all drugs were purchased from Sigma.
Assay for delayed neuronal death. Unopposed neuronal hyperactivity was induced by replacing the kynurenate–Mg2+ medium in chronically blocked cultures with one without the blockers for a variable time between 15 min and 24 hr. At 24 hr after the initiation of hyperexcitability, delayed neuronal death was determined by ethidium homodimer staining (2 μm for 5 min at room temperature). Neuronal injury allows ethidium homodimer to penetrate and intercalate into DNA to yield a bright red fluorescence. The stained culture dishes were examined under an epifluorescence microscope (Olympus IX50), and images of random fields were captured (Dage RC300; Dage-MTI, Michigan City, IN; Scion Imager V3.0; Scion Corporation, Frederick, MD) under both phase-contrast and fluorescent lighting. Phase-contrast images, which do not reveal cell death, were captured first to avoid bias in sampling. Live and dead neurons (100 total) were counted off-line from the captured images to determine the proportion of dead cells.
Immunocytochemistry. Cells grown on glass coverslips were fixed in ice-cold acid methanol (95% methanol and 5% acetic acid) for 10 min and permeabilized in PBS containing 0.2% Triton X-100 (PBST). The cells were blocked in PBST with 10% serum for 10 min, and all subsequent reactions were performed in PBST with 2% serum. Rabbit anti-ρ1 polyclonal antibody raised against an epitope of human ρ1 peptide conjugated with keyhole limpet hemocyanin (QRQRREVHEDAHK) (Hackam et al., 1997) was prepared and affinity purified by Genemed Synthesis (South San Francisco, CA). Double immunohistochemical staining for GFP and ρ1 subunit proteins was accomplished by overnight simultaneous primary antibody exposure [1:200 mouse anti-GFP (Boehringer Mannheim, Indianapolis, IN) and 1:50 rabbit anti-ρ1 antibodies] at 4Co followed by exposure for 1 hr at room temperature (RT) to anti-mouse IgG-FITC and anti-rabbit IgG-rhodamine secondary antibody (both at 1:200). Captured images were pseudocolored for the final figures using Adobe Photoshop.
Western blot analysis. HEK293 cells were transduced with the same virus used for neuronal transduction. After 24 hr the majority of the cells were brightly fluorescent but had not yet developed a cytopathic appearance. At this time the cells were scraped off with cold PBS, pelleted, and resuspended in ice-cold lysis buffer (1% NP-40, 50 mm NaCl, 30 mm Na pyrophosphate, 30 mm NaF, and 10 mm Tris HCl, pH 7.6) containing 1× proteinase inhibitor cocktail (Boehringer Mannheim). After a 30 min incubation in the lysis buffer on ice and centrifugation at 14,000 rpm for 20 min at 4 Co, the supernatant was aspirated and quantified for protein and then separated by a 10% SDS-PAGE. After an overnight transfer onto nitrocellulose and block in 5% milk in TBS plus Tween (TBST) for 1 hr at RT, paired lanes of control and transfected samples were probed with mouse anti-GFP antibody (1:500) and rabbit anti-ρ1antibody (1:100), all in 2% milk-TBST, for 2 hr at RT. ρ1-Antibody-specific blocking peptide (25 μg/ml) was added to the primary antibody for the blocking peptide experiment. After three washes and a secondary antibody exposure [goat anti-mouse IgG-HRP (Bio-Rad, Hercules, CA) or cat anti-rabbit IgG-HRP (Santa Cruz Biotechnology, Santa Cruz, CA); both at 1:1000] for 1 hr at RT, the membrane was reacted with a chemiluminescent substrate (NEN Life Sciences, Boston, MA), and an image was obtained by exposing an x-ray film. The final figure was obtained by digitizing the x-ray film and composing the images using Adobe Photoshop.
Recombinant adenovirus expression of GABA ρ1 subunits in cultured hippocampal neurons produces functional GABACreceptors
We constructed two recombinant adenoviruses (Ads). The control Ad-GFP expressed only the GFP reporter; the other, Ad-ρ1-GFP, coexpressed both GFP and the GABAC ρ1 subunit. Hippocampal neuron cultures were transduced with a range of concentrations of the two viruses. After 24 hr at 1 × 10− 5 pfu/1 ml, between 10 and 20% of cells exhibited GFP fluorescence with no apparent effect on the morphology of live neurons (Fig.1 A). This was the highest concentration at which neurons showed no gross morphological abnormalities and few non-neuronal cells (e.g., glia) were transduced. Double-immunocytochemical staining of Ad-ρ1-GFP-transduced neurons (ρ1-GFP-neurons) demonstrates ρ1 and GFP immunoreactivity in the same cells (Fig. 1 B). The GFP immunoreactivity is distributed diffusely throughout the cell in both ρ1-GFP-neurons and control Ad-GFP-transduced neurons (GFP-neurons). In contrast, ρ1immunoreactivity was found only in ρ1-GFP-neurons. Comparable ρ1 immunoreactivity in dendrites could not be discerned with certainty.
To confirm the expression of proteins of the expected molecular mass, we transduced HEK293 cells with the same viral constructs. Figure1 C, a Western blot of the membrane fraction harvested 24 hr after viral transduction, demonstrates an anti-ρ1 antibody immunoreactive band at 60 kDa consistent with the expected molecular mass of the ρ1 subunit. Identification of this band is confirmed by its disappearance in the presence of a specific anti-ρ1-antibody-epitope blocking peptide. A separate protein band at 25 kDa, corresponding to the GFP reporter protein, can also be seen in the soluble fraction; it is not sensitive to the blocking peptide.
Electrophysiological properties of ρ1-GFP-neurons, control GFP-neurons, and nonfluorescent neurons were measured by whole-cell patch clamp. Control GFP-neurons were activated by GABA but not bycis-4-aminocrotonic acid (CACA), a GABAC-receptor-selective agonist (Fig.1 D). In contrast, both GABA and CACA evoked currents in 16 of 17 fluorescent ρ1-GFP-neurons. As expected, the CACA-evoked current was kinetically distinct from the GABA-evoked current; activation and deactivation were slow, and there was no significant desensitization. The CACA-evoked current in ρ1-GFP-neurons demonstrated a linear current–voltage (I–V) relationship with a reversal potential shift (2.4 ± 0.7 mV with symmetric chloride concentrations and −58 ± 1.4 mV with a low internal chloride concentration) consistent with a current carried by the chloride ion (Fig. 1 E). These properties are characteristic of current mediated by GABAC receptors (Bormann and Feigenspan, 1995). CACA failed to evoke current in nonfluorescent neurons (n = 14) from the same culture dish (data not shown). We conclude that ρ1-GFP-neurons possess functional GABAC receptors.
Kinetic properties of GABAA receptors are not modified by expression of GABAC receptors in the same neuron
It is possible that the simultaneous expression of both GABAA and GABAC receptors in the same neuron might alter their kinetic properties. Alternatively, the GABAA and GABACreceptors may function independently and maintain their respective kinetic properties even when coexpressed together. We examined the time course of GABA-gated currents in transduced neurons. In control neurons, with only GABAA receptors, the time course of GABA-gated current was well fit by the expected monoexponential activation, biexponential desensitization, and monoexponential deactivation functions [Figs.2, left column, top(no virus) and middle (GFP) rows, 3]. GABA-gated current in ρ1-GFP-neurons, with both GABAA and GABAC receptors, is shown in Figure 2 (left column, bottom row). The slower desensitization and deactivation in comparison with those of the GABAA-only response are readily apparent. The GABAA-selective antagonist bicuculline blocked responses to GABA in control neurons and changed the kinetics of the ρ1-GFP-neurons to the slowly activating and deactivating current typical of GABAC receptors (Fig. 2, middle column). In contrast, the GABAC-selective antagonist imidazole-4-acetic acid (I4AA) had no effect on control neuron kinetics but returned the kinetics of the ρ1-GFP-neurons to that of GABAA alone (Fig. 2, right column). Figure 3 shows current traces at expanded time scales to emphasize the kinetic differences between GABAA and GABAC receptor-gated currents. Currents from nontransduced, GFP-neuron, and ρ1-GFP-neuron in the presence of I4AA superimposed, qualitatively demonstrating no kinetic effect of viral transduction in itself. Current from a ρ1-GFP-neuron in the presence of bicuculline demonstrated distinct kinetic signatures. Both GABAA (Fig. 3 D) and GABAC (Fig. 3 E) receptor-mediated components of ρ1-GFP-neuron were well described by a triexponential activation/desensitization and monoexponential deactivation kinetic model. Parameters of the kinetic fits to the pharmacologically separated currents within the resolution of our limited drug application system are shown in Table1.
GABAC receptors inhibit hyperexcitable hippocampal neurons
Ordinarily, hippocampal neurons grown in standard culture conditions exhibit modest levels of spontaneous activity; we observed 0.97 ± 0.32 action potentials/sec, as shown in Figure4 A (n = 8). Neurons cultured under conditions of chronic kynurenate blockade of glutamate receptors and chronic Mg2+blockade of synapses were hyperexcitable (Furshpan and Potter, 1989). After removal from chronic blockade, these neurons exhibited spontaneous action potential bursts, typically 300–500 msec long, riding on a wave of depolarization reminiscent of the paroxysmal depolarization shifts seen in epileptic foci in vivo andin vitro (Fig. 4 B). The mean action potential rate was 5.5 ± 1.03 sec− 1; the rate during bursts approached 20 sec− 1. This seizure-like electrical hyperexcitability occurred in 11 of 12 neurons examined.
Hyperexcitable neuron cultures were transduced with either Ad-ρ1-GFP or the control Ad-GFP. After 2–6 d, whole-cell patch-clamp recording was established in neurons expressing GFP. To avoid shifting the chloride reversal potential to a more positive value and thus artifactually enhancing excitability, a low chloride ion internal solution was used. After an initial 1–2 min settling period, action potential firing patterns were stable for at least 20 min for both control and transduced neurons (Fig.4 E). Action potential rates were determined from a 5 min data collection window after stabilization. Typically, control GFP-neurons exhibited the paroxysmal action potentials characteristic of nontransduced hyperexcitable neurons (Fig. 4 C). In contrast, ρ1-GFP-neurons exhibited low-frequency spontaneous action potentials without evidence of bursting (Fig. 4 D).
Pooled data from several experiments demonstrate large differences in spontaneous action potential rates: nontransduced, 5.4 ± 1.0 sec− 1; after Ad-GFP, 5.8 ± 0.9 sec− 1; and after Ad-ρ1-GFP, 0.48 ± 0.08 sec− 1 (Fig.4 F). The spontaneous action potential frequency of hyperexcitable neurons treated with Ad-ρ1-GFP was comparable with that of control neurons. Thus Ad-ρ1-GFP transduction reverses neuronal hyperexcitability in the kynurenate–Mg2+-blockade-induced model of epilepsy.
Both GABAA and GABAC receptors participate in suppression of hyperexcitability in Ad-ρ1-GFP-transduced hippocampal neurons
We then evaluated the effects of selectively activating GABAA and GABAC receptors in hyperexcitable neurons. Figure5 A (left) shows that nontransduced neurons (top), GFP-neurons (middle), and ρ1-GFP-neurons (bottom) all responded to GABA (30 μm) with membrane hyperpolarization and acute inhibition of spontaneous action potentials. In nontransduced and control neurons, this presumably occurred via activation of constitutively present GABAA receptors, whereas in ρ1-GFP-neurons, GABA activated both GABAA and GABAC receptors. Figure 5 A (right) shows that the GABAC-selective agonist CACA (100 μm) had no effect on nontransduced (top) or GFP-neurons (middle), but in ρ1-GFP-neurons (bottom), CACA caused membrane hyperpolarization and suppressed action potentials in those few neurons with spontaneous firing.
We also examined the effect of blockade of GABAAand GABAC receptors on ρ1-GFP-transduced hyperexcitable neurons. As expected, selective blockade of GABAA receptors with bicuculline failed to increase spontaneous firing. Surprisingly, selective blockade of GABAC receptors with I4AA also failed to increase spontaneous firing. That is, I4AA did not reverse the suppression of hyperexcitability that followed Ad-ρ1-GFP transduction. After coapplication of both bicuculline and I4AA or application of the nonselective GABA receptor antagonist picrotoxin (50 μm) alone, blocking both GABAA and GABACreceptors, the seizure-like electrical activity returned (Fig.5 B). Thus it appears that Ad-ρ1-GFP-induced expression of GABAC receptors somehow enhances endogenous GABAA receptor-mediated inhibition of neuronal excitability.
To examine further this apparent enhancement of GABAA receptor activity, we studied receptor-specific current density after Ad-ρ1-GFP transduction. Figure6 A shows GABA-evoked current density (normalized by cell capacitance) from nontransduced neurons (gray bars), GFP-neurons (black bars), and ρ1-GFP-neurons (hatched bars) at 2 d intervals after viral transduction. The total current density evoked by 100 μm GABA did not change for control nontransduced neurons and GFP-neurons. In ρ1-GFP-neurons, where the GABAC receptor number should be increasing with time since transduction, total current density increased, as expected. Pharmacological isolation of the GABAC component of the total current density (by bicuculline blockade of GABAA receptors) confirmed this time-dependent increase in the GABAC contribution (Fig. 6 B,triangles). Interestingly, pharmacological isolation of the GABAA component (by I4AA blockade of GABAC receptors) also showed a time-dependent increase in the current density (Fig. 6 B,squares). In fact, the contribution of GABAA receptors to the increase in total current density was approximately equal to that of GABACreceptors. Again the presence of GABAC receptors seems to enhance GABAA receptor function.
It is possible that the apparent enhancement of GABAA receptor function is the result of the reduction in neuronal activity caused by transduction. It is well known that neuronal activity (Penschuck et al., 1999) or depolarization as a surrogate of neuronal activity (Gault and Siegel, 1997, 1998) regulates many neurotransmitter receptors, including the GABAA receptors. If the change in neuronal activity alone is responsible, then similar changes in activity in nontransduced neurons (by inhibition of electrical activity by TTX or suppression of synaptic transmission by elevated Mg2+) should mimic the action of ρ1-GFP-virus. If the changes are nonspecific effects of viral transduction, the control GFP-virus should mimic the effects of the ρ1-GFP-virus.
To preclude possible interactions between the chronic kynurenate–Mg2+ growth condition and the GABAA current density enhancement, the effects of TTX, Mg2+, and viral transduction were examined in neurons grown in otherwise normal culture medium. Neither culturing the neurons for 6 d in TTX or Mg2+-supplemented medium nor GFP-virus transduction increased the GABAA current density over that of control nontreated neurons (Fig. 6 C).
Ad-ρ1-GFP transduction suppresses hyperexcitability-induced delayed neuronal death
Prolonged unopposed excitation kills neurons (Abele et al., 1990;Murray et al., 1998). Regardless of the precise mechanism, because Ad-ρ1-GFP transduction of hyperexcitable neurons decreases spontaneous activity, it may also decrease hyper-excitability-associated excitotoxic delayed neuronal death. To determine baseline excitotoxic neuronal death rates, we subjected hyperexcitable neurons to paroxysmal electrical activity by replacing the kynurenate–Mg2+ blocking medium with a medium without blockers for 30 min and then restoring the blocking medium. After 24 hr, 45–52% of the neurons were dead, as determined by staining with ethidium homodimer. In control cultures subjected only to sham wash, with no period of paroxysmal activity, only 1–5% of the neurons were dead (Fig. 7 A). Increasing the duration of hyperexcitability increased delayed neuronal death (Fig. 7 B), consistent with a previous report (Murray et al., 1998). In three separate experiments, 2 d after transduction, hyperexcitable ρ1-GFP-neurons were subjected to the same 30 min period of kynurenate–Mg2+ removal. After 24 hr, 54 ± 2% of the neurons were dead, not significantly different from baseline. However, of the ∼20% of neurons exhibiting GFP fluorescence and that are thus certain to have been successfully transduced and to express the GABAC receptor, only 1–5% were dead (Fig. 7 C,D). Successful transduction with Ad-ρ1-GFP provides nearly complete protection from hyperexcitability-induced excitotoxic delayed neuronal death in this model.
A recombinant adenovirus, Ad-ρ1-GFP, designed to express both the GABA ρ1 subunit and GFP was created. The coexpression of GFP allowed for an easy visual identification of successfully transduced neurons. Immunocytochemical study of individual GFP-positive neurons and a Western blot of virally transduced HEK293 cells demonstrated the de novo expression of ρ1 protein. The punctate pattern of the membrane immunoreactivity could not be discerned, but because hippocampal neurons abundantly express the microtubule-associated protein-1B (MAP-1B) (data not shown), aggregation into receptor patches is expected (Hanley et al., 1999). The distinctive pharmacology of GABAC receptors (insensitive to bicuculline, sensitive to picrotoxin, selectively activated by CACA, and selectively inhibited by I4AA) allowed for a definitive confirmation of the presence of this receptor in Ad-ρ1-GFP-transduced hippocampal neurons. As expected, ρ1-GFP-neurons demonstrated CACA-activated chloride-permeable channels with the threefold slower activation and fivefold slower deactivation kinetics of functional GABAC receptors. Of note is our preliminary observation indicating a lack of bicuculline-resistant mIPSC in ρ1-GFP-neurons (Cheng and Yang, 2000). Although we are unable to distinguish between the lack of functional versus physical presence of synaptic GABAC receptors, data thus far are consistent with the expression of abundant somatic but not synaptic GABAC receptors.
Although previous studies have suggested the presence of a bicuculline-resistant GABAC-like receptor in the hippocampus during the first 2 weeks of postnatal development (Strata and Cherubini, 1994) and ρ2 subunit mRNA has been detected in postnatal day 5 and adult rat hippocampus by Northern blot and reverse transcription-PCR analyses (Boue-Grabot et al., 1998), we observed no evidence of GABA-evoked current mediated by endogenous GABAC receptors in control neurons. In contrast to previous reports of no desensitization of ρ1 subunits expressed in Xenopusoocytes (Cutting et al., 1991), our ρ1-GFP-neurons exhibited slow but significant desensitization during GABA application. Significant desensitization of the GABAC receptors has also been seen when the human ρ1 subunit is expressed in eukaryotic hosts (Filippova et al., 1999). These observations suggest a difference in processing of the receptor protein between the eukaryotic and theXenopus expression systems. However, within the limitation of the rather slow perfusion used by us, the GABAC receptors expressed in hippocampal neurons retain the fundamental pharmacological and kinetic properties of GABAC receptors that have been well described previously.
In neurons rendered hyperexcitable by chronic kynurenate–Mg2+ treatment, Ad-ρ1-GFP transduction dramatically altered the action potential firing pattern and reduced the frequency of spontaneous action potentials. The bursting pattern of paroxysmal action potential firing disappeared, and the overall action potential rate decreased to control levels. Application of CACA caused a hyperpolarizing shift in membrane potential, confirming the expression of GABAC receptors in these cells. Because forced expression of GABAC receptors abolished the hyperexcitability, we expected selective blockade of the GABAC receptors by I4AA to return the neurons to the hyperexcitable state. Surprisingly, I4AA alone had little effect on the spontaneous action potential rate. Bicuculline blockade of endogenous GABAA receptors in these same neurons also did not reestablish the hyperexcitable state. A technical reason for failure of the selective antagonists to reestablish hyperexcitability is unlikely. Although a high concentration of GABA could compete off the competitive antagonists, the proper summation of pharmacologically separated component currents equals the total current density (Fig. 6), indicating that under our experimental conditions, both bicuculline and I4AA completely blocked their respective receptors. Blockade of both GABAC and GABAA receptors by coapplication of both selective antagonists or application of picrotoxin alone was able to render the neurons once again hyperexcitable, but even these neurons had a lower frequency of spontaneous action potentials than did nontransduced or control Ad-GFP-transduced cells. These observations suggest that suppression of hyperexcitability by Ad-ρ1-GFP is not caused solely by the expression of additional inhibitory receptors but depends, at least in part, on other, unknown, alterations in neuronal excitability.
Neuronal hyperexcitability in the chronic kynurenate–Mg2+ model is caused by enhancement of both NMDA and AMPA/kainate glutamate receptors (Van den pol et al., 1996). This mechanism is consistent with the recently demonstrated activity-dependent decrease in membrane targeting of the AMPA receptor GluR1subunit (Lissen et al., 1998). Similarly, blocking neuronal electrical activity by TTX results in elongation of dendritic spines (Papa and Segal, 1996) and a decrease in GABAA α1 and α2 subunit density (Penschuck et al., 1999). Chronic activation of the GABAC receptor, which clamps the membrane potential to the chloride equilibrium potential, would be expected to result in the loss of electrical activity. By following this logic, the resulting electrical silence would upregulate AMPA receptors and downregulate GABAA receptors, increasing network excitability. This is the opposite of what we observed. Moreover, because the effect of chronic kynurenate–Mg2+ blockade is virtually to abolish spontaneous electrical activity, activation of GABAC receptors could have little additional effect on electrical activity in this culture system. An interesting possibility is that under our culture condition, stimulation of the GABAC receptors resulted in membrane depolarization overcoming the glutamate blockade. If this were the case, enhanced electrical activity could have resulted in the upregulation of GABAA receptors.
In Ad-ρ1-GFP neurons cultured in standard medium, pharmacological separation of the total current evoked by GABA into components mediated by the GABAA and the GABACreceptors demonstrated an unexpected induction of GABAA receptor-mediated current density. Although it is possible that this effect is simply the result of reduced neuronal activity caused by expression of GABACreceptors, neither inhibition of electrical activity (by TTX) nor inhibition of synaptic transmission (by elevated Mg2+) was sufficient to enhance current density (Fig. 6 C). Additionally, electrical silence has been reported to reduce GABAA receptor α1 and α2 subunit expression when examined by in situ hybridization (Penschuck et al., 1999), and excessive depolarization (by elevated K+ or glutamate stimulation of cerebellar granule neurons) increases GABAA δ subunit transcription (Gault and Siegel, 1997, 1998). Both of these effects are in the wrong direction and fail to account for our observations. Induction of GABAA receptors by Ad-ρ1-GFP must be mediated by signals other than membrane potential.
A recent study indicates that Ig-neuregulin, a member of the epidermal growth factor superfamily that activates receptor tyrosine kinase, selectively increases GABAA receptor expression via induction of the β2 subunit (Rieff et al., 1999). Insulin, another growth factor well known to activate receptor tyrosine kinase, rapidly recruits GABAA receptors from the cytoplasmic to postsynaptic domains, increasing the amplitude of GABAA receptor-mediated mIPSC (Wan et al., 1997). It is possible that GABAC receptor expression-mediated induction of GABAA receptors involves this or similar intracellular signal transduction pathways. GABAC receptor-dependent post-translational modification of GABAA receptors is another possible mechanism for this effect.
Regardless of the precise mechanism, Ad-ρ1-GFP reversed neuronal hyperexcitability and resulted in a reduction in delayed excitotoxic neuronal death. Our results are consistent with a previous observation that GABAergic neocortical neurons are resistant to NMDA receptor-mediated injury (Tecoma and Choi, 1989). Exogenous administration of diazepam, an allosteric GABA potentiator, also has been shown to reduce postischemic and traumatic neuronal death in vivo (Schwartz et al., 1995; O'Dell et al., 2000). Because generalized augmentation of GABAergic inhibition, whether pharmacologically or by gene targeting, may impair normal synaptic plasticity (Levkovitz et al., 1999), there may be a strong advantage for a regional rather than a general enhancement of inhibition. Delivery of adenovirus through direct stereotactic injection might allow expression of virally transduced GABACreceptors in a sharply restricted subset of neurons, augmenting endogenous GABA receptor-mediated neuroprotection without the side effects of conventional drug therapy.
We focused on a forced expression of inhibitory GABA receptors to suppress hyperexcitability, but other mechanisms could be used. For example, increased GABAergic inhibition by enhancement of presynaptic release in transgenic mice overexpressing superoxide results in resistance to systemic kainic acid-induced seizures (Levkovitz et al., 1999). Overexpression of a chloride transporter, forcing the chloride reversal potential to become more negative (Staley et al., 1996), is another approach. Excitability can be reduced by viral expression of inwardly rectifying potassium channels with a consequent increase in the threshold for action potential generation (Ehrengruber et al., 1997; Johns et al., 1999). Because activation of the classical MAP kinases (Murray et al., 1998) may mediate delayed neuronal death, expression of a dominant-negative MAP kinase mutant might render neurons resistant to excitotoxic death. Finally, a deeper understanding of the genetic basis of epilepsy could lead to a mechanism-based gene therapy. For example, defects in channel proteins such as GIRK2, GluRB, or Kv1.1 have been implicated as possible causes of epilepsy (Noebles, 1996). Targeted corrections of such defects might be possible in the future.
This work was supported by National Institutes of Health Grants GM 57578 and GM 52325. We thank Sundeep Malik and Nancy Ward for assistance with molecular constructs.
Correspondence should be addressed to Dr. Jay Yang, Department of Anesthesiology, University of Rochester Medical Centere, 601 Elmwood Avenue, Rochester, NY 14642. E-mail:.