Synergistic Roles for G-protein γ3 and γ7 Subtypes in Seizure Susceptibility as Revealed in Double Knock-out Mice*

Background: Specificity of G-protein function may be determined by a specific αβγ composition. Results: Combinatorial disruption of γ3 and γ7 produces a severe seizure phenotype not observed with either gene alone. Conclusion: This reflects distinct roles for γ3 and γ7 in Gi/o- and Golf-signaling pathways that modulate seizure susceptibility. Significance: The γ subunits direct the assembly of distinct G-protein αβγ heterotrimers that specify diverse receptor actions. The functions of different G-protein αβγ subunit combinations are traditionally ascribed to their various α components. However, the discovery of similarly diverse γ subtypes raises the possibility that they may also contribute to specificity. To test this possibility, we used a gene targeting approach to determine whether the closely related γ3 and γ7 subunits can perform functionally interchangeable roles in mice. In contrast to single knock-out mice that show normal survival, Gng3−/−Gng7−/− double knock-out mice display a progressive seizure disorder that dramatically reduces their median life span to only 75 days. Biochemical analyses reveal that the severe phenotype is not due to redundant roles for the two γ subunits in the same signaling pathway but rather is attributed to their unique actions in different signaling pathways. The results suggest that the γ3 subunit is a component of a Gi/o protein that is required for γ-aminobutyric acid, type B, receptor-regulated neuronal excitability, whereas the γ7 subunit is a component of a Golf protein that is responsible for A2A adenosine or D1 dopamine receptor-induced neuro-protective response. The development of this mouse model offers a novel experimental framework for exploring how signaling pathways integrate to produce normal brain function and how their combined dysfunction leads to spontaneous seizures and premature death. The results underscore the critical role of the γ subunit in this process.

Proper functioning of the central nervous system requires the coordination of several hundred receptors whose actions may be mediated by a similarly large number of distinct G-protein ␣␤␥ heterotrimers. Identifying the specific G-protein ␣␤␥ subunit combinations functioning in particular signaling pathways has been a challenge. Although specificity of G-protein function was originally ascribed to the various ␣ subtypes, there is a growing recognition that diverse ␤␥ dimers may impart an additional level of selectivity (1)(2)(3). Compared with the five ␤ subtypes, the 12 ␥ subtypes are more structurally diverse, suggesting the in vivo specificity observed among different ␤␥ dimers is most likely due to the ␥ component (2,4,5). Providing a rigorous test of this hypothesis, we produced Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ mice, which lack the closely related ␥ 3 and ␥ 7 subunits. Subsequent characterization of these animals revealed distinct neurological phenotypes, reflecting their roles in different receptor signaling pathways (6 -8). Offering a mechanistic basis for their diverse roles, biochemical analyses of these animals identified a critical role for the ␥ subunit in directing the assembly of distinct G i/o and G olf heterotrimers (6 -8). Taken together, these results support the notion that even closely related ␥ subtypes have distinct signaling roles and biological functions in the context of the animal.
In this study, we sought to extend these findings by exploring a novel interaction between signaling pathways requiring ␥ 3 and ␥ 7 subunits in brain. Suggesting this possibility, Gng7 Ϫ/Ϫ mice exhibit a nearly 40% up-regulation of the ␥ 3 protein in the striatum (8). The increased ␥ 3 abundance could reflect a compensatory mechanism aimed at replacing the lost ␥ 7 protein that is required for the G olf -dependent signaling pathway. Alternatively, this change could reflect an adaptive mechanism arising from interaction between G i/o -and G olf -dependent signaling pathways that converge on a common neurological process. To distinguish between these possibilities, we produced Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice and characterized them at the behavioral, neurological, electrophysiological, cellular, and biochemical levels. Collectively, these results showed that double knock-out mice exhibit a progressive seizure disorder and premature death that is not observed for either single knock-out strain alone on the same genetic background. We speculate the severity of the phenotype results from simultaneous disruption of G i/o -and G olf -dependent signaling pathways in different neuronal populations that normally operate together to limit seizure initiation, seizure propagation, or seizure-induced damage.

EXPERIMENTAL PROCEDURES
Mice and Husbandry-Animal use was approved by the Geisinger Institutional Animal Care and Use Committee. Every effort was made throughout the study to minimize usage, pain, and discomfort of the animals. The generation of Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ single knock-out mice was described previously (6,7). On a mixed genetic background (129, FVB, and B6), the Gng3 Ϫ/Ϫ mice experienced more handling-induced seizures (24%) compared with their littermate controls (8%) (7). However, after backcrossing onto the C57BL/6J (B6) background (The Jackson Laboratory, Bar Harbor, ME) for five or more generations, the Gng3 Ϫ/Ϫ mice showed no signs of seizure activity or premature death compared with their littermate controls (7). On either a mixed (129, BALB/c, and B6), or a B6 background, the Gng7 Ϫ/Ϫ mice did not display any evidence of seizure activity (6). Accordingly, to minimize any effect of the genetic background, the Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ mice were maintained on a B6 background (ՆN7 backcross) in this study. For experimental purposes, Gng3 ϩ/Ϫ or Gng7 ϩ/Ϫ mice were intercrossed to produce single knock-out and control groups, whereas Gng3 ϩ/Ϫ Gng7 Ϫ/Ϫ were intercrossed to generate double knock-out and control groups. Immediately after weaning, mice were genotyped and assigned to experimental groups that were similarly matched for age and sex.
Survival and Video Surveillance-Mice of different genotypes were incorporated into the survival study as they became available. The mice were housed in polycarbonate cages in ventilated racks (Thoren Caging Systems, Inc., Hazelton, PA) on a 14-h light and 10-h dark cycle, with the temperature maintained between 21 and 23°C. The mice were allowed free access to water and standard chow (Mouse Diet 9F, Purina Mills, St. Louis), which contains 38.5% starch, 9% fat, 20% protein, and 3% fiber. To investigate the possible impact of a ketogenic diet on survival, mice were provided a TestDiet 8053 (Purina Mills), which contains 0.0% carbohydrate, 70% fat, 13.6% protein, and 8.3% fiber. For the survival study, the date and proximate cause of death of the animals were determined by close monitoring by facility staff and by video surveillance of home cages outfitted with infra-red CCTV cameras to continuously monitor the mice therein (ProVideo CVC-320WP, Amityville, NY). Signals were processed into a quad format with an EverPlex 4BQ (Ever-Focus Electronics Corp., San Marino, CA) and recorded with a time lapse video cassette recorder (HS-1280U, Mitsubishi Digital Electronics America, Inc., Irvine, CA). If a mouse was found dead in its cage, the previous 24 h of video were reviewed to determine the proximate cause of death. In 19 of 21 deaths recorded, seizures were found to immediately precede death. In the few cases in which animals were euthanized for humane reasons (e.g. severe dermatitis), mice were not included in the calculations. Survival curves were plotted, and median life spans and 95% confidence intervals (CIs) 3 were calculated for each genotype. To determine the effects of genotype or sex differences on life span, a univariate analysis was performed using JMP 6.0 (SAS Institute, Carey, NC).
Electroencephalography-As shown by video surveillance, Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice experienced tonic-clonic seizures immediately preceding death. To look for neurological abnormalities associated with seizure activity, we performed electroencephalography on double knock-out, single knock-out, and wild type littermates. Under anesthesia, mice (12-16 weeks old) were implanted with epidural screw electrodes (Plastics One, Roanoke, VA) in five locations, frontopolar, right and left frontal, and right and left posterior, on the mouse skull that were connected to an electrode pedestal. The frontal electrodes were 1-1.5 mm anterior to the coronal suture; the mean distance between the frontal and parietal electrodes was 5.1 mm, and the distance between the frontal electrodes was 3.6 mm. After a 2-week period for recovery, electroencephalography (EEG) recordings were made on mice at varying intervals over the next 16 weeks, using a Nicolet Bravo electro-encephalograph (Nicolet, Madison, WI). Over the course of the study, 9 of 15 Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice died compared with 0 of 8 wild type mice, 0 of 9 Gng3 Ϫ/Ϫ single knock-out mice, and 0 of 14 Gng7 Ϫ/Ϫ single knock-out mice.
All EEG recordings were retained in digital format. Prior to analyses, EEG recordings were filtered with a 1 Hz high pass filter, a 35 Hz low pass filter, and a 60 Hz notch filter. Multiple analyses were performed. First, the average spike rate was computed for each mouse from the right frontal-right posterior and left frontal-left posterior derivations, although the other derivations were used to help identify artifacts. Spike counts were measured in multiple 30-min blocks, and the results were averaged to obtain a single mean spike rate from each mouse. The Kruskal-Wallis test was used to determine whether there was a difference between the mean spike rates among mice from different genotypes. Second, the power spectrum at each frequency was evaluated and averaged from at least three 30-min EEG segments for each mouse. The mean power for each frequency (Ͻ1, 1-2, 2-6, 6 -10, and Ͼ10 Hz) was then determined. A repeated measures ANOVA test was used to determine whether there was any difference in the power spectrum among mice of different genotypes. Third, the interhemispheric coherence, which is a measure of the degree of synchrony between electrical activity in the right and left hemispheres, was also computed. For each 1-s epoch, the product of the Fourier transform of the EEG activity at each frequency from the right and left hemisphere (right frontal-right posterior and left frontal-left posterior) was computed and averaged over all epochs in the 30-min file. The coherence measure for each frequency was then calculated as the square of the average product divided by the product of the mean power in each derivation. The coherence value ranges between 0, when there is no synchrony, to 1, when there is complete synchrony. Finally, GABA B agonist-induced EEG changes were assessed among mice of different genotypes by quantifying the power in each frequency band described above for 5-min clips taken at base line, 20 and 40 min, and 7 and 10 h post-injection (10 mg/kg intraperitoneal injection of baclofen). A repeated measures ANOVA test was performed, using genotype as the between subjects variable and the mean power and post-injection time as within subjects variables.
Brain Dissection-All dissections were completed within 5-10 min of death (9, 10) by making a vertical slice 0.5-1 mm caudal to the olfactory bulbs and a second vertical slice just rostral to the optic chiasm. The intervening section was placed with the caudal face up, and the nuclei accumbens was dissected with a 1-mm micropunch (Fine Science Tools, Foster City, CA) centered over each anterior commissure. The prefrontal cortex was dissected superior to the corpus callosum near the midline, and the caudate nuclei were dissected with a 2-mm micropunch inferior to the corpus callosum, bilaterally. After removal of the hypothalamus with tweezers, the caudal portion of the brain was placed dorsal side up. The cerebellum and pons were removed by a vertical slice between the superior and inferior colliculi. A slice was then made at a 45°angle from the dorsal caudal end down toward the ventral rostral end. The enterorhinocortical regions of the ventral portion were dissected. Finally, the ventral midbrain was isolated from the remaining ventral portion, by trimming the dorsal midbrain with a transverse slice. The various brain regions were placed in individual tubes, frozen immediately with liquid nitrogen, and stored at 80°C until used for RNA or protein analyses described below.
RNA Analyses-The distribution of Gng3 and Gng7 transcripts was determined by RT-PCR analysis. The Mouse Multiple Tissue cDNA Panel (Clontech, Palo Alto, CA) was used as a PCR template to amplify Gng3, Gng7, and Gapdh, using the indicated primers shown in supplemental Table 1. PCRs were performed using the Advantaq Plus PCR kit (Clontech). The cycling conditions were 38 cycles of 94°C for 30 s and 68°C for 2 min, followed by a final extension of 68°C for 5 min. Reactions were run in a PTC-100 Programmable Thermal Controller (MJ Research, Watertown, MA). Aliquots were removed from each sample at 22,24,26,30,34, and 38 cycles and visualized on 2% agarose gels containing ethidium bromide.
In parallel, the relative abundance of Gng3 and Gng7 transcripts was assessed by qPCR analysis. Total RNA was isolated from dissected brain regions from wild type and knock-out mice using TRIzol (Invitrogen). From 1 g of RNA, the cDNA template was prepared using Moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI). The qPCR standards were constructed by subcloning Gng3-and Gng7specific PCR products into PCR II Topo vector (Invitrogen), using indicated primers shown in supplemental Table 1. Plasmid standards were quantitated spectrophotometrically and were serially diluted to contain 10 1 to 10 7 DNA molecules. Melt curve and agarose gel electrophoresis were used to confirm single product amplification. Standard curves were evaluated for linearity (r ϭ 0.99 -1.0) and amplification efficiency (Ͼ90%). For each qPCR, duplicate samples of 50 ng of total RNA equivalents of brain region cDNA or plasmid standards were amplified for 40 cycles using gene-specific primers designed to span intron junctions (supplemental Table 1). Reactions were performed using iQ SYBR Green supermix (Bio-Rad) and run on the iCycler device with version 3.1 software (Bio-Rad).
Cellular Localization Strategies-The cellular distribution of ␥ 3 and ␥ 7 reporter proteins was assessed. For this purpose, transgenic (Tg) mice, in which expression of enhanced green fluorescent protein (GFP) is driven by the Gng3 promoter (Tg(Gng3-GFP)HK208Gsat mice) or the Gng7 promoter (Tg(Gng7-GFP)FG220Gsat mice), were obtained from Mutant Mouse Regional Resource Center, University of California, Davis (stock numbers 015490-UCD and 011393-UCD). Because GFP expression in the Tg(Gng7-GFP)FG220Gsat mice was not sufficient to achieve single cell resolution, we produced a line of KI(Gng7-␥ 3 -IRES-GFP) mice, in which the endogenous Gng7 locus was used to independently drive ␥ 3 and GFP expression. For this purpose, a targeting vector was designed that contained a modified Gng7 locus replacing the protein coding exons of ␥ 7 with the ␥ 3 cDNA, an internal ribosome re-entry site (IRES), and a GFP cDNA.
Adenylyl Cyclase Assay-Frozen striatal punches were homogenized on ice with a motorized pestle (Kimble Chase, Vineland, NJ) in HME with proteinase inhibitors (20 mM HEPES, pH 8.0, 2 mM MgCl 2 , 1 mM EDTA, 1 mM benzamidine, 0.1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 20 M leupeptin, 1.4 M pepstatin, 27 M tosyl-L-lysine chloromethyl ketone, 28 M tosyl-L-phenylalanine chloromethyl ketone) and then repeatedly passed through a 25-gauge needle. Nuclei and unbroken cells were pelleted by low speed centrifugation (350 ϫ g) for 5 min, and membranes were collected by ultracentrifugation at 250,000 ϫ g for 1 h, resuspended in HME with proteinase inhibitors, and then stored at Ϫ80°C. Protein concentrations were determined with Coomassie Plus (Pierce). Adenylyl cyclase activity was determined by incubating membrane protein ( In Vivo Responses to Baclofen-The muscle-relaxing effect of baclofen was assessed on an ENV-576 M Rota-Rod Treadmill (Med Associates, Inc., St. Albans, VT). Mice of different genotypes were placed on the Rota-Rod, which was then started at a constant 16 rpm. After acclimating mice to the instrument for 2 consecutive days, mice were given an intraperitoneal injection of saline (5 ml/kg) on the 3rd day followed by an intraperitoneal injection of a maximally effective dose of baclofen (10 mg/kg) on the 4th day. Thirty minutes after injection, mice were placed on the Rota-Rod. The results were recorded as the time spent on the instrument. Any mice that had not fallen from the Rota-Rod were removed after 5 min. The temperature lowering effect of baclofen was measured with a rectal thermometer by recording the temperature of mice before and 30 min after injection of baclofen.
Patch Clamp Electrophysiology-Gng3 ϩ/Ϫ mice were intercrossed to prepare hippocampal neurons from individual embryonic day 18 mice (13). After plating onto 12-mm polylysine-treated coverslips, patch clamp analyses were performed on neurons in culture between 11 and 14 days without prior knowledge of the genotype. Currents from G-protein-sensitive inwardly rectifying potassium (GIRK, Kir3) channels were measured as described previously (14). Briefly, neurons were constantly voltage-clamped at Ϫ60 mV, recorded using a Multiclamp700B amplifier, digitized with a Digidata 1322B, sampled at 4 kHz, low pass-filtered at 1 kHz, and collected using pClamp9.2 (all from Molecular Devices). Series resistance and cell capacitance were automatically compensated and monitored at the beginning and end of each experiment. Potassium currents were monitored by switching from a low potassium solution containing 140 mM NaCl, 4 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 20 mM HEPES, and 10 mM glucose, pH 7.4, to a high potassium solution containing 84 mM NaCl, 60 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 20 mM HEPES, and 10 mM glucose. The adenosine receptor agonist, 5Ј-N-ethylcarboxamidoadenosine (NECA; 2 M), the somatostatin receptor agonist (1 M), or baclofen (100 M) were dissolved in the bath solution and applied using an automated perfusion system. Current amplitudes were measured at Ϫ60 mV. TertiapinQ (120 nM), a specific peptide inhibitor of Kir3 channels, and barium (3 mM) were used to measure the residual inwardly rectifying current. The tertiapinQ-sensitive current is defined as basal current. Any current activated by a given receptor ligand above the basal current is defined as agonist-induced current.
Statistics-For behavioral and biochemical studies, sample statistics and Student's t tests were computed using Excel (Microsoft). Data are presented as means Ϯ S.E. Univariate survival curves were calculated and compared with a log-rank 2 test using JMP (SAS Institute, Cary, NC). EEG results were compared with the Kruskal-Wallis test, a nonparametric ANOVA, using Statistica (StatSoft, Tulsa, OK). EEG results were also compared using one-and two-way ANOVAs with Bonferonni post-tests using Prism 5.0 (GraphPad Software, San Diego). Any significant effects have been reported in the text or the accompanying figure legend.

RESULTS
Severe Phenotype of Double Knock-out Mice-The Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice were used to investigate the impact of combined loss of the ␥ 3 and ␥ 7 subunits. The effectiveness of this strategy was confirmed by immunoblot analysis of striatal membranes, revealing complete loss of the ␥ 3 and ␥ 7 proteins in brains obtained from double knock-out mice (Fig. 1A). Although born at the expected Mendelian frequency, Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice showed high mortality that was not observed for either single knock-out line (Fig. 1B). In all, 51 of 52 double knock-out mice died before 1 year of age, with a median survival of 75 days (95% CI, 68 -81 days). There was no significant effect of sex on life span. Because 32% of the double knock-out mice displayed handling-induced seizures at a median of 18 days before death (range, 6 -130 days), we suspected that the double knock-out mice were dying as a result of recurrent seizures.
Seizures Are the Probable Cause of Death-To examine the events surrounding their deaths, four mouse cages were subjected to continuous surveillance with infrared video cameras. From these recordings, we ascertained that 19 of 21 Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice experienced seizures for ϳ1 min immediately prior to their deaths. Seizure-induced deaths were observed at various times throughout the day and were not associated with Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype any particular activity (i.e. sleeping, walking, eating, or grooming). Typically, seizures were characterized by a progression from wild running to tonic-clonic convulsion to tonic hindlimb extension that ended in death. Further supporting seizures as the proximate cause of death, administration of a ketogenic diet, which has been used as an effective treatment for refractory seizures (15)(16)(17), significantly prolonged the life span of double knock-out mice (Fig. 1C). Double knock-out mice on a ketogenic diet displayed significantly longer life spans than their wild type and single knock-out littermates on a regular diet (compare Fig. 1, B and C). However, female double knockout mice on a ketogenic diet had a median survival of 334 days (95% CI, 239 -430 days), compared with 154 days for male double knock-out mice on a ketogenic diet (95% CI, 113-185 days). Taken together, these results are most consistent with the deaths of double knock-out mice resulting from seizure activity and that administration of a ketogenic diet to suppress their seizure activity improved their viability.
Abnormal Electrical Activity in Knock-out Mice-To identify neurophysiological abnormalities in double knock-out mice, we compared video-EEG recordings from four groups of mice as follows: Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ (n ϭ 14); Gng3 Ϫ/Ϫ (n ϭ 9); Gng7 Ϫ/Ϫ (n ϭ 14); and wild type mice (n ϭ 7). The data from three 30-min clips were analyzed for each mouse. Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice exhibited several EEG abnormalities. First, representative EEG tracings ( Fig. 2A) and quantitative analysis (Fig. 2B) showed that the frequency of interictal epileptiform discharges was strongly influenced by genotype, as determined by the Kruskal-Wallis test ( 2 ϭ 12.9, df ϭ 3, p Ͻ 0.005). In particular, the interictal spike frequency was highest in the double knock-out mice, intermediate in the single knock-out mice, and lowest in the wild type mice. Because interictal spike frequency is an indicator of increased seizure risk (18), the finding that double knock-out mice exhibited more interictal discharges was consistent with their seizure phenotype. Second, double knock-out mice displayed a significantly lower interhemispheric coherence (Fig. 2C). Because a lower inter-hemispheric coherence is commonly observed in neurological disorders (19 -21), the observation that double knock-out mice displayed reduced inter-hemispheric connectivity was also consistent with a neurological phenotype. Finally, a Spearman rank correlation analysis showed a strong inverse correlation between the inter-hemispheric coherence and the spike frequency (r ϭ Ϫ0.38, p Ͻ 0.02). Indeed, the combination of these two EEG changes by themselves allowed the identification of double knock-out mice with 75% accuracy without prior knowledge of their genotypes.
Functional Redundancy of Closely Related ␥ Subtypes within the Same Signaling Pathway as a Possible Basis for the Double Knock-out Phenotype-We first considered the possibility that the closely related ␥ 3 and ␥ 7 subtypes are substituting for one another in the same signaling pathway. Both the Gng3 and Gng7 transcripts were predominantly expressed in brain (Fig. 3A). Although the Gng3 transcript was widely expressed throughout brain, the Gng7 transcript was almost exclusively restricted to the striatum, including the caudate-putamen and nucleus accumbens, along with the enterorhinocortical region, including the hippocampus (Fig. 3B). Because their expression intersected primarily in the striatum, we focused on exploring functional interactions between the ␥ 3 and ␥ 7 subtypes in this region as the most likely basis for the double knock-out phenotype. Previously, the ␥ 7 subunit was shown to be required for both adenosine A 2A receptor (A 2A R) and dopamine D 1 receptor FIGURE 1. Gene targeting strategy. A, validation of gene targeting strategy by immunoblotting cholate-extracted membrane protein from dorsal striatum of single knock-out (Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ ), double knock-out (Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ ), and wild type (wt) mice, verifying loss of appropriate ␥ 3 and/or ␥ 7 subunit(s); Ras is shown as a loading control. B, on the regular diet, survival of Gng3 Ϫ/Ϫ mice at an N10 backcross to B6 (Ͼ1 year) is normal. In contrast, survival of Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice (75 days, 95% CI, 68 -81 days) is severely reduced compared with their Gng3 ϩ/ϩ Gng7 Ϫ/Ϫ littermates (Ͼ1 year) (log rank 2 ϭ 120.7, p Ͻ 0.0001). There was no significant difference between male and female double knock-out mice in terms of their reduced survival (log rank 2 ϭ 0.3, df ϭ 1, p ϭ 0.6). C, on a ketogenic diet, survival of Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice (239 days, 95% CI, 152-292 days) was still reduced compared with their Gng3 ϩ/ϩ Gng7 Ϫ/Ϫ littermates (Ͼ1 year) (log rank 2 ϭ 57.5, p Ͻ 0.0001). However, survival of Gng3 Ϫ/Ϫ ;Gng7 Ϫ/Ϫ mice on ketogenic diet was improved relative to those on regular diet (log rank 2 ϭ 37.8, p Ͻ 0.0001). In this case, there was a significant difference between male and female double knock-out mice in terms of overall survival (log rank 2 ϭ 9.5, df ϭ 1, p ϭ 0.002).
(D 1 R) signaling in striatum (6,8). Here, we demonstrated that the ␥ 3 subunit was not required for either of these signaling pathways. Gene targeted loss of the ␥ 3 protein did not affect stimulation of adenylyl cyclase by the A 2A R agonist CGS21680 (Fig. 4A) nor by dopamine or the D 1 R selective agonist (Ϯ)-6chloro-7,8-dihydroxy-1-phenyl-2,3,4,5-tetrahydro-1H-3-ben zazepin hydrobromide (Fig. 4C). Likewise, combined deletion of both ␥ 3 and ␥ 7 proteins did not block activation of adenylyl cyclase by dopamine to a greater extent than loss of the ␥ 7 protein alone (Fig. 4C). Finally, deletion of the ␥ 3 subunit did not affect adenylyl cyclase stimulation by forskolin (Fig. 4B), and a combined loss of both ␥ 3 and ␥ 7 proteins subunits did not produce a greater effect than loss of the ␥ 7 protein alone (Fig.  4D).
Because the forskolin response is potentiated by the presence of the stimulatory G-proteins (22), we next investigated how gene targeted loss of the ␥ 3 or the ␥ 7 subunit affected levels of the G olf or G s protein. We showed previously that the ␥ 7 subunit was required for G olf assembly in striatum (6,8). Confirming and extending this result, gene-targeted loss of the ␥ 7 subunit markedly suppressed ␣ olf and ␤ 2 levels that were modestly reduced further by loss of both the ␥ 3 and ␥ 7 proteins (Fig. 4E). Attesting to the specific nature of these changes, no significant reductions in ␣ s , ␣ i3 , or ␤ 1 levels were observed in either single or double knock-out mice (Fig. 4E). Likewise, we demonstrated previously that the ␥ 3 subunit was not required for G olf assembly in striatum but was linked to suppression of ␣ i3 and ␤ 2 levels  ). B, mean spike rate was increased in the Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice (n ϭ 14) relative to wild type mice (n ϭ 7), and Gng3 Ϫ/Ϫ (n ϭ 9) and Gng7 Ϫ/Ϫ mice (n ϭ 14) had intermediate spike rates (one-way ANOVA, F 3,3 ϭ 4.6, p ϭ 0.008, **, p Ͻ 0.05 for comparison with wild type in Tukey test). C, interhemispheric coherence was reduced in Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice compared with all other genotypes (one-way ANOVA df ϭ 3, F ϭ 4.7, p ϭ 0.007, **, p Ͻ 0.05 for comparison with all other genotypes in Tukey tests). Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype in cortex (7). Confirming and extending these findings, genetargeted deletion of the ␥ 3 subunit resulted in significant suppression of ␤ 2 protein and a trend toward decreased ␣ i3 content that were not reduced further by loss of both the ␥ 3 and ␥ 7 proteins (Fig. 4F). Taken together, these results indicated the more severe phenotype of double knock-out mice was associated with circumscribed changes in the ␣ olf , ␣ i3 , and ␤ 2 levels first reported for the single knock-out mice (6 -8) and was not the result of global deficits in multiple ␣ and ␤ proteins.

Functional Cross-talk between ␥ 3 -and ␥ 7 -dependent Signaling Pathways within the Same Neuronal Population as a Possible Mechanism
for Double Knock-out Phenotype-Next, we considered the possibility that the ␥ 3 and ␥ 7 subtypes are acting in separate signaling pathways within the same striatal cell type. The striatum is comprised of 90% medium spiny projection neurons and 10% large aspiny interneurons (23). Employing an innovative expression profiling technique, Doyle et al. (24) showed that Gng3 and Gng7 mRNA transcripts are expressed in In contrast, adenylyl cyclase activity in the dorsal striatum of Gng7 Ϫ/Ϫ and Gng3 Ϫ/Ϫ ;Gng7 Ϫ/Ϫ mice is reduced to a similar extent at base line and in response to dopamine, the D 1 dopamine-selective agonist (6-chloro-PB) (C), or forskolin (Forsk) (D). (*, p Ͻ 0.05 for comparison with Gng3 Ϫ/Ϫ by t test, n ϭ 6.) Immunoblot of cholate-extracted proteins from dorsal striatum normalized to Ras and expressed as percentage of levels in Gng3 Ϫ/Ϫ mice (E) shows ␣ olf and ␤ 2 are markedly reduced in striatal membranes from Gng7 Ϫ/Ϫ mice and are further reduced in Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice, although ␣ s , ␣ i3 , and ␤ 1 are not reduced in Gng7 Ϫ/Ϫ mice and unchanged or slightly increased in Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice. (*, p Ͻ 0.05 for comparison with Gng3 Ϫ/Ϫ by t test, **, p Ͻ 0.05 for comparison with Gng7 Ϫ/Ϫ or Gng3 Ϫ/Ϫ by t test, n ϭ 3). F shows ␤ 2 is significantly reduced in cortical membranes from Gng3 Ϫ/Ϫ mice (*, p Ͻ 0.01 for comparison with Gng7 Ϫ/Ϫ by t test, n ϭ 3), although ␣ i3 and ␤ 1 are not changed. No further reduction of ␤ 2 is observed in cortical membranes from Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice. Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype MARCH 2, 2012 • VOLUME 287 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 7127 both striatal cell populations (Fig. 5A). Because mRNA levels of G-protein subunits might not reflect their protein levels (25), we extended this analysis to determine the cellular distribution of the ␥ 3 and ␥ 7 reporter proteins. For this purpose, corticostriatal slices from transgenic mice expressing GFP under control of the Gng3 promoter were used. To visualize GFP expressing neurons, a low magnification image of a representative slice from these mice is shown in Fig. 5B, and a higher magnification image of the same field is shown in Fig. 5C. The cortex contained numerous green cells, and the dorsal striatum (caudateputamen) showed only a few GFP-positive neurons that represented Ͻ1% of striatal cells. The scarcity of GFP-positive neurons in dorsal striatum is suggestive of cholinergic interneurons that account for only a small fraction of striatal cells (26). To identify the cholinergic interneurons, a high magnification image of the same field stained with the ChAT antibody is shown in Fig. 5D, and the extent of overlap between GFP and ChAT expression patterns is revealed in Fig. 5E. The finding that ChAT staining showed extensive overlap with GFP expression confirmed expression of the ␥ 3 reporter protein in cholinergic interneurons and the surrounding neuropil (Fig. 5, B-E). These results revealed that the ␥ 3 reporter protein was expressed in most cortical neurons, along with striatal interneurons of the cholinergic type.
Subsequently, corticostriatal slices from knock-in mice expressing GFP under control of the endogenous Gng7 locus were examined. Fig. 6A confirmed the expected GFP expression pattern in the striatum that recapitulated the endogenous ␥ 7 expression revealed by in situ hybridization (Allen Brain Atlas). To visualize GFP-expressing neurons, a low magnification image of a representative slice is shown in Fig. 6B, and a higher magnification image of the same field is shown in Fig.  6C. In contrast to the cortex that was devoid of any green cells, the dorsal striatum (caudate-putamen) contained numerous GFP-positive neurons that account for the majority of striatal cells. The preponderance of GFP-positive neurons in the dorsal striatum is indicative of medium spiny neurons that account for ϳ90% of striatal neurons (27). To identify the medium spiny neurons, a high magnification view of the same field stained with dopamine-and adenosine-regulated 32-kilodalton phos-  Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype phoprotein (DARPP-32) antibody is shown in Fig. 6D, and coincidence between the GFP and DARPP-32 expression patterns is revealed in Fig. 6E. The finding that GFP expression overlapped with DARPP-32 staining confirmed expression of the ␥ 7 reporter protein in medium spiny projections neurons and the surrounding neuropil (Fig. 6, B-E). Because medium spiny neurons receive both glutamatergic inputs from cortical neurons and cholinergic inputs from striatal interneurons, these results revealed for the first time that the ␥ 3 and ␥ 7 reporter proteins are largely segregated between different neuronal subpopulations that contact each other. Furthermore, these data argue against cross-talk between G i/o -and G olf -signaling pathways requiring the ␥ 3 and ␥ 7 subunits in the same neuronal population as the basis for the double knock-out phenotype.
Functional Interaction between Signaling Pathways in Different Neuronal Populations as the Most Likely Explanation for the Double Knock-out Phenotype-Finally, we considered the possibility that G i/o -and G olf -signaling pathways are acting in different neuronal populations operating within a circuit to regulate a common neurological process. Although a requirement for the ␥ 7 subunit in G olf assembly and D 1 R and A 2A R signaling is now established (6,8), a role for the ␥ 3 subunit in a particular receptor signaling pathway has not been fully elucidated. Suggesting a role in GABA B R signaling, the ␥ 3 protein is expressed in cortex, hippocampus, and striatum in a similar pattern to the GABA B receptor (28). Moreover, gene targeted deletion of the ␥ 3 protein confers susceptibility to seizures in an analogous fashion to loss of the GABA B receptor (29 -31). To explore a possible role for the G-protein ␥ 3 subunit acting downstream of this receptor, we assessed the ability of a GABA B R-specific agonist to induce delta waves on EEG recordings from wild type (n ϭ 3), Gng3 Ϫ/Ϫ (n ϭ 4), Gng7 Ϫ/Ϫ (n ϭ 4), and Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice (n ϭ 3). Delta waves can be readily identified as slow frequency, high amplitude waves. Baclofen effectively induced delta waves on EEG recordings from wild type mice that appeared by 40 min and disappeared by 7 h post-injection (Fig.  7A). In contrast, baclofen showed an impaired ability to induce delta waves on EEG tracings from both Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ mice (Fig. 7A). Individual comparisons showed a significant time by band by genotype interaction (ANOVA, F 48,169 ϭ 4.19, p Ͻ 0.00001), indicating that baclofen produced the greatest effect in wild type mice, comparably impaired effects in both Gng3 Ϫ/Ϫ mice and Gng7 Ϫ/Ϫ mice, and a significantly worsened effect in Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice (Fig. 7B). The greater impairment of the delta wave response seen in double knock-out mice offers further support for a functional interaction between G-protein ␥ 3 -and ␥ 7 -dependent signaling pathways that modulate the delta wave response.
To further probe a requirement for G-protein ␥ 3 subunit acting downstream of the GABA B receptor, we compared additional baclofen-mediated responses (31) among Gng3 Ϫ/Ϫ and Gng7 Ϫ/Ϫ mice. To assess the muscle-relaxing effect of baclofen, we measured the time that mice spent walking on the Rota-Rod apparatus following injection of baclofen. Both wild type and Gng7 Ϫ/Ϫ mice showed the muscle-relaxing effect of baclofen, as demonstrated by decreased time spent walking on the Rota-Rod (Fig. 8A). In fact, both groups of mice employed the unusual strategy of staying on the Rota-Rod by wrapping their legs around the bar and spinning (spinning mice were given a time of 0 s). In marked contrast, Gng3 Ϫ/Ϫ mice did not show the muscle-relaxing effect of baclofen, as demonstrated by both the increased time spent walking on the Rota-Rod (Fig. 8A) and failure to exhibit the unusual spinning behavior. Next, we measured the temperature lowering effect of baclofen (31). Again, both wild type and Gng7 Ϫ/Ϫ mice exhibited the temperature lowering effect of baclofen. In contrast, this response was markedly reduced in Gng3 Ϫ/Ϫ mice (Fig. 8B). Taken together, these results revealed for the first time that loss of the G-protein ␥ 3 subunit was selectively associated with impaired GABA B R responsiveness at both the neurological and behavioral levels.
To directly investigate a requirement for the ␥ 3 subunit in GABA B R signaling, we measured GABA B R-mediated activation of GIRK currents in neurons from wild type mice and Gng3 Ϫ/Ϫ mice. Basal GIRK currents were only slightly reduced in Gng3 Ϫ/Ϫ mice, and the fractions of neurons exhibiting basal activity between the two genotypes were similar. Consistent with results from previous studies (32,33), baclofen-activated GIRK channels were observed in all 27 hippocampal neurons tested from wild type mice (Fig. 8C). In contrast, baclofenstimulated GIRK channels were seen in only in 7 of the 25 neurons from Gng3 Ϫ/Ϫ mice. The difference in the fraction of Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype MARCH 2, 2012 • VOLUME 287 • NUMBER 10 responders in wild type versus Gng3 Ϫ/Ϫ mice was significant (p ϭ 0036, Fisher's exact test). In addition, the amplitude of baclofen-induced currents was significantly attenuated in the Gng3 Ϫ/Ϫ mice compared with the wild type mice (p Ͻ 0.01, unpaired t test). Finally attesting to the specificity of this effect, the adenosine and somatostatin receptor agonists (NECA; somatostatin) activated GIRK channels in similar proportions of neurons from wild type and Gng3 Ϫ/Ϫ mice (NECA, 11/27 for WT, and 12/25 for Gng3 Ϫ/Ϫ ; somatostatin 9/26 for WT and 13/25 for Gng3 Ϫ/Ϫ ). Activation of GIRK channels in a fraction of wild type neurons by these agonists is consistent with a previous report (32). In addition, the amplitudes for NECA-and somatostatin-activated currents in hippocampal neurons from the wild type and Gng3 Ϫ/Ϫ mice were not different. Collectively, these data pointed to a specific defect in a post-synaptic GABA B R signaling pathway in a significant portion of ␥ 3 -deficient neurons, which presumably reflects the heterogeneous nature of these cells in culture (34).
Finally, to confirm defective GABA B R signaling was due to loss of the G-protein ␥ 3 and not the receptor itself, we quantified the GABA B R levels in wild type, single, and double knock-out mice, using a monoclonal antibody that recognized both alternatively spliced forms of the R1 subunit (28). Wild type neurons express both the GABA B1a and GABA B1b splice variants (Fig. 9, A and B). Moreover, there were no significant differences in abundance of these forms between wild type and ␥ 3 -deficient tissue. These data demonstrate that loss of the G-protein ␥ 3 subunit rather than down-regulation of the GABA B receptor was responsible for the defective baclofen responses observed in Gng3 Ϫ/Ϫ knock-out mice. Thus, combined neurological, behavioral, and electrophysiological results point to a G i/o protein containing ␥ 3 acts downstream of one or more GABA B splice variants in a cell-specific manner (35).

DISCUSSION
Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ mice show premature lethality. Because seizures trigger cardiac or respiratory arrest (36) and seizure FIGURE 8. Behavioral and electrophysiological responses to baclofen. A, wild type and Gng7 Ϫ/Ϫ mice showed marked impairment in their ability to walk on a Roto-Rod following injection of baclofen (10 mg/kg), although Gng3 Ϫ/Ϫ had less of an impairment. B, body temperature dropped by Ͼ3°C in wild type mice, Ͼ2°C in Gng7 Ϫ/Ϫ mice, but Ͻ1°C in Gng3 Ϫ/Ϫ mice in response to baclofen (10 mg/kg). (*, p Ͻ 0.05 for comparison with wild type mice, **, p Ͻ 0.01 for comparison with wild type or Gng7 Ϫ/Ϫ mice, N Ն 8). C, baclofen-induced GIRK currents were attenuated in neurons isolated from Gng3 Ϫ/Ϫ mice compared with those from wild type littermates (**, p Ͻ 0.01, unpaired t test). No difference was observed in basal, NECA-induced, or somatostatin-induced GIRK currents. suppression by the ketogenic diet (37)(38)(39) is very effective in prolonging their life span, we believe that the premature mortality of double knock-out mice is most likely caused by their severe seizure disorder. Thus, this mouse model offers a novel experimental system for understanding how combinatorial defects in signaling pathways can converge to produce polygenic forms of epilepsy. Furthermore, the finding that these mice display a severe seizure phenotype that is not observed for either single knock-out on the same genetic background provides strong evidence for a functional interaction between G-protein ␥ 3 -and ␥ 7 -dependent signaling pathways that normally limit seizure susceptibility, seizure propagation, or seizure-induced damage. Below, we compare the expression profiles, G-protein ␣ partner preferences, and receptor requirements among the different genotypes to reveal a mechanistic explanation for the severity of the double knock-out phenotype.
Role of ␥ 3 in GABA B R Signaling-The signaling pathway(s) dependent on the G-protein ␥ 3 subunit are not known. We show that the ␥ 3 reporter protein is present in cortical neurons that make glutamatergic contacts onto striatal neurons. This expression pattern mirrors that reported for the GABA B receptor (40) responsible for inhibition of glutamate release in striatum (41). Likewise, we demonstrate that the ␥ 3 reporter protein is localized to striatal interneurons that integrate synaptic inputs over large areas within the striatum (42). This is consistent with the localization reported for the GABA B receptor (40). Taken together, the similarity of their cellular expression profiles raise the possibility that a specific GABA B receptor variant may utilize a G-protein containing the ␥ 3 subunit to modulate neuronal excitability and impact striatal function (35,43).
Further supporting this possibility, Gng3 Ϫ/Ϫ mice show a similar loss-of-function phenotype to that reported for Gabbr1 Ϫ/Ϫ subunit-specific mice. In this regard, the GABA B receptor functions as an obligate heterodimer; the GABA B1 subunit, which is encoded by the Gabbr1 gene, contains the ligand-binding site and localization motif, although the GABA B2 subunit, which is encoded by the Gabbr2 gene, mediates coupling to the G-protein(s) (35,43). Complete ablation of either the Gabbr1 (29,30) or Gabbr2 gene (31) produces a severe seizure disorder resulting in death. However, more subtle phenotypes result from individual ablation of alternatively spliced forms of the Gabbr1 gene (44,45) that are thought to convey distinct functions through their differential subcellular localizations (44). In this regard, Gabbr1a Ϫ/Ϫ mice, which retain predominantly post-synaptic GABA B1b,2 receptors, display a mild seizure phenotype that does not affect viability, whereas Gabbr1b Ϫ/Ϫ mice, which preserve mostly pre-synaptic GABA B1a,2 receptors, exhibit no apparent seizure phenotype. Similar to the subunit-specific Gabbr1 Ϫ/Ϫ phenotypes, Gng3 Ϫ/Ϫ mice exhibit a mild seizure phenotype on the seizuresensitive FVB background but no obvious seizure defect on the seizure-resistant B6 background.
Confirming the G-protein ␥ 3 subunit is acting downstream of the GABA B receptor, we identified defective GABA B R signaling in Gng3 Ϫ/Ϫ mice. In the post-synaptic setting, the GABA B1b,2 receptor reportedly couples through a G i/o -protein to activate a specific GIRK channel (33). Pointing to a specific role in this process, baclofen-induced GIRK activation is lost in a significant proportion of hippocampal neurons derived from Gng3 Ϫ/Ϫ mice. Ruling out other possible explanations for this defect, there are no significant differences in GABA B receptor abundance or effector regulation by other receptor agonists. Therefore, loss of the G-protein, secondary to genetic deletion of the ␥ 3 subunit, appears to be the most likely explanation for the defective GIRK activation seen in this study. Suggesting possible G-protein ␣ partners, the gene-targeted loss of the ␥ 3 subunit produces coordinate suppression of the ␣ i3 , ␣ o , and ␤ 2 subunits in certain brain regions (7). Taken together, these results are most consistent with the post-synaptic GABA B1b,2 receptor acting through G-protein ␣ i/o ␤ 2 ␥ 3 trimer to open GIRK channels, causing reduced neuronal excitability. When this pathway is disrupted by either genetic inactivation of the GABA B1 receptor (44), the G-protein ␥ 3 subunit (this study), or the GIRK channel (46), animals are prone to developing seizures depending on the presence or absence of other modifier genes in the strain background.
Role of ␥ 7 in D 1 R and A 2a R Signaling-The signaling pathway(s) dependent on G-proteins containing the ␥ 7 subunit are now known. The ␥ 7 reporter protein is preferentially expressed in striatal projection neurons (also called medium spiny neurons). As shown previously (6,8), Gng7 Ϫ/Ϫ mice lacking the ␥ 7 subunit exhibit impaired assembly of a specific G-protein ␣ olf ␤ 2 ␥ 7 trimer, defective D 1 R-and A 2A R-stimulated adenylyl cyclase activation, and altered locomotor behaviors (47). However, these knock-out mice do not exhibit any evidence of spontaneous seizure activity or premature lethality (6). Moreover, mice lacking either the A 2A R (48,49) or the D 1 R (50, 51) are not described as having seizures, although loss of the D 1 R-expressing cells themselves has been shown to trigger a progressive seizure disorder (52). Hence, it is not clear how defective A 2A R or D 1 R signaling contributes to the severe seizure phenotype of Gng3 Ϫ/Ϫ Gng7 Ϫ/Ϫ double knock-out mice. Based on the available evidence, we speculate that a G-protein ␣ olf ␤ 2 ␥ 7 trimer acting downstream of the A 2A R or D 1 R may confer a neuroprotective effect. Supporting such a possibility, seizure activity has been shown to increase adenosine levels and A 2A R signaling (53), although the consequences are controversial. Consistent with a neuroprotective effect, A 2A R stimulation attenuates brain damage induced by kanaic acid-induced excitoxicity (54) or striatal lesion (55). Conversely, arguing against a neuroprotective action, A 2A R blockade also reduces brain damage (49,56). Clearly, more studies will be needed to resolve this issue. Providing some insight into these conflicting results, a recent paper suggests that A 2A R activation may switch between neuroprotective and neurodegenerative states depending on existing levels of glutamate (57). Alternatively, D 1 R activation may be responsible for conferring a neuroprotective effect (58). Additional studies will be needed to directly investigate a possible role for the G-protein ␥ 7 subunit in these processes.
Convergent Roles of ␥ 3 and ␥ 7 in Separate Signaling Pathways-The severe seizures and increased interictal spiking seen in double knock-out mice point to a novel functional interaction between G i/o -and G olf -dependent signaling pathways requiring ␥ 3 and ␥ 7 in mediating neuronal excitability and/or protection. As a working model, we speculate that a post-synaptic GABA B receptor utilizes Gng3 ؊/؊ Gng7 ؊/؊ Mice Have Severe Seizure Phenotype MARCH 2, 2012 • VOLUME 287 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 7131 a G i/o protein containing the ␥ 3 subunit to regulate neuronal excitability (43), whereas the A 2A or D 1 receptor requires the G olf protein containing the ␥ 7 subunit to confer a neuroprotective response (49,54,59). In such a scenario, simultaneous disruption of both signaling pathways could account for the strong seizure phenotype that dramatically reduces the life span of double knockout mice to only 75 days. Based on these data, the G-protein ␥ 3 and ␥ 7 subtypes can be added to a growing list of susceptibility genes that may act synergistically to contribute to human seizure disorders of polygenic origin. In this regard, evidence showing the GABA B , A 2A adenosine, and D 1 dopamine receptors utilize specific G-protein ␣␤␥ heterotrimers offers a new interface for more selective intervention in seizure disorders.
Likewise, the greater blockade of the delta wave response seen in double knock-out mice points to a similar functional interaction between G i/o -and G olf -dependent signaling pathways requiring ␥ 3 and ␥ 7 in mediating wakefulness. In this regard, GABA B R blockade has been shown to reduce delta waves (31), providing a likely explanation for the reduced response seen in Gng3 Ϫ/Ϫ mice. Moreover, D 1 R antagonists have been reported to suppress the amplitude of delta waves (60,61), offering a possible explanation for the decreased delta wave response seen in Gng7 Ϫ/Ϫ mice. Finally, disruption of both GABA B R and D 1 R signaling would be entirely consistent with the complete blockade of the delta wave response seen in double knock-out mice. At this point, where and how these signaling pathways converge is not known.
In summary, we report a new mouse line that reveals for the first time that in vivo disruption of G i/o and G olf signaling pathways produces a severe seizure phenotype. This is particularly interesting in that it is one of the few mouse models that recapitulate the polygenic basis of many forms of human epilepsy.
To determine the relevance of this mouse model to the clinical condition, future work will explore whether similar defects in these signaling pathways are observed in surgically resected epileptic tissue from patients undergoing treatment for refractory seizures. Notably, the demonstration that individual or combinatorial deletion of the closely related ␥ 3 and ␥ 7 subunits produces distinct and identifiable seizures reinforces the growing recognition that the nature of the ␥ component plays a critical role in the signal transduction process (6 -8).