Search tips
Search criteria 


Logo of wtpaEurope PMCEurope PMC Funders GroupSubmit a Manuscript
Nat Neurosci. Author manuscript; available in PMC 2007 November 23.
Published in final edited form as:
Published online 2007 June 17. doi:  10.1038/nn1927
PMCID: PMC2092503

From synapse to behavior: rapid modulation of defined neuronal types with engineered GABAA receptors


In mammals, identifying the contribution of specific neurons or networks to behavior is a key challenge. Here we describe an approach that facilitates this process by enabling the rapid modulation of synaptic inhibition in defined cell populations. Binding of zolpidem, a systemically active allosteric modulator that enhances the function of the GABAA receptor, requires a phenylalanine residue (Phe77) in the γ2 subunit. Mice in which this residue is changed to isoleucine are insensitive to zolpidem. By Cre recombinase-induced swapping of the γ2 subunit (that is, exchanging Ile77 for Phe77), zolpidem sensitivity can be restored to GABAA receptors in chosen cell types. We demonstrate the power of this method in the cerebellum, where zolpidem rapidly induces significant motor deficits when Purkinje cells are made uniquely sensitive to its action. This combined molecular and pharmacological technique has demonstrable advantages over targeted cell ablation and will be invaluable for investigating many neuronal circuits.

A classical approach to the study of brain function is selective lesioning. Unfortunately, the interpretation of data from such studies can be confounded by compensatory changes, whereby unrelated systems are recruited to alleviate, if only partially, any deficit. A complementary method involves reversibly silencing, albeit with little or no cell-type selectivity, the activity of a pathway or nucleus through cooling or stereotaxic drug administration (for example, see refs. 1,2). Reversible approaches have advantages over permanent lesioning. First, the effects of acute regional inactivation cannot be easily overcome by compensatory changes, because the inactivated system is altered only briefly (reviewed in refs. 3,4). Second, we can learn, in principle, not only ‘where’, but also ‘when’ and for ‘how long’ a brain region is involved in a given function3. Similar considerations apply to the functional dissection of brain circuit components. Thus, to determine how specific cell types influence network properties and contribute to animal behavior, a method is needed for reversibly inactivating a single neuronal type in a specific brain area5. This approach poses two challenges: first, there are many cells of each type; and second, neuronal classes are often dispersed as sparse populations in large volumes (such as subtypes of GABAergic interneurons in the hippocampus).

The effects on brain function of drugs acting on GABAA receptors show that it is not necessary to silence a cell type to alter network properties or behavior: a modest modulation of inhibitory postsynaptic currents (IPSCs) is sufficient6-8. GABAA receptors, which are present on all neurons in the mammalian brain, are ligand-gated anion-permeable channels and produce fast inhibition. They are commonly formed from two α, two β and a single γ2 subunit (encoded by Gabrg2)6-8. The α1βγ2, α2βγ2 and α3βγ2 subunit combinations account for most GABAA receptors; of these, the main subtype is α1β2γ2 (refs. 6-8). Benzodiazepines, benzodiazepine-site ligands (such as the imidazopyridine zolpidem) and anesthetics are positive allosteric modulators of GABAA receptors6-10. These drugs cause a conformational alteration in the GABAA receptor such that GABA-induced anion flux is enhanced11. Elegant studies of ‘knock-in’ mice, carrying amino acid changes at drug-binding sites in the α or β subunits, show how subtype-specific modulation of GABAA receptors differentially influences mouse behavior6,12-15. These studies are ‘subtractive’: by changing genes encoding α or β subunits across the whole brain and observing the diminished effect of benzodiazepines, zolpidem or anesthetics on the mouse, the contributions of GABAA receptor subtypes to specific behaviors (related to anxiety, sedation, anesthetic actions or motor control) can be deduced6,12-15.

To develop a method for functional dissection of circuit components, we considered reversing this GABAA receptor knock-in strategy by generating zolpidem-insensitive mice and then genetically imposing zolpidem sensitivity on a selected cell type. We chose zolpidem as the ligand because of the way in which it binds to the GABAA receptor. One GABAA receptor subunit, the γ2 subunit, is expressed almost universally in brain16,17 and is incorporated into most GABAA receptors. Zolpidem binding occurs throughout the brain17,18 inside a pocket created between an α subunit (α1, α2 or α3) and the γ2 subunit19. The γ2 subunit contributes an essential residue (Phe77) to this pocket20,21. As shown with cloned GABAA receptors, if Phe77 of γ2 is mutated to isoleucine, the receptors function normally with respect to activation by GABA but they are no longer modulated by zolpidem20,21. Pharmacological, electrophysiological and behavioral tests have shown that mice with an engineered γ2 Ile77 mutation are insensitive to zolpidem22-24. The Ile77 mutation does not affect the action of GABA at GABAA receptors: in γ2 Ile77 mice, miniature IPSCs (mIPSCs) have normal rise times, amplitudes and decay kinetics22,23.

We therefore considered that, by cell type selectively replacing (‘swapping’) the Ile77 γ2 subunit in zolpidem-insensitive mice with the zolpidem-sensitive (Phe77) variant, we would be able to use zolpidem to modulate reversibly the activity of any defined neuronal population. To test the feasibility of this method, we chose to use Purkinje cells, which offer straightforward genetic targeting25 and a quantifiable behavioral readout—namely, motor performance on the rotarod. Purkinje cells provide the only output of the cerebellar cortex. Dendritic and perisomatic synaptic inhibition, provided by stellate and basket cells, respectively, alters the intrinsic activity of Purkinje cells26 and shapes their response to excitatory synaptic input27,28. Purkinje cells express α1β2γ2- and α1β3γ2-type GABAA receptors29,30 that underlie fast GABAergic IPSCs.

Here we describe how we have made GABAA receptors on Purkinje cells uniquely sensitive to zolpidem and then have used this drug to produce rapid and reversible behavioral effects through selective modulation of inhibitory synaptic input. To emphasize the advantages of this approach, we have also generated a static (long-term) knockout of fast GABAergic input onto Purkinje cells by producing cell type-selective deletion of their γ2 subunits, and have compared the effects of the two manipulations on the same behavioral tests under identical conditions.


Explanation of the strategy applied to Purkinje cells

The strategy was to produce mice devoid of zolpidem-sensitive GABAA receptors by engineering the γ2 subunit gene (Gabrg2) such that the codon for Phe77 encoded isoleucine. The gene was also engineered to contain loxP sites, allowing its inactivation by Cre recombinase (Fig. 1). Coexpression of the zolpidem-sensitive γ2 Phe77 (hereafter γ2F77) subunit and Cre recombinase in Purkinje cells should thus lead to a subunit swap, such that Cre-positive, γ2F77-positive Purkinje neurons would not contain γ2 Ile77 (hereafter γ2I77) subunits, which would have been deleted by Cre recombinase (Fig. 1). We termed these Purkinje γ2 swap mice ‘PC-γ2-swap’ mice: only their Purkinje cells should express zolpidem-sensitive GABAA receptors.

Figure 1
Strategy to restrict zolpidem sensitivity to selected neural types, using Cre recombinase to drive a γ2I77 to γ2F77 subunit swap. The genotype, the pentameric GABAA receptors formed from α, β and γ2 (γ2F77 ...

Deletion of the GABAA receptor γ2 subunit in Purkinje cells

The γ2I77lox line served as both a loxP-flanked (floxed) allele for cell type-specific ablations of GABAA receptor γ2 subunits (γ2I77lox × cell type-specific promoter driving Cre) and the foundation line for producing cell type-selective zolpidem sensitivity (γ2I77lox × cell type-specific promoter driving Cre × cell type-specific promoter driving γ2F77 subunit ‘swapping’; see Fig. 2, and Supplementary Methods and Supplementary Fig. 1 online).

Figure 2
Summary of mouse genotypes. In situ hybridization autoradiographs show the expression of various GABAA receptor γ2 subunit mRNA in brain sections from γ2I77lox, PC-Δγ2 and PC-γ2-swap adult mice. The corresponding ...

The γ2 subunit targets GABAA receptors to postsynaptic sites and confers full (25-30 pS) single-channel conductance on the receptor31,32. To compare and to contrast the dynamic modulation of GABAergic input onto Purkinje cells with the long-term removal of such input, we also generated mice with a Purkinje-cell-specific deletion of γ2, and thus a deletion of synaptic GABAA receptors. Mice with Cre recombinase expressed selectively in Purkinje cells (L7Cre mice33; see Supplementary Fig. 2 online) were crossed with γ2I77lox mice, producing a Purkinje cell-selective ‘static’ knockout of synaptic responsiveness to GABA (these mice were termed ‘PC-Δγ2’ mice; Fig. 2).

Purkinje cell-restricted γ2F77GFP expression

For the zolpidem-sensitive γ2F77 subunit swap, we tagged the γ2F77 subunit at its amino terminus with enhanced green fluorescent protein, eGFP (γ2F77GFP; see Supplementary Methods and ref. 34) to distinguish it from the endogenous γ2I77 subunit. By using the L7 Purkinje cell-specific promoter25, mice were produced with expression of γ2F77GFP restricted to Purkinje cells. One of the L7γ2F77GFP lines (of six screened) gave strong expression of the mRNA specifically in Purkinje cells, as detected by in situ hybridization with a GFP-specific probe. We then made triple crosses (γ2I77lox × L7Cre × L7γ2F77GFP (see Methods)) to generate PC-γ2-swap mice (Fig. 2). At the cellular level, we first established that the γ2F77GFP subunit in these mice was targeted to GABAergic synapses, and that IPSCs onto Purkinje cells were uniquely sensitive to zolpidem.

γ2F77GFP is enriched at GABAergic synapses

Primary GFP fluorescence and GFP immunoreactivity in PC-γ2-swap mice was selectively localized to Purkinje cells throughout the cerebellum (Supplementary Fig. 3 online). The rest of the brain in PC-γ2-swap mice (or littermate control mice) had no GFP immunoreactivity. Purkinje cells of PC-γ2-swap mice contained granular GFP immunoreactivity in both their cell body (excluding the nucleus) and their proximal dendrites, presumably corresponding to γ2F77GFP associated with endo-membranes (Fig. 3). Strongly immunopositive puncta and 0.5-1-μm segments of membrane were present in the soma and the main dendrites (Fig. 3). Scattered GFP-positive puncta were also present throughout the molecular layer and presumably corresponded to smaller dendrites of Purkinje cells (Fig. 3).

Figure 3
Synaptic expression of the γ2F77GFP subunit in Purkinje cells of adult PC-γ2-swap mice. A confocal microscopic image shows immunoreactivity for GFP (green), the α1 subunit of the GABAA receptor (blue) and GAD (red) in the cerebellum. ...

To confirm that the γ2F77GFP subunits were located in GABAergic synapses, we performed triple immunolabeling (Fig. 3). The GFP-positive segments always colocalized with the endogenous α1 subunit of the GABAA receptor. Immunoreactivity for the α1 subunit was also present as short segments and puncta in the molecular layer outside the GFP patches, corresponding to GABAergic synapses on other cerebellar neurons. Most GFP patches, both on Purkinje cells and in the neuropil, were apposed to glutamic acid decarboxylase (GAD)-immunoreactive structures, representing GABAergic nerve terminals (Fig. 3).

Restriction of zolpidem sensitivity to Purkinje cells

To determine the zolpidem sensitivity of GABAergic synaptic transmission in the different mouse lines, we recorded mIPSCs from Purkinje cells and molecular layer interneurons (presumptive stellate cells) in slices prepared from adult mice. As reported previously22, the properties (amplitude, rise time and decay) of mIPSCs in γ2I77 mice were similar to those of mIPSCs in γ2F77/F77 and C57BL/6 mice, confirming that the γ2F77I point mutation was a neutral substitution. The γ2I77lox mice generated here differ from previously reported γ2F77I mice22 only by the inclusion of a loxP site located 5′ to exon 4 (Supplementary Fig. 1). Accordingly, mIPSCs recorded in both Purkinje cells (amplitude, 57.5 ± 5.2 pA; 10-90% rise time, 296 ± 11 μs; decay (τw; see Supplementary Methods), 2.7 ± 0.2 ms; charge transfer, 165.4 ± 13.8 fC; n = 11 cells) and stellate cells (61.3 ± 10.4 pA; 328 ± 16 μs; 2.6 ± 0.1 ms; 194.4 ± 33.3 fC; n = 5) had properties similar to those of the corresponding mIPSCs reported previously22.

In situ hybridization data suggested that our Cre-loxP strategy eliminated the γ2 subunit from Purkinje cells, a condition necessary to facilitate the Ile77 to Phe77 swap (Fig. 2). To confirm this elimination, we also recorded from Purkinje cells of PC-Δγ2 mice and their γ2I77lox/γ2I77lox littermates. Whereas mIPSCs were present in every control Purkinje cell (n = 21), they were absent from all Purkinje cells of PC-Δγ2 mice (n = 10; Fig. 4a), verifying the effective deletion of the γ2 subunit and the elimination of fast GABAergic synaptic input onto Purkinje cells.

Figure 4
Potentiation of GABAA receptor-mediated mIPSCs by zolpidem in Purkinje cells of PC-γ2-swap mice. (a) Continuous 20-s segments of whole-cell recordings (-70 mV) from Purkinje cells of γ2I77lox (postnatal day 140), PC-Δγ2 ...

In Purkinje cells of PC-γ2-swap mice, mIPSC amplitude (46.4 ± 7.2 pA), 10-90% rise time (307 ± 8 μs), τw (3.1 ± 0.2 ms) and charge transfer (158.6 ± 21.7 fC) of mIPSCs (n = 8) were not significantly different from their corresponding measures in γ2I77lox mice (see above; all P > 0.05). These data verify effective deletion of the γ2I77 subunit and its replacement with the γ2F77GFP subunit. Although basic mIPSC properties in the two lines were indistinguishable, modulation of mIPSCs by zolpidem was, as predicted, markedly different. Zolpidem had no effect on the amplitude or kinetics of mIPSCs in Purkinje cells of γ2I77lox mice; in the PC-γ2-swap mice, by contrast, it significantly raised the amplitude (49.7 ± 9.9 to 62.0 ± 12.1 pA), prolonged τw (3.2 ± 0.2 to 5.3 ± 0.2 ms) and increased the charge transfer (178.9 ± 29.0 to 377.5 ± 13.6 fC) of Purkinje cell mIPSCs (n = 5, all P < 0.05; Fig. 4b,c). The restoration of zolpidem sensitivity was restricted to Purkinje cells: zolpidem had no effect on the properties of mIPSCs recorded from stellate cells in slices from the same mice (Supplementary Fig. 4 online).

In female PC-γ2-swap mice, transgene expression was mosaic and about 50% of Purkinje cells expressed the transgene. Most probably, the L7γ2F77GFP transgene integrated on an X chromosome. Accordingly, some Purkinje cells from females showed a complete lack of mIPSCs (equivalent to that seen in the PC-Δγ2 mice). In cells where the γ2F77GFP transgene was expressed, however, mIPSCs were normal and showed zolpidem sensitivity comparable to that of mIPSCs in γ2F77/F77 or C57BL/6 mice22. In male mice, at least 90% of Purkinje cells swapped their γ2 subunits, and only male mice were used for behavioral tests involving zolpidem.

Dynamic modulation of inhibition produces motor deficits

What role does GABAergic inhibition of Purkinje cells have in motor coordination? Unexpectedly, mice with no fast synaptic inhibition onto Purkinje cells (PC-Δγ2 mice) had no motor disabilities and showed neither ataxia nor tremor. They performed the rotarod task to the same standard as their littermate controls (Fig. 5a). Given this observation, we considered what would be the behavioral effect of acute modulation of fast GABAergic input onto Purkinje cells when PC-γ2-swap mice were given zolpidem. Without drugs, PC-γ2-swap mice showed no motor disabilities: they learned the rotarod task to the same standard as their littermate controls. On the first day of training, a small difference in latency to fall was detected, but this difference disappeared during subsequent training days (Fig. 5a).

Figure 5
Motor performance on a rotarod and horizontal beam after systemic administration of zolpidem. (a) The rotarod was accelerated from 5 to 30 r.p.m. over 180 s. Data for γ2I77lox littermate control (n = 20), PC-γ2-swap (n = 20) and PC-Δγ2 ...

Intraperitoneal administration of zolpidem to PC-γ2-swap mice produced no sedative or overt ataxic effect. Within 1-5 min of drug injection, mice were tested on the rotarod. Zolpidem significantly reduced the ability of trained PC-γ2-swap mice to stay on the rotating rod (Fig. 5b; repeated measures analysis of variance (ANOVA), F1,84 = 10.4, P < 0.001; zolpidem treatment × mouse line interaction, F1,84 = 47.6, P < 0.001; Newman-Keuls post hoc test, P < 0.001). With the initial dose of 3 mg/kg (body weight) of zolpidem, the effect was highly significant (Fig. 5b) and almost saturated. No greater effect was produced by a cumulative dose of 9 mg/kg of zolpidem. In an additional experiment, a single dose of 12 mg/kg of zolpidem shortened the latency to fall in the PC-γ2-swap mice from 144 ± 30 s to 96 ± 48 s (P < 0.05), indicating that the lowest doses used produced the maximal motor-impairing effect. These effects of zolpidem were produced without sedation. This observation was expected, because the γ2I77 subunit, retained in all cells other than Purkinje cells, prevents the normal effects of zolpidem on, for example, sleep regulating centers13,22. The γ2I77lox littermates were all unaffected.

To further test fine motor coordination and balance, we used the ‘walking beam’ test, whereby mice have to cross a suspended rod. Without zolpidem, there were no differences between PC-γ2-swap mice and their γ2I77lox littermates (repeated measures ANOVA, F1,148 = 2.01, P > 0.05, data not shown). After administration of zolpidem (12 mg/kg), however, PC-γ2-swap mice needed significantly more time to traverse the beam (treatment × mouse line interaction, repeated measures ANOVA, F1,26 = 9.84, P < 0.01; Newman-Keuls post hoc test, P < 0.001 for mouse line). By contrast, their γ2I77lox littermates were not affected (Fig. 5c).

The behavioral effects produced by zolpidem in PC-γ2-swap mice could be blocked by flumazenil (Ro 15-1788), which antagonizes the binding of zolpidem to GABAA receptors containing the γ2F77 subunit24. PC-γ2-swap mice were pre-treated with vehicle or 15 mg/kg of flumazenil and then given a single dose of 6 mg/kg of zolpidem. Immediately afterwards they were tested on the rotarod. The latency to fall in the vehicle pre-treated group (n = 6) shortened from 173 ± 13 s to 137 ± 13 s (one-way ANOVA and Newman-Keuls post hoc test, P < 0.05), but stayed the same in the group pre-treated with flumazenil (170 ± 13 s; one-way ANOVA, P > 0.05 n = 4).


Rapid modulation avoids adaptational change

Motor coordination on the accelerating rotarod is a complex task, involving many brain areas, cell types and synapses. Here, to demonstrate the feasibility of a reversible method for manipulating defined synapses, we have isolated the contribution of GABAA receptor-mediated synaptic input onto Purkinje cells to this task. Dynamically modifying GABAergic input, through acute systemic administration of zolpidem to PC-γ2-swap mice, produced a result different from that of static long-term removal of the input (through knockout of synaptic GABAA receptors in PC-Δγ2 mice). Under the same test conditions, static deletion did not reduce performance on the rotarod, whereas fast and reversible enhancement of the same GABAergic input produced acute debilitating effects. These contrasting results reflect, in the chronic intervention, the adaptive abilities of neurons and networks to engage alternative strategies to achieve similar ends (for example, see refs. 4,35,36). Nevertheless, chronic interventions can be informative. When no compensation occurs, ablating a component can demonstrate its necessity for a particular process. Indeed, PC-Δγ2 mice do have profound deficits in the consolidation of cerebellar motor memories, indicating that fast inhibition onto Purkinje cells provides essential functions that cannot be compensated for after chronic ablation (P. Wulff et al., unpublished data). Thus, pursuing both types of study (chronic ablation and reversible intervention) gives a fuller picture of how a ‘component’ (here, a defined synaptic input to a target cell) contributes to a ‘system’ (here, cerebellar motor coordination).

Several methods to study network components, using either fast (millisecond) or relatively slow (hours to days) interventions, are available. Light-activated ion channels, receptors or pumps enable ultrafast pinpoint activation or inhibition of neurons37,38. Although excellent for use in vitro or in vivo for transparent species, these approaches are likely to be less applicable to the study of behavior and distributed populations of cells in vivo in mammals. The Drosophila allatostatin receptor system, which couples to mammalian potassium GIRK channels, allows reversible silencing of neurons within seconds and, like light-activated channels, is not limited to a particular vertebrate species39,40. The agonist peptide allatostatin, however, requires intracerebral injection and its uneven diffusion in the brain may cause difficulties for in vivo studies. Methods that block synaptic transmission have very slow kinetics41,42, but these approaches seem to be useful for long-term (days) synaptic silencing41,42. By contrast, our strategy does not silence or activate cells. Instead, it confers the ability to modulate reversibly the action of an endogenous neurotransmitter and thus it more subtly interferes with network function. These various methods are complementary, potentially enabling a full range of physiological processes to be investigated.

We have shown that zolpidem, by promoting a restricted and specific increase in fast GABAergic inhibition from molecular layer interneurons onto Purkinje cells, decreases the ability of mice to carry out complex motor tasks. To gauge the success of the strategy, it is important to know the extent of motor deficit that one might expect to see after such Purkinje cell-specific modulation. Zolpidem is a potent hypnotic, acting on GABAA receptors in brain areas that regulate sleep13. We have shown previously that in wild-type mice zolpidem produces severe sedation and deficits in rotarod performance22; given this sedation, we cannot estimate what fraction of the performance deficit originates from modulation of the cerebellum and motor centers. However, comparable data from studies on benzodiazepine-insensitive mice, notably α1H101R mice, are available. In cerebellar circuits α1βγ2 is the predominant GABAA receptor type29,30, and in a wild-type cerebellum zolpidem will modulate IPSCs onto most cell types. From studies of mice (α1H101R) with α1βγ2 receptors made insensitive to diazepam12,14,15, the net ‘contribution’ of modulating α1βγ2-type GABAA receptors to rotarod behaviors (as judged by decreased latencies to fall) is about 25% (this reflects modulation of α1 receptors throughout the brain, not just in the cerebellum14,15). We found here that zolpidem, acting on α1βγ2 receptors in Purkinje cells alone, induced a 33% decrease in latency to fall, broadly fitting with results from α1H101R mice14,15. The effects seem to be greater than those obtained after blockade of GABA release from Purkinje cell terminals in the deep cerebellar nuclei (15% decrease in latency to fall with the rod rotating at 80 r.p.m.; ref. 41).

Methodological considerations

Our approach is designed to enable zolpidem, through the cell type-selective modulation of GABAA receptor-mediated synaptic currents, to be used in behaving animals to alter the activity of defined neuronal populations. For this strategy, active GABAergic synapses are needed. The method requires that targeted neurons express α1, α2 or α3 subunits together with any β subunit and the γ2 subunit. Many cell types do indeed express predominantly α1βγ2- or α2βγ2-type receptors; for example, various subtypes of hippocampal GABAergic interneuron have zolpidem-sensitive GABAA receptors10,23. In Purkinje cells, zolpidem modulation primarily affects synaptic inhibition22; these cells do not show a tonic GABAA receptor-mediated current30,43. Other cell types, however, may express zolpidem-sensitive extrasynaptic GABAA receptors (containing α1, α2 or α3, β and γ2 subunits) with GABA-dependent or GABA-independent tonic activity7,44. If such cells were targeted by our strategy, the activity of these receptors would also be enhanced by zolpidem.

Zolpidem can be used both in slices and in behaving animals; it can be administered by intraperitoneal (i.p.) injection or by stereotaxic injection into the brain. Zolpidem binds preferentially to α1-containing GABAA receptors (α1β2γ2 or α1β3γ2; inhibition constant, Ki = 20 nM), but at higher concentrations it will modulate receptors containing α2 or α3 subunits (α2βγ2 or α3βγ2; Ki 400 nM). Because zolpidem can be given at doses of up to 30 mg/kg in γ2I77 mice with no effects22, it can be used to modulate cell types that express only α2- and α3-containing GABAA receptors. Zolpidem tartrate is the preferred preparation because of its maximal water solubility. An advantage of our method is that zolpidem has a rapid and short-acting effect: zolpidem maximally occupies its binding sites in mouse brain within minutes of i.p. injection and has a half-life of 20 min (ref. 45).

In addition to zolpidem, various other GABAA receptor modulatory drugs, such as the β-carbolines DMCM and β-CCM, the imidazobenzodiazepine bretazenil, the cyclopyrrolone zopiclone, and the quinolines PK 8165 and PK 9084, depend on the Phe77 residue of the γ2 subunit for binding20,21,24. On neuronal membranes isolated from γ2I77 mouse brains, the binding affinities of these drugs are greatly reduced24. Some of these drugs, for example DMCM and β-CCM, are inverse agonists at the benzodiazepine binding site of GABAA receptors and decrease receptor function11,20,21. The usual effect of DMCM administration (seizures) is abolished in γ2I77 mice46. Thus, in principle, both zolpidem and an inverse agonist can be used to bidirectionally modulate fast inhibition onto neurons that have undergone a γ2 subunit swap.

For its full potential to be realized, the method that we describe will require mouse Cre lines with suitable cell type-specific expression. The brain region- and cell type-specific gene expression patterns identified in, for example, the Allen Brain Atlas project will eventually aid this aim47. We used triple crosses of mouse lines, but the procedure could be streamlined by using Cre-mediated switching of an Ile77 to a Phe77 exon or viral delivery of a γ2F77-ires-Cre transgene into γ2I77lox brains. We anticipate that the unique features of our method will be invaluable for investigating the contribution of specific neurons to brain function and behavior.


Transgenic mouse production and breeding

All procedures were done in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986, had ethical approval from the Regierungspräsidium Karlsruhe or were approved by the Laboratory Animal Committee of the University of Helsinki.

Production of the γ2I77lox and L7γ2F77GFP lines and mouse genotyping are described in the Supplementary Methods. Mice homozygous for the γ2I77lox gene and heterozygous for the L7γ2F77GFP transgene were crossed with mice homozygous for γ2I77lox and heterozygous for an L7Cre transgene33. Littermates of the following genotypes were used for the experiments: γ2I77lox/γ2I77lox/L7γ2GFP/L7Cre (PC-γ2-swap), γ2I77lox/γ2I77lox/L7Cre (PC-Δγ2) and γ2I77lox/γ2I77lox (littermate controls).

In situ hybridization

We carried out in situ hybridization as described16. The probe sequences to exon 4 of the γ2 subunit and GFP are given in the Supplementary Methods.

Triple labeling immunohistochemistry

We used the following primary antibodies: affinity-purified rabbit antibody to the α1 subunit48 (0.5 μg/ml), guinea-pig antiserum to EGFP (1:100; a gift from R. Tomioka, RIKEN, Japan) and sheep antibody to GAD (1:1,000; ref. 49). See Supplementary Methods for detailed protocol and image acquisition.


Whole-cell recordings were made from neurons in parasagittal slices of cerebellar vermis prepared from adult male and female mice. A bicarbonate-buffered ‘external’ solution and a CsCl-based ‘internal’ solution (near symmetrical Cl- concentration) were used. mIPSCs were recorded at -70 mV in the presence of 5 μM CNQX (6 cyano-7-nitroquinoxaline-2,3-dione disodium), 10 μM d-AP5 and 0.5-1.0 μM tetrodotoxin. In all tests, synaptic currents recorded under such conditions were completely blocked by the GABAA receptor antagonist SR 95531 (gabazine, 10-20 μM; data not shown). All recordings were made at near physiological temperature (34-38 °C), except those from PC-Δγ2 mice and their littermates, which were made at room temperature (22-25 °C). For full details of the recording and analysis, see Supplementary Methods.

Zolpidem (Tocris Bioscience) was dissolved in dimethyl sulfoxide (DMSO) at a concentration of 50 mM and was used at 1 μM (final DMSO 0.002%). Zolpidem was applied to each slice only once, and its effect on mIPSCs was determined only after equilibration for at least 2 min.

Statistical tests were done with GraphPad Prism software (Prism 3.0, GraphPad Software Inc.). Except where noted, treatment groups and mouse lines were compared with two-tailed paired or unpaired Student’s t-tests, as appropriate. Where data were non-normally distributed (Shapiro-Wilk test), non-parametric tests were used (Wilcoxon Matched pairs or Mann-Whitney U-test). The level of significance was set at P < 0.05.

Mouse behavioral experiments

Male and female γ2I77lox (n = 20; all bred as littermates from the compound crosses, see above), PC-γ2-swap (n = 20) and PC-Δγ2 (n = 5) mice aged 4-10 months (20-45 g) were used in tests. Basic behavioral characterization was done as described in the Supplementary Methods.

Rotarod and horizontal beam tests were done as described50. The mice were given an i.p. injection of zolpidem (Sanofi-Synthelabo AB) 1-5 min before the rotarod and walking beam tests (1-4 trials). Zolpidem tartrate was crushed from a tablet, suspended in physiological saline and injected at 10 ml/kg (body weight). Drug dosing was either acute (one bolus injection) or cumulative (3 plus 3 plus 3 mg/kg, 5 min between injections). We used cumulative dosing of zolpidem with repeated rotarod testing in male mice. There was a minimum washout period of 3 d between pharmacological tests. Full details are given in the Supplementary Methods.

Statistical analyses of motor behavior were performed with SPSS 12.0.1 (SPSS Inc.) or GraphPad Prism software. Treatment groups and mouse lines were compared with either repeated measures ANOVA or one-way ANOVA, followed by Newman-Keuls post hoc test or Dunnett’s test. In all statistical tests, the significance level was set at P < 0.05.

Supplementary Material

Supplementary Figures and Supplementary Methods


We thank E. Sigel for pointing out the γ2I77 mutation; J. Oberdick for the L7 expression cassette; S.J. Moss for the γ2F77GFP plasmid; M. Meyer for the L7Cre line; R. Tomioka and E. Mugnaini for antibodies to EGFP and GAD, respectively; H. Monyer for discussion and support; F. Zimmermann for the transgene and stem cell injections; I. Preugschat-Gumprecht for help with mouse genotyping; D. Andersson, T. Karayannis and M. Capogna for contributing to initial electrophysiological recordings; and S. Brickley, M. Capogna, S.G. Cull-Candy, C. De Zeeuw, N. Franks, T. Klausberger and Z. Nusser for comments on the manuscript. This work was funded by the VolkswagenStiftung (grant I/78 554 to W.W., E.R.K., W.S. and P.S.), the Deutsche Forschungsgemeinschaft (grant WI 1951/2 to W.W. and P.W.), the UK Medical Research Council (grant G0501584 to W.W.), the J. Ernest Tait Estate (to W.W. and T.G.), a Heidelberg Young Investigator Award (to P.W.), the Academy of Finland (to E.R.K. and A.-M.L.), the Sigrid Juselius Foundation (to E.R.K. and E.L.), the Institute Pasteur-Fondazione Cenci Bolognetti (to M.R.), the Austrian Federal Government (W.S.), the Medical University Vienna (W.S.) and a Wellcome Trust Programme Grant (to M.F.).



The authors declare no competing financial interests.


1. Riedel G, et al. Reversible neural inactivation reveals hippocampal participation in several memory processes. Nat. Neurosci. 1999;2:898–905. [PubMed]
2. Lomber SG. The advantages and limitations of permanent or reversible deactivation techniques in the assessment of neural function. J. Neurosci. Methods. 1999;86:109–117. [PubMed]
3. Pereira de Vasconcelos A, et al. Reversible inactivation of the dorsal hippocampus by tetrodotoxin or lidocaine: a comparative study on cerebral functional activity and motor coordination in the rat. Neuroscience. 2006;141:1649–1663. [PubMed]
4. Wulff P, Wisden W. Dissecting neural circuitry by combining genetics and pharmacology. Trends Neurosci. 2005;28:44–50. [PubMed]
5. Crick F. The impact of molecular biology on neuroscience. Phil. Trans. R. Soc. Lond. B. 1999;354:2021–2025. [PMC free article] [PubMed]
6. Rudolph U, Mohler H. Analysis of GABAA receptor function and dissection of the pharmacology of benzodiazepines and general anesthetics through mouse genetics. Annu. Rev. Pharmacol. Toxicol. 2004;44:475–498. [PubMed]
7. Farrant M, Nusser Z. Variations on an inhibitory theme: phasic and tonic activation of GABAA receptors. Nat. Rev. Neurosci. 2005;6:215–229. [PubMed]
8. Whiting PJ. GABAA receptors: a viable target for novel anxiolytics? Curr. Opin. Pharmacol. 2006;6:24–29. [PubMed]
9. Perrais D, Ropert N. Effect of zolpidem on miniature IPSCs and occupancy of postsynaptic GABAA receptors in central synapses. J. Neurosci. 1999;19:578–588. [PubMed]
10. Thomson AM, Bannister AP, Hughes DI, Pawelzik H. Differential sensitivity to Zolpidem of IPSPs activated by morphologically identified CA1 interneurons in slices of rat hippocampus. Eur. J. Neurosci. 2000;12:425–436. [PubMed]
11. Campo-Soria C, Chang Y, Weiss DS. Mechanism of action of benzodiazepines on GABAA receptors. Br. J. Pharmacol. 2006;148:984–990. [PMC free article] [PubMed]
12. Rudolph U, et al. Benzodiazepine actions mediated by specific γ-aminobutyric acidA receptor subtypes. Nature. 1999;401:796–800. [PubMed]
13. Crestani F, Martin JR, Mohler H, Rudolph U. Mechanism of action of the hypnotic zolpidem in vivo. Br. J. Pharmacol. 2000;131:1251–1254. [PMC free article] [PubMed]
14. Crestani F, Martin JR, Mohler H, Rudolph U. Resolving differences in GABAA receptor mutant mouse studies. Nat. Neurosci. 2000;3:1059. [PubMed]
15. McKernan RM, et al. Sedative but not anxiolytic properties of benzodiazepines are mediated by the GABAA receptor α1 subtype. Nat. Neurosci. 2000;3:587–592. [PubMed]
16. Wisden W, Laurie DJ, Monyer H, Seeburg PH. The distribution of 13 GABAA receptor subunit mRNAs in the rat brain. I. Telencephalon, diencephalon, mesencephalon. J. Neurosci. 1992;12:1040–1062. [PubMed]
17. Duncan GE, et al. Distribution of [3H]zolpidem binding sites in relation to messenger RNA encoding the α1, β2 and γ2 subunits of GABAA receptors in rat brain. Neuroscience. 1995;64:1113–1128. [PubMed]
18. Niddam R, Dubois A, Scatton B, Arbilla S, Langer SZ. Autoradiographic localization of [3H]zolpidem binding sites in the rat CNS: comparison with the distribution of [3H]flunitrazepam binding sites. J. Neurochem. 1987;49:890–899. [PubMed]
19. Sancar F, Ericksen SS, Kucken AM, Teissere JA, Czajkowski C. Structural determinants for high-affinity zolpidem binding to GABAA receptors. Mol. Pharmacol. 2007;71:38–46. [PubMed]
20. Buhr A, Baur R, Sigel E. Subtle changes in residue 77 of the γ subunit of α1β2γ2 GABAA receptors drastically alter the affinity for ligands of the benzodiazepine binding site. J. Biol. Chem. 1997;272:11799–11804. [PubMed]
21. Wingrove PB, Thompson SA, Wafford KA, Whiting PJ. Key amino acids in the γ subunit of the γ-aminobutyric acidA receptor that determine ligand binding and modulation at the benzodiazepine site. Mol. Pharmacol. 1997;52:874–881. [PubMed]
22. Cope DW, et al. Abolition of zolpidem sensitivity in mice with a point mutation in the GABAA receptor γ2 subunit. Neuropharmacology. 2004;47:17–34. [PubMed]
23. Cope DW, et al. Loss of zolpidem efficacy in the hippocampus of mice with the GABAA receptor γ2 F77I point mutation. Eur. J. Neurosci. 2005;21:3002–3016. [PubMed]
24. Ogris W, et al. Affinity of various benzodiazepine site ligands in mice with a point mutation in the GABAA receptor γ2 subunit. Biochem. Pharmacol. 2004;68:1621–1629. [PubMed]
25. Smeyne RJ, et al. Local control of granule cell generation by cerebellar Purkinje cells. Mol. Cell. Neurosci. 1995;6:230–251. [PubMed]
26. Häusser M, Clark BA. Tonic synaptic inhibition modulates neuronal output pattern and spatiotemporal synaptic integration. Neuron. 1997;19:665–678. [PubMed]
27. Mittmann W, Koch U, Häusser M. Feed-forward inhibition shapes the spike output of cerebellar Purkinje cells. J. Physiol. (Lond.) 2005;563:369–378. [PubMed]
28. Santamaria F, Tripp PG, Bower JM. Feed-forward inhibition controls the spread of granule cell-induced Purkinje cell activity in the cerebellar cortex. J. Neurophysiol. 2007;97:248–263. [PubMed]
29. Laurie DJ, Seeburg PH, Wisden W. The distribution of 13 GABAA receptor subunit mRNAs in the rat brain. II. Olfactory bulb and cerebellum. J. Neurosci. 1992;12:1063–1076. [PubMed]
30. Fritschy JM, Panzanelli P, Kralic JE, Vogt KE, Sassoe-Pognetto M. Differential dependence of axo-dendritic and axo-somatic GABAergic synapses on GABAA receptors containing the α1 subunit in Purkinje cells. J. Neurosci. 2006;26:3245–3255. [PubMed]
31. Gunther U, et al. Benzodiazepine-insensitive mice generated by targeted disruption of the γ2 subunit gene of γ-aminobutyric acid type A receptors. Proc. Natl. Acad. Sci. USA. 1995;92:7749–7753. [PubMed]
32. Schweizer C, et al. The γ2 subunit of GABAA receptors is required for maintenance of receptors at mature synapses. Mol. Cell. Neurosci. 2003;24:442–450. [PubMed]
33. Barski JJ, Dethleffsen K, Meyer M. Cre recombinase expression in cerebellar Purkinje cells. Genesis. 2000;28:93–98. [PubMed]
34. Kittler JT, et al. Analysis of GABAA receptor assembly in mammalian cell lines and hippocampal neurons using γ2 subunit green fluorescent protein chimeras. Mol. Cell. Neurosci. 2000;16:440–452. [PubMed]
35. Brickley SG, Revilla V, Cull-Candy SG, Wisden W, Farrant M. Adaptive regulation of neuronal excitability by a voltage-independent potassium conductance. Nature. 2001;409:88–92. [PubMed]
36. Marder E, Goaillard JM. Variability, compensation and homeostasis in neuron and network function. Nat. Rev. Neurosci. 2006;7:563–574. [PubMed]
37. Lima SQ, Miesenbock G. Remote control of behavior through genetically targeted photostimulation of neurons. Cell. 2005;121:141–152. [PubMed]
38. Zhang F, et al. Multimodal fast optical interrogation of neural circuitry. Nature. 2007;446:633–641. [PubMed]
39. Gosgnach S, et al. V1 spinal neurons regulate the speed of vertebrate locomotor outputs. Nature. 2006;440:215–219. [PubMed]
40. Tan EM, et al. Selective and quickly reversible inactivation of mammalian neurons in vivo using the Drosophila allatostatin receptor. Neuron. 2006;51:157–170. [PubMed]
41. Karpova AY, Tervo DG, Gray NW, Svoboda K. Rapid and reversible chemical inactivation of synaptic transmission in genetically targeted neurons. Neuron. 2005;48:727–735. [PubMed]
42. Yamamoto M, et al. Reversible suppression of glutamatergic neurotransmission of cerebellar granule cells in vivo by genetically manipulated expression of tetanus neurotoxin light chain. J. Neurosci. 2003;23:6759–6767. [PubMed]
43. Wall MJ, Usowicz MM. Development of action potential-dependent and independent spontaneous GABAA receptor-mediated currents in granule cells of postnatal rat cerebellum. Eur. J. Neurosci. 1997;9:533–548. [PubMed]
44. McCartney MR, Deeb TZ, Henderson TN, Hales TG. Tonically active GABAA receptors in hippocampal pyramidal neurons exhibit constitutive GABA-independent gating. Mol. Pharmacol. 2007;71:539–548. [PubMed]
45. Benavides J, et al. In vivo interaction of zolpidem with central benzodiazepine (BZD) binding sites (as labeled by [3H]Ro 15-1788) in the mouse brain. Preferential affinity of zolpidem for the omega 1 (BZD1) subtype. J. Pharmacol. Exp. Ther. 1988;245:1033–1041. [PubMed]
46. Leppa E, et al. Agonistic effects of the β-carboline DMCM revealed in GABAA receptor γ2 subunit F77I point-mutated mice. Neuropharmacology. 2005;48:469–478. [PubMed]
47. Lein ES, et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature. 2007;445:168–176. [PubMed]
48. Zezula J, Fuchs K, Sieghart W. Separation of α1, α2 and α3 subunits of the GABAA-benzodiazepine receptor complex by immunoaffinity chromatography. Brain Res. 1991;563:325–328. [PubMed]
49. Oertel WH, Schmechel DE, Mugnaini E, Tappaz ML, Kopin IJ. Immunocytochemical localization of glutamate decarboxylase in rat cerebellum with a new antiserum. Neuroscience. 1981;6:2715–2735. [PubMed]
50. Korpi ER, et al. Cerebellar granule-cell-specific GABAA receptors attenuate benzodiazepine-induced ataxia: evidence from α6-subunit-deficient mice. Eur. J. Neurosci. 1999;11:233–240. [PubMed]