We used quantitative electron-microscopic analysis of immunogold labelling to investigate translocation of GABAA receptors in HEK 293 cells stably expressing the α1, β2 and γ2 subunits of rat GABAA receptors9, the most common subunit combination of native GABAA receptors in the mammalian CNS10. We found that, under control conditions, only a small number of immunogold-labelled β2 subunits could be observed at the membrane surface, and that most labelling was localized intracellularly (Fig. 1a,1). Treatment with insulin dramatically increased the number of gold-labelled GABAA receptor subunits in the plasma membrane (Fig.1b,1); the change could be detected as early as 10 min after insulin treatment (data not shown). An increase in receptor synthesis was not involved, as even 60 min of insulin treatment did not affect the total cellular expression of β2 subunits (Fig. 1c). Pretreatment of cells with genistein, a membrane-permeable inhibitor of protein-tyrosine kinases11,12, prevented the insulin-induced translocation without altering basal GABAA receptor distribution (Fig.1c), suggesting a requirement for activation of the insulin receptor tyrosine kinase13. To determine whether the insulin-induced receptor translocation leads to an increase in the number of functional GABAA receptors on the plasma-membrane surface, we recorded GABAA receptor-mediated whole-cell currents in these cells. Bath application of insulin (0.5 µM) produced a significant increase in the amplitude of the currents, an effect not observed in cells pretreated with genistein (Fig. 1d). These results demonstrate a rapid increase in the number of functional GABAA receptors on the plasma-membrane surface as a result of receptor translocation by activation of insulin-receptor tyrosine kinase.

Figure 1: Insulin induces a rapid translocation of GABAA receptors from intracellular compartments to the plasma membrane and potentiates.
figure 1

GABAA receptor-mediated currents in HEK 293 cells stably transfected with cDNAs encoding rat GABAA receptor α1, β2 and γ2 subunits. Electron micrographs of ultrathin cryosections labelled with immunogold using a monoclonal antibody against the rat β2/β3 GABAA receptor subunits show subcellular distribution of the β2 subunits in control (a) and after a 60-min insulin (0.5 µM) stimulation (b). Immunogold-labelled receptor proteins are indicated by arrowheads; scale bars, 0.1 µm. c, Quantification of membrane and total cellular expression of β2 subunits. Cells were incubated for 60 min in the extracellular solution (control) or in solution supplemented with either 0.5 µM insulin or 50 µM genistein (Gen.). For genistein pretreatment (Gen. + insulin), cells were first treated with 50 µM genistein for 10 min, and then incubated for 60 min in the genistein solution supplemented with 0.5 µM insulin. Values are means ± s.e. from 3 experiments (25 cells per experiment) and were analysed using the unpaired Student t-test (** denotes P < 0.01). d, Recordings of representative individual GABAA receptor-mediated whole-cell currents induced by puffing GABA (100 µM, 100 ms) onto the transfected cells at a holding membrane potential of −60 mV before (control) and 10 min after addition of 0.5 µM insulin. For genistein treatment, cells were incubated in 50 µM genistein before the start of whole-cell recording and exposed to the drug throughout recording period. Insulin increased the amplitudes of the currents by 40% (−376.1 ± 106.6 pA for control versus −526.3 ± 145.3 pA 10 min after insulin, P < 0.05, n = 8), and this effect was prevented by genistein pretreatment (−263 ± 83.7 versus −261 ± 81.6 pA; P > 0.05, n = 6).

To determine which subunit(s) is required for insulin-induced receptor translocation, HEK 293 cells were transiently transfected with rat GABAA receptor β2, γ2 and the epitope-tagged α1FLAGsubunit complementary DNAs in various combinations. Confocal scanning sections double-stained for surface and for entire cell staining demonstrated that a large proportion of α1β2 transfectants was surface-labelled (Fig. 2). However, among cells transfected with α1γ2, or α1 or β2 alone, few were surface-labelled. Insulin (0.5 µM, 10 min) increased the proportion of surface-labelled cells in α1β2 and β2 by 29% and 103%, respectively, but not in α1γ2 or α1 transfected cells (Fig. 2). Thus the β2 subunit seems to be required for the insulin-induced increase in receptor surface expression.

Figure 2: The β-subunit is required for insulin-mediated translocation of recombinant GABAA receptors transiently expressed in HEK 293 cells.
figure 2

A, B, Confocal optical sections (1 µm) of cells transfected with α1FLAGβ2 (A) or α1FLAGγ2 (B) before (a and b) and after (c and d) 10-min insulin treatment (0.5 µM). Double-immunofluorescence staining was done by first staining (under non-permeabilized conditions) for the FLAG epitope tag inserted in the extracellular domain (near the N terminus) of α1 (b and d, red) followed by permeabilized staining for the main intracellular loop of the same subunit (a and c, green). C, Effects of insulin on cell surface expression. For cells transfected with α1FLAG, α1FLAGγ2 or α1FLAGβ2 subunit combinations, the rate of surface expression was calculated as the number of surface-labelled cells (labelled with anti-Flag antibody) as a percentage of total transfected cells (labelled with polyclonal anti-α antibody). For cells transfected with the β-subunit alone, surface-labelled cells and transfected cells were identified by staining with antibody to β2/β3 subunits under non-permeabilized and permeabilized conditions, respectively. Each combination was repeated in 3 experiments, and 500–1,000 transfected cells were counted in each combination in each experiment.

We next investigated whether insulin can effect the translocation of native CNS GABAA receptors in situ in cultured mouse hippocampal neurons. When native GABAA receptors were stained under permeabilized conditions with anti-β2/β3 antibody, basal levels of plasma membrane-associated receptors were observed in all cells. However, large numbers of the receptors were found to be intracellularly localized through the cell soma (Fig. 3A,3). Insulin (0.5 µM, 10 min) increased the ratio of the receptor labelling on the plasma membrane to that in the cytoplasmic compartment (Fig. 3A, 3 and 3), providing evidence for translocation of intracellularly localized receptors to the plasma membrane. Optical sections stained under non-permeabilized conditions demonstrated that insulin increased the labelling on the plasma-membrane surface (Fig. 3B), and that this increase in the surface labelling occurred preferentially on the proximate dendrites (Fig. 3B, 3 and 3), sites where GABAergic synapses have been shown to be predominantly localized14. These results are consistent with a recruitment by insulin of GABAA receptors to the postsynaptic domains near the sites of presynaptic release. This possibility was further examined, using electron microscopy, in the CA1 neurons of control and insulin-treated hippocampal slices. Under control conditions, most immunogold-labelled β2/β3 subunits were found to be located along synaptic, dendritic and somatic membranes, with the highest concentration of receptors in the postsynaptic side of the synaptic junctions, consistent with the synaptic enrichment of the GABAA receptors previously reported15. Treatment of slices with insulin (0.5 µM, 10 min) significantly increased the receptor density on both postsynaptic and dendritic membranes, but not on the somatic membrane (Fig. 3C and 3, 3). These results demonstrate that recruitment of GABAA receptors to the postsynaptic domains by insulin is not unique to neurons in primary culture, but can also occur in mature neurons in situ.

Figure 3: Insulin causes membrane translocation and clustering of native GABAA receptors in CNS neurons.
figure 3

Control (a) and insulin-treated (b; 0.5 µM, 10 min) cultured mouse hippocampal neurons (AC) were stained with the β2/β3 monoclonal antibody. A, Optical sections (1 µm) of the cell soma stained under permeabilized conditions show that insulin induces a relocation of intracellular GABAA receptors to the plasma membrane. B, Optical sections scanned through the cell soma show that, under non-permeable conditions, GABAA receptors expressed on the membrane surface are selectively labelled and insulin increases the surface labelling (a and b). The same neurons shown in a and b are below in stacked serial optic sections of the entire cell body. Under basal conditions (c), numerous small clusters of GABAA receptors are distributed diffusely over the entire cell, including remote processes. After insulin treatment (d), receptors are concentrated in large clusters on the proximate dendrites. Scale bars, 10 µm for A and B. C, Insulin increases synaptic GABAA receptor density of mature CNS neurons in situ. Electron micrographs show immunogold-labelled GABAA receptor β2/β3 subunits in the postsynaptic junctions (arrowheads) in the hippocampal CA1 region from a control slice (a) and a slice treated with 0.5 µM insulin for 10 min at 35 °C (b). Scale bars, 0.1 µm. D, Effect of insulin on surface expression of native GABAA receptors. a, Quantification of the ratio of fluorescence intensity of β2/β3 subunit staining on the membrane to that in cytoplasm from cultured neurons stained under permeabilized conditions (n = 25 neurons). b, Quantification of fluorescence intensity on the membrane surface from cultured neurons stained under non-permeabilized conditions (n = 45 neurons). c, Quantification of immunoparticle densities associated with the β2/β3 subunits of native GABAA receptors in mature hippocampal CA1 neurons in situ, categorized by membrane region; n = 25, 25 and 100 for somatic, dendritic and synaptic membranes, respectively. * P < 0.05, ** 0.01 and *** 0.001.

To determine whether the newly recruited surface receptors are functional, we examined effects of insulin on currents evoked by exogenously applied GABA in cultured neurons. Bath application of insulin (0.5 µM) potentiated the currents at both near- and sub-saturated GABA concentrations (26.4 ± 5.3% at 200 µM, 15.6 ± 7.9% at 10 µM, n = 5 neurons; Fig. 4A). Thus insulin increases the maximum response, providing support for an increase in the number of functional membrane GABAA receptors. To provide evidence for recruitment of functional GABAA receptors at postsynaptic junctions, we also investigated modulation by insulin of GABAA receptor-mediated mIPSCs. Within 10 min, insulin caused an increase of approximately 30% in the amplitude of the mIPSCs (−37 ± 4.2 control versus −47 ± 2.9 pA after insulin, P < 0.01, n = 5; Fig. 4B, 4), but caused no change in either the rise time (10% to 90% of the peak, 2.7 ± 0.1 versus 2.6 ± 0.2 ms, P > 0.05; Fig. 4B, 4) or the decay time constant (τ = 28.3 ± 3 versus 26.8 ± 2 ms, P > 0.05; Fig. 4B, 4). Thus the results are consistent with an increased number of active postsynaptic receptors. Insulin was also found to increase mIPSC frequency (Fig. 4B, 4); although this could possibly be due to a higher rate of quantal release16,17, it could also be fully accounted for by the recruitment of mIPSCs, previously below the threshold of detection, owing to an increase in the number of active receptors18. Finally, in CA1 hippocampal neurons in brain slices prepared from adult rats, insulin enhanced the amplitude of the mIPSCs by 23% (20.1 ± 1.9 versus 24.6 ± 2.8 pA, P < 0.01, n = 4) but not their time course (τ = 21.2 ± 2.6 versus 22.6 ± 4.1 ms, P > 0.05). These results suggest that the insulin-induced increase in the number of functional postsynaptic GABAA receptors occurs in mature neurons in situ, indicating that the receptor relocation is likely to be of physiological importance.

Figure 4: Potentiation of GABAA receptor-mediated current responses by insulin in cultured hippocampal neurons.
figure 4

A, Insulin potentiates whole-cell GABA currents in perforated patch configuration at a holding membrane potential of −60 mV. Currents were induced in the same neuron by alternate perfusion of GABA at saturated and non-saturated concentrations through a multi barrel fast perfusion system. The doses were determined from full dose–response curves constructed for three neurons under control conditions; 10 and 200 µM are doses producing approximately half and maximum responses, respectively (data not shown). a, Superimposed sample current traces before and 10 min after 0.5 µM insulin application. b, Normalized peak currents (Iinsulin/Icontrol) from 5 individual neurons. Insulin was applied in the bath during the period indicated by the bar. B, Whole-cell recordings of GABAA receptor-mediated mIPSCs at a holding potential of −60 mV before (control) and 8 min after addition of insulin (0.5 µM). a, Examples of consecutive traces of mIPSCs. b, Averages of 100 individual mIPSCs. The decay of mIPSCs before and after insulin were well fitted by a single exponential. c, Histograms showing amplitude distributions of a 3-min recording of mIPSCs. The histograms were fitted using an equation described previously30. Bath application of insulin for 5–10 min increased the amplitude of mIPSCs without altering their time course. Note that insulin also increased the frequency of mIPSCs (0.94 ± 0.17 versus 1.6 ± 0.25 events s−1, n = 5, P < 0.01)

Our results have demonstrated that insulin can regulate the surface expression of GABAA receptors and thereby modulate GABAA-receptor-mediated synaptic inhibition. Insulin and insulin receptors are expressed at high levels in discrete regions within the CNS, and insulin is released by depolarization in cultured CNS neurons13. Insulin has been shown to act as a neuromodulator of many brain functions, such as food intake13. The regulation of food intake by brain insulin may be mediated in part through GABAA receptors, as insulin-induced hyperphagia in freely moving rats can be blocked by antagonists of GABAA receptors applied in the ventromedial hypothalamic region19. The translocation of GABAA receptors by insulin may also be important outside the CNS. In pancreatic islet, GABA is expressed at high levels in insulin-secreting β-cells20, and functional GABAA receptors are present in glucagon-secreting α-cells21. It has been reported that glucose inhibition of glucagon secretion is mediated through the activation of GABAA receptors on the α-cells, and this does not seem to be associated with any detectable increase in GABA release21. Because secretion of insulin in the islet increases in the presence of high concentrations of glucose, our results suggest that insulin may mediate glucose feedback regulation of glucagon release by increasing the activity of GABAA receptors in α-cells. Thus the observed modulation of GABAA receptors by insulin may be important for the regulation of food intake and glucose homoeostasis under physiological conditions, as well as in diseases such as diabetes and obesity.

Receptor translocation has important implications for synaptic function. Evidence from both experimental and modelling studies supports the view that GABA released from a single vesicle nearly saturates postsynaptic GABAA receptors at CNS synapses3,6,7. Therefore, an increase in quantal transmitter release would have little effect on the amplitude of the inhibitory synaptic potential. In contrast, the insertion of additional functional GABAA receptors into the postsynaptic membrane would serve as an effective means of enhancing inhibitory synaptic transmission. Thus the rapid insulin-induced translocation of GABAA receptors to the postsynaptic domain of inhibitory synapses provides a mechanism for the generation of synaptic plasticity in these synapses22, and may also have implications for the proposed recruitment of silent synapses following the long-term potentiation of glutamate-receptor-mediated synaptic transmission18,23,24,25.


Immunogold staining and electron-microscopic examination. HEK cells stably transfected with cDNAs encoding the α1, β2 and γ2 subunits of rat GABAA receptors (4D4 cells)9 were provided by D. Carter. Transcription of all three subunit mRNAs in the cell line was confirmed by reverse transcription–polymerase chain reaction (RT–PCR) (data not shown). Cells of passages 14–19 were plated in 100-mm culture dishes and used at 80% confluence. Cells were rinsed in PBS and incubated in the extracellular solution (control) or in the extracellular solution supplemented with 0.5 µM insulin for 10–60 min. Vibratome hippocampal slices (300 µm) were prepared from adult Sprague-Dawley rats. After a recovery period of 1 h artificial cerebrospinal fluid (ACSF)26 at 35 °C, slices were incubated in ACSF containing 0.5 µM insulin for 10 min. After treatment, cells or slices were immediately fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.2) for 1 h. Pelleted cells were then embedded in gelatin. Embedded cells or dissected CA1 region of the slices were infused with 2.3 M sucrose for several hours and frozen in liquid nitrogen. Sections (100 nm) were then cut and immunogold-labelled using the monoclonal antibody bd-17 against rat β2/β3 subunits and 10-nm gold particles conjugated to goat anti-mouse IgG. In some cases, primary antibody was omitted, or replaced with a monoclonal antibody (bd-24) that recognizes human, but not rat, GABAA-receptor α-subunit27, to show the specificity of β2/β3 labelling. Non-transfected 293 cells were also stained with anti-β2/β3 antibody as a negative control. Images were captured with a CCD camera in the transmission electron microscope, and gold particles were quantified using the morphometric program NIH Image 1.59.

Cell transfection and neuronal cultures. The rat GABAA-receptor subunit cDNAs, pcDNAI-β2 and the pcDNA3-γ2 (short form of γ2), were provided by C. Kaufman and D. Gunnersen (Laboratory of Neuroscience, NIDDK). The Flag-tagged rat α1 subunit cDNA (pcDNA3-α1FLAG) was provided by J. H. Steinbach. This was generated by inserting the Flag sequence (KDYKDDDDKL) in the extracellular N-terminal domain between amino acids 33 and 34 of the translated sequence. This epitope tag did not alter either expression or function of the recombinant GABAA receptor, and the tagged α1 subunit protein was recognized on the surface of intact cells using the M2 monoclonal anti-FLAG antibody28. HEK 293 cells were plated onto collagen-coated 22-mm glass coverslips set in standard 35-mm culture dishes and transfected with 2 µg of each GABAA-receptor subunit plasmid using Lipofectamine (GIBCO) according to the manufacturer's protocol. Cultured mouse hippocampal neurons were grown as previously described29.

Immunofluorescent confocal microscopy. HEK 293 cells at 48 h post-transfection or neurons following 8–20 days in culture were fixed with 4% paraformaldehyde in PBS for 10 min and permeabilized with 0.25% Triton X-100 in PBS for 10 min, or fixed without permeabilization. GABAA receptors were stained with bd-17 monoclonal antibody recognizing rat β2/β3 subunits (15 µg ml−1) and with FITC-conjugated secondary antibody. For double-labelling experiments, cells were first stained with monoclonal anti-FLAG epitope (16 µg ml−1) and rhodamine-conjugated anti-mouse antibody without permeabilization. After a further 10 min fixation and 10 min permeabilization, cells were stained with a polyclonal antibody recognizing the rat α-subunit (1 : 200) and with FITC-conjugated anti-rabbit antibody. Specific staining was tested by either replacing the monoclonal primary antibody with an unrelated antibody, the monoclonal anti-α1 subunit of the human GABAA receptor (clone bd-24)27, or by omitting the primary antibody in the case of the polyclonal antibody. Subcellular localization of immunofluorescently labelled receptors was examined with a Leica TCS-4D confocal microscope, and fluorescence intensity was quantified using the Scion ImagePC software.

Electrophysiology. Whole-cell recordings of stably transfected HEK 293 cells (4D4 cells) and cultured neurons were made using an Axopatch-1B or -1D amplifier (Axon Instruments, Foster City, CA) at room temperature (20–22 °C). Culture medium was exchanged completely with standard extracellular solution 1–2 h before recording. The extracellular solution contained (in mM): 140 NaCl, 1.3 CaCl2, 5.4 KCl, 1 MgCl2, 25 HEPES, 33 glucose, 0.0005 TTX (pH 7.4 using NaOH, 320–335 mosm). The intracellular solution contained (in mM): 120 CsCl, 35 CsOH, 2 MgCl2, 1 CaCl2, 11 EGTA, 2 TEA, 10 HEPES, 4 ATP, pH 7.3. For perforated-patch recording, 100 µg ml−1nystatin (Sigma) was added to the intracellular solution. Standard whole-cell recordings of CA1 neurons were performed in adult rat hippocampal slices perfused with ACSF at 35 °C in a submerged recording chamber. For recording of mIPSCs, 2 mM Mg2+(plus AP5 for neurons in slices) and 20 µM CNQX were added to the extracellular solution to inhibit NMDA and AMPA receptor-mediated excitatory synaptic currents. The mIPSCs were abolished by the addition of 10 µM GABAA receptor antagonist bicuculine. mIPSCs were first recorded on videotape and later acquired and analysed off-line using the SCAN program (Strathclyde Software).

Drugs and materials. Insulin (purified pancreatic bovine, Zn2+-free) was provided by C. C. Yip (Banting & Best Institute, Toronto, Canada). Monoclonal antibody bd-17, which recognizes extracellular epitopes on both β2 and β3 subunits of the rat GABAA receptor27, was purchased from Boehringer Mannheim Biochem (Canada). The polyclonal antibody against the main intracellular loop between transmembrane domains III and IV of rat α1 subunit of the GABAA receptor was purchased from PharMingen (San Diego). The monoclonal antibody to the Flag epitope (M2) was from Eastman Kodak Scientific Imaging System (New Haven).