Long-term depression (LTD) is an activity-dependent weakening of synaptic efficacy at individual inhibitory synapses, a possible cellular model of learning and memory. Here, we show that the induction of LTD of inhibitory transmission recruits activated calcineurin (CaN) to dephosphorylate type-A GABA receptor (GABAARs) via the direct binding of CaN catalytic domain to the second intracellular domain of the GABAAR-γ2subunits. Prevention of the CaN–GABAA receptor complex formation by expression of an autoinhibitory domain of CaN in the hippocampus of transgenic mice blocks the induction of LTD. Conversely, genetic expression of the CaN catalytic domain in the hippocampus depresses inhibitory synaptic responses, occluding LTD. Thus, an activity-dependent physical and functional interaction between CaN and GABAA receptors is both necessary and sufficient for inducing LTD at CA1 individual inhibitory synapses.
In CA1 neurons of the hippocampus, field stimulation of Schaffer-collateral fibers evokes diphasic excitatory–inhibitory synaptic currents: a fast EPSC mediated by AMPA receptors followed by a fast IPSC (Lu et al., 2000). A brief high-frequency stimulation (tetanus) of the Schaffer-collateral fibers produces long-term potentiation (LTP) of EPSCs and concomitantly long-term depression (LTD) of IPSCs (Andersen and Lømo, 1968; Stelzer et al., 1987; Lu et al., 2000). These coordinately regulated bidirectional changes of the excitatory and inhibitory synaptic strength are considered to be a cellular model of learning and memory (Bliss and Collingridge, 1993; Paulsen and Moser, 1998). Although we now know a great deal about the molecular steps contributing to the induction of LTP at excitatory synapses (Malenka and Nicoll, 1999;Malinow et al., 2000; Ali and Salter, 2001; Lisman and Zhabotinsky, 2001), the molecular mechanisms underlying the induction of LTD at CA1 individual inhibitory synapses have not been identified (Abraham et al., 1987; Thompson and Gahwiler, 1989; Thompson, 1994; Aizenman et al., 1998).
Fast IPSCs in CA1 neurons are mediated predominantly by type-A GABA receptors (GABAARs) (MacDonald and Olsen, 1994). The induction of LTD of GABAAreceptor-mediated IPSCs (GABAAR-IPSCs) requires activation of excitatory NMDA receptors (Stelzer et al., 1987). The biochemical links between the NMDA receptors and the pathway for inducing LTD of GABAAR–IPSCs involve calcineurin (CaN) (Lu et al., 2000). CaN, or Ca2+/calmodulin-dependent phosphatase 2B, consists of a 61 kDa catalytic domain (CaN-A) and a 19 kDa regulatory subunit (CaN-B) (Cohen, 1989). The enzymatic activity of CaN in CA1 neurons of the hippocampus is increased by activation of NMDA receptors (Lu et al., 2000). Also, there is extensive evidence showing that CaN plays a critical role in the activity-dependent changes of excitatory synaptic transmission in the hippocampus (Winder and Sweatt, 2001). Genetic inhibition of endogenous CaN in the forebrain (CaN transgenic mice), by expressing the auto-inhibitory domain of CaN, showed that LTP at CA1 excitatory synapses induced by subsaturating but not saturating tetanic stimulation was enhanced both in vitro and in vivo (Malleret et al., 2001). In another line of experiments, transgenic mice (CN98 transgenic mice) were generated in which a catalytic domain of CaN (CaN-Aα) was expressed in the forebrain (Winder et al., 1998). In these animals, NMDA receptor-dependent LTP induced by weak tetanus in the CN98 mutant mice was no different from that seen in control mice, but LTP elicited by stronger tetanus was reduced in CN98 mutant mice. These data suggest that manipulations of CaN activity alter the ability of CA1 excitatory synapses to induce LTP.
Because CaN downregulates GABAAreceptor function (Chen and Wong, 1995), activated CaN may interact with the synaptic GABAA receptors for NMDA receptor-dependent LTD at CA1 inhibitory synapses. A direct test of this hypothesis has not yet been undertaken. Thus, we used double whole-cell patch-clamp recordings to induce LTD at CA1 individual inhibitory synapses. By taking advantage of a combined molecular genetic and biochemical approach, we demonstrate that NMDA receptor-dependent physical and functional interaction between CaN-A and GABAAR-γ2 subunit (GABAARγ2S) fulfills necessary and sufficient conditions for inducing LTD of inhibitory transmission.
Materials and Methods
Double whole-cell patch-clamp recordings and LTD induction. Hippocampal slices (300 μm) were prepared from 30 ±2-d-old CaN transgenic control or mutant mice for Figures 1-4 and from 34 ±3-d-old CN98 transgenic mice for Figures 6 and 7. CaN mutant mice were generated by expressing the autoinhibitory domain in the C terminus of CaNα in the forebrain with the doxycycline-dependent reverse tetracycline-controlled transactivator system, resulting in a 35–45% decrease in CaN activity. For all experiments, doxycycline (Mutual Pharmaceutical, Philadelphia, PA) was administered at 6 mg/g food at least 1 week before experimentation. Control mice were treated with dox only, as described in detail previously (Malleret et al., 2001). The slices were prepared as described previously (Lu et al., 1998, 2000). All procedures were in compliance with and approved by the University Animal Care and Use Committee, University of Calgary. For double whole-cell patch-clamp recordings from CA1 interneuron and pyramidal cell pairs, hippocampal slices were visualized with infrared (IR) illumination and a differential interference contrast (DIC) Axioskop 2FS plus equipped with Hamamatsu C2400–07E optics (see Fig.1 A). A whole-cell recording (tight-seal >1 GΩ) with patch electrode (3–5 MΩ) was initially obtained from a CA1 interneuron at the border of stratum radiatum and lacunosum-moleculare (LM). Subsequently, the second whole-cell recordings (tight-seal >10 GΩ) were established from a CA1 pyramidal cell. The synaptic connections between CA1 interneurons and pyramidal cells were also identified by post hoc morphological analysis (see Fig.1 A). Single spikes in interneurons triggered unitary IPSCs in pyramidal cells that were blocked by the GABAA receptor antagonist bicuculline (10 μm). An extracellular stimulating electrode was placed at the CA1 Schaffer-collateral fibers. LTD was induced by tetanus (two 100 Hz stimuli lasting 1 sec at an intertrain interval of 10 sec) of the Schaffer-collateral fibers. The unitary GABAAR–IPSCs were filtered at 5 kHz with a low-pass filter. Data were digitized at a frequency of 10 kHz and stored on-line using the pclamp8 system. The input resistance and series resistance in postsynaptic pyramidal cells were monitored using prevoltage steps (−2 mV, 100 msec) at 5 min intervals throughout the period of the experiment. Series resistance ranged from 9 to 12 MΩ. Input resistance was 328 ± 29 MΩ. For current-clamp mode, the intracellular solution contained (in mm): 115 K+-gluconate, 7.5 K+Cl, 27.5 K+-methylsulfate, 10 HEPES, 0.2 EGTA, 2 Mg-ATP, 0.3 guanosine triphosphate, and 0.1% biocytin, pH 7.4, 296 mOsm. For voltage-clamp recordings, the low Cl− solution contained (in mm): 142.5 Cs-gluconate, 7.5 CsCl, 10 HEPES, 0.2 EGTA, 2 Mg-ATP, and 0.3 guanosine triphosphate, pH 7.4, 296 mOsm, and the high Cl− solution contained (in mm): 150 CsCl, 10 HEPES, 0.2 EGTA, 2 Mg-ATP, 0.3 guanosine triphosphate, and 0.1% biocytin, pH 7.4, 296 mOsm.
Coimmunoprecipitation, affinity purification (“pull-down”), and Western blotting. The CA1 region was microdissected as described previously (Lu et al., 1998). Four CA1 regions from control or LTD (5 or 30 min after induction of LTD) were pooled and homogenized in ice-cold lysis buffer containing 50 mmTris-HCl, pH 7.6, 150 mm NaCl, 1% NP-40, 2 mm EDTA, 1 mm sodium orthovanadate, and proteinase inhibitor mixture (Sigma, St. Louis, MO) (5 μl/100 mg tissue). After clearing debris by centrifuging at 14,000 × g at 4°C, protein concentration in the extracts was determined by Bradford assay (Bio-Rad, Hercules, CA). The extracts (∼500 μg protein) were incubated with nonspecific IgG (2 μg) or polyclonal mouse anti-CaN-A (2 μg; PharMingen, San Diego, CA) with or without 10 μg anti-CaN-A immunizing antigen (peptide I457-P482; PharMingen) overnight at 4°C, followed by the addition of 40 μl of Protein G-Sepharose (Sigma) for 3 hr at 4°C. In the immunoprecipitations with polyclonal rabbit antibody against GABAAR-α1 subunit (anti-α1, 2 μg; Upstate BioTechnology, Lake Placid, NY), the anti-α1 was cross-linked to the protein G-Sepharose. This previous cross-linking of the anti-α1 permits the elution of antigen only to prevent interference of the IgG subunits in blotting (He et al., 1995). Cross-linking was performed with 0.5% glutaraldehyde (30 min, 25°C). The reaction was then terminated and washed four times with lysis buffer. The extracts (∼500 μg protein) were then incubated with the cross-linked anti-α1 with or without 10 μg immunizing antigen (a peptide corresponding to residues 1–15 of the α1 subunit) (Upstate Biotechnology). Immunoprecipitates were washed four times with lysis buffer and denatured with SDS sample buffer and separated by 12% SDS-PAGE. Proteins were transferred onto nitrocellulose membranes using a Bio-Rad mini-protein-III wet transfer unit overnight at 4°C. Transfer membranes were then incubated with blocking solution [5% nonfat dried milk dissolved in TBST buffer (pH 7.5, 10 mm Tris-HCl, 150 mm NaCl, and 0.1% Tween 20)] for 1 hr at room temperature, washed three times, and incubated with monoclonal rabbit primary antibody against CaN-A (1:1000), polyclonal rabbit antibodies against GABAAR-α1 subunit (anti-α1, 1:1000; Alpha Diagnostic), GABAAR-β2 subunit (anti-β2, 1:1000; Alpha Diagnostic), or GABAAR-γ2 subunit (anti-γ2, 1:2000; Alpha Diagnostic) for 1 hr at room temperature. Membranes were washed three times with TBST buffer and incubated with the appropriate secondary antibodies (1:1000 dilution) for 1 hr followed by washing four times. Signal detection was performed with an enhanced chemiluminescence kit (Amersham Biosciences, Arlington, IL). The lanes marked “input” were loaded with 10% of the starting material used for immunoprecipitation. The precipitated bands were semiquantified using “the normalizing method” of the Densitometer Quantity One (see Quantity One User Guide, Bio-Rad). The intensities of the lanes marked input in each gel were normalized as 100%. Each of the other bands in the same gel was then expressed as the percentage of the respective input.
Glutathione S-transferase (GST) fusion proteins of GABAAR-α1 334–420 (GST-α1), -β2 327–451(GST-β2), -γ2S 317–442(GST-γ2S), -γ2S 332–442(*GST-γ2S), and -γ2S 317–332(short form) were prepared from bacterial lysates as described in detail previously (Liu et al., 2000). The extracts (∼200 μg of proteins) were incubated with the indicated GST fusion proteins (∼100 μg of protein), with or without 200 μg of γ2 peptide, overnight at 4°C, and then for another 3 hr at 4°C after 20 μl of Protein G-Sepharose (Sigma) was added. Beads were washed five times with lysis buffer. Eluted proteins were incubated in sample buffer (final concentration 5% SDS) and subjected to SDS-PAGE (12% gel). Transferred proteins were revealed by Western blot. In the experiments of Figure 3, E and F, the transferred membranes were incubated with monoclonal mouse primary antibody against α-adaptin (1:1000; CN Biosciences) and monoclonal mouse antibodies against Src (1:1000; Upstate Biotechnology) for 1 hr at room temperature. Signal detection was performed with an enhanced chemiluminescence kit (Amersham Biosciences).
In vitro binding assays.CaN420 (10 μg/ml) or CaN-B subunit (10 μg/ml) was incubated overnight at 4°C in 0.5 ml containing 40 mm Tris-HCl, pH 7.5, 0.5 mmCaCl2, 150 mm 2-mercaptoethanol, 0.2 mg/ml BSA, 40 μl glutathione-Sepharose beads (Pharmacia), and 10 μg GST-α1, GST-β2, GST-γ2S, or *GST-γ2S. In some assays, as indicated in Figure 3, 10 μg/ml γ2-peptide or scrambled γ2-peptide (Biosynthesis Inc.) was included. The amino acid sequence of scrambled γ2-peptide was LDHSYFKVNDRDKPKK; it was created by random ordering of the sequence of γ2-peptide, LHYFVSNRKPSKDKDK, corresponding to the 317–332 residues of the GABAA receptor γ2 subunit. The beads were washed five times with 200 μl PBS containing 0.1% Triton X-100 and eluted twice with 20 μl glutathione elution buffer. Eluted proteins were incubated in sample buffer (final concentration 5% SDS) and subjected to SDS-PAGE (12% gel). Transferred proteins were revealed by Western blot. Expression and purification of CaN420 and CaN-B were described in detail previously (Perrino et al., 1995).
The γ2 subunit phosphorylation assay. Rabbit polyclonal anti-γ2pSer327antibodies were raised against phospho-γ2peptide (317LHYFVSNRKP(p)SKDKDK332). The resulting antisera were then affinity purified with phospho-γ2 peptide immobilized on Affigel 10 (Bio-Rad). This antibody is competitively blocked with antigen peptide (see Fig. 5 B). The GABAA receptors were immunoprecipitated by previous cross-linked polyclonal rabbit anti-γ2from CA1 hippocampal extracts, as described above. Precipitated GABAA receptors and the in vitrophosphorylated GST-γ2S and the GST-γ2S mutant (Ser327-Ala) (see below) were subjected to SDS-PAGE (12% gel). Transferred proteins were incubated with rabbit polyclonal anti-γ2pSer327(1:500) for 1 hr at room temperature. Signal detection was performed with an enhanced chemiluminescence kit (Amersham Biosciences).
The GST-α1, GST-β2, GST-γ2S, or GST-γ2Smutant (Ser327-Ala) or GST alone (200 μg/each), was incubated with 1 μg/ml protein kinase C (PKC), catalytic fragment (BioMol Research Laboratories), protein kinase Aα catalytic subunit (BioMol Research Laboratories), and 0.4 mm [γ32P]-ATP (1000 cmp/pmol), or the same concentration of ATP (see Fig. 5 A), in 20 mm HEPES, pH 7.5, 10 mm MgCl2, 0.5 mm CaCl2, 10% glycerol, 5 μg/ml diolein, for 5 min at 30°C. The products were incubated with 10 μg of beads for 1 hr at room temperature and washed five times with 200 μl of PBS. The 32P-labeled GST-α1, GST-β2, GST-γ2S beads were then resuspended in phosphatase assay buffer, which contained 40 mmTris-HCl, pH 8.0, 0.1 m NaCl, 0.4 mg/ml bovine serum albumin, 1 mm dithiothreitol, and 0.45 μm okadaic acid and incubated at 30°C for 1 min in buffer. In Figure 5 A, the phosphorylated GST-γ2S and the GST-γ2Smutant (Ser327-Ala) were subjected to SDS-PAGE (12% gel).
Induction of LTD of the unitary GABAAR–IPSCs at CA1 inhibitory synapses
LTD of the unitary GABAAR–IPSCs was recorded at CA1 interneuron–pyramidal cell synapses, using double whole-cell patch-clamp recordings with an IR-DIC optic system (Fig.1 A). As shown in Figure1 B, 30 min after tetanus, the amplitude of the unitary GABAAR–IPSCs decreased to 68.2 ± 7.4% (mean ± SEM) of baseline (n = 10 cells/5 control mice). This decrease was maintained over the course of 1 hr recordings.
To determine whether LTD was presynaptic or postsynaptic in origin, we estimated the unitary GABAAR–IPSC variability by computing the inverse of the square of the coefficient variance (CV-2 =M 2/ς2) (Edwards et al., 1990; Silver et al., 1998), where M is the mean unitary GABAAR–IPSCs, and ς is the variance about M. In accordance with previous observations (Nusser et al., 1998), the distribution of the unitary GABAAR–IPSCs at the baseline (baseline noise: −3 ± 4.1 pA; n = 10 cells/5 control mice) has several clearly distinguishable peaks. The distribution was fitted by the sum of multiple Gaussian functions with one peak centered at 0 mV (failures) and other skewed peaks. The induction of LTD produced a shift in the distribution of the unitary GABAAR–IPSCs toward smaller amplitude values, with no change in the number of the failures (Fig.1 B). The mean unitary GABAAR–IPSC amplitude (M) is reduced after the induction of LTD, whereas the CV obtained by the method ofM 2/ς2is unchanged (Fig. 1 B). The data indicate that a decrease in the postsynaptic GABAA receptor function may underlie LTD at CA1 inhibitory synapses.
To determine whether a decrease in the channel conductance (γ) of the synaptic GABAA receptors or a decrease in the number (N o) of synaptically activated channels contributes to LTD, we performed nonstationary fluctuation analysis (non-SFA) (Traynelis et al., 1993; De Koninck and Mody, 1994;Otis et al., 1994; Auger and Marty, 1997) for the experiments in Figure1 B (n = 10 cells/5 animals). We first analyzed the kinetic properties of the averaged unitary GABAAR–IPSCs. As can be seen in Figure1 C, no changes of the unitary GABAAR–IPSCs rise times or their decay time constants (τ) were observed after the induction of LTD. The average responses were then scaled to the peak and subtracted from individual unitary GABAAR–IPSCs. The variance of the fluctuation around mean was calculated and plotted against mean currents (Fig. 1 C). The data points were fit by a parabolic function (ς2 =iI m −I m 2/N o), where ς2 is the variance,I m is the mean current, iis the single channel current, and N ois the number of open channels of synaptic GABAAreceptors. At the baseline, estimated γ was 24.2 ± 3.6 pS, which was not significantly different from that (23.6 ± 3.8 pS) during LTD (p > 0.50;n = 10 cells/5 mice). Estimated γ is close to previously reported values of 20–32 pS for estimates of γ derived from noise analysis in hippocampal granule cells (De Koninck and Mody, 1994; Nusser et al., 1998). In contrast, a significant decrease inN o during LTD was observed; the synapses have on average 51 ± 7.2 (mean ± SEM;n = 10) open channels on the baseline, and this number decreased to 36 ± 4.8 (mean ± SEM; n = 10 cells) after the induction of LTD. The data suggest that the average number of open channels of the synaptic GABAAreceptor channels is reduced during LTD.
LTD recruits CaN-A to form a complex with GABAA receptors
To explore the mechanisms underlying reduction in the number of open GABAA receptor channels during LTD, we explored the physical interaction of endogenous CaN and GABAA receptors. We immunoprecipitated extracts of control (Fig. 2 A) and LTD-CA1 slices using antibodies against either CaN-A (anti-CaN-A) or the GABAA receptor-α1subunit (anti-α1). We found that 5 and 30 min after the induction of LTD, GABAAreceptor-α1, -γ2, and -β2 subunits, the predominant GABAA receptor subunits expressed in hippocampus (McKernan and Whiting, 1996), were coimmunoprecipitated with anti-CaN-A. Conversely, CaN-A was immunoprecipitated with anti-α1 (Fig. 2 B). A nonspecific IgG did not immunoprecipitate either CaN-A or GABAA receptors. In control CA1 extracts, however, no coimmunoprecipitation of CaN-A or GABAA receptors was observed with either antibody (Fig. 2 A), indicating that the induction of LTD recruits CaN-A into the GABAA receptor complex.
The major intracellular loops of the GABAAreceptor subunits contain many consensus phosphoserine/threonine residues (Moss et al., 1992; Brandon et al., 1999, 2000), which may be targeted by CaN-A. To investigate this possibility, we constructed GST fusion proteins encoding the second intracellular loops α1 334–420, β2 327–451, and γ2S 317–442 (short form) of the GABAA receptors. These fusion proteins precipitated CaN-A from the LTD-CA1 extracts but not from controls (Fig. 3 A). The data suggest that synaptic activity drives activated CaN into the GABAA receptor complex. The GABAA receptor fusion proteins may bind indirectly to the activated CaN and directly to their respective subunit to pull down CaN-A from the CaN-A-GABAAreceptor complex in the CA1 extracts.
We then determined whether CaN-A directly binds to GABAA receptors. We generated a constitutively expressed recombinant CaN catalytic fragment (CaN420) that exhibits stable Ca2+-independent phosphatase activity (Perrino et al., 1995). The GST fusion proteins encoding α1 334–420, β2 327–451, γ2S 317–442, and γ2L 317–446subunits of the GABAA receptors were incubated with either CaN420 or CaN-B. As shown in Figure 3 B, the CaN420, but not CaN-B, bound to GST-γ2S 317–442and γ2L 317–446, but not to GST alone, or α1- or β2-peptide fusion proteins. A GST–GABAA receptor-γ2Sdeletion mutant (GST-γ2S 332–442or *GST-γ2S) failed to bind CaN420, indicating the importance of residues 317–332 within the γ2 subunit for direct interaction with CaN-A. This was confirmed with a synthesized γ2-peptide encoding residues 317–332 of the γ2 subunit, which prevented binding of CaN420 to GST-γ2S (Fig. 3 B). Consistent with this, we also found that CaN-A can bind to GST-γ2 317–332(Fig. 3 C) and that γ2-peptide interfered with GABAA receptor-CaN-A association in the LTD CA1 extracts (Fig. 3 D).
It is known that synaptic GABAA receptor function is regulated by tyrosine kinase Src and adaptin α and β subunits of AP2 (Kittler et al., 2000; Brandon et al., 2001) and that Src- and adaptin-γ2 subunit interaction increases the synaptic GABAA receptor activity. Thus, CaN may have acted as a competitive inhibitor of these molecules to downregulate the GABAA receptor function. To investigate this, we examined whether γ2-peptide interferes with Src- and adaptin-GABAA receptor association. Consistent with previous studies, we found that GABAAreceptor fusion proteins precipitated the endogenous Src (Fig.3 E) and α-adaptin (Fig. 3 F) in the CA1 extracts. In the presence of γ2 peptide, α-adaptin but not Src can still be precipitated. The data suggest that Src may also bind to residues 317–332 of the γ2 subunit.
CaN-A-GABAA receptor complex formation is necessary for LTD
To test whether this activity-dependent interaction between CaN-A and GABAA receptors is essential for the induction of LTD at CA1 individual inhibitory synapses, we examined the consequence of blocking CaN-A-GABAA receptor complex formation. First, we blocked endogenous CaN by expressing a peptide corresponding to the autoinhibitory domain in the C terminal of CaN-Aα in the hippocampus of transgenic mice (Malleret et al., 2001). No interaction between CaN-A and GABAA receptors could be observed 5 and 30 min after tetanus in the CaN mutant mice (Fig. 4 A), and the effect of the tetanus on the unitary GABAAR–IPSCs was completely abolished in mutant mice in that the unitary GABAAR–IPSCs 30 min after tetanus was 93 ± 10.6% of baseline (Fig.4 B) (n = 10 cells/5 animals). Second, blockade of the CaN-A-GABAA receptor complex formation by an NMDA receptor antagonist, AP-5 (Fig. 4 C), prevented the induction of LTD of the unitary GABAAR–IPSCs (Fig. 4 D). Third, we applied 10 μmγ2-peptide directly into the CA1 postsynaptic pyramidal cells and found that it abolished LTD of the unitary GABAAR–IPSCs (Fig. 4 E). A peptide with the same amino acid composition, but in random order, scrambled γ2-peptide, served as a control, and did not prevent induction of LTD. Thus, NMDA receptor-dependent interaction of activated CaN and GABAA receptors was required for induction of LTD at CA1 inhibitory synapses.
CaN-A and LTD dephosphorylates GABAA receptor γ2 subunits
The CaN-A–GABAA receptor complex formation may strategically position the CaN catalytic domain to dephosphorylate synaptic GABAA receptors. It is now known that CaN-A directly interacts with residues 317–332 within the γ2 subunit that contains a phosphoserine (pSer327) residue. We therefore developed a phosphospecific antibody to pSer327γ2 peptide (anti-γ2pSer327) to analyze LTD-dependent changes in GABAAreceptor γ2 subunit phosphorylation in CA1 neurons. Anti-γ2pSer327 was specific for pSer327 in γ2 subunit, reacting with phosphorylated wild-type GST-γ2S but not with the mutant GST-γ2S(Ser327-Ala) (Fig.5 A). Next, we immunoprecipitated GABAA receptors from the CA1 extracts. Blot analysis of the immunoprecipitates with anti-γ2pSer327detected multiple reactive bands, but only the one corresponding to the 51 kDa was selectively blocked by preabsorption with the pSer327-γ2-peptide antigen (Fig. 5 B), demonstrating that GABAA receptor γ2 subunit is phosphorylated under basal conditions. After induction of LTD of the unitary GABAAR–IPSCs in the CA1 slices, we found that the immunoreactivity to anti-γ2pSer327 was significantly decreased at 5 min (74.2 ± 13.2% over control;p < 0.05; n = 4) and 30 min (70.9 ± 10.1% over control; p < 0.05; n = 5; normalized by immunoreactivity to a general anti-γ2) (Fig. 5 C). This net dephosphorylation of GABAA receptor γ2 subunit was not caused by a tetanus-induced decrease in total protein, because blot analysis showed that the amount of γ2 subunit was unchanged (Fig.4 D).
To assess whether activated CaN is responsible for LTD-dependent dephosphorylation of the GABAA receptor γ2 subunit, we examined phosphorylation of the γ2 subunits in the CaN mice. No change in the basal level of γ2 subunit phosphorylation was observed in either the CaN control or the mutant mice. However, tetanic stimulation failed to reduce the immunoreactivity to anti-γ2pSer327 in the CaN mutants, as shown in Figure 5 D, suggesting that CaN-A is necessary for the GABAA receptor dephosphorylation in situ. Moreover, blockade of NMDA receptors by AP-5 inhibited tetanus-induced dephosphorylation of the GABAA receptors in CaN control mice. We subsequently studied the consequence of blocking other protein phosphatases (PPs) by applying 1 μm calcyculin A, an inhibitor of PP1/2A but not of CaN (Cohen and Cohen, 1989). We found that calyculin A had no effect on the decrease in the immunoreactivity to anti-γ2pSer327 in CaN control mice (Fig. 5 D).
To identify further specific γ2 subunit dephosphorylation, but not other subunits of the GABAA receptors, we labeled the GST-α1, -β2, or -γ2 subunit with32P in vitro. CaN420 specifically caused a decrease in the level of32P labeling of the GST-γ2S only, and CaN-A immunoprecipitated from LTD-CA1 slices (CaNltd) produced similar results (Fig.6). In addition, in the presence of γ2-peptide, neither CaN420 nor CaNltd affects32P labeling of the GST-γ2S (Fig. 6), indicating that GABAA receptor γ2Sresidues 317–332 represent the interacting site of the endogenous CaN-A. Taken together, the above results indicate that activity-dependent CaN-A-GABAA receptor complex formation enables CaN-A to dephosphorylate the GABAA receptor γ2Ssubunit that leads to the induction of LTD of the unitary GABAAR–IPSCs.
Overexpression of CaN-A reduces mGABAAR–IPSCs
We next explored whether the CaN-A-GABAAreceptor complex formation is sufficient for the induction of LTD. We expressed the CaN catalytic domain, CaN-Aα, in the hippocampus of transgenic mice (CN98 mutant mice) (Mansuy et al., 1998; Winder et al., 1998). We affinity precipitated extracts from the CN98 mutant and control CA1 slices using GST-α1 334–420, -β2 327–451, and -γ2S 317–442. We observed that these fusion proteins precipitated CaN-A from the CN98 mutant CA1 extracts but not from control (Fig.7 A), showing that overexpressed CaN-A physically interacts with GABAA receptors.
We then examined whether an interaction of the overexpressed CaN-A and synaptic GABAA receptor causes a depression of the GABAAR–IPSCs. Pharmacologically isolated (in the presence of 10 μm CNQX and 1 μm TTX), spontaneously miniature GABAAR–IPSCs (mGABAAR-IPSCs) in CA1 pyramidal cells were measured. A significant decrease in mean amplitude of mGABAAR–IPSCs, but not in their frequency, was observed in the CN98 mutant mice (Fig. 7 B) (63.9 ± 6.6% of the controls; n = 10 cells/10 animals;p < 0.05). Mean intervals of mGABAAR–IPSCs in the CN98 mutant and control mice were 49.2 ± 7.2 and 52.6 ± 6.9 msec (n= 10 cells/5 animals; p > 0.05), respectively. The distribution of the unitary GABAAR–IPSCs shifted to smaller-amplitude values in the CN98 mutant mice, whereas theCV was unchanged (Fig. 7 B). Using peak-scaled, non-SFA, the properties of synaptically activated GABAA receptor channels in CN98 mice were determined. A mean γ of 23.6 ± 4.3 pS (mean ± SEM;n = 10 cells/10 animals) was obtained in CN98 control mice. This value of γ shows no difference from that in CN98 mutant mice. The synapses in CN98 mice have on average 48 ± 6.1 (mean ± SEM; n = 10 cells/5 animals) open channels, and this number decreased to 32 ± 4.2 (mean ± SEM; n = 10 cells/10 animals) in CN98 mutant mice, indicating that a decrease in the number of the synaptically activated GABAA receptor channels is responsible for the reduced mGABAAR–IPSCs in CN98 mutant mice.
CaN-A-induced responses and LTD occlude each other
If a reduction of the GABAAR–IPSCs by CaN-A mimics the features of tetanus-induced LTD, then LTD and the GABAAR–IPSC reduction by CaN-A may mask each other. This was initially investigated by comparing the current–variance relationship of spontaneously occurring IPSCs (sIPSCs; without TTX) 30 min after tetanus in CN98 control mice with that in mutant mice (Fig.8 A). Estimated sIPSC rise times in control and the CN98 mutant mice were 0.22 ± 0.02 and 0.24 ± 0.03 msec, respectively. The decay time constants were 3.18 ± 0.21 msec in control compared 3.21 ± 0.18 msec in the CN98 mutant mice. The channel conductance (γ) of the synaptic GABAA receptors in the CN98 mutant mice was 24.1 ± 3.2 pS, similar to that (22.9 ± 2.8 pS) in control mice (Fig. 8 A). In contrast, the number (N o) of open channels of the synaptic GABAA receptors was reduced in CN98 mutant mice; the synapses in control mice have on average 53 ± 8.2 (mean ± SEM) open channels, and this number decreased to 39 ± 6.4 (mean ± SEM) in the CN98 mutant mice. No reduction of theN o after tetanus was observed in CN98 mutant mice, whereas tetanus did produce a decrease in the number of synaptically activated GABAA receptor channels in CN98 control mice. Thus, CaN-A caused a decrease in the number of the synaptic GABAA receptor channels that appeared to occlude the tetanus effect.
Second, we examined the dependence of LTD on the concentration of exogenous CaN-Aα. A consistent threshold for the depression of the unitary GABAAR-IPSCs was detected by application of CaN420 (2 μg/ml), with a maximum depression near 10 μg/ml. CaN420 (10 μg/ml) was then applied directly into the CA1 pyramidal cells through the patch electrodes. The amplitude of the unitary GABAAR–IPSCs decreased to 58.4 ± 6.7% of baseline (n = 6 recordings). In contrast, heat-inactivated CaN420 (iCaN420) had no effect (Fig.8 B). Moreover, application of CaN420 did not alter the reversal potentials, as shown by the current–voltage curves for peak amplitude of the unitary GABAAR–IPSCs in Figure 8 B. The data demonstrate that activated CaN reduces the peak amplitude of unitary GABAAR– IPSCs, with no change in driving force.
In other experiments, tetanus produced a reduction of the unitary GABAAR–IPSCs, but there was no further decrease when CaN420 was applied intracellularly (Fig. 8 C). On the other hand, in cells not conditioned by tetanus, perfusion of CaN420 at the same time after beginning recording caused a decrease in the unitary GABAAR–IPSCs that reached a stable level at 63.8 ± 7.9% of the baseline (n = 6 cells). Thus, the CaN-A-induced reduction of the unitary GABAAR–IPSCs and LTD occluded each other, indicating an overlapping mechanism of action.
Our analysis of the molecular mechanisms contributing to LTD of the unitary GABAAR–IPSCs revealed a novel postsynaptic event: an activity-dependent physical and functional interaction between CaN-A and the GABAAreceptor-γ2 subunit permits a rapid and sustained reduction of inhibitory synaptic strength at individual CA1 synapses. Furthermore, our results indicate that the CaN–GABAA receptor complex formation occurs only at synapses conditioned by activation of postsynaptic NMDA receptors. We determined that activated endogenous CaN was recruited into the GABAA receptors at CA1 inhibitory synapses via a direct binding with the γ2S subunit, because a synthesized γ2-peptide that encodes residues 317–332 of the γ2S subunit prevented CaN interaction with GABAA receptors. Because intracellular application of the γ2-peptide directly into the postsynaptic neurons inhibited the induction of LTD at CA1 individual inhibitory synapses, the most parsimonious explanation for our results is that activation of NMDA receptors results in increased Ca2+ entry, which activates CaN, and drives the activated CaN to dephosphorylate postsynaptic GABAA receptors, leading to a downregulation of GABAA receptor function.
Although we did not directly rule out presynaptic release properties during LTD, similar values of the coefficient variance of the evoked unitary IPSCs and spontaneous IPSCs before versus after the induction of LTD were observed. Using peak-scaled, non-SFA, our data showed that the reduced number of the synaptically activated GABAA receptor channels accounts for LTD at CA1 inhibitory synapses. Thus, taken together, our biochemical and electrophysiological data are not consistent with the presynaptic locus of NMDA receptor-dependent LTD of the unitary GABAAR–IPSCs in CA1 neurons.
CaN is known to play a role in the induction of LTD of excitatory transmission in CA1 neurons (Kirkwood and Bear, 1994; Mulkey et al., 1994). Consistent with these pharmacological studies, a recent study in the knock-out mice that lack CaN regulatory domain B1 showed that the induction of LTD at CA1 excitatory synapses was impaired (Zeng et al., 2001). The mechanisms by which the CaN-dependent signaling pathway participates in LTD at CA1 excitatory synapses are thought to involve the CaN-dependent activation of PP1 by inactivating I-1, a PP1 inhibitor (Lisman and Zhabotinsky, 2001). Consistent with this idea, the postsynaptic injection of I-1 peptides that mimics the phosphorylated, activated state of I-1 blocks the induction of LTD of excitatory transmission in CA1 pyramidal cells (Mulkey et al., 1994). From our results using GST fusion proteins of the subunit intracellular domains, we suggest that activated CaN is directly targeted to the phospho-Ser327 of the γ2S subunit. Therefore, low-frequency stimulation activates the CaN-dependent PP1 pathway that leads to the induction of LTD of excitatory transmission, whereas a brief high-frequency stimulation drives an activated CaN directly onto the GABAA receptors, leading to the induction of LTD at CA1 inhibitory synapses.
Among the mechanisms proposed for modification of GABAA receptor activity, one of the simplest is a change in the number of GABAA receptors in the postsynaptic membrane. There is extensive evidence showing that the GABAA receptor γ2 subunit plays a critical role in postsynaptic membrane trafficking as well as in synaptic targeting of GABAA receptors. For example, the number of synaptic GABAA receptors was decreased in cerebral cortex of mice lacking the γ2 subunit (Essrich et al., 1998). The γ2 subunits of GABAAreceptors interact differentially with diverse intracellular molecules including GABARAP (Nymann-Andersen et al., 2002), α-adaptin (Kittler et al., 2000), and Src (Brandon et al., 2001), all of which have been implicated in synaptic targeting of GABAA receptors. For example, in cultured hippocampal neurons, the α-adaptin-γ2 subunit association disrupts the clathrin-dependent GABAAreceptor endocytosis. Therefore, it is possible that CaN-A interferes with the α-adaptin-γ2 subunit interaction to regulate the number of functional GABAA receptors in the inhibitory postsynaptic sites. However, our observation show that the γ2-peptide that prevents CaN-A-GABAA receptor complex formation did not change the α-adaptin-GABAA receptor association, indicating that CaN acts via direct CaN-A–γ2 subunit interactions rather than as an adaptin substrate. Interestingly, clathrin-mediated endocytosis is involved in the expression of cerebellar LTD at excitatory synapses (Wang and Linden, 2000). It will be of importance to explore whether CaN regulation of GABAA receptors is mediated by the clathrin-dependent GABAA receptor internalization. The internal versus surface expression of GABAA receptors in the postsynaptic sites needs to be examined to determine CaN-mediated GABAAreceptor trafficking to and from the cell surface that is therefore likely to be an important mechanism for expression of LTD at inhibitory synapses. In addition to adaptin, there has been a report that tyrosine kinase Src modulates neuronal as well as recombinant GABAA receptors to enhance receptor channel activity (Moss et al., 1995). Src kinase also phosphorylates both Y365 and Y367residues of the γ2L (long form) subunit (Moss et al., 1995). By specifically blocking the association of Src with the GABAA receptor γ2 subunit (Fig. 3 E), CaN may act as a competitive inhibitor of Src to downregulate GABAA receptor function. Clearly, further studies will be needed to clarify the interaction of CaN and Src kinase with regard to GABAA receptor function.
There is extensive evidence showing that PKC phosphorylates GABAA receptors and increases the peak amplitude of mIPSCs recorded from hippocampal granule cells (Poisbeau et al., 1999) (but see Connolly et al., 1999). PKC also enhances the channel activity of recombinant GABAA receptors expressed in the L929 cell line (Lin et al., 1994, 1996). A recent study shows that activation of PP1 decreases in phosphorylation of Ser657/660 on the catalytic domain of PKCα and PKCβII (Thiels et al., 2000). This decrease in PKC phosphorylation is associated with a decrease in PKC activity. Thus, CaN may act through the PP1 pathway to downregulate PKC leading to the downregulation of GABAA receptor function for inducing LTD at CA1 inhibitory synapses. Because the GABAA receptor subunit combination and the receptor density expressed in CA1 pyramidal cells are heterogeneous (Pettit and Augustine, 2000), the effect of PKC phosphorylation of the GABAA receptor on its channel activity may depend on the receptor types under study. Therefore, whether the CaN–PKC pathway is involved in LTD at CA1 inhibitory synapses needs to be clarified in hippocampal slices. Further studies will use the strategies and protocols established in this study to determine how protein kinases are targeted to the synaptic GABAA receptors in synaptic plasticity. The final illustration of the cellular and molecular mechanisms underlying bidirectional regulation of GABAA receptors mediated by kinases and phosphatases in synaptic plasticity can teach us much about what kinds of molecules are needed to build arrays of the GABAA receptor regulations.
CaN is reported to modulate the channel kinetics of the GABAA receptors in cultured neurons (Jones and Westbrook, 1997). However, we found that there were no changes of the unitary GABAAR–IPSC rise times or their decay time constants after activation of CaN. The different observations could be attributed to the fact that in previous studies, reduced decay times of IPSC were caused by inhibition of endogenous CaN, which is generally dependent on the basal level of CaN activity. In hippocampal slices, however, we found that there were no physical and functional interactions of CaN and GABAAreceptors at basal CA1 inhibitory synapses. Therefore, it is possible that enzymatic regulation of synaptic receptors in cultured neurons is different from that in acute isolated neurons in the slices. Consistent with this idea, two other previous studies show that the activated but not basal level of CaN reduces the peak amplitude of whole-cell GABAA receptor currents in acute isolated hippocampal neurons (Stelzer and Shi, 1994; Chen and Wong, 1995).
Much of the difficulty in assigning changes at the level of individual inhibitory synapses is that the IPSCs, originating from feed-forward (Alger and Nicoll, 1982) as well as feed-back activation of the CA1 interneurons (Lacaille and Schwartzkroin, 1988) by the Schaffer-collateral stimulation, are polysynaptic and thus obscure any link between properties of IPSCs and the efficacy of individual inhibitory synapses. Conventionally, the GABAAR–IPSCs in CA1 pyramidal neurons are studied under conditions in which excitatory glutamate receptors are blocked. This manipulation, however, fails to account for NMDA receptor-dependent intracellular events that are essential for both GABAA receptor regulation and the initiation of sustained change in inhibitory synaptic strength. Here, we established double whole-cell patch-clamp recordings by which we monitored the unitary GABAAR–IPSCs between identified pairs of CA1 interneurons and the innervated pyramidal cells from hippocampal slices. Using both genetic manipulations and biochemical assays, we determined the mechanisms for excitatory activity-dependent LTD at inhibitory synapses. These data demonstrate a previously unknown molecular mechanism by which individual GABAergic synapses alter their efficacy in CA1 neurons of the hippocampus. Because inhibitory synaptic strength is critical for the control of networks within the brain (Ben-Ari and Represa, 1990), our results may suggest a CaN-dependent cellular substrate of learning and memory.
↵* J.W. and S.L. contributed equally to this work.
This work was supported by the Canadian Institute for Health Research (Y.M.L), Heart and Stroke Foundation, Canada (Y.M.L), Alberta Heritage Foundation for Medical Research (Y.M.L), Canada Foundation for Innovation (Y.M.L), and Alberta Foundation for Innovation and Science (Y.M.L). We thank Dr. Brian Perrino for the construct of CaN420. We thank Drs. John F. MacDonald, Wayne Giles, Keith Sharkey, and Brian MacVicar, for critical comments on this manuscript.
Correspondence should be addressed to Dr. YouMing Lu, Department of Physiology and Biophysics, Faculty of Medicine, University of Calgary, Calgary, Canada, T2N 4N1. E-mail:.