GABA and the GABAB receptor agonist baclofen activated a potassium conductance in acutely dissociated hippocampal CA3 neurons. Baclofen-activated current required internal GTP, was purely potassium selective, and showed strong inward rectification. As with acetylcholine-activated current in atrial myocytes, external Cs+ blocked inward but not outward current in a highly voltage-dependent manner, whereas Ba2+ blocked with no voltage dependence. Unlike the cardiac current, however, the baclofen-activated current showed no intrinsic voltage-dependent relaxation. With fast solution exchange, current was activated by baclofen or GABA with a lag of ∼50 msec followed by an exponential phase (time constant ∼225 msec at saturating agonist concentrations); deactivation was preceded by a lag of ∼150 msec and occurred with a time constant of ∼1 sec. GABA activated the potassium conductance with a half maximally effective concentration (EC50) of 1.6 μm, much lower than that for activation of GABAA receptor-activated chloride current in the same cells (EC50 ∼25 μm). At low GABA concentrations, activation of the GABAB current had a Hill coefficient of 1.4–2.1, suggesting cooperativity in the receptor-to-channel pathway. Although the maximal conductance activated by GABABreceptors is much smaller than that activated by GABAAreceptors, its higher sensitivity to GABA and slower time course make it well suited to respond to low concentrations of extra-synaptic GABA.
Two major postsynaptic conductances activated by GABA exist in central neurons. One is a chloride conductance activated by GABAA receptors and is present in virtually all central neurons. The second is a less widely distributed potassium conductance activated by GABAB receptors (Gähwiler and Brown, 1985; Newberry and Nicoll, 1985; Dutar and Nicoll, 1988). In neurons that possess both receptors postsynaptically, the requirements for activation are often strikingly different, with GABAB-mediated inhibitory postsynaptic potentials requiring stronger or more sustained stimulation than GABAA-mediated responses (Dutar and Nicoll, 1988; Otis and Mody, 1992). The reasons for this are not clearly understood.
In hippocampal pyramidal neurons, the GABAB-activated potassium conductance is inwardly rectifying (Gähwiler and Brown, 1985) and mediated by G-proteins (Andrade et al., 1986; Thalmann, 1988;Thompson and Gähwiler, 1992a) and can be blocked by external Ba2+ ions (Newberry and Nicoll, 1985; Misgeld et al., 1989). These properties are shared by transmitter-activated potassium currents present in various other neurons (North et al., 1987; North, 1989; Nicoll et al., 1990) as well as in cardiac atrial myocytes, where detailed characterization has been possible (Hartzell, 1988). A family of cDNAs encoding subunits of G-protein-activated inward-rectifier potassium (GIRK) channels has been described (Dascal et al., 1993; Kubo et al., 1993; Lesage et al., 1994; Doupnik et al., 1995; Krapivinsky et al., 1995b). Both native channels and cloned channels seem to be activated by binding of βγ subunits of G-proteins (Logethetis et al., 1987; Reuveny et al., 1994; Wickman et al., 1994; Huang et al., 1995; Inanobe et al., 1995; Krapivinsky et al., 1995b; Oh et al., 1995). G-protein α subunits may modulate the activation (Schreibmayer et al., 1996) or help mediate coupling to specific transmitter receptors (Huang et al., 1995).
Detailed characterization of the GABAB-activated potassium channels in hippocampal neurons has been hindered by lack of a suitable single-cell preparation. Experiments using tissue slices are limited by the difficulty of making rapid solution changes. For unknown reasons, the conductance is absent or minimal in conventional hippocampal tissue culture preparations (Harrison, 1990; Yoon and Rothman, 1991; Pfrieger et al., 1994), although it is preserved in organotypic cultures (Gähwiler and Brown, 1985). Even such basic properties as the dose–response relationship for GABA and kinetics of current activation are unknown. Knowledge of such properties should help in understanding the differences in synaptic activation of GABAA and GABAB receptors.
Acutely isolated cells have been used previously to study serotonin-activated potassium current in dorsal raphe neurons (Penington et al., 1993a). Stimulated by this, we found that robust GABAB receptor-activated potassium currents can be recorded in a preparation of acutely dissociated hippocampal CA3 neurons that permits rapid solution changes. We characterized the kinetics, rectification, ionic block, and GABA dependence of the current. The results show that the slow kinetics of GABAB-mediated inhibitory postsynaptic currents are primarily attributable to kinetics of receptor-to-channel coupling and that activation of GABAB receptors requires very low levels of extracellular GABA, perhaps near background levels.
MATERIALS AND METHODS
Preparation of freshly dissociated neurons.Hippocampi from 7- to 12-d-old Long Evans rats were dissected in ice-cold, oxygenated dissociation solution containing (in mm): 82 Na2SO4, 30 K2SO4, 5 MgCl2, 10 HEPES, 10 glucose, and 0.001% phenol red indicator, pH 7.4. Slices were cut 400 μm thick and incubated for 9 min at 37°C in dissociation solution containing 3 mg/ml protease (Type XXIII, Sigma, St. Louis, MO). Enzyme solution was then replaced with dissociation solution containing 1 mg/ml trypsin inhibitor and 1 mg/ml bovine serum albumin, and the slices were allowed to cool to room temperature under an oxygen atmosphere. As cells were needed, slices were withdrawn, and the CA3 region was dissected out and triturated to release individual cells. Cells were placed in the recording chamber in Tyrode’s solution containing (in mm): 150 NaCl, 4 KCl, 2 CaCl2, 2 MgCl2, 10 glucose, and 10 HEPES, pH 7.4 with NaOH. Cells were used within 6–8 hr of slice preparation.
CA3 pyramidal neurons were identified morphologically, based on size and shape. Cells identified as pyramidal neurons had a large pyramidal-shaped cell body (12–16 μ width, 20–36 μ length) with a thick apical dendritic stump (4–6 μ width, 18–24 μ length) and, in some neurons, one to four basal dendritic stumps (1–4 μ width, 6–14 μ length). Average cell capacitance was 23 ± 5 pF (n = 125; range, 15–38 pF). Despite the presence of dendritic stumps, the cells behaved as if they were electrotonically compact: capacity transients settled with a single time constant of 200–350 μsec.
Whole-cell voltage-clamp recordings. Patch pipettes were pulled from 100 μl Boralex micropipettes (Dynalab, Rochester, NY.). Pipette resistances ranged from 2 to 5 MΩ when filled with internal solution containing (in mm): 108 KH2PO4, 4.5 MgCl2, 9 HEPES, 9 EGTA, 14 mm creatine phosphate (Tris salt), 4 mmMg-ATP, and 0.3 mm GTP (Tris salt), pH adjusted to 7.4 with 135.4 mm KOH. The creatine phosphate ATP and GTP were added from 10× concentrated aliquots stored at −70°C. To prevent nucleotide hydrolysis, the final internal solution was kept on ice after phosphates were added.
The liquid junction potential between the internal solution and Tyrode’s solution (in which the current was zeroed before obtaining a seal) was −12 mV, measured as described by Neher (1992). Membrane potentials were corrected for this junction potential.
Seals were formed and the whole-cell configuration was obtained in bath Tyrode’s solution. The cell was then bathed by a gravity-driven, constant stream of external solution flowing through microcapillary perfusion pipes positioned directly in front of the cell. The perfusion pipes consisted of a linear array of 12 microcapillary tubes of internal diameter 200 μ or 250 μ, glued together side by side and fed from separate reservoirs. Solutions were changed by moving the perfusion pipes. In some experiments, the cell was lifted from the bottom of the chamber, and solution changes were made under computer control by using a solenoid attached to the recording pipette to rapidly move the cell between adjacent pipes. The speed of solution exchanges made by this method (used for the experiments in Figs. 8, 9, 10, 11, 12, 13) was measured by the time course of current relaxation when the GABAB-activated conductance was activated by steady application of 10–100 μm baclofen and the cell was moved between solutions containing 60 mm and 16 mmK+. The time constant of solution change varied between 3 and 20 msec in different cells.
External recording solutions consisted of modified Tyrode’s solution with 4–60 mm KCl, with KCl substituted for an equimolar amount of NaCl. Tetrodotoxin (TTX) was included at 2–3 μm in all solutions to block sodium currents. R(+)-baclofen, GABA, GTP-γ-S, GDP-β-S, and ω-conotoxin MVIIC were stored as concentrated aliquots at −70°C or −20°C and diluted into recording solution on the day of the experiment. Whole-cell currents were recorded with an Axopatch 200A patch-clamp amplifier, filtered at 2 kHz, digitized at 20–50 KHz, and stored using a BASIC-FASTLAB analog/digital interface and software (Indec Systems, Sunnyvale, CA) or a Digidata interface and PClamp6 software (Axon Instruments, Foster City, CA).
Current–voltage curves. Current–voltage curves were determined using voltage ramps from −172 mV to +8 mV, 100 msec in duration. To smooth the voltage signal, it was low-pass filtered at 0.5 kHz (4-pole Bessel filter) before being applied to the patch-clamp amplifier. The voltage was corrected for the delay resulting from the filtering. Baclofen-induced current was obtained by subtracting ramp currents before and after application of baclofen, as shown in Figure3. In hippocampal CA3 neurons, baclofen inhibits voltage-dependent calcium current as well as activating potassium current. This effect overlaps with the baclofen-activated K current at potentials positive to −40 mV, where calcium current begins to be activated. We initially used Cd2+ to block calcium currents, but found that the inward rectifier K current was also reduced (∼50% block with 100 μm Cd2+). Nimodipine and nicardipine, which block L-type calcium current, were also found to partially block the baclofen-activated K current (∼60% block by 10 μmnimodipine and ∼60% block by 3 μm nicardipine). The peptide ω-conotoxin MVIIC inhibits at least three components of calcium current in CA3 neurons (McDonough et al., 1996), and the calcium current that remains has little or no sensitivity to short applications of baclofen (K. J. Swartz and B. P. Bean, unpublished observations). ω-Conotoxin MVIIC (10 μm) was therefore included in all external solutions in experiments measuring reversal potentials.
All statistics are given as mean ± SEM.
Initial experiments were carried out with baclofen rather than GABA to avoid activating GABAA receptors. Baclofen activated inwardly rectifying potassium current in 401 of 402 neurons tested. Figure 1 shows the dose dependence of the baclofen-activated current. Substantial current was activated by baclofen concentrations of 500 nm and above. The dose–response relationship for baclofen could be fit fairly well by the logistic equation with an EC50 of 3 μmand a Hill coefficient of 1 (Fig. 1 B). Despite the presence in the cells of large chloride conductances that could be activated by GABAA receptors (see Figs. 13, 14), baclofen up to 1 mm had perfect selectivity for the GABAB-mediated response. With applications of 100 μm or higher, the baclofen-activated current began to decline or desensitize after ∼1 sec (Fig. 1 A).
Dependence on GTP
Our standard pipette solution included 300 μm GTP. To test the GTP dependence of the baclofen response, we omitted pipette GTP or replaced it with nonhydrolyzable GTP and GDP analogs (Fig.2 A,B). If GTP was omitted from the pipette solution, the baclofen-induced current fell to 28 ± 11% (n = 8) of the initial current after 12 min of dialysis. If the GDP analog GDPβS was included in the pipette solution (at 800 μm), the baclofen-induced current disappeared faster, falling to <5% in 4–12 min of dialysis in 12 of 14 cells tested. When the poorly hydrolyzable GTP analog GTPγS (300 μm) was included in the pipette solution, a small inward current began to appear after 2–3 min of dialysis, even in the absence of baclofen. If baclofen was then applied, a large current was activated promptly and irreversibly (Fig. 2 B). A second baclofen application produced no further response. Similar irreversible responses were obtained in 11 of 11 cells tested with GTPγS. These experiments suggest that the coupling between GABAB receptors and potassium channels is dependent on GTP acting via G-proteins.
In hippocampal slices, baclofen activation of potassium current can be prevented by intracerebral injections of pertussis toxin (Andrade et al., 1986; Andrade and Nicoll, 1987; Thalmann, 1988; Thompson and Gähwiler, 1992a). We tested the ability of pertussis toxin to disrupt the receptor-channel coupling in dissociated neurons, but with inconclusive results. With inclusion of the catalytic subunit (A-protomer, 5 μg/ml, preactivated by 20 mm DTT) in the pipette solution, along with its substrate NAD (1 mm), the baclofen-activated current declined slowly with time, with a 30–55% reduction in current (n = 4) after up to 45 min of dialysis. We used the same procedure in rat atrial myocytes and also found only slow decline in the acetylcholine-activated current, even though the atrial muscarinic response in various species is completely sensitive to pertussis toxin applied overnight or to excised patches (Pfaffinger et al., 1985; Kurachi et al., 1986a,b; Ito et al., 1991). We conclude that under our conditions the procedure is not adequate for testing the pertussis toxin sensitivity of the transduction pathway.
When applied for short times, the sulfhydryl alkylating agentN-ethyl-maleimide (NEM) specifically eliminates responses mediated by pertussis toxin-sensitive G-proteins, whereas it spares responses by other G-proteins (Nakajima et al., 1990; Wollmuth et al., 1995). Figure 2 C shows the effect of 50 μm NEM on the baclofen-activated current in hippocampal neurons. With a 1 min extracellular application, the baclofen response was abolished completely. This result was obtained in four of the four cells tested. Although this result is far from a definitive test of mediation of the response by pertussis toxin-sensitive G-proteins, the rapid and complete effects of NEM raise the possibility that lower concentrations could be used as a tool to modify receptor-to-channel coupling in a titratable manner.
Figure 3 shows the current–voltage relationship for the baclofen-activated K current under quasi-physiological ionic conditions, with 4 mm external K+ and 243 mm internal K+. In contrast to serotonin-activated current in acutely dissociated dorsal raphe neurons (Penington et al., 1993a), it was easy to record currents with physiological external K+ concentrations. Current–voltage curves were obtained using a voltage ramp from −172 mV to +8 mV. The ramp was kept relatively short (100 msec) to minimize the time at voltages positive to −50 mV, where large voltage- and time-dependent potassium currents are activated. The baclofen-induced current reversed at −90 mV and was strongly inwardly rectifying, with inward current at −150 mV (carried by 4 mm K+), approximately three times larger than the outward current at −50 mV (carried by 243 mm K+). Despite the powerful rectification, however, the baclofen-induced outward current was substantial compared with basal currents at potentials of −80 to −40 mV, the range of typical neuronal resting potentials. In the cell of Figure 3, the zero current potential (which would correspond to the resting potential if the cell were not voltage-clamped) was hyperpolarized from −44 mV to −53 mV by baclofen. Thus, the outward potassium current is large enough to have a significant hyperpolarizing effect.
Figure 4 shows the dependence of the baclofen-activated current on external potassium concentration, determined in a single neuron. As potassium was increased from 4 to 60 mm, the inward current increased dramatically, and the reversal potential shifted almost exactly as predicted by the Nernst equation for a purely potassium-selective conductance. As is characteristic of inwardly rectifying potassium channels (Hille, 1992a), the voltage dependence of rectification depended onV m-E k as external potassium was altered. At strongly negative voltages (below approximately −140 mV), inward current saturated. This saturation disappeared in Na-free external solutions (not shown). The saturation is likely attributable to voltage-dependent block by external Na+ ions, as has been described for other inward rectifiers (Ohmori, 1978; Standen and Stanfield, 1979; Harvey and Ten Eick, 1989).
Block by external Cs+and Ba2+
Baclofen-induced current was blocked by external Cs+and Ba2+ ions. With 4 mm external K+, Cs+ blocked inward current in a highly voltage-dependent manner and had no effect on outward current at concentrations up to 3 mm (Fig.5 A). At any given voltage, the concentration dependence of Cs+ block could be fit well by a Langmuir isotherm (Fig. 5 B), with a half-blocking concentration that increased exponentially from ∼30 μm at −162 mV to ∼200 μm at −122 mV, corresponding to an e-fold change every 10 mV. In another series of experiments using 16 mmexternal potassium, the half-blocking concentration of Cs+was slightly lower than for 4 mm K+ at voltages positive to −150 mV (as if the cesium ion was “trapped” by the higher external K+), but the IC50 at voltages negative to −150 mV saturated at ∼40 μmCs+ (Fig. 5 C).
Block by Ba2+ was very different. It showed no voltage dependence, with outward current blocked equally as well as inward current (Fig. 6). At all voltages, the dose dependence of inhibition could be fit well by a Langmuir isotherm with an IC50 of 12 μm (Fig. 6 B). The experiment in Figure 6 A was carried out with a fast ramping protocol that in principle might allow too little time for relaxation of voltage-dependent block; however, there was also no voltage dependence evident when it was tested with voltage steps lasting 2–5 sec (not shown). In contrast to Cs+ block, a fourfold increase in external K+ concentration had no effect on block by Ba2+, regardless of voltage (Fig.6 C).
Lack of time-dependent gating
Many inwardly rectifying potassium channels display time- and voltage-dependent components of current (for review, see Hille, 1992a). Such voltage-dependent relaxations are seen with the potassium conductance activated by muscarinic receptors in most cardiac atrial cells (Noma and Trautwein, 1978; Sakmann et al., 1983; Simmons and Hartzell, 1987; but see Carmeliet and Mubagwa, 1986). We tested for voltage-dependent relaxations using long hyperpolarizations (Fig.7). In contrast to the results in cardiac cells, there was no resolvable time-dependent relaxation in the baclofen-induced current. This result was obtained both in physiological K+with a step from −72 mV to −132 mV (Fig. 7 A) and in 60 mm K+ with a step from −12 mV to −132 mV (Fig. 7 B).
Kinetics of activation and deactivation
The kinetics of current activation on exposure to baclofen depended on the baclofen concentration. Maximal current occurred at ∼3-5 sec with application of 5 μm baclofen and at ∼1 sec with application of 100 μm baclofen. In experiments in which rapid solution changes were made under computer control, current was seen to rise with a sigmoidal time course (Fig.8 A). The final two thirds of the rising phase could be fit well by an exponential, with a time constant that declined from ∼600 msec at 5 μm baclofen to ∼250 msec at 100 μm baclofen (Fig. 8 A). In the collected results shown in Figure 8 B, it can be seen that increasing the baclofen concentration from 100 μm to 1 mm produced little further acceleration of kinetics, suggesting that the asymptotic value of 225 msec reflects a rate-limiting step subsequent to binding of agonist to receptor. On removal of baclofen, the decline of current was also sigmoidal. After an initial delay, the decline of current could be fit well with a single exponential (Fig. 8 A). The time constant of decay ranged from ∼450 msec to 2 sec among individual cells; the average time constant was 1.1 ± 0.1 sec (n = 38).
Figure 9 shows with higher resolution the sigmoidicity of activation and deactivation. The cell was moved between solutions with a computer-driven solenoid. The time course of solution change at the membrane of the cell was determined by the relaxation of current when the cell was moved between two solutions, both containing baclofen but having different K+ concentration (thin traces); solution exchange was complete in ∼10 msec. On application of agonist, there was a lag of ∼50 msec before any significant activation of current (Fig. 9 A). On removal of agonist, there was no significant decline in current for ∼150 msec. The lag in activation varied in different cells from 40 to 60 msec and that for deactivation from 150 to 200 msec.
The lags in activation and deactivation are consistent with a multi-step pathway involving agonist binding to receptor, activation of G-protein, and opening of K+ channels. This suggests a complex temporal relationship between the presence of agonist and activation of current, especially for short applications of agonist. As shown in Figure 10, the current greatly outlasted the presence of agonist with short exposures. With a 200 msec baclofen application (Fig. 10 A), current continued to rise after removal of agonist, so that peak current occurred approximately 150 msec later. With an even shorter application (for 60 msec), the conductance change did not even start until free agonist was no longer present (Fig. 10 B), and it reached a peak 220 msec later.
In the experiments with short applications of agonist, we found that the deactivation of current was dependent on the length of agonist exposure. This effect is shown in Figure 11, in which baclofen was applied for different durations to a single cell. After the initial delay, the decline of current could be fit well by a single exponential in all cases. As the application length was increased, deactivation became progressively slower, changing from a time constant of 267 msec with a 200 msec application to a time constant of 721 msec after a 5 sec application. This effect was seen consistently.
Kinetics with GABA as agonist
We also examined currents activated by GABA, the natural agonist. The GABAB response was isolated by recording with 100 μm picrotoxin to block the GABAA response. Control experiments showed that 100 μm picrotoxin had no effect on current activated by 50 μm baclofen. Figure12 compares in the same cell current activated by saturating concentrations of baclofen and GABA. Both the magnitude of current and the kinetics of activation and deactivation were essentially identical when elicited by GABA or baclofen. At saturating GABA concentrations, the activation time constant was 232 ± 17 msec (n = 13), and the deactivation time constant was 1.0 ± 0.2 sec (n = 14), very close to the values of 225 msec and 1.1 sec obtained with baclofen.
GABAB- and GABAA-activated currents compared
Both GABAB and GABAA receptors are present in the cell bodies of hippocampal CA3 neurons. We used the ability to apply well defined agonist concentrations to directly compare their sensitivity to GABA. The experiments in Figure13 compare the kinetics and sensitivity of the two types of current, activated by 3 μm and 1 mmGABA. The GABAB response was recorded with 100 μm picrotoxin, and the GABAA current was recorded with 1 mm Ba2+ to block GABAB currents. The GABAB-induced current was more sensitive to GABA, with 3 μm GABA activating more than half the maximal current activated by a saturating concentration of 1 mm GABA. As expected, the GABAB response was slower than the GABAA response, especially at 1 mm GABA, where the rise of the GABAA response was as fast as the solution exchange. The deactivation of the GABAB response was also much slower than the GABAA response. In addition, desensitization of the GABAB response was far slower than that of the GABAA response. Figure 13 C compares the magnitude and current–voltage relationship of the two responses when studied with physiological solutions and activated by saturating agonist concentrations. The GABAA response reversed significantly positive to the GABAB response and was far larger. The peak outward current elicited at −50 mV averaged 1280 ± 350 pA for the GABAA response and 50 ± 3 pA for the GABAB response.
Figure 14 shows a detailed comparison of the sensitivity of the GABAA and GABAB responses to GABA. GABA was much more potent in activating the GABABresponse (EC50 1.6 μm) than the GABAA response (25 μm). Figure14 B shows the result of experiments examining the apparent stoichiometry of the GABAB response at low GABA concentrations (0.1 −0.3 μm GABA). The current increased supralinearly with GABA concentration in all cells examined, with a Hill coefficient between 1.4 and 2.1 for individual cells (mean 1.66 ± 0.16, n = 4). Interestingly, the Hill coefficient was consistently higher when the response to this range of concentrations was smaller (relative to the current elicited by 100 μm GABA). This suggests that the variability in Hill coefficient between cells arises from variability in the position on the dose–response curve of the 0.1–0.3 μm range tested. Such variability could result from different ratios of receptor to G-protein or potassium channels. The GABAA response also had a supralinear response at GABA concentrations (in this case, 1–5 μm) that activated 0.003–0.2 of the maximal conductance. Both the relative potency of a given concentration and the Hill coefficient (1.70 ± 0.03; range, 1.6–1.8; n = 8) showed less variability than for the GABAB response, consistent with more direct coupling of ligand binding to channel activation.
The data on GABA dose dependence (Fig. 14) and current–voltage characteristics (Fig. 13 C) of the two responses can be considered together to estimate the current that each response is capable of generating at low GABA concentrations with physiological ionic conditions. GABA at 0.3 μm activates ∼0.1 of the maximal GABAB current (an average of +50 pA at −50 mV), yielding a 5 pA outward current. We could not directly measure any activation of GABAA current by 0.3 μm GABA, but extrapolating the relationships in Figure 14 C predicts fractional activation of 0.0004. Even with the maximal current of +1280 pA for the peak GABAA response at −50 mV, the predicted current of 0.5 pA is lower than that for GABAB current. Thus, even though GABAA receptors can generate a far larger maximal current in our hippocampal neurons, the higher sensitivity of the GABAB response suggests that it carries more current at low GABA concentrations.
Block by Cs+ and Ba2+
All inwardly rectifying potassium channels are blocked by external Cs+ and Ba2+ ions, but the properties of block differ among different types of channels. For example, block by Ba2+ is highly voltage dependent in some inward rectifiers (Standen and Stanfield, 1978; Hille, 1992a; Takano and Ashcroft, 1996). Ba2+ block of the GABAB receptor current was not voltage dependent, however. This corresponds to the behavior of acetylcholine-activated current in cardiac cells (Carmeliet and Mubagwa, 1986). Similarly, the potency of Ba2+ block in hippocampal neurons (EC50 12 μm, Fig. 6) is virtually identical to that in cardiac tissue (2–18 μm) (Carmeliet and Mubagwa, 1986). Block by Cs+ is also virtually identical in hippocampal neurons and cardiac tissue, in both potency and strong voltage dependence (compare Fig. 7 and Argibay et al., 1983).
Comparison with cloned channels
At least four members of the GIRK family of cDNAs, GIRK1, GIRK2, GIRK3, and CIR, are expressed in the hippocampus (Kobayashi et al., 1995; Lesage et al., 1995; Karschin et al., 1994, 1996; Ponce et al., 1996; Spauschus et al., 1996). Different GIRK proteins combine to form multimeric channels (Duprat et al., 1995; Kofuji et al., 1995;Krapivinsky et al., 1995a; Lesage et al., 1995; Spauschus et al., 1996). Native G-protein-activated channels in heart include GIRK1 and CIR subunits (Krapivinsky et al., 1995a). The subunit composition of native G-protein-activated channels in hippocampal neurons is not yet known, but the single-channel properties of GIRK1/GIRK2 channels (Velimirovic et al., 1996) resemble those of G-protein-gated channels in neurons (Miyake et al., 1989; Penington et al., 1993b; Oh et al., 1995; Grigg et al., 1996). The rectification properties that we found in CA3 neurons, with substantial outward current, correlate better with GIRK1/GIRK2 (Velimirovic et al., 1996) or GIRK1/CIR (Krapivinsky et al., 1995a; Spauschus et al., 1996) channels than with monomeric GIRK1 channels, which rectify so strongly that there is essentially no outward current (Dascal et al., 1993; Kubo et al., 1993). So far, none of the multimeric cloned channels demonstrate block by Ba2+similar to that in native CA3 neuron channels (EC50 12 μm with no voltage dependence). Both GIRK1/CIR (Krapivinsky et al., 1995a) and GIRK1/GIRK2 (Velimirovic et al., 1996) channels have much lower Ba2+ sensitivity (EC50∼500 μm), and block of both is strongly voltage dependent (Spauschus et al., 1996; Velimirovic et al., 1996). Thus, the voltage dependence and sensitivity of Ba2+ block of native channels are quite different than any combination of subunits yet described. Possibly, native channels have additional subunits to those so far known.
The GABAB receptor-activated current showed no time-dependent relaxations with voltage steps (Fig. 7), which was different from muscarinic receptor-activated current in cardiac myocytes (Noma and Trautwein, 1978; Simmons and Hartzell, 1987; Clark et al., 1990). GIRK1 forms channels that show prominent voltage-dependent relaxations, regardless of whether current is activated by various G-protein-linked receptors or by coexpressed βγ G-protein subunits (Kubo et al., 1993; Doupnik et al., 1995;Krapivinsky et al., 1995a). Similar voltage-dependent gating of other inward-rectifier channels results from time- and voltage-dependent block and unblock by intracellular spermine or a related polyamine (Ficker et al., 1994; Lopatin et al., 1994; Fakler et al., 1995), and polyamines can interact similarly with cloned GIRK1 channels (Yamada and Kurachi, 1995). It is possible that hippocampal neurons have lower concentrations of such polyamines than oocytes or cardiac myocytes. In principle, endogenous polyamines may have been dialyzed out of the neurons by our whole-cell recording, but the baclofen-induced current in undialyzed hippocampal neurons studied with sharp microelectrode recording also showed no obvious time dependence during hyperpolarizations (Gähwiler and Brown, 1985). The difference from myocytes could also reflect different subunit composition of the channels, which influences the relaxations (Kofuji et al., 1995; Lesage et al., 1995; Velimirovic et al., 1996).
Our experiments provide the first description of the kinetics of GABAB receptor-activated current with rapid application and removal of agonist. As with the kinetics of GABAB receptor inhibition of calcium current (Pfrieger et al., 1994), activation and deactivation take hundreds of milliseconds, far slower than for channels directly activated by ligand binding. The limiting rate of channel activation at high ligand concentrations was ∼4 sec-1, which can be compared with 4000 sec−1 for GABAA receptor chloride channels (Maconochie et al., 1994). Reasonable candidates for the rate-limiting step in GABAB receptor activation of potassium channels include GDP/GTP exchange, diffusion of activated G-protein to the channel, and channel activation by activated G-protein. Consistent with a multi-step pathway, the main rising phase of current was preceded by a lag of ∼50 msec; similar lags have been seen for potassium current activated by acetylcholine in myocytes (∼80 msec; Inomata et al., 1989) and by noradrenaline in submucosus plexus neurons (∼60 msec;Surprenant and North, 1988). The time constant for the main phase of the off response (0.2–1.0 sec, depending on duration of exposure) is also similar to that for other G-protein-mediated conductances: 0.3–2.0 sec for atrial cells (Breitwieser and Szabo, 1988; Friel and Bean, 1990) and 0.15–0.2 sec for submucosus plexus neurons (Surprenant and North, 1988). The off response may well be limited by the rate of GTP hydrolysis, estimated by Breitwieser and Szabo (1988) to be ∼1 sec−1 in atrial cells. The dependence of the off rate on the duration of agonist application has not been noted previously for G-protein-activated K+ currents, and its origin is unclear. The effect could arise if the K+ channels can bind more βγ subunits than are needed to produce channel activation. A requirement for unbinding of multiple βγ subunits could contribute to the lag in deactivation as well.
Comparison with synaptically activated currents
The current activated by short pulses of agonist is almost identical in kinetics to the GABAB-mediated synaptic currents recorded by Otis et al. (1993) in granule neurons. The current in Figure 10 B activated by a 60 msec application of baclofen could be fit very well (dashed line) with m4h kinetics (Otis et al., 1993). The fit gave τm of 108 msec and τh of 345 msec, similar to their average values of 112 and 282 msec (for the main component of decay) for synaptic currents at 22–23°C. The comparison suggests that the slow kinetics of synaptic currents can be accounted for entirely by the time course of receptor-to-channel coupling.
Because the synaptically activated GABAB conductance does not show inward rectification (Thalmann, 1988; Otis et al., 1993), the possibility was raised that the rectification of baclofen-induced current reflects inadequately voltage-clamped cells in slice recordings (Otis et al., 1993). Because our recordings were performed with cells that can be voltage-clamped beyond reproach, this possibility can be ruled out. The reason for the nonrectification of synaptically activated current remains a puzzle; perhaps intracellular levels of either polyamines or Mg2+ are lower near dendritic receptors.
Dose–response and GABAB versus GABAA sensitivity
Our results provide the first measurements of the sensitivity of GABAB receptor-activated current to well defined low concentrations of GABA. They show that the EC50for activation of potassium current is much lower (1.6 μm) than for activation of GABAA receptor chloride channels (25 μm) in the same cells. Most likely, GABAB receptors have significantly higher affinity for GABA than do GABAA receptors, although the functional EC50 may be lower if there are “spare” GABAB receptors. In any case, the high sensitivity to low GABA concentrations supports the idea that GABABreceptor-activated current may be readily activated by diffuse “spill-over” of GABA from synapses (Thompson and Gähwiler, 1992b; Isaacson et al., 1993; Mody et al., 1994). In fact, the current was significantly activated by GABA in the range of 0.1–0.4 μm, within the range estimated for basal extracellular GABA (0.8 μm, Lerma et al., 1986; 0.2 μm,Tossman et al., 1986). This suggests that only minimal increases are needed.
We found that the dose–response relationship at low GABA concentrations is nonlinear, with a Hill coefficient of 1.7. To our knowledge this is the first demonstration of a nonlinear relationship in agonist activation of a G-protein-coupled conductance [cf.Breitwieser and Szabo, 1988; Inomata et al., 1989; Hille, 1992b; seeDestexhe and Sejnowski (1995) for a model incorporating stronger supralinearity]. The supralinearity implies cooperativity at some stage in the coupling pathway. Unlike the GABAA current, which has an identical Hill coefficient, the cooperativity is unlikely to arise at the receptor level. A likely step for cooperativity is the activation of channels by βγ subunits; in atrial mycocyte channels, activation by various βγ subunits yielded an average Hill coefficient of 1.5 (Krapivinsky et al., 1995b), although values as high as 3 have been found for particular βγ combinations (Ito et al., 1992; Krapivinsky et al., 1995b).
It is a long-standing observation that GABAB-mediated inhibitory postsynaptic potentials require stronger or more sustained stimulation than GABAA-mediated responses do. Our results show that this occurs despite a much higher sensitivity of GABAB receptors to GABA. The failure of GABABreceptors to be activated by synaptic stimulation capable of activating GABAA receptors argues strongly for different localization of the two receptor types, with GABAA receptors located immediately postsynaptically and GABAB receptors located primarily extrasynaptically, as has been proposed (Thompson and Gähwiler, 1992b; Isaacson et al., 1993). The high sensitivity of GABAB receptors makes them well suited to respond to extrasynaptic GABA.
This work was supported by National Institutes of Health Grant HL35034. We thank David Cardozo for advice, and Gabriela Greif, Craig Jahr, Stefan McDonough, Indira Raman, Gary Westbrook, and John Williams for comments on this manuscript.
Correspondence should be addressed to Deborah Sodickson, Volen Center, Brandeis University, 415 South Street, Waltham, MA 02254.
Bruce Bean’s present address: Department of Neurobiology, Harvard Medical School, 220 Longwood Avenue, Boston, MA 02115.