Anesthetics appear to produce neurodepression by altering synaptic transmission and/or intrinsic neuronal excitability. Propofol, a widely used anesthetic, has proposed effects on many targets, ranging from sodium channels to GABAA inhibition. We examined effects of propofol on the intrinsic excitability of hippocampal CA1 neurons (primarily interneurons) recorded from adult rat brain slices. Propofol strongly depressed action potential production induced by DC injection, synaptic stimulation, or high-potassium solutions. Propofol-induced depression of intrinsic excitability was completely reversed by bicuculline and picrotoxin but was strychnine-insensitive, implicating GABAA but not glycine receptors. Propofol strongly enhanced inhibitory postsynaptic currents (IPSCs) and induced a tonic GABAA-mediated current. We pharmacologically differentiated tonic and phasic (synaptic) GABAA-mediated inhibition using the GABAA receptor antagonist SR95531 (gabazine). Gabazine (20 μM) completely blocked both evoked and spontaneous IPSCs but failed to block the propofol-induced depression of intrinsic excitability, implicating tonic, but not phasic, GABAA inhibition. Glutamatergic synaptic responses were not altered by propofol (≤30 μM). Similar results were found in both interneurons and pyramidal cells and with the chemically unrelated anesthetic thiopental. These results suggest that suppression of CA1 neuron intrinsic excitability, by these anesthetics, is largely due to activation of tonic GABAA conductances; although other sites of action may play important roles in affecting synaptic transmission, which also can produce strong neurodepression. We propose that for some anesthetics, suppression of intrinsic excitability, mediated by tonic GABAA conductances, operates in conjunction with effects on synaptic transmission, mediated by other mechanisms, to depress hippocampal function during anesthesia.
GABAA receptors are GABA-operated chloride channels that produce inhibition in at least two modes: via relatively short-lived (“phasic”) fast direct inputs at synapses (e.g., inhibitory postsynaptic potentials; referred to here as GABAphasic) and via induction of a sustained, background form of inhibition (“tonic” GABAA conductances; referred to here as GABAtonic) (Mody 2001). While important roles for GABAphasic are well established, roles of GABAtonic are poorly understood, and the molecular identity and location of receptors underlying GABAtonic are unknown in most brain regions with the notable exception of cerebellar granule cells (Brickley et al. 1996; Hamann et al. 2002). Recent evidence indicates that GABAtonic is present in many brain areas and has an important role in controlling the input-output properties of at least some neurons (Brickley et al. 1996; Hamann et al. 2002; Semyanov et al. 2003). There is also emerging evidence that GABAtonic, like GABAphasic, is activated/enhanced directly or indirectly by many clinically used drugs and some endogenous compounds, including some anesthetics: propofol (Bai et al. 2001); midazolam (Bai et al. 2001; Yeung et al. 2003); vigabatrin (Overstreet and Westbrook 2001); and neuroactive steroids (Stell et al. 2003).
Anesthetics may act by altering fast synaptic input (i.e., enhancing inhibition or suppressing excitation) and/or suppressing the intrinsic excitability of neurons (i.e., depressing directly the ability of neurons to fire action potentials). This study is primarily focused on gaining a better understanding of the relative role (if any) of GABAtonic in producing depression of intrinsic neuronal excitability in neurons in intact acute brain slices, where the neuronal properties should be similar to in vivo. In the case of propofol, a widely used anesthetic, there are many sites of action that may affect intrinsic excitability, including effects on phasic (Trapani et al. 2000) and tonic (Bai et al. 2001) inhibition mediated by GABAA receptors (GABAA-Rs); sodium channels (Ratnakumari and Hemmings 1997; Rehberg and Duch 1999); potassium channels (Magnelli et al. 1992); cation channels (Ih) (Higuchi et al. 2003); calcium channels (Guertin and Hounsgaard 1999; Inoue et al. 1999); glutamate channels (Orser et al. 1995); and glutamate release (Ratnakumari and Hemmings 1997). Assuming that propofol suppresses intrinsic excitability, two critical questions emerge: is GABAtonic enhancement by anesthetics of sufficient magnitude to affect intrinsic excitability and given the existence of other proposed sites of propofol action that would affect intrinsic excitability, are effects on GABAtonic relatively dominant, relatively unimportant, or one contributor among many? To our knowledge, there has been no direct investigation of the role of anesthetic enhancement of GABAtonic in affecting intrinsic neuronal excitability and no attempts to determine the relative role of this mechanism.
Propofol acts on a wide variety of GABAA receptor subtypes to enhance GABA action and/or directly activate the receptors (Trapani et al. 2000; Williams and Akabas 2002). A recent analysis of mice with a point mutation in the GABAA-R beta3 subunit indicates that beta3-containing receptors play an important role in mediating propofol effects in vivo (Jurd et al. 2003), and receptors containing this mutation have very little potentiation by propofol (Siegwart et al. 2002). In a general sense, these results with beta3 mutants provide further strong evidence for an important role for GABAA-Rs in producing propofol anesthesia in vivo.
The hippocampus has been widely used for studies of anesthetics (MacIver 1997; Nicoll and Madison 1982) and appears to be an important locus for anesthetic action (Ma et al. 2002). In this study, we primarily investigated CA1 interneurons because these cells appear to play large roles in controlling CA1 network activity (Fricker and Miles 2001; Klausberger et al. 2003); hence, relatively small effects on these neurons are expected to have large effects on circuit rhythms, which are known to be disrupted during anesthesia (MacIver et al. 1996).
Brain slice preparation methods have previously been described in detail (Nishikawa and MacIver 2000). In short, standard transverse hippocampal slices (400–500 μm) from mature Sprague-Dawley rats (P28–P40, most P33-36) were prepared using a vibratome. All procedures conform to Stanford University and National Institutes of Health guidelines.
Standard visualized slice procedures were used (Nishikawa and MacIver 2000). All interneuron recordings were from nonpyramidal CA1 neurons at or near the border of stratum radiatum and s. lacunosum-moleculare; “giant” neurons (Gulyas et al. 1998) were avoided. Interneurons were highly variable in cell size and shape. Interneurons had relatively nonaccommodating action potentials but varied in resting potential and threshold current required to produce spiking; there was no spontaneous spiking. Input resistances were often ∼350–500 MΩ but varied from ∼200 to ∼850 MΩ. However, there were no obvious differences between cells in response to propofol or thiopental: every interneuron displayed a decrease in both spiking and input resistance. Throughout current-clamp experiments, sets of current steps were applied repetitively at fixed intervals (typically, 6 steps/set: 2 hyperpolarizing, 1 at 0 pA, and 3 depolarizing; 1 set/min). In long current-clamp experiments, we found that neurons tolerated protocols using small numbers of depolarizing current steps repetitively applied (3/set) better than protocols involving sets with large numbers of depolarizing steps. We did not apply tonic currents to adjust resting potentials. Whole cell recordings from pyramidal cells were from visualized neurons in s. pyramidale. All experiments were conducted at room temperature (22–24°C) using a submersion chamber and >95% of tubing was Teflon to minimize drug binding. Using continuous perfusion of artificial cerebrospinal fluid (ACSF) at 2–3 ml/min complete bath replacement took ∼2 min as measured by dye exchange. Each slice was used for only a single experiment. The following external ACSF was used (in mM): 124 NaCl, 3.5 KCl, 1.25 NaH2PO4, 2 MgSO4, 2 CaCl2, 26 NaHCO3, and 10 glucose; it was bubbled with 95% O2-5% CO2 to reach pH 7.4. In some voltage-clamp experiments, this solution was supplemented with 5 mM CsCl (to block Ih) and/or 50 μΜ CdCl2 (to block calcium currents) to improve voltage clamping. When cadmium was used, the NaH2PO4 was eliminated. For high-K+-evoked spiking (Fig. 5), normal ACSF was supplemented with 6.5 mM KCl (replacing NaCl) to lead to a total of 10 mM KCl. We used standard whole cell methods (pipette resistance 4–8 MΩ). For all current-clamp recordings (and voltage-clamp recordings where indicated), a KGluconate-based internal solution was used containing (in mM) 100 KGluconate, 10 EGTA, 5 MgCl2, 40 HEPES, 2 Na2ATP, and 0.3 NaGTP (pH 7.2 with KOH). For voltage-clamp experiments, this solution was supplemented with QX-314 (1–5 mM) to block sodium currents and, in experiments on evoked inhibitory postsynaptic currents (IPSCs; i.e., Fig. 4), the concentration of MgCl2 was changed to 3 mM. KCl internal was 100 KCl, 10 EGTA, 5 MgCl2, 40 HEPES, 2 Na2ATP, 0.3 NaGTP, and 1 QX-314 (pH 7.2 with KOH). For recording of cell-attached spiking (Fig. 4), the pipette was filled with ACSF, a loose seal was formed on a visually identified interneuron, and voltage-clamp mode was used [Vcommand = 0 mV; (Bieda and Copenhagen 2000)]. In most experiments on synaptically evoked spiking and glutamatergic inputs (Fig. 4), the CA1–CA3 connection and the CA1-subiculum connection were cut, and a cut was often made at the border of s, pyramidale and s, radiatum to further isolate activity. For recording of extracellular spiking from CA1 pyramidal cells (Fig. 5), an ACSF-filled pipette (6 MΩ) was placed in s. pyramidale. All voltage values are corrected for electrode offset and junction potential. All recordings were established for >15 min before recording baseline data, and recordings showing unstable response properties (>15% variability) were not used. To stimulate synaptic inputs, a bipolar tungsten electrode was placed in stratum radiatum, as previously described (Nishikawa and MacIver 2000).
All compounds were from Sigma/RBI, except propofol, a kind gift of Zeneca Pharmaceuticals. Propofol was stored frozen as a 100 mM stock in DMSO under nitrogen gas in individual aliquots. Thiopental was made fresh daily as a 12.5 mM stock in distilled water and NaOH.
Data collection and analysis
Whole cell and synaptically evoked spiking data (Figs. 1–4) were collected using PClamp 8.0 (Axon Instruments, Union City, CA) and analyzed using custom procedures written (by M. C. Bieda) in Igor Pro 3.15/4.0 (Wavemetrics, Oswego, OR). Extracellular spiking from stratum pyramidale was collected using DataWave (DataWave Technologies, Long Mount, CO); analysis was performed using this program's spike discrimination and rate meter analyses; results were verified manually. The two-tailed Student's t-test or repeated-measures ANOVA with Tukey test were used to evaluate statistical significance (P ≤ 0.05; GraphPad Prism software, Prism, San Diego, CA).
Determination of the appropriate concentration of propofol for use in in vitro experiments, particularly brain slice experiments, has been quite difficult, with previous studies using from 0.2 μΜ (Antkowiak 1999) to 500 μΜ (Wakasugi et al. 1999). We chose a nominal concentration of 30 μΜ for the majority of experiments based on several considerations, including 1) the excellent correlation across a range of anesthetics (e.g., volatile and barbiturates) between in vitro anesthetic concentrations that suppress CA1 field population spikes (fPS) and clinically relevant in vivo anesthetic concentrations (MacIver 1997) coupled with our previous finding that 30 μΜ propofol suppresses the CA1 fPS to a similar degree as inhalational anesthetics at clinical concentrations (Turnquist et al. 2002); 2) the fact that the concentration of propofol in the slice at the end of our typical drug application period (40 min) is much lower than the applied bath concentration (5–30 μΜ) because steady-state propofol effects require >300 min of application (see discussion for details); and 3) the desire to use an adequate concentration of propofol to assay for several possible previously described sites of action, which exhibit somewhat lower affinities for propofol [e.g., sodium channels (Ratnakumari and Hemmings 1997; Rehberg and Duch 1999)]. Given these considerations, the actual effect-site concentration of propofol in these experiments is unknown, and it is possible that the effect-site concentration may be closer to levels used clinically for neuroprotection than levels used for anesthesia (i.e., 2–5 μM). Because further analysis of this question relies, in part, on the results presented in this report, we defer this topic to the discussion.
Propofol strongly depresses interneuron excitability
Whole cell current-clamp recordings from CA1 interneurons near the border of s. lacunosum-moleculare were performed to assess the effect of propofol on cell excitability, as measured by spiking in response to current injection (“intrinsic excitability”). Propofol (30 μM), in a mostly reversible manner, strongly depressed both the ability of positive current steps to induce action potential production (data not shown; n = 5 cells; spiking (spikes/s): control: 23.6 ± 1.3, propofol: 8.2 ± 1.4, wash: 17.5 ± 2.2; P < 0.001 control vs. propofol) and the neuronal input resistance (equivalent to an increase in input conductance; n = 5 cells, data not shown; conductance (nS): control: 2.4 ± 0.5, propofol: 3.63 ± 0.5, wash: 2.9 ± 0.6; P < 0.05 control vs. propofol). Effects on spiking and input resistance developed slowly and roughly in parallel during a 35-min propofol application. For comparisons between cells, the effect of propofol on current injections leading to ∼20 spikes/s in control was analyzed. This approach provided a means of normalizing responses from cells with differing responses to current injections. Propofol produced a small hyperpolarization (n = 5 cells; −3.0 ± 1.9 mV; not significant), which we did not attempt to compensate for with current injection. Hence, excitability suppression may be primarily due to shunting via conductance increases, although effects on other processes (e.g., sodium channels) could also play a role. This depression of intrinsic excitability is in strong contrast to our previous results with inhalational anesthetics (Nishikawa and MacIver 2000), which do not alter current-evoked discharge in these neurons.
There is extensive evidence that propofol strongly enhances GABAA receptor function (Trapani et al. 2000). Application of the GABAA channel antagonist picrotoxin (50 μΜ), in the continued presence of propofol, effectively reversed the effects of propofol on spiking and input conductance (Fig. 1, A–C). These findings suggest that propofol's depression of intrinsic excitability is largely due to activation of a GABAA-R-operated chloride channel; however, picrotoxin has been shown to depress other currents, including some cation channels (e.g., Das et al. 2003). It is also possible that this current is glycine mediated or comes from a picrotoxin-sensitive non-ligand-gated chloride channel.
To determine more directly whether propofol induced a chloride conductance, as would be expected with activation of GABAA-Rs, we used an ionic replacement approach in voltage clamp experiments (Fig. 1, D–F). Propofol produced a conductance increase with both low- and high-chloride internal solutions, and we found that at –60 mV, propofol produced an outward current with a low chloride (KGluc; ECl = approximately −65 mV) internal solution and a large inward current with a high chloride (KCl; ECl = approximately −5 mV) internal solution. These effects were largely reversed by picrotoxin. Overall (n = 5 cells in each condition), propofol produced a small outward current in KGluc-loaded cells (33 ± 11 pA) and a large inward current in KCl-loaded cells (−198 ± 51 pA). With KCl internal solutions, we used short (∼20 min) applications of propofol because, in initial experiments, we observed significant cell swelling with our typical (40- to 60-min application) propofol protocol. It is important to note that these effects were not observed with KGluconate internal solutions, which more closely reflect the internal chloride concentration of neurons in the intact brain. Therefore the quantified propofol-induced increases in inward current and conductance for KCl experiments are underestimates. To determine approximate minimum applied levels of propofol that activate this chloride conductance, we applied propofol at different concentrations for 20–30 min (KCl internal, Vh = −60 mV), to allow comparison to current measurements at 30 μΜ, and quantified the increase in current. We found significant currents at both 5 and 10 μΜ using this protocol (Fig. 1G; n = 4 each concentration; 5 μΜ: −28 ± 7 pA; 10 μΜ: −61 ± 8 pA). For comparison purposes, the data for 30 μΜ propofol, which used an identical protocol, is replotted here (from Fig. 1F, left). These results suggest that propofol at applied concentrations as low as 5 μΜ also activate this chloride conductance. Although glycine receptors are present in young adult rat (3–4 wk old) hippocampal CA1 pyramidal cells and interneurons (e.g., Chattipakorn and McMahon 2002), propofol suppression of spiking and enhancement of conductance was found to be insensitive to the glycine receptor antagonist strychnine and sensitive to GABAA-R antagonist bicuculline, indicating that these propofol effects were due to GABAA-R but not glycine receptors (Fig. 1H). In voltage-clamp experiments testing strychnine and bicuculline, we used a 1:1 mixture of Kgluconate and KCl internal solutions to avoid the problems encountered with the KCl solution in our previous voltage-clamp experiments (see above text regarding Fig. 1, D–F). Taken together, these results support the hypothesis that propofol activates a GABAA-R associated chloride conductance at applied concentrations as low as 5–10 μΜ.
Differential pharmacology of GABAA conductances
A number of studies have examined block of GABAphasic and GABAtonic by SR95531 (Bai et al. 2001; Brickley et al. 2001; Hamann et al. 2002; Nusser and Mody 2002; Overstreet and Westbrook 2001; Semyanov et al. 2003; Stell and Mody 2002). All of these studies agree that 20 μM SR95531 is sufficient to completely block GABAphasic; however, some of these studies find that 20 μM SR95531 will strongly block GABAtonic, whereas others find that GABAtonic is insensitive to 20 μM SR95531. We examined the ability of this relatively high concentration of SR95531 (20 μM) to discriminate GABAphasic and GABAtonic with the logic that if a high concentration were sufficient to differentiate the populations, we could be assured of blocking GABAphasic, whereas mostly preserving GABAtonic.
To test this differential pharmacology, we evoked monosynaptic IPSCs [recorded using 17.2 μΜ 6-cyano-7-nitroquinoxalene-2,3-dione (CNQX) and 100 μΜ 2-amino-5-phosphonovaleric acid (APV) to block glutamate receptors] in neurons using s. radiatum stimulation. To avoid problems associated with large propofol-induced chloride currents (see Fig. 1E and preceding text discussion), we used a low-chloride (KGluc) internal solution. Assuming that SR95531 can be used to differentiate phasic and tonic GABAA conductances (Bai et al. 2001), we expected to observe a propofol-induced enhancement of the IPSC and the increase in holding currents we had already observed. This increase in holding current should be composed of both summating spontaneous IPSCs [which are augmented in propofol (Bai et al. 2001)] and a component due to GABAtonic. Thus we expected that SR95531 would block a portion of the increase in holding current. Picrotoxin should block the remaining increase in holding current.
As expected, we found that propofol both enhanced the evoked IPSC and produced a background tonic current (Fig. 2). Additive application of SR95531 (20 μΜ) completely blocked IPSCs but only partially reduced the tonic current, whereas additive application of picrotoxin fully blocked the tonic current. Although the data appear to show that SR95531 only partially inhibits the tonic current for a short period of time (a “transient” block), a small sustained block of the tonic current could be masked by the slow rise of the tonic current, implying that in actuality SR95531 may be producing a sustained partial block of the tonic current. Most cells showed a smaller “transient” effect of SR95531 on the holding current than seen in Fig. 2B. All cells (n = 6; 1 cell was lost before application of SR95531) showed a similar pattern of effects with the exception of variable changes in IPSC amplitude. On average, the maximal (and transient) effect of SR95531 reduced the total holding current by ∼12% [Ihold (total): propofol: 67 ± 9 pA; transient minimum in Pro + SR: 59 ± 10 pA], which corresponds to a decrease of ∼20% in the net propofol-induced current. However, by the end of the SR95531 application period (15–20 min), the holding current had recovered to the same amplitude as in propofol alone (propofol: 67 ± 9 pA, propofol + SR: 68 ± 13 pA). Overall, statistically significant differences were found at the P < 0.01 (or better) level for holding current (Ihold) with the following comparisons: Con versus +Pro, Con versus +SR, +Pro versus +Pic, and +SR versus +Pic, and no significant (P > 0.05) differences were found for +Pro versus +SR and Con versus +Pic. In contrast to the effect on the eIPSC, propofol's enhancement of Ihold had a delayed onset and failed to saturate during the application period. This is consistent with a model in which there is slow propofol entry into the slice (Hollrigel et al. 1996; Turnquist et al. 2002) and a higher affinity of GABAphasic versus GABAtonic for propofol. The SR95531-mediated partial and transient inhibition and the decrease in variance in Ihold due to SR95531 may reflect blockade of background summating synaptic spontaneous IPSCs. Every individual cell showed a complete block of GABAphasic (both for spontaneous and evoked IPSCs) and at most a transient and partial block of GABAtonic by SR95531. These findings demonstrate that SR95531 and picrotoxin can be used to differentiate phasic and tonic GABAA conductances in hippocampal CA1 neurons.
Role of GABAtonic in propofol neurodepression
In whole cell current-clamp experiments (Fig. 3), SR95531 (20 μΜ, 15 min) failed to reverse the effects of propofol; in fact, the depression of excitability continued to increase during SR95531 application. Additive application of picrotoxin fully reversed propofol effects. Overall, statistically significant differences were found at the P < 0.001 level for both conductance and spiking for the following: Con versus +Pro, Con versus +SR, +Pro versus +Pic, and +SR versus +Pic, and no significant (P > 0.05) differences were found between +Pro versus +SR and Con versus +Pic. Hence, the effects of propofol on intrinsic excitability appeared to be either mostly or entirely due to enhancement of GABAtonic.
We next sought to determine whether propofol suppressed synaptically evoked spiking in a similar manner. To avoid potential whole cell dialysis effects and the forcing of intracellular chloride concentration to pipette concentration, we generally employed a cell-attached patch mode to monitor action potential production (Bieda and Copenhagen 2000). Electrical stimuli in s. radiatum were adjusted so that 50–90% of trials resulted in spike production under control conditions [probability of spiking—P(s); percentage calculated using 20–30 consecutive trials in each condition].
In a representative experiment (Fig. 4A), propofol decreased P(s) in a picrotoxin-sensitive manner. Because similar results were found in all cells tested using either whole cell recording mode (n = 5: Con: 66 ± 6%; Pro: 6 ± 3%; Pro + Pic: 96 ± 3%) or cell-attached recording mode (n = 5: Con: 69 ± 4%; Pro: 2 ± 1%; Pro + Pic: 94 ± 2%), these data sets are accumulated together in Fig. 4C (left). These data suggest that propofol depressed synaptically evoked action potential discharge via GABAA conductances and that whole cell dialysis does not significantly affect propofol's actions.
Given the powerful enhancement of IPSC charge transfer by propofol (see Fig. 4), propofol's depressant effect in these experiments could occur primarily via enhancement of GABAphasic. If propofol primarily acts via GABAphasic, then SR95531 should largely reverse the depression by propofol, and further additive application of picrotoxin should have little effect. In a representative experiment (Fig. 4B), propofol's actions were largely SR95531 insensitive and picrotoxin sensitive. In all cells tested (n = 5; Fig. 4C, right), propofol strongly and consistently depressed P(s). Overall, statistically significant differences were found at the P < 0.05 (or better) level for the following: Con versus +Pro, Con versus +SR, +Pro versus +Pic, and +SR versus +Pic, and no significant (P > 0.05) differences were found with +Pro versus +SR and Con versus +Pic, indicating that effects were overall SR95531 insensitive and picrotoxin sensitive. In addition, in all cases (5/5), application of picrotoxin in the presence of SR95531 increased P(s), and in 80% (4/5) of cells, P(s) in picrotoxin was much greater than in SR95531 (means for these 4 cells: Con: 70 ± 5%; +Pro: 9 ± 5%; +SR: 18 ± 7%; +Pic: 90 ± 6%). These strong picrotoxin effects coupled with the weak and statistically nonsignificant SR95531 effects indicate that, under our conditions, suppression of synaptically evoked spiking by propofol occurs largely via activation of GABAtonic. However, these results must be interpreted with caution; for in vivo conditions, where the ratio and timing of inhibitory and excitatory postsynaptic potentials (IPSPs and EPSPs) could be different, GABAphasic may play a much larger, even dominant role.
Propofol has been described as suppressing glutamate exocytosis (Lingamaneni et al. 2001; Ratnakumari and Hemmings 1997) and potentially could affect AMPA/K receptors on interneurons. The protocols employed in Fig. 4, A–C, could fail to reveal these effects because a large increase in P(s) by elimination of GABAergic inhibition could overwhelm a small decrease in P(s) produced by a weaker suppression of glutamatergic transmission. To test this possibility, we evoked pure glutamatergic EPSCs and EPSPs in interneurons using stratum radiatum stimulation and picrotoxin (100 μΜ) to block GABAA-mediated responses. We found that propofol (30 μΜ, 40 min) did not affect the amplitude of glutamatergic synaptic responses (EPSCs or EPSPs; Fig. 4, D–F). Therefore these results suggest that propofol does not affect glutamate release or AMPA/K receptors on these cells.
Propofol suppresses CA1 pyramidal cell excitability
A selective strong suppression of inhibitory interneurons should be strongly proconvulsant, which is not consistent with anesthetic actions. This effect would be prevented by a propofol suppression of excitatory neurons in the circuit. Hence, we tested the effects of propofol on CA1 pyramidal cells. Propofol induces a tonic GABAA-R-mediated current in CA1 pyramidal cells (Whittington et al. 1996). Using a KCl-based solution in voltage clamp (n = 3; protocol as in Fig. 1E), we found that propofol (30 μΜ) induced a conductance (control: 4.2 ± 0.8 nS; propofol: 7.5 ± 0.6 nS; P < 0.05) that was largely picrotoxin-sensitive (to 5.2 ± 0.9 nS). Propofol induced a significant inward current (at −60 mV, −167 ± 21 pA) consistent with a chloride conductance. These results indicate that propofol also suppressed CA1 pyramidal cell excitability via activation of GABAA conductances.
Hippocampal pyramidal cells often show strong spike adaptation due to activation of calcium-activated potassium channels (Madison and Nicoll 1986); the activity of these channels would probably be greatly affected by whole cell dialysis in our long (∼80 min) experiments. To avoid this problem, we induced spiking by continuously applying an external solution containing 10 mM potassium and performed extracellular multiunit recordings in the CA1 s. pyramidale. GABAA receptor activity was isolated by applying a combination of CNQX/APV/CGP55845A to block glutamate and GABAB receptors. Bicuculline (10 μΜ) fully reversed propofol (10 μΜ) suppression of pyramidal cell excitability (propofol: 19 ± 5% of control; bicuculline + propofol: 104 ± 8% of control; data not shown), implicating GABAA conductances. In all subsequent experiments, we added SR95531 to all solutions to block GABAphasic and used 10 μΜ propofol. Propofol produced a picrotoxin-sensitive depression of multi- and single-unit activity that slowly developed during lengthy (60 min) applications but did not appear to affect the size or shape of action potentials, suggesting that activation of GABAtonic by propofol can strongly reduce the intrinsic excitability of CA1 pyramidal neurons (Fig. 5).
Role for GABAtonic in thiopental neurodepression
To determine whether other anesthetics might be able to suppress intrinsic excitability via GABAtonic, we tested thiopental, a widely used intravenous barbiturate anesthetic. Thiopental has been reported to induce a sustained GABAA-R-mediated current in pyramidal cells (Lukatch and MacIver 1996; Whittington et al. 1996). Thiopental at a concentration of 80 μM produces deep anesthesia in rats (MacIver et al. 1996) and depresses CA1 fPS to a similar extent as volatile anesthetics at clinical concentrations as well as 30 μM propofol (MacIver 1997; Turnquist et al. 2002). As with propofol, we found that thiopental (80 μM) strongly but slowly depressed both neuronal excitability and input resistance in a largely SR95531-resistant and picrotoxin-sensitive fashion [n = 5 cells; data not shown; spiking (Hz): Con: 20.5 ± 1.2, thiopental: 6.4 ± 1.4, thiopental + SR: 4.7 ± 3.2, thiopental + SR + picrotoxin: 21.1 ± 1.8; conductance (nS): Con: 2.5 ± 0.3, thiopental: 3.7 ± 0.4, thiopental + SR: 4.1 ± 0.4, thiopental + SR + picrotoxin: 1.9 ± 0.3]. Overall, statistically significant differences were found at the P < 0.001 level for both conductance and spiking with the following comparisons (TP, thiopental): Con versus +TP, Con versus +SR, +TP versus +Pic, and +SR versus +Pic, and no significant (P > 0.05) differences were found with +TP versus +SR. A small, but significant (P < .05) difference was observed between Con and +Pic. These results indicate that thiopental, like propofol, depressed neuronal excitability and enhanced input conductance by activating GABAtonic.
These data represent, to our knowledge, the first direct demonstration that anesthetics can suppress intrinsic neuronal excitability as measured by action potential production, by activating GABAtonic. We found that, under our conditions, suppression of intrinsic excitability by propofol was completely reversed by blockade of GABAtonic; hence, this conductance appears to play a dominant role in producing this anesthetic-induced suppression of intrinsic excitability in these neurons. However, for in vivo conditions, effects on sites that alter synaptic input (in particular, direct enhancement of GABAphasic as in Fig. 2) will play important roles, and assessing the relative importance, overall, of GABAtonic in producing anesthesia warrants further study. Generally, suppression of intrinsic excitability should have a variety of effects, including suppression of global firing activity, a reduced probability of long-term potentiation induction and depression of rhythmic neuronal activity. Disruption of rhythmicity by propofol or thiopental has been observed in vivo (MacIver et al. 1996; Tomoda et al. 1993) and in brain slices (Lukatch and MacIver 1996; Whittington et al. 1996); these effects may be due to effects on GABAtonic, but effects on GABAphasic would also certainly contribute. We anticipate that future directions relating to this work will focus on evaluating the role of GABAtonic in producing anesthesia in vivo, generalizing the results in this study to other anesthetics, other brain regions (especially spinal cord, thalamus, and neocortex) as well as determining the properties of GABAtonic.
Previous studies have described several sites of action for propofol (see introduction). Our propofol concentrations (5, 10, and 30 μM) and experimental protocols (DC injection, synaptically evoked spiking, high-K+-evoked spiking; use of both whole cell and extracellular recording) were designed to allow detection of effects of these previously proposed sites for propofol depression of intrinsic excitability. None of these other sites appeared to play an important role in producing propofol effects on intrinsic excitability but might contribute to actions on synaptic transmission. Alternatively, roles of other mechanisms in suppressing intrinsic excitability may occur with higher propofol concentrations and/or in other brain regions. Our results suggesting a powerful role for GABAA conductances are consistent with recent in vivo evidence showing that a point mutation in the beta-3 GABAA subunit can markedly attenuate propofol anesthesia in mice (Jurd et al. 2003). Our data strongly and clearly implicated GABAtonic in mediating propofol's suppression of intrinsic excitability. In the case of synaptically evoked spiking, exact discrimination of relative roles of phasic and tonic GABAA conductances in propofol action is complicated by the expected synergistic action of propofol at these two sites, variability in the amount of synaptic inhibitory input into interneurons, and effects from the timing of action potential production during EPSP-IPSP sequences. Further research will be needed to understand better the roles of GABAphasic, GABAtonic, and other mechanisms in producing anesthesia in vivo. Interestingly, sedation, which occurs with low concentrations of propofol (much lower than surgical concentrations of propofol), appears to be SR95531-sensitive and may occur primarily via enhanced GABAphasic (Nelson et al. 2002). However, because high concentrations of SR95531 can also block GABAtonic (Nusser and Mody 2002; Yeung et al. 2003), linking sedation to actions on GABAphasic is not straightforward.
Propofol action at GABAtonic
Several distinct mechanisms potentially underlie propofol enhancement of tonic GABAA conductances. Both propofol potentiation of GABA action at GABAA receptors and direct propofol activation of GABAA receptors have been extensively documented (Trapani et al. 2000; Williams and Akabas 2002) and can occur at quite low propofol concentrations (≤ 1 μΜ) in dissociated CA1 pyramidal cells (Hara et al. 1993, 1994). Propofol enhancement of GABA release (Hollrigel et al. 1996; Murugaiah and Hemmings 1998) could increase the tonic GABA concentration and thereby increase tonic GABAA conductance. Differentiation of these mechanisms will require a better understanding of GABAtonic, but there is currently a limited understanding of the activation properties, function, and identity of the receptors producing GABAtonic. A recent report (Semyanov et al. 2003) indicates that in agreement with our results, GABAtonic exists in both CA1 pyramidal cells and CA1 interneurons. The low background concentration of GABA (0.8–2.9 μΜ) (Lerma et al. 1986) and the sustained nature of the current imply that the receptors producing GABAtonic are probably relatively high affinity and relatively nondesensitizing.
A point mutation in the GABAA-R beta3 subunit (N265M) nearly abolishes propofol action at beta3-containing receptors (Siegwart et al. 2002). Mice with this beta3 (N265M) receptor show much reduced sensitivity to propofol (Jurd et al. 2003), indicating an important role for beta3 mechanisms. However, the continued (but depressed) sensitivity to propofol of these mice in the loss of righting reflex assay suggests that non-beta3 receptors probably also play a role. Mice with a beta2 mutation (N265S) still show propofol effects (Reynolds et al. 2003), and propofol still potentiates GABA action at receptors containing beta2 subunits with mutations of this residue (N265M) (Siegwart et al. 2003), so these results are difficult to interpret for propofol. Hippocampal GABAtonic cannot be the same as cerebellar granule cell extrasynaptic GABAA conductances because the cerebellar α6-containing receptors are not found in hippocampus (Sperk et al. 1997; Wisden et al. 1992) and are blocked by SR95531 (Brickley et al. 2001). The receptor-channels producing GABAtonic in CA1 may differ from those producing GABAphasic in conductance (Bai et al. 2001) and kinetic and modulatory properties (Banks and Pearce 2000). In CA1 pyramidal cells, the tonic conductance is produced by receptors containing alpha5 subunits but not delta subunits; these receptors are most likely alpha5beta3gamma2 (Caraiscos et al. 2004). Hence, in our experiments on pyramidal cells, propofol is probably acting via these receptors. For interneurons, the composition of the GABAA-Rs is much less clear. Interneurons are highly variable, express a variety of GABAA subunits (at least alpha1, beta2, delta, gamma2) (Sperk et al. 1997), and their GABAA-R subunit composition has not been studied as extensively as that in pyramidal cells. However, alpha5 and beta3 subunits are not prominent in interneurons (Sperk et al. 1997), so the interneuron tonic conductance probably differs in molecular composition from that of pyramidal cells. Some interneurons do contain delta subunits (Sperk et al. 1997), which confer high GABA sensitivity to extrasynaptic/perisynaptic receptors that produce tonic currents in dentate granule cells and cerebellar granule cells (Stell et al. 2003). However, because these subunits do not appear to be found in all interneurons (Sperk et al. 1997) but we observed strong propofol effects in all interneurons, it is probable that, as in pyramidal cells, gamma subunits may be important components of the GABAA-Rs underlying the propofol-induced tonic conductance in some interneurons.
We are not aware of an expressed GABAA-R that is SR95531 insensitive and picrotoxin sensitive. GABAtonic may simply represent a monoliganded state of receptors with the same molecular composition as phasic receptors for some interneurons (Bai et al. 2001) or the unbinding rate for SR95531 may be too slow to permit phasic receptors to become unblocked during the short time that GABA is present in a synaptic cleft (Jones et al. 1998). Currently, it is difficult to discriminate between these possibilities.
GABAtonic in other cells and with other anesthetics
We found that the excitability of both GABAergic hippocampal interneurons and glutamatergic CA1 hippocampal pyramidal cells was strongly suppressed by propofol activation of GABAtonic, suggesting that other neuronal types may be affected similarly. Several lines of evidence suggest that GABAtonic may be a general locus of action for some, but not all, types of general anesthetics. We found that thiopental, a barbiturate general anesthetic, chemically unrelated to propofol, also strongly suppressed interneuron intrinsic excitability via GABAtonic; however, anesthesia produced by thiopental in vivo probably involves a strong synaptic component involving other mechanisms. Addition of various anesthetics, including barbiturates [methohexital (Zhang et al. 1998); pentobarbital (Uchida et al. 1996)], steroid anesthetics [alphaxalone (Ueno et al. 1997)], and other anesthetics [etomidate (Uchida et al. 1996)], produces currents or hyperpolarizations that are bicuculline sensitive and relatively insensitive to SR95531. Hence, several other general anesthetics may target GABAtonic; however, this has not been shown directly and the relative importance of GABAtonic for depression of intrinsic excitability for these other anesthetics is unknown. The benzodiazepine midazolam only weakly potentiates GABAtonic (Bai et al. 2001), but it will be important to determine whether this site plays a role in anesthetic-mediated cell death (Jevtovic-Todorovic et al. 2003). In contrast to these results, investigation of volatile anesthetics (e.g., halothane and isoflurane) has failed to reveal cellular effects consistent with activation of GABAtonic (Nishikawa and MacIver 2000).
Concentration of propofol
Measuring directly the free concentration of propofol in the brain during various states of anesthesia is not currently possible and is complicated by propofol's high protein binding (Servin et al. 1988), lipophilicity (Tonner et al. 1992), and preferential partitioning into brain tissue (Shyr et al. 1995). Previous brain slice studies have used from 0.2 μΜ (Antkowiak 1999) to 500 μΜ (Wakasugi et al. 1999). We found propofol effects at 5 and 10 μΜ (Figs. 1G and 5). There is strong evidence that the propofol concentration in a brain slice is significantly less than applied perfusate concentrations. First, our experiments consistently show strong evidence for slow propofol entry—propofol effects developed slowly and often continued to increase after 40–60 min of application time (e.g., Figs. 2B, top, and 5). Second, suppression of CA1 fPS amplitude by 5 μΜ propofol continues to increase even after >210 min of application (Turnquist et al. 2002) in contrast to the rapid onset and stabilization of effects in isolated neurons (e.g., <0.5 min) (Bai et al. 2001; Hara et al. 1993). Third, slices often exhibit a high diffusional/binding barrier for entry of compounds (Dunlap et al. 1994), which should be particularly important for propofol given its strong protein binding and high lipid solubility. Finally, a previous study (Hollrigel et al. 1996) also concluded that propofol enters brain slices slowly and that actual concentrations in the slice are almost certainly lower than perfusate concentrations, over the time course of our studies. Hence, the actual free effect site concentration of propofol acting within 30–40 min in our experiments is probably much lower than the applied concentration. An in vivo study in rats (Shyr et al. 1995) indicates that clinically relevant active site concentrations during anesthesia may be ≥10 μΜ, a concentration which activated GABAtonic (Figs. 1G and 5) in our experiments. We did not test concentrations higher than 30 μΜ because this concentration already produced profound neurodepression. In part to address concentration issues, we also investigated thiopental, for which there is better pharmacodynamic and pharmacokinetic data. Thiopental was studied using 80 μΜ, a concentration of thiopental previously described as corresponding to deep anesthesia in rats (MacIver et al. 1996). This anesthetic also depressed neuronal excitability with a similar time course in our experiments.
In conclusion, enhanced GABAtonic is likely to play an important role in the CNS depression that is associated with propofol and thiopental anesthesia. The depression of intrinsic neuronal excitability produced by these anesthetics would lead to decreased discharge frequencies of both pyramidal cells and inhibitory interneurons—disrupting the circuit functions of hippocampal neurons. This could contribute to the loss of recall, suppression of sensorimotor integration, and hypnosis that is produced by anesthetics.
This research was supported by National Institute of General Medical Science Grant 54767.
We acknowledge excellent technical support from F. Monroe and P. Turnquist. We thank Dr. Martin Angst for aid in statistical analysis and helpful discussion regarding pharmacokinetic and pharmacodynamic reasoning and Dr. James Trudell for advice on use of thiopental.
Present address for M. Bieda: Santa Fe Institute, 1399 Hyde Park Rd., Santa Fe, NM 87501.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by the American Physiological Society