Dual Actions of Volatile Anesthetics on GABA^AIPSCs : Dissociation of Blocking and Prolonging Effects

Young rats (14 to 42 days old) were decapitated under enflurane anesthesia, and the heads were immersed immediately in cold (4 [degree sign]C) artificial cerebrospinal fluid (ACSF; composed of 127 mM NaCl, 1.21 mM KH²PO⁴, 1.87 mM KCl; 26 mM NaHCO³, 2.17 mM CaCl², 1.44 mM MgSO⁴, and 10 mM glucose) saturated with 95% oxygen and 5% carbon dioxide. A block of tissue containing both hippocampi was removed with the brain immersed in ACSF, and the tissue was glued to a vibratome tray with cyanoacrylate glue. Slices (400 [micro sign]m) were cut and submerged at 35 [degree sign]C for 1 h before being transferred to the recording chamber, which was perfused at 3 ml/min with ACSF saturated with 95% oxygen and 5% carbon dioxide at 24 [degree sign]C.

Cells in stratum pyramidale of CA1 were visualized using a video camera (Hamamatsu C2400; Hamamatsu, Japan) connected to an upright microscope (Zeiss Axioskop; Thornwood, NJ) equipped with an infrared bandpass filter (Chroma D775/220; Brattlesboro, VT), a long working-distance water-immersion objective (Zeiss Achroplan 40X, 0.75 numerical aperture), and differential interference contrast optics (Nomarski). Whole-cell recordings were obtained at room temperature (24 [degree sign]C) using an Axopatch 1D patch-clamp amplifier and pClamp software (Axon Instruments, Foster City, CA). Data were filtered at 5 kHz, sampled at 10 kHz (Digidata 1200), and stored on a Pentium-based PC (Microsoft, Redmond, WA). Patch pipettes were fabricated from borosilicate glass (Garner KG-33; Garner Glass, Claremont, CA; 1.7-mm outer diameter, 1.1-mm inner diameter) using a Flaming-Brown two-stage puller (model P-87; Sutter Instruments, Novato, CA), fire polished, and coated with Sylgard to reduce electrode capacitance. Tight-seal whole-cell recordings were obtained using standard techniques. Patch pipettes had open-tip resistances of 2-4 M Omega when filled with the recording solution (composed of 140 mM CsCl, 10 mM NaCl, 10 mM HEPES, 10 mM BAPTA, 2 mM MgATP, and 5 mM QX-314, pH 7.3). Access resistances were typically 10-20 M Omega and were compensated by 60% to 80%. Cells were held at -60 mV. The mIPSCs were isolated by bath application of 20 [micro sign]M CNQX and 40 [micro sign]M D,L-APV to block AMPA and N-methyl-D-aspartate-mediated (NMDA) currents, and by the inclusion of CsCl and QX-314 in the patch pipette to block GABA^B-mediatedcurrents. The remaining currents were blocked completely by bath application of 10 [micro sign]M bicuculline (not shown).

APV, CNQX, TTX, and bicuculline were prepared at X50 to 100 stock solutions in 0.9% saline and applied using syringe pumps (model 55-1111; Harvard Apparatus, Natick, MA) set to flow at 1% or 2% of the ACSF flow rate of achieve the desired bath concentrations.

The ACSF was bubbled with enflurane, isoflurane, or halothane using calibrated vaporizers, and the gas-phase concentrations were monitored continuously using a gas monitor (Multigas Monitor 602, Criticare Systems, Waukesha, WI). Teflon tubing was used to eliminate loss of anesthetic between the ACSF reservoir and the recording chamber. To determine the extent of loss of agent between the outflow of the perfusion line and the tissue (a distance of about 1 cm), liquid phase measurements were performed using the observed sensitivity of Ca^2+-sensitiveelectrodes to volatile agents. [19]The Ca²⁺anesthetic-sensitive electrode (MI-600, Microelectrodes, Bedford, NH) was calibrated using solution flowing in the perfusion line and responded linearly for the concentration range used in these experiments (Figure 1). After we obtained three or four calibration points at different anesthetic concentrations, we placed the electrode in the center of the recording chamber, in approximately the same position as a typical slice, with the objective lowered into the bath and the flow rate adjusted to 3 ml/min. The resulting change in voltage in response to a single anesthetic concentration was then measured, and the actual concentration in the recording chamber was calculated from the regression line to the calibration data (Figure 1). This measurement was made for each agent at least once, and for enflurane three times, and yielded consistent losses of 10% to 15%. An average value of 12% was used to correct gas phase concentrations before calculating aqueous phase concentrations.

Aqueous phase concentrations (C^Aq) were calculated from gas phase concentrations (P) using the formula Equation 1where [small lambda, Greek] is the Ostwald solubility coefficient, which varies with temperature T. [20]The measured value of [small lambda, Greek] for isoflurane is 1.04 at 25 [degree sign]C, and for halothane it is 1.20 at 25 [degree sign]C. [21]The solubility coefficient for enflurane has not been measured at 25 [degree sign]C but was estimated to be 1.08 at 25 [degree sign]C based on the relation between lambda and the temperature for isoflurane. [21,22] 

Aqueous median effective concentration (EC⁵⁰) values in rats (i.e., the aqueous concentration corresponding to the minimum alveolar concentration (MAC) and taking into account temperature and solubility [20]) for each anesthetic at room temperature were estimated by assuming that the observed relation between temperature and EC⁵⁰for halothane, [23]Equation 2also applies to isoflurane and enflurane. Using this relation, EC⁵⁰values at 25 [degree sign]C for halothane, enflurane, and isoflurane were estimated as 0.25, 0.30, and 0.58 mM, respectively. We refer to these concentrations as MAC^Aq.

Volatile agents were applied for 10 min in initial experiments to maximize the concentration range applied to individual cells. In some cases it was noted that anesthetic effects had not reached steady state in that time period, but exponential fits to the peak amplitude time series data extrapolated to steady state indicated that the responses were within 74% to 98% of their steady state level. In all of these cells (n = 4), the concentrations of isoflurane and enflurane applied were matched, and there were no significant differences when comparing agents for parameters measured at the end of 10 min in anesthetic versus the extrapolated steady state values (P > 0.1 by a paired Student's t test). In the remaining experiments, applications were 15-20 min, which was always sufficient to reach steady state. All effects of volatile agents were observed to be fully reversible, unless the recording was lost before the wash (n = 6 cells) or the cell was exposed to a different concentration of the same agent without washing (n = two applications in two cells).

Analysis of mIPSC frequency and amplitude to distinguish presynaptic from postsynaptic effects of experimental manipulations has been used successfully in many studies during the past decade. [24,25]By analogy, with results at the neuromuscular junction, [26,27]mIPSCs are assumed to represent the spontaneous fusion of individual vesicles or quanta of the neurotransmitter with the presynaptic membrane. [28]Direct experimental evidence in support of the quantal hypothesis at central synapses [29]argues that this assumption is valid. From this hypothesis it follows that only actions on the presynaptic nerve terminal can alter the frequency of these spontaneous fusion events, and thus changes in the frequency of mIPSCs are indicative of presynaptic drug actions. To interpret changes in the mean amplitude of the responses to these spontaneous fusion events, the assumption is made that the receptors apposed to the release site are saturated by the neurotransmitter released. Alterations in mIPSC peak amplitude can then reasonably occur only by changes in the responsiveness of the postsynaptic receptors. There are substantial experimental and theoretical data to support this assumption at many central synapses, [30]but it should be noted that there are synapses at which mIPSC amplitude depends on neurotransmitter concentration. [31]Additional experiments performed using exogenous GABA applications did not rely on this assumption and confirmed that the peak postsynaptic response is reduced by volatile agents in a manner predicted by the observations on mIPSCs.

GABA was applied focally under visual control using a picospritzer (General Valve; Fairfield, NJ). Puffer pipettes were fabricated from sharp electrodes that were pulled and broken to tip diameters of approximately 0.5 [micro sign]m and filled with GABA (100 mM, pH 3). A relatively high agonist concentration was used to mimic the high concentration believed to occur after synaptically released GABA. However, the concentration actually achieved at the postsynaptic receptors was undoubtedly less than the concentration in the pipette. After whole-cell access was established with the recording electrode, the puffer pipette was placed adjacent (within 5 [micro sign]m) to the cell. Holding currents of 100-500 nA were used to prevent GABA from leaking from the tip. Brief (3-5 ms) pressure pulses (5-20 pounds per square inch) were used to elicit currents that decayed with kinetics that were similar to synaptic currents.

Data were analyzed on a Pentium-based personal computer using ClampFit (Axon Instruments), Origin (Micro-Cal, Northampton, MA), and StatMost (DataMost, Los Angeles, CA). Data were filtered off-line at 2 kHz. Spontaneous events were analyzed using an automated event detection algorithm that measured IPSC amplitude, 10-90% rise times (t^rise), and the time to 63% decay (t^decay). [32]The t^decaymeasurement was discarded if the next event occurred during the decay tail ("stacked" events). Because t (^rise) much < t^decay, the amplitude measurement was reliable even for stacked events. During the analysis, the detection algorithm constantly updated the baseline and thus was unaffected by changes in baseline holding current. The amplitude threshold was set as 3 [middle dot][small sigma, Greek]^noise, where [small sigma, Greek]^noisewas measured during periods of no visually detectable events. Under control conditions, [small sigma, Greek]^noisewas typically <3 pA, and the detection algorithm successfully detected more than 98% of fast-rising mIPSCs. In those cases in which the anesthetic application caused an increase in [small sigma, Greek]^noise, the final 3 min of the application of the highest concentration of volatile agent were analyzed to determine the threshold, and this value was used to analyze all the data for that particular cell. When more than one drug was applied to the same cell, the threshold was selected from the data with the largest value of [small sigma, Greek]^noiseand used to analyze all of the data from that cell.

Isoflurane consistently caused a greater increase in [small sigma, Greek]^noisethan did enflurane, which could skew the detected mIPSCs toward larger amplitude events and thus cause an overestimation of the difference in blocking effect between the two agents. However, this did not appear to be a significant problem in our data, for two reasons. First, in four of five cells to which 0.6 mM isoflurane was applied, and one of five cells to which 1.2 mM isoflurane was applied, the equivalent concentration of enflurane was also applied, and thus the same threshold was used to analyze data in response to both agents. The remaining enflurane data were reanalyzed with the threshold set to be the original threshold multiplied by the ratio of the average increase in [small sigma, Greek]^noisein response to isoflurane divided by the average increase in response to enflurane (i.e., 1.9/1.25 for 1.2 mM enflurane). Although reanalyzing the data with these higher thresholds caused a small decrease (<12%; i.e., the largest change was from 52% to 46% block) in the estimate of the blocking effect in response to high concentrations of enflurane, the difference in blocking effects between isoflurane and enflurane was not significantly affected.

To characterize the decay kinetics of fast IPSCs, a subset of events also was selected for exponential curve fitting. Events were selected only if no other event occurred within 250 ms of the peak. The IPSCs typically were described best by two exponential components. To quantify the overall decay time, we computed the weighted time constant [small tau, Greek]^wt=(A¹[small tau, Greek]¹+ A²[small tau, Greek]²)/(A¹+ A²), where A¹is the amplitude of the ith component. This measure was especially useful at high concentrations of anesthetic, because for the noisiest data [small tau, Greek]¹and [small tau, Greek]²could vary widely depending on the initial values of the fit parameters, whereas [small tau, Greek]^wtwas observed to be relatively stable.

Effects on peak amplitude and interevent interval were evaluated using the Kolmogorov-Smirnoff test, with a significance level of P < 0.01. Exponential curves were fit to IPSC decays using the Levenberg-Marquardt non-linear least-squares algorithm. Dose-response data were fit by Hill equations (y =[1 - y sub [infinity][middle dot][1 +(x/k^1/2)ⁿ](^-1)+ y sub [infinity], where x is the concentration of anesthetic, y sub [infinity] is the saturating response, n is the Hill coefficient, and k^1/2is the concentration yielding a half-maximal response; i.e., EC⁵⁰or the median inhibitory concentration [IC⁵⁰]), also using the Levenberg-Marquardt algorithm. Uncertainties in the parameter estimates are presented as the square roots of the diagonal elements of the covariance matrices. These fits were used solely to quantify the data and allow easier comparison between anesthetic agents. Because of the limited concentration range tested, any physical interpretation of these parameters should be made with care. Statistical comparisons between anesthetic agents at single concentrations were made using paired Student's t tests. Comparisons between agents across multiple concentrations were made using two-way analysis of variance with repeated measures, with the anesthetic as the between-subject factor and concentration as the within-subject factor. All data are presented as the mean +/− SD.

These data are based on recordings from 24 pyramidal cells located in the CA1 region of the hippocampus. The mIPSCs occurred at rates of 1-10 Hz at room temperature and were easily distinguished from baseline noise (Figure 2). Nearly all events were fast rising (t^rise< 1 ms) and fast decaying ([small tau, Greek]^wt[tilde operator] 25 ms) and likely correspond to the GABA^AIPSC evoked by stimulating in stratum pyramidale (GABA^A,fast). [32,33]Recordings were often stable for more than 1 h (e.g., Figure 2A), with little rundown in mIPSC amplitude (<5% per h) and no change in IPSC decay.

Volatile anesthetics reduce the peak amplitudes of evoked IPSCs in hippocampal [1,2]and cerebellar neurons. [3]This effect may be due to decreases in the release probability or receptor response, or both. We used the modulation of mIPSCs to distinguish between these possibilities. Changes in mIPSC amplitude are consistent with a postsynaptic action of these agents, whereas changes in mIPSC frequency represent presynaptic effects.

Both isoflurane and enflurane reduced the peak amplitude of mIPSCs (Figure 2), suggesting that these agents act postsynaptically to block GABA (^A) IPSCs, such as by directly blocking the receptors. However, the concentration dependence of the blocking effect was agent specific, as enflurane reduced the amplitude of mIPSCs to a greater extent than did isoflurane (two-way nested analysis of variance: F[3,32]= 11.9; P < 0.05;Figure 3A). The effect of enflurane was significant even at the lowest concentration tested, whereas isoflurane had no significant effect at concentrations less than 0.6 mM (Student's t test). The dose-response data for enflurane and isoflurane were fit by Hill equations with similar Hill coefficients, but different values for IC⁵⁰and y sub [infinity](enflurane: n = 2.3+/−0.4, IC⁵⁰= 0.50+/−0.06 mM, y sub [infinity]= 0.37+/−0.06; Iso: n = 2.6+/−0.8, IC⁵⁰= 0.79+/−0.24 mM, y sub [infinity]= 0.56+/−0.13;Figure 3A). At 0.6 mM, halothane blocked mIPSC amplitude to a lesser extent than did enflurane or isoflurane, although the latter effect was not significant (P = 0.058, by the Student's t test;Figure 3A).

To compare the blocking effects at equivalent anesthetic potencies, we replotted the data in terms of concentration relative to the estimated MAC^Aqvalues at 25 [degree sign]C (halothane: 0.25 mM; isoflurane: 0.30 mM; enflurane: 0.58 mM). [23]In this case, the difference in blocking effect between isoflurane and enflurane was even more striking (Figure 3B). Even at 2 MAC^Aq, isoflurane blocked mIPSC peak amplitudes by less than 15%, whereas an equivalent block by enflurane was seen at 0.5 MAC^Aq. At 2.1 MAC^Aq, enflurane blocked the mIPSC amplitude by more than 55%. If the ratio between MAC-awake and the MAC for enflurane is similar to that of isoflurane (i.e., approximately 0.4 [34]), then the reduction in peak mIPSCs was not substantial for either agent for concentrations corresponding to those equivalent to or less than the MAC-awake.

The conclusion that the effects on mIPSC amplitude are indicative of postsynaptic actions relies on the assumptions that mIPSCs are the responses to single quanta of neurotransmitters, and that these packets of transmitter are sufficient to saturate the postsynaptic receptors. Although there is substantial evidence to support these assumptions (see Materials and Methods), they have not been demonstrated conclusively for these cells. To confirm that enflurane and isoflurane can act directly on the postsynaptic receptors to reduce current amplitude, we investigated the modulation of the responses of CA1 pyramidal cells to puff applications of exogenous GABA.

Briefly, high concentration GABA puffs elicited rapidly decaying inward currents. Bath application of enflurane (0.6 mM;Figure 4A) or isoflurane (0.6 mM;Figure 4B) substantially reduced the amplitude of these GABA currents. Consistent with the results on mIPSCs, enflurane had a greater effect than isoflurane did on peak amplitude (enflurane: 70%+/− 19% block [n = 3]; isoflurane: 34%+/− 16%[n = 4]; P < 0.05 by the Newman-Keuls' test). The lack of quantitative agreement between the magnitude of block observed here and that observed in experiments on mIPSCs may reflect differences in the time course and concentration of transmitter at the postsynaptic receptors.

We have shown that volatile anesthetics reduce mIPSC amplitude, indicating that at least part of the blocking effect observed for evoked IPSCs results from actions on postsynaptic receptors. It is still possible that part of this effect is caused by decreases in release probability, such as by direct actions on synaptic vesicle proteins. The frequency of mIPSCs is a measure of the probability of quantal release from presynaptic terminals apposed to the pyramidal cell of interest. The mIPSC rate was measured as the interevent interval in the absence and presence of volatile agent. The mean interevent interval varied across cells (range, <1 Hz to >10 Hz), but controls for different drug treatments did not differ significantly (P > 0.05, by the Newman-Keul's test). As expected for a Poisson process, the interevent interval distributions were well fit by single exponential functions (data not shown).

At 0.3 mM, enflurane caused a small decrease in the mIPSC interevent interval (i.e., an increase in the mIPSC rate) that was not significant (279 +/− 149 ms to 259 +/− 137 ms; P = 0.07, by the paired Student's t test), whereas at 0.6 mM there was no consistent effect across cells (two cells showed an increase in rate, three cells a decrease, and in one there was no change;Figure 5A). At 1.2 mM there was a consistent decrease in rate, but the pronounced blocking effect may have reduced a sizable number of events to values less than baseline. Isoflurane at 0.3 mM caused a consistent but small increase in the mIPSC rate, but this effect was not significant (256 +/− 158 to 233 +/− 145 ms; P = 0.08, Figure 5B). At 0.6 mM, the effect of isoflurane was more pronounced, reducing the average interevent interval from 423 +/− 266 ms to 346 +/− 259 ms (P < 0.05, by the paired Student's t test). As with enflurane, at 1.2 mM isoflurane consistently reduced the mIPSC rate, but the combination of block and increased baseline noise may have led to many events that were less than baseline values. The effect of halothane at 0.6 mM was similar to that of isoflurane at 0.3 mM (277 +/− 88 to 225 +/− 104 ms; P < 0.05, by the paired Student's t test;Figure 5C). For none of the volatile agents at <or= to 0.6 mM was a decrease in the mIPSC rate observed, and thus the substantial block by enflurane of the evoked IPSC²cannot be secondary to a reduction in release probability.

The most common measure of enhancement of synaptic GABA^Aresponses by general anesthetics is the prolongation of the decay phase of the current, an effect previously studied for several volatile agents. [1-3,35]We wanted to compare the ability of equimolar concentrations of a volatile agent to increase the [small tau, Greek]^wtof mIPSCs. The ability of volatile agents to prolong GABA^Asynaptic currents may be a consequence of the mechanism of block, as for local anesthetic block of nicotinic acetylcholine receptors, [16]or the two effects may represent separate modulatory processes via actions at multiple sites. We have tried to distinguish between these two possibilities by comparing the concentration dependence of a prolongation for isoflurane and enflurane. We have already shown that enflurane is more potent at blocking mIPSCs than is isoflurane. If the blocking and prolonging effects share a common effector mechanism, then enflurane should also be more potent at prolonging IPSCs. In addition, within the agent, we would expect the concentration dependence for blocking and prolonging the currents to be similar.

As expected, mIPSC decay times were prolonged substantially by all three volatile agents (Figure 6). At 0.3 mM, isoflurane and enflurane more than doubled the weighted time constant (enflurane: control 24.1 +/− 5.2 ms, drug 48.9 +/− 5.5 ms, n = 5; isoflurane: control 22.2 +/− 1.6 ms, drug 55.5 +/− 11.8 ms, n = 5), and this prolongation increased to more than four times at 1.2 mM (enflurane: control 23.7 +/− 3.7 ms, drug 98.8 +/− 22 ms, n = 5; isoflurane: control 25.3 +/− 6.2 ms, drug 113.5 +/− 28.7 ms, n = 5;Figure 6B). There was no significant difference between enflurane and isoflurane in their effects on mIPSC decay time (two-way nested analysis of variance: F(3,32)= 1.15, P = 0.29). At 0.6 mM, however, halothane was significantly less efficacious than either isoflurane or enflurane, prolonging mIPSCs by less than 2.5 times (control 28.2 +/− 3.9 ms, drug 72.9 +/− 7.4 ms; P < 0.05 compared with isoflurane or enflurane using the Student's t test). Hill Equation fitsto the dose-response curves yielded similar parameters for the two agents (enflurane: y [infinity]= 4.4 +/− 0.6, EC⁵⁰= 0.39 +/− 0.09 mM, n = 2.8 +/− 1.4; isoflurane: y [infinity]= 4.7 +/− 0.5, EC⁵⁰= 0.32 +/− 0.05 mM, n = 2.7 +/− 1.2;Figure 6B). For both drugs, there was greater enhancement at 0.15 mM than is predicted by the fit, possibly indicating a second enhancement process at low concentrations. This may also explain the large uncertainties with the Hill coefficients. At equivalent MAC^Aqfractions, the prolongation dose-response curve for enflurane was shifted relative to the curve for isoflurane by approximately the difference in MAC (^Aq) values for the two agents (EC⁵⁰: isoflurane, 1.1 MAC^Aq; enflurane, 0.68 MAC^Aq;Figure 6C). Enflurane prolonged mIPSCs to a significantly greater extent than did isoflurane at 1 MAC^Aq(by the unpaired Student's t test, P < 0.05).

In comparing Figure 3A and Figure 6B, it is immediately apparent that although the concentration dependencies of block by enflurane and isoflurane are coincident, this is not true for the prolonging effect. In addition, within agent the blocking versus prolonging effects do not have identical concentration dependence. In the case of enflurane, the EC⁵⁰for prolonging the current and the IC⁵⁰for blocking the current (0.5 mM) differed by about 30%, but for isoflurane the half-maximal concentration for blocking the current exceeded the value for prolonging the current by more than two times. This point is illustrated more clearly in Figure 7, where the increase in decay time relative to control is plotted versus the percentage block of the peak mIPSC amplitude for cells exposed to varying concentrations of enflurane (open symbols) and isoflurane (filled symbols). The two populations are clearly distinct, as indicated by the two linear least-squares fits to the two sets of data, and for a given level of enhancement, isoflurane is less likely to block the receptors than is enflurane.

To evaluate the net effects of enflurane, isoflurane, and halothane on inhibitory currents in pyramidal cells, we measured the total charge transfer during an average mIPSC by calculating the integral of the current trace. Figure 8A shows the normalized data plotted as a function of absolute anesthetic concentration. Despite the substantial blocking effects observed (Figure 3A), all three agents caused a net increase in inhibition at each concentration tested. The dose-response curves peaked at 0.6 mM for isoflurane and enflurane, (isoflurane: Q^Drug/QCtrl= 3 +/− 0.6; enflurane: 2.2 +/− 0.6). As illustrated in Figure 8B, this peak in the normalized charge-transfer dose-response curve occurs at 1.1 MAC^Aqfor enflurane but at 2 MAC^Aqfor isoflurane. Interestingly, there is no difference between enflurane and isoflurane at <or= to 1 MAC^Aq. Thus there is good correspondence between the two agents at anesthetic concentrations sufficient to affect cortically mediated processes such as response to a command (MAC-awake) and explicit and implicit learning. [34,36] 

Volatile anesthetics have been shown to activate GABA^Areceptors in hippocampal neurons in the absence of synaptically or exogenously applied GABA, [1,11,37]presumably by binding to the receptor and acting as a weak agonist. We observed that application of isoflurane, enflurane, and halothane typically caused an increase in the standard deviation of baseline noise ([small sigma, Greek]^noise;Figure 9). Isoflurane caused the most prominent increase in [small sigma, Greek]^noise(Figure 9A and Figure 9B), whereas the effect was more variable and less pronounced in response to enflurane and halothane (Figure 9C). The increase in [small sigma, Greek]^noisewas typically associated with an inward shift in the holding current of 50 to 150 pA. With resting input conductances of 5-10 nS, this represents an increase in conductance of 8% to 50%. This effect resulted from an increase in the basal activity of GABA^Areceptors, because the application of picrotoxin blocked both the increase in noise and the change in holding current (n = 3; data not shown). It is possible that volatile anesthetics could produce these effects by increasing the sensitivity of GABA^Areceptors to ambient GABA, because GABA is present in the extracellular space at submicromolar levels. With these data, we cannot distinguish between a direct activation and activation secondary to an increased affinity for GABA. It should be noted, however, that when expressed as a MAC^Aqfraction, the dose dependencies for isoflurane and enflurane in producing this effect are dissimilar (Figure 9D), suggesting a minor role for this process in producing behavioral effects.

We presented data on the concentration dependence of the modulation of mIPSCs by two structurally related volatile anesthetics, isoflurane and enflurane, and on the modulation by a single concentration of halothane. We found that enflurane, isoflurane, and halothane decreased the peak amplitudes of mIPSCs but had only modest effects on mIPSC frequency. Isoflurane and enflurane also reduced the peak amplitudes of responses to puff-applied GABA. These data suggest that the primary mechanism for the observed decreased in peak evoked IPSCs by these agents [1,2]is a direct action on the postsynaptic receptors. At equimolar concentrations, the blocking potency of enflurane was significantly greater than that of isoflurane (Figure 3A), but enflurane and isoflurane prolonged the decay times of mIPSCs to similar extents (Figure 6B). The dissociation observed across agents between blocking and prolonging effects is consistent with actions by anesthetics at multiple sites on GABA^Areceptors.

The data presented here show that the block of IPSCs by volatile agents is a result, at least in part, of actions on the postsynaptic receptors. The observation that enflurane does not completely block mIPSCs, even at a saturating concentration [3](see also Figure 3A), argues against a channel-blocking mechanism but suggests instead that enflurane acts allosterically to inhibition channel opening. In addition, preliminary studies on isolated receptors have shown that inhibition of channel opening does not require previous activation, [38]as would be expected for an open channel blocker.

We suggest that the relation between the blocking and prolonging effects of volatile anesthetics on mIPSCs has mechanistic implications and is most consistent with a multiple-site hypothesis for anesthetic action. This argument can be summarized as follows. Suppose that both the blocking and prolonging effects are modeled by altering a single anesthetic-sensitive kinetic step. By systematically varying this kinetic parameter to simulate the anesthetic actions at different drug concentrations, we would obtain a relation between the blocking and prolonging effects (see Figure 7). If another anesthetic modulates the same single kinetic step (although not necessarily with the same potency or efficacy), then varying this same kinetic parameter will lead to the same relation between blocking and prolonging. Our data show that the same relation is not obtained, which means that either a different single kinetic step is affected by each drug or that multiple kinetic steps are affected by both agents, but in different proportions. We believe that it is unlikely that each drug affects a single, but different, kinetic step, because of the similarity of the normalized IPSCs in the presence of enflurane compared with isoflurane (Figure 6A). Also inconsistent with a single site is the observation that isoflurane prolongs the current to a significant degree at 0.3 mM but does not block the current significantly at this concentration (Figure 3A and Figure 6B). Thus we conclude that multiple kinetic steps are involved. Our data do not exclude the possibility that multiple processes contribute to the prolongation of mIPSCs and that some part of the prolongation is secondary to a blocking effect.

Studies of nicotinic acetylcholine receptors, which show significant sequence and structural homology to GABA^Areceptors, [39]may provide some insight into the mechanism of modulation by volatile anesthetics. These agents profoundly block the responses of nicotinic acetylcholine receptors to high concentrations of agonist [18,40]but do not prolong the responses to synaptically released acetylcholine. [41]Isoflurane produces a characteristic flickering between open and closed states that is similar to the sequential open channel block observed at these receptors for lidocaine derivatives. [16]However, modeling studies of single-channel currents and macroscopic currents in response to rapid applications of agonist suggest that sequential and cyclic (i.e., where the blocker can bind and unbind from the closed receptor) blocking mechanisms are inadequate to account for the data. Dilger et al. [18]suggested that two parallel modulatory processes exist, for example, mediated by two isoflurane binding sites, in which isoflurane binds to one site to produce the block and binds to another site to change the affinity of the receptor for agonist.

We speculate that a similar model applies to volatile anesthetic modulation of GABA^Areceptors. One effector mechanism would produce the blockade of these receptors, as evident in the recordings of synaptic currents (Figure 2and Figure 3A) and in the responses of cells to high concentrations of exogenously applied GABA [10,11](Figure 4). The second effector mechanism would produce the observed increase in decay time of synaptic currents (Figure 6), and probably the enhanced response to low concentrations of agonist observed in several studies. [11,42]This could be achieved, for example, by decreasing the rate of dissociation of agonist from the receptor, [43]as proposed for the modulation of nicotinic acetylcholine receptors by ethanol. [17]This model parallels the suggestion that the blocking and direct gating actions of sevoflurane likely represent two distinct mechanisms. [11] 

Previous studies have shown that enflurane and isoflurane increase the amplitude and prolong the decay of the responses to relatively slow applications of low concentrations of GABA. [44]In contrast to our results, isoflurane has been found to be more potent than enflurane in these effects. [42]One likely explanation for this discrepancy is that the modulation of responses to slow, low concentrations of GABA may combine both blocking and enhancing actions. Thus a more apt comparison may be with the effects on charge transfer during a mIPSC, which combines both blocking and enhancing effects, and for which isoflurane was observed to be more protent than enflurane in its modulatory effect (Figure 8).

In the current study, 0.6 mM enflurane blocked fast mIPSCs by 40%, with little effect on mIPSC frequency, indicating that influrane does not act on the presynaptic release machinery to reduce IPSC amplitude. However, the evoked GABA^A,fast was blocked by about 60% at a similar concentration, [2]suggesting that part of the block of the evoked IPSC is due to other presynaptic effects, such as increased firing thresholds of presynaptic cells or fibers. This conclusion is also supported by the findings of Antkowiak and Heck, [3]who showed that [tilde operator] 0.6 mM enflurane blocks spontaneous IPSCs (i.e., action potential-dependent and -independent IPSCs) by approximately 55% but blocks mIPSCs by only approximately 35%. Thus, the effects of volatile anesthetics on circuit behavior are likely to be more complex than predicted solely from their modulation of GABA^Areceptors alone.

In addition to their modulatory effects on responses to GABA, general anesthetics also have been shown to directly activate GABA^Areceptors by binding to sites distinct from the GABA binding site. [45]We observed enhanced basal activity of GABA^Areceptors in response to all three agents, although the effect was most pronounced for isoflurane (Figure 9). This activity could result from either direct activation of the receptors by volatile agents, as has been reported previously for hippocampal neurons, [11,37]or indirectly by increasing the sensitivity to or the level of ambient GABA. [46]It is difficult to distinguish between these possibilities with the data presented here. It should be noted that previous studies using synapatosomal preparations and microdialysis failed to demonstrate increases in GABA release or concentration. [8]However, the increase in mIPSC frequency observed in our experiments for isoflurane at 0.6 mM is consistent with enhanced release.

The behavioral state of “anesthesia” encompasses several behavioral effects, and it is obviously difficult to relate all aspects to a single electrophysiologic measure. Enhancement of GABAergic neurotransmission is a reasonable hypothesis for some of the behavioral effects mediated by cortical processes, such as the effects on awareness, learning, and seizure threshold, but the relevant parameter to measure this enhancement is unclear. If the behavioral effects of isoflurane and enflurane were mediated by the same set of processes, then we would expect that the parameters relevant for mediating these behavioral effects would have similar dose dependencies relative to a behavioral measure of anesthetic potency such as MAC or MAC-awake.

Previously we showed that fast mIPSCs observed in CA1 pyramidal cells are likely to originate at the same synapses that produce fast GABA^AIPSCs evoked by stimulating stratum pyramidale (GABA^A,fast³²). These somatic IPSCs were postulated to control the spike output of pyramidal cells, because their blockade with furosemide, a subtype-specific GABA^Areceptor antagonist, [47]leads to population bursts in response to single afferent shocks. [33]Consistent with this hypothesis, the proconvulsant volatile agent enflurane blocked GABA^A,fast by about 60% at 0.5 mM, but halothane, which is not a proconvulsant agent, had little effect on the amplitude of GABA^A,fast. [2]Modulation of a different synaptic current, a slow, dendritic GABA^AIPSC evoked by stimulation in stratum lacunosum-moleculare, was proposed to mediate the amnestic and sedative properties of volatile anesthetics. This hypothesis was based on the equivalent effects of halothane and enflurane on GABA^A,slow [2]and on the putative role of this IPSC in modulating the induction of long-term potentiation at dendritic excitatory synapses on CA1 pyramidal cells. [48] 

Our findings suggest at least the possibility that enhancement of GABA^A,fast may underlie some of the behavioral effects of these agents, and thus they are not entirely consistent with the hypothesis just described. Although at equivalent MAC^Aqfractions isoflurane is substantially less effective than enflurane at blocking mIPSC peak amplitude, for MAC^Aqfractions <or= to 1, enflurane and isoflurane are equally effective at increasing the net charge transfer during the average mIPSC. Thus, as with the issue of blocking versus enhancement at the molecular level, the question arises about the functionally relevant measure of inhibitory strength when inhibition in a network of cells such as CA1 is considered. Is the peak amplitude of inhibitory currents relevant or their net charge transfer? The answer to this question may depend on the time course of the process being inhibited. For example, in the case of IPSCs competing with fast rising excitatory synaptic currents to control spike output, the peak amplitude of the inhibitory current may be more important than the duration or total charge transfer. In the case of modulation of the level of summed dendritic excitation that may trigger second messenger cascades and more slowly rising somatic voltages, the total charge transfer may be more relevant parameter. Thus, the proconvulsant property of enflurane may indeed be related to its greater blocking potency, but other behavioral effects may be mediated by enhanced total charge transfer of both GABA^A,fast and GABA^A,slow.

The authors thank the reviewers for valuable comments and Phil Shils and Christine Krekel for technical support.