INTRODUCTION

TOP ABSTRACT INTRODUCTION Methods RESULTS DISCUSSION REFERENCES

Synchronous neuronal activity, which is seen in extracellular field recordings as gamma oscillations (roughly 20 to 80 Hz), has been ascribed a role in setting the temporal relationships that may be critical for various aspects of information processing, storing, and retrieval (Ritz and Sejnowski 1997; Singer and Gray 1995). A widely accepted paradigm of gamma oscillations suggests that GABA_A receptors are crucially involved in determining their frequency (Jefferys et al. 1996; Traub et al. 1998; Wang and Buzsaki 1996).

In view of the proposed relationship between gamma rhythmicity, cognitive functions, and states of consciousness of the brain (Baldeweg et al. 1998; Llinas and Ribary 1993; Singer 1993; Traub et al. 1998), it is not surprising that a number of studies have focused on pharmacological modulation of the temporal and spatial properties of gamma oscillations, in particular those properties that are seen after the administration of psychoactive drugs, e.g., diazepines, propofol, barbiturates, morphine, and ketamine (Faulkner et al. 1998; Whittington et al. 1996, 1998). However, little is known about the actions of intrinsic physiological modulators of neuronal functions on gamma activity.

Starting from the classical studies of 19th century physiologist Jean-Marie Charcot and his pupils on "hystero-epileptic fits," it has been widely recognized that changes in the efficacy of breathing have a profound effect on brain excitability (for review see Tombaugh and Somjen 1998). Hyperventilation, i.e., excess breathing, which leads to a decrease in the partial pressure of CO₂ (pCO₂) (hypocapnia), is known to alter cortical sensory responses (Huttunen et al. 1999) and, in susceptible persons, hyperventilation can precipitate seizures (Foerster 1924), a finding that is still in clinical use for the detection of petit mal epilepsy (Niedermeyer and Lopes da Silva 1993). On the other hand, hypercapnia, an increase in pCO₂, usually has the opposite effect and leads to a decrease in neuronal excitability (Balestrino and Somjen 1988; Caspers et al. 1987). In view of the well-recognized and central role of CO₂ as a physiological modulator of nervous activity, it is striking that there are almost no mechanistic data on the modes of action of changes in pCO₂ at the neuronal network level in the brain.

Alterations in respiratory activity, if not balanced by simultaneous changes in the production of CO₂, accompany profound changes in pCO₂ and have a direct effect on the acid-base status of the brain parenchyma. A wealth of evidence has been accumulated during the last few decades that demonstrates that pH is a powerful physiological modulator of neuronal activity; this has led to the hypothesis that H⁺ ions have an important signaling function in brain tissue. Changes both in extracellular and in intracellular pH exert a strong influence on gross neuronal excitability (see, for example, Jarolimek et al. 1989 and Lee et al. 1996; for reviews see Ballanyi and Kaila 1998; Chesler and Kaila 1992; Kaila and Chesler 1998).

In the present work, we examined the effects of pCO₂ changes in vitro on a model of hippocampal gamma oscillations recently described by Fisahn et al. (1998). In the model, persistent gamma activity is induced by cholinergic excitation by the muscarinic agonist carbachol. This model relies on reciprocal inhibitory and excitatory connections in the CA3 network and is independent of N-methyl-D-aspartate (NMDA) receptors. A critical role for GABA_Aergic mechanisms has been demonstrated for carbachol-induced gamma oscillations (Buhl et al. 1998; Fisahn et al. 1998) and recent computational studies suggest a role for gap junctions as well (Kopell et al. 2000; McMahon et al. 1998; Traub and Bibbig 2000; Williams and Kauer 1997).

We found that the intra- and extracellular alkalosis that takes place in hypocapnia (1% CO₂) has a dual effect on carbachol-induced gamma oscillations. Intracellular alkalosis during hypocapnia resulted in increased oscillation frequency and decreased integrated gamma-band amplitude whereas extracellular alkalosis was related to an enhancement of the temporal stability of gamma oscillations. We further provide evidence to show that the temporal stabilization seems to be mainly attributable to an increase in the amplitude of GABA_A receptor-mediated inhibitory postsynaptic currents (IPSCs).

	Methods

TOP ABSTRACT INTRODUCTION Methods RESULTS DISCUSSION REFERENCES

Hippocampal slices 450 µm thick were prepared from adult Wistar rats of either sex. After deep ether or pentobarbital anesthesia, the rats were decapitated and their brains were quickly removed. Horizontal slices were cut with a Vibroslice (Campden Instruments, Loughborough, UK) and transferred to an interface-type recording chamber where they were kept at 34°C, gassed with humidified mixtures of O₂ with 5% (control solution) or 1% (hypocapnic solution) CO₂, and continuously perfused with physiological solution containing (in mM) 129 NaCl, 1.25 NaH₂PO₄, 1.8 MgSO₄, 1.6 CaCl₂, 3 KCl, 21 NaHCO₃, 10 glucose, pH 7.4. In solutions containing trimethylamine (TriMA), NaCl was replaced by an equivalent amount of TriMA (pH 7.4).

Before recordings were made, slices were allowed to recover in the standard 5% CO₂ solution for at least 1 h after preparation. Field activity was recorded at one or two sites in the dendritic layer of area CA3 with a NeuroData IR 183 (NeuroData Instruments, New York, NY) using low resistance, saline-filled patch pipettes with an open diameter of about 5 µm. A CED 1401 interface (Cambridge Electronic Design, Cambridge, UK) was used for storage. Data were band-pass filtered between 1.5 and 300 Hz and were digitized at 2 kHz.

Field-potential oscillations were recorded and characterized in epochs of 8 s. The gamma-band amplitude was computed as an integral of the 20-80 Hz segment of the amplitude spectrum. This approach was taken because it gives a measure that 1) is linearly related to the oscillation amplitude, 2) is not sensitive to inter-experiment variability in oscillation frequency, and 3) is not sensitive to the slight inter-experiment variability in the oscillation waveform shape, which depends on exact electrode position. The main oscillation frequency was taken as the peak frequency of the amplitude spectrum of the autocorrelation function (ACF). Temporal stability was indexed by the decay time constant  of an exponential fitted to the amplitude envelope of the ACF. The ACF amplitude A(t) was computed with the Hilbert-transform as follows: A(t) = mod [s(t) + is'(t)], where i is the imaginary unit, s(t) denotes the ACF, and s'(t) is its Hilbert-transform. In other words, the Hilbert-transform was used to find the imaginary part of an oscillation, the real part of which is the ACF. Treating the oscillation in the complex plane allows for continuous evaluation of the oscillation amplitude (rather than only at the oscillation peaks given by the real part). In general, the modulus mod (z) (i.e., the oscillation amplitude) of a complex number z, where z = x + iy, is given by mod (z) = . To further obtain a frequency-independent measure of temporal stability, we normalized the decay time constant with cycle duration T (given by the reciprocal of main oscillation frequency) in each experiment.

Extracellular pH (pH_o) was measured with double-barreled H⁺-sensitive microelectrodes. The electrodes were manufactured as described in Arens et al. (1992) and had a response slope of 52-56 mV for a unit change in pH. Intracellular recordings were made with sharp microelectrodes (40-100 M resistance) containing a mixture of K⁺-acetate (1.5 M), K⁺-methyl-sulfonate (1 M), KCl (6 mM), and 50 µM of the lidocaine derivative 2(triethylamino)-N-(2,6-dimethylphenyl)-acetamide (QX-314). Recordings were performed in voltage clamp mode with an SEC10L amplifier (NPI Electronic, Tamm, Germany). GABA_A receptor-mediated inhibitory postsynaptic currents (IPSCs) were pharmacologically isolated by antagonists of ionotropic glutamate receptors and GABA_B receptors. NMDA receptors were blocked with (±)-2-amino-5-phosphonopentanoic acid (APV, 60 µM), non-NMDA receptors were blocked with 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzol(f)quinoxaline-7-sulfonamide (NBQX, 10 µM), and GABA_B receptors were blocked with the highly potent blocker CGP 55845A (2 µM). The GABA_A receptor-mediated IPSCs were evoked with bipolar stimulation electrodes placed in the hilus of the dentate gyrus (0.1 ms pulse duration) every 10-20 s and the cells were held at their original resting potential, which was approximately 60 mV (63 to 57 mV), with electrodes containing QX-314. The stimulation strength was adjusted to obtain approximately 70% of the maximum response in each cell and was then kept constant throughout an experiment. Five to ten consecutive IPSCs were averaged prior to the estimation of maximum amplitude and decay kinetics. Decay was approximated with a least-squares fitted monoexponential function (IgorPro).

For optical imaging of neuronal pH (pH_i), slices were transferred to a submerge-type chamber (0.4 ml, perfused at 1.5 ml/min) with a coverslip bottom and were anchored with platinum bars. The acetoxymethyl ester form of the pH-sensitive dye 2',7'-bis-(2-carboxyethyl)-5-(and-6)-carboxyfluorescin (BCECF-AM) (Molecular Probes) was dissolved at 1 M in DMSO. Dye was injected, at a final concentration of 300-500 µM in saline, into the recording area (stratum pyramidale) with glass capillary pipettes (6-12 µm wide) by application of slight pressure. Loading was monitored visually and, after obtaining a sufficiently high level, the pipette was removed. The fluorescence imaging system (Photon Technology International, Lawrenceville, NJ) consisted of an inverted microscope (Nikon Diaphot, 20× objective), a monochromator/chopper unit, and an intensified CCD camera (IC-200). BCECF image pairs (pH sensitive at 495 nm excitation and pH insensitive at 440 nm excitation) were acquired with Axon Imaging Workbench software (Axon Instruments, Foster City, CA). Acquisition was performed at a rate of 0.7 Hz and each cycle contained 16 averaged video frames with both excitation wavelengths. pH_i calibration was performed with the null-point method (Eisner et al. 1989). The examined cells were visually identified as pyramidal neurons as judged by their morphology and location.

APV was purchased from Tocris, carbachol was purchased from RBI, and QX-314 was purchased from Sigma. NBQX was a gift from NOVO Nordisk (Gentofte, Denmark) and CGP 55845A was a gift from Ciba Geigy (Basel, Switzerland). All numerical data are expressed as mean ± SE. The significance levels were estimated with Student's t-test for paired data.

	RESULTS

TOP ABSTRACT INTRODUCTION Methods RESULTS DISCUSSION REFERENCES

Bath application of the cholinergic agonist carbachol (5-10 µM) readily induced field potential oscillations in the stratum radiatum of area CA3 within 5-10 min (n = 20 slices, Fig. 1A). The oscillations gave rise to a narrow peak in amplitude spectra that remained stable in each experiment (Fig. 1B), which indicates the presence of only minor frequency fluctuations. Addition of the NMDA receptor antagonist APV (60 µM) left the oscillations unaltered (n = 5 slices, data not shown). To block NMDA receptor-mediated long-term synaptic plasticity, we coapplied carbachol and APV in all subsequent experiments (Fisahn et al. 1998).

View larger version (39K): [in this window] [in a new window]   Fig. 1. Effects of hypocapnia (1% CO2) on carbachol-induced gamma oscillations. A: representative field potential recordings from the stratum radiatum in the CA3 region. The middle trace (1% CO2) was taken 9 min after the switch from 5 to 1% CO2. B, left: amplitude spectra for the control, 1% CO2, and washout conditions in this experiment. Right: autocorrelation function (ACF) amplitude envelopes (see METHODS) for the three conditions. Note the slower decay of the ACF in 1% CO2. C: pooled data from 16 recordings (in 12 slices) indicate that 1% CO2 decreases integrated amplitude (from 20 to 80 Hz), increases oscillation frequency, and prolonges ACF decay times (/T; see METHODS for details). Asterisks denote significance levels: *P < 0.05; **P < 0.01; ***P < 0.001.

Effects of hypocapnia on carbachol-induced gamma oscillations

We induced hypocapnia by switching the gassing mixture from 5 to 1% CO₂. This was associated with a reversible decrease in the gamma-band amplitude (20 to 80 Hz) from 0.19 ± 0.03 to 0.16 ± 0.02 mVHz (P = 0.04, n = 16; Fig. 1, B and C). The oscillation frequency was increased from 31.9 ± 0.8 to 34 ± 1 Hz (P = 0.047; Fig. 1, B and C). In parallel with these relatively modest changes, the decay time of ACFs was strongly prolonged (Fig. 1B). We quantified the ACF collapse by fitting an exponential to the amplitude envelope of the ACF and normalized the decay time constant  with the duration of one oscillatory cycle T to obtain a frequency-independent measure /T (see METHODS). Hypocapnia induced a dramatic and reversible enhancement of /T from 4 ± 1 to 8 ± 1 (P = 0.000007, Fig. 1C). The prolongation of ACF decay times, as indexed by the /T enhancement, indicates that intrinsic phase-correlations strengthened and, therefore, the temporal stability of the oscillations increased.

Is the temporal stabilization mediated by a change in pH_o or pH_i?

Recordings with extracellular pH electrodes showed that a switch from 5 to 1% CO₂ induced a rise in pH_o of approximately 0.5 pH units, which took about 4 min to reach 90% of the steady-state level and 2 min to recover (n = 5 slices; Fig. 2A). This alkaline shift gives a good estimate of the rate of equilibration of CO₂ within the interstitial space of the slice (Voipio 1998).

View larger version (16K): [in this window] [in a new window]   Fig. 2. Hypocapnia increases both interstitial and intracellular pH. A: representative recording of interstitial pH with a pH electrode placed in the dendritic layer. B: fluorescence recordings of intracellular pH from two representative pyramidal cells (see METHODS for details). Trimethylamine (TriMA, 20 mM) induces an intracellular alkaline shift comparable with hypocapnia.

Fluorescence recordings of pH_i indicated that hypocapnia induced a monophasic alkaline shift in neuronal pH_i of 0.2 ± 0.013 pH units from the control level and that the alkalosis did not show recovery before switching back to 5% CO₂ (n = 29 neurons in 7 slices; Fig. 2B). We did not observe an undershoot of pH_i upon the reintroduction of 5% CO₂, which further indicated that the alkaline shift did not activate mechanisms regulating pH_i (i.e., uptake of acid equivalents during the alkaline load).

The results of pH_i and pH_o measurements therefore indicated that both pH_i and pH_o undergo a monophasic increase in 1% CO₂. To elucidate which of them was responsible for the temporal stabilization of carbachol-induced gamma oscillations, we evoked a rise in pH_i with the weak membrane-permeant base TriMA (see, for example, Eisner et al. 1989). As was expected, 20 mM TriMA produced an intracellular alkaline shift that had an amplitude of 0.13 ± 0.010 pH units (n = 13 neurons in 3 slices; Fig. 2B). TriMA mimicked hypocapnia by reversibly decreasing the mean gamma-band amplitude from 0.24 ± 0.05 to 0.17 ± 0.03 mVHz (P = 0.003, n = 7 slices; Fig. 3) and by increasing the oscillation frequency from 37 ± 3 to 39 ± 3 Hz (P = 0.02). Strikingly unlike hypocapnia, TriMA decreased /T from 2.1 ± 0.3 to 1.2 ± 0.1 (P = 0.04). This suggests that the enhancement of the temporal stability in 1% CO₂ is mediated by the rise in pH_o rather than in pH_i.

View larger version (37K): [in this window] [in a new window]   Fig. 3. Increasing intracellular pH with TriMA decreases amplitude and increases oscillation frequency but does not prolong ACF decay. A: representative field potential recordings from control conditions, after 15 min of TriMA (20 mM) application, and after washout of TriMA. B: amplitude spectra and ACF amplitude envelopes for this recording. C: pooled data from 7 recordings (in 7 slices) show that TriMA decreases integrated amplitude (from 20 to 80 Hz) and increases oscillation frequency. Unlike hypocapnia, TriMA decreases ACF decay time.

Effects of hypocapnia on resting conductances and on IPSCs

In intracellular recordings from CA3 pyramidal cells, neuronal input resistance was not affected by hypocapnia (control, 78 ± 13 M; hypocapnia, 74 ± 14 M; P = 0.4, n = 9 cells). We then evoked GABA_A receptor-mediated IPSCs with single-pulse stimulation under the blockage of AMPA receptor-, NMDA receptor-, and GABA_B receptor-mediated transmission. Hypocapnia led to an increase in the amplitude of the IPSC from 420 ± 50 to 440 ± 50 pA which on washout decreased to 370 ± 40 pA (for hypocapnia vs. control, P = 0.03; for hypocapnia vs. the mean of control and washout, P = 0.01; Fig. 4). Assuming that the decrease in IPSC amplitude from control to washout reflects a rundown of the GABA_A receptor response, the relative hypocapnia-induced amplitude enhancement was 10 ± 3% (P = 0.004). The decay-time constant, however, was not affected (15.9 ± 3.2 to 15.8 ± 2.7 ms, P = 0.47).

View larger version (11K): [in this window] [in a new window]   Fig. 4. Hypocapnia enhances inhibitory postsynaptic current (IPSC) amplitude but does not affect the decay time constant. A: representative intracellular recording from a CA3 stratum pyramidale neuron showing pharmacologically isolated GABAA receptor-mediated currents evoked by single-pulse stimulation. B: normalization with the peak amplitude demonstrates how 1% CO2 does not change the decay kinetics of the current. Note that, for visualization, the traces are slightly shifted.

Hypocapnia-induced /T enhancement is reversed by bicuculline

We then asked whether the selective increase in IPSC amplitude could underlie the temporal stabilization observed in hypocapnia. The GABA_A receptor antagonist bicuculline (1 µM) decreases the GABA_A receptor-mediated IPSC amplitude but does not affect its decay time constant and therefore provides a means for selectively antagonizing the effect of hypocapnia on GABA_A receptor-mediated transmission. As before, we first induced gamma oscillation with carbachol and then switched to 1% CO₂ to induce temporal stabilization. We then applied 1 µM bicuculline. Previous work has shown that this concentration decreases IPSCs by 14 ± 4% in CA3 pyramidal neurons (Traub et al. 1993). The bicuculline application had a tendency to decrease the mean gamma-band amplitude and to increase the oscillation frequency, although the effects were not significant in our study (n = 8 recordings in 5 slices; Fig. 5). Bicuculline therefore did not counteract the effects of hypocapnia on oscillation amplitude and frequency. It did, however, progressively decrease /T from the onset of application and, after ~15 min, reversibly returned /T to near control levels (P = 0.2), thereby largely reversing the stabilizing effect of hypocapnia (Fig. 5).

View larger version (39K): [in this window] [in a new window]   Fig. 5. Reversing hypocapnia-induced IPSC amplitude increase with bicuculline returns ACF decay times to near control values but does not reverse amplitude or frequency effects. A: representative field potential recordings from control conditions, after 9 min in 1% CO2, after 13 min of bicuculline (1 µM) application, and after washout of bicuculline. B: amplitude spectra and ACF amplitude envelopes for this recording. C: pooled data from 8 recordings (in 5 slices) indicate that bicuculline does not reverse amplitude or frequency effects of hypocapnia but does, to a large extent, attenuate prolonged ACF decay time.

Comparison of the effects of hypocapnia and up-modulators of GABA_A receptors on gamma oscillations

A unique aspect of the hypocapnia-induced enhancement of IPSC amplitude was the absence of any increase in the decay-time constant. In this respect, the actions of 1% CO₂ differ from those of pharmacological up-modulators of GABA_A receptors, such as benzodiazepines and barbiturates, which increase both IPSC amplitude and duration (Poncer et al. 1996; Roepstorff and Lambert 1994; Segal and Barker 1984). To see whether a joint increase in IPSC duration and amplitude would lead to temporal stabilization, we tested the actions of midazolam (MDZ) and pentobarbital (PB) on carbachol-induced gamma oscillations.

MDZ, a benzodiazepine (20 µM), and PB (10 µM) increased the gamma-band amplitude from 0.16 ± 0.03 to 0.19 ± 0.04 mVHz (P = 0.009, n = 12 recordings in 6 slices) and from 0.20 ± 0.04 to 0.22 ± 0.04 mVHz (P = 0.004, n = 12 recordings in 8 slices; Fig. 6), respectively. MDZ and PB also decreased the oscillation frequency from 33 ± 1 to 29.4 ± 0.8 Hz (P = 0.001) and from 30.7 ± 0.6 to 26.5 ± 0.4 Hz (P = 0.0002), respectively. MDZ decreased /T slightly and not significantly (from 6 ± 1 to 5 ± 1, P = 0.08) whereas PB decreased /T more dramatically from 4.4 ± 0.9 to 2.7 ± 0.4 (P = 0.03). Therefore, when compared with hypocapnia, both MDZ and PB had exactly opposite effects to the amplitude, frequency, and temporal stability of gamma oscillations.

View larger version (36K): [in this window] [in a new window]   Fig. 6. Effects of midazolam (MDZ) and pentobarbital (PB) on gamma oscillations. A: representative field potential recordings from control conditions and after MDZ or PB application. B: amplitude spectra and ACF amplitude envelopes for these recordings. C: pooled data from 12 recordings (MDZ, 6 slices; PB, 8 slices) indicate that both MDZ and PB enhance integrated amplitude and decrease oscillation frequency. MDZ did not affect and PB decreased ACF decay time.

	DISCUSSION

TOP ABSTRACT INTRODUCTION Methods RESULTS DISCUSSION REFERENCES

In terms of concentration-effect relationships, the H⁺ ion (usually quantified as pH) is a potent intrinsic modulator of brain function at the molecular and cellular levels, acting at nanomolar concentration changes to bring about significant alterations in the activity patterns of individual neurons and neuronal circuitries (see, for example, Jarolimek et al. 1989 and Kaila and Ransom 1998). Because the CO₂/HCO buffer is the main pH buffer system within the brain (Chesler 1990), a major determinant of brain pH in vivo is the regulation of pCO₂ by breathing. Hence, by regulating respiration, the brain is capable of exercising control over its own excitability (Balestrino and Somjen 1988).

In previous studies (unpublished data; cf. Huttunen et al. 1999), we observed that a 3-min period of voluntary hyperventilation by healthy human subjects can, in some cases, lead to a decrease in end-tidal pCO₂ from the control level of approximately 38 mmHg (which corresponds to the 5% CO₂ in the present experiments) to a value lower than 2% (see also Brian 1998). Hence, in this respect, the design of the present experiments can be considered to be physiologically relevant. As to pathophysiological phenomena, the present study might shed light on the controversial mechanisms and diagnostics related to the "hyperventilation syndrome," a common type of panic disorder (see, for example, Dager et al. 1995).

Recently, studies of oscillatory activity patterns generated in the CNS have experienced a strong renaissance (Bullock 1997). In particular, gamma oscillations (usually seen as coherent population activity in the 20-80 Hz range) have attracted attention because they have been postulated to play a central role in the temporal coding of sensory information (Gray 1994; Ritz and Sejnowski 1997; Singer 1993) as well as in higher cognitive functions (Baldeweg et al. 1998; Llinas and Ribary 1993; Steriade 1999). The aim of the present work was to examine the influence of a decrease in pCO₂ (hypocapnia) on an in vitro model of gamma activity that was based on exogenous activation of muscarinic receptors in hippocampal slices (Fisahn et al. 1998).

We found that hypocapnia exerted two mechanistically distinct effects. First, hypocapnia induced a decrease in gamma-band amplitude and an increase in oscillation frequency. Second, there was a dramatic increase in the temporal stability of the gamma oscillations. The amplitude and frequency effects were closely mimicked by intracellular alkalinization with TriMA but were not reversed by bicuculline or mimicked by the GABA up-modulators MDZ and PB. It therefore seems to be evident that, during hypocapnia, intracellular alkalinization was the primary factor leading to amplitude decrease and frequency increase. It is well known that gap-junctional conductance is enhanced at elevated levels of pH_i (Church and Baimbridge 1991; Spray et al. 1981). In this light, it is possible that the frequency increase in hypocapnia was caused by an enhanced level of interneuronal excitation that was mediated by an increase in the efficacy of gap-junctional coupling (Galarreta and Hestrin 1999; Gibson et al. 1999; Tamás et al. 2000; Traub and Bibbig 2000).

Robust temporal stabilization of gamma oscillations during hypocapniathe main finding of the present studywas unlikely to be mediated by an increase in pH_i. In light of the synchronizing effect of somatic/dendritic gap junctions between interneurons (Tamás et al. 2000), we were surprised to find that TriMA, in contrast with hypocapnia, decreased temporal stability. This finding excludes the possibility that intracellular alkalinization could underlie the hypocapnia-induced stabilization. We found that hypocapnia enhanced the amplitude of GABA_A receptor-mediated IPSCs but did not affect the current's decay-time constant. This selective amplitude enhancement could be counteracted with a low concentration of the GABA_A receptor antagonist bicuculline. Bicuculline decreased the ACF decay times to close to control levels and thereby largely reversed the hypocapnia-induced temporal stabilization of gamma oscillations. The idea that an exclusive increase in IPSC amplitude is required to obtain the stabilizing effect was further supported by the finding that the GABA_A receptor up-modulators MDZ and PB, which increase both the IPSC amplitude and the decay-time constant, did not induce temporal stabilization but had the opposite effect. Taken together, these results indicate that an increase in IPSC amplitude, mediated by a rise in pH_o, plays a major role in the temporal stabilization of gamma oscillations during hypocapnia.

How then could the hypocapnia-mediated change in postsynaptic inhibition lead to temporal stabilization? One possibility is that an increase in IPSC amplitude, which is not associated with a prolongation of its decay-time constant, leads to a sharper onset of, and especially to a more abrupt termination of, postsynaptic inhibitory action. This could narrow the time window for the "escape" of postsynaptic excitation which in turn would sharpen the population discharges of interneurons and pyramidal cells and thereby decrease the amount "noisy" fluctuations in population behavior. The present data do not distinguish between pre- and postsynaptic effects of pH_o that are involved in the enhancement of the IPSC. Both are possible because certain GABA receptors are up-modulated by an increase in pH_o (Huang and Dillon 1999; but see Pasternack et al. 1996) whereas an increase in pH_o is also known to enhance voltage-gated Ca²⁺ currents (Tombaugh and Somjen 1998), which may enhance the release of GABA from presynaptic terminals. In addition, a shift in GABA_A receptor reversal potential may contribute to IPSC enhancement (Kaila 1994).

In conclusion, our results indicate that changes in brain pCO₂ can have strong effects on gamma oscillations. The effects are mediated by changes in interstitial as well as intracellular pH. In particular, respiratory alkalosis, which is similar to that associated with hyperventilation in vivo, has a profound influence on the temporal stability of gamma oscillations. We therefore suggest a novel kind of pCO₂/pH modulation where extracellular alkalosis leads to an increase in the efficacy of entrainment of pyramidal cell firing by the local inhibitory network.