|Home | About | Journals | Submit | Contact Us | Français|
Cortical pyramidal neurons alter their responses to input signals depending on behavioral state. We investigated whether changes in somatic inhibition contribute to these alterations. In layer 5 pyramidal neurons of rat visual cortex, repetitive firing from a depolarized membrane potential, which typically occurs during arousal, produced long-lasting depression of somatic inhibition. In contrast, slow membrane oscillations with firing in the depolarized phase, which typically occurs during slow-wave sleep, produced long-lasting potentiation. The depression is mediated by L-type Ca2+ channels and GABAA receptor endocytosis, whereas potentiation is mediated by R-type Ca2+ channels and receptor exocytosis. It is likely that the direction of modification is mainly dependent on the ratio of R- and L-type Ca2+ channel activation. Furthermore, somatic inhibition was stronger in slices prepared from rats during slow-wave sleep than arousal. This bidirectional modification of somatic inhibition may alter pyramidal neuron responsiveness in accordance with behavioral state.
In sensory cortex, pyramidal neurons send output signals to other cortical and subcortical areas, contributing to perception and behavior. They receive sensory afferent input from the thalamus as well as input from other cortical areas. Our degree of consciousness as well as physiological responsiveness to sensory stimulation is greatly reduced when we sleep. Sensory inputs are adjusted depending on arousal state and reduced at the level of the thalamus during sleep (McCormick and Bal, 1997), which may at least in part explain the reduction in sensory responsiveness when we sleep. There is accumulating evidence suggesting that even during sleep the brain is involved in some sort of information processing, such as consolidation of memory acquired during arousal, although we lack consciousness and any memory of that mental activity (Walker and Stickgold, 2004; Hobson, 2005). Thus, signal processing in cerebral cortex may change drastically according to behavioral state.
Consistent with this hypothesis, cortical pyramidal neurons show different intrinsic firing patterns depending on behavioral state, although the mean firing rate does not change much (Steriade et al., 1993, 2001). During quiet wakefulness, these neurons maintain a depolarized membrane potential and fire repetitively, whereas, during slow-wave sleep, they undergo slow membrane potential oscillations (<1 Hz) and fire on the depolarizing phase of the oscillations. The significance of these distinct firing patterns is not known.
We hypothesized that the distinct firing patterns of pyramidal neurons contribute to alterations in cortical signal processing depending on behavioral state. The firing pattern of postsynaptic cells could produce plastic changes in inhibitory connections to them (Gaiarsa et al., 2002), and somatic inhibition of pyramidal neurons is capable of regulating neuronal output very effectively. Thus, in the present study, we tested whether these different patterns of firing modulate somatic inhibition onto pyramidal neurons. The present results show that somatic inhibition undergoes up- and downregulation depending on whether their firing pattern mimics slow-wave sleep or an aroused state, respectively. It is likely that these modifications result from trafficking of GABAA receptors into and out of synaptic membranes, and the direction of modification depends on the activation of different subtypes of voltage-gated Ca2+ channels.
We examined inhibitory postsynaptic currents (IPSCs) generated at synapses that terminate on the soma of visualized layer 5 pyramidal neurons in slices from rat visual cortex. Using whole-cell patch-clamp recording and pharmacological blockade of excitatory synaptic transmission, we recorded IPSCs as inward currents at −70 mV, which was near the resting membrane potential, with an internal electrode solution that contained a high concentration of Cl−(75 mM). To selectively activate somatic synapses, presynaptic fibers were stimulated at weak intensities with a sharp glass micropipette placed near the soma (“s” in Figure 1A, top left). To confirm stimulus selectivity, we applied bicu-culline methiodide (BMI, 100 μM) iontophoretically (5–50 nA) to the soma near the stimulus site or to the apical dendrite slightly (30–50 μm) distal to the stimulus site via another sharp glass pipette. In an example case shown in Figure 1A, BMI application to the soma totally abolished IPSCs evoked by stimuli at intensities less than twice the threshold value required to evoke IPSCs, whereas BMI application to the apical dendrite reduced IPSCs only slightly. In six cells examined, IPSCs evoked by stimuli at 1.5 times the threshold intensity were completely abolished by BMI applied to the soma, but they were reduced only slightly by BMI applied to the dendrite (87.5% ± 3.9% of control, n = 6). These observations suggest that IPSCs evoked by stimuli weaker than 1.5 times the threshold intensity are mostly of somatic origin. Thus, we used stimuli 1.2–1.5 times the threshold intensity required to evoke IPSCs in the following experiments.
To test whether repetitive firing can modify inhibitory synaptic strength, we applied brief depolarizing current pulses (n = 1500, 20 Hz for 5 s, repeated at intervals of 10 s) in current-clamp mode, each of which initiated one action potential, while the test stimulation eliciting IPSCs was stopped. When these action potentials were evoked at a depolarized membrane potential (−60 mV), which mimics the firing pattern of pyramidal cells during quiet wakefulness (Steriade et al., 2001), long-lasting depression of IPSCs was induced (Figures 1B and 1D). The magnitude of the depression was a 46.5% ± 9.51% (n = 9) decrease from the baseline level at 30–40 min after repetitive firing (Figure 1F). This depression is similar to the long-lasting depression of inhibitory synaptic strength previously demonstrated using intracellular recording with sharp glass microelectrodes (Kurotani et al., 2003). When repetitive firing was elicited from the resting membrane potential (about −70 mV), similar long-lasting depression was also induced (Figures 1D and 1F). In contrast, when action potentials were elicited from a hyperpolarized membrane potential (−90 mV), long-lasting potentiation occurred instead (Figures 1C, 1E, and 1F; average magnitude, 113% ± 48.2% increase, n = 9). Long-lasting potentiation and depression were not accompanied by a change in series resistance, input resistance, or 10%–90% rise time of IPSCs for any experimental group (p > 0.7; see Figures S1 and S2 available online). When the membrane potential was maintained at −60 or −90 mV but without repetitive firing, no change in IPSC amplitude was observed (Figures 1D, 1E, and 1F).
Although high-frequency firing of sensory cortical neurons was demonstrated during quiet wakefulness using extracellular unit recording as well as intracellular recording with sharp glass microelectrodes, a much lower frequency of firing was shown by a recent study using a whole-cell recording method, which provides higher-quality intracellular recordings compared with the method using sharp glass microelectrodes (Margrie et al., 2002). These authors suggested that sampling bias may at least partly explain this discrepancy because extracellular unit recording easily overlooks cells showing very low-frequency firing. An extracellular recording study using microwires chronically implanted, which likely detect even very low-frequency firing, demonstrated that sensory cortical neurons fired at about 2 Hz in the quiet state of awake rats (Fanselow and Nicolelis, 1999). Thus, we tested whether such low-frequency firing is also effective to induce modifications in IPSCs. The same number of action potentials elicited at a low frequency (2 Hz instead of 20 Hz) could produce similar long-lasting changes, although the effect was weaker (Figure S3), suggesting that these modifications can be induced by repetitive firing in a wide range of frequency and hence in a majority of cells.
It is unlikely that the induction of these changes we observe depends on synaptic transmission, because we stopped test stimulation during the repetitive firing, and polysynaptic inhibitory responses do not occur due to the pharmacological blockade of excitatory synaptic transmission. Therefore, we call these modifications long-lasting depression and potentiation to distinguish them from long-term depression (LTD) and potentiation (LTP), which are induced by high-frequency stimulation of presynaptic fibers in the same neurons (Komatsu and Iwakiri, 1993).
We next examined the effect of membrane potential oscillation around the resting membrane potential on inhibitory synaptic strength. Long-lasting potentiation was induced (average magnitude, 99.0% ± 29.0% increase, n = 14; Figures 2A, 2C, and 2D) when the membrane potential was oscillated at a low (0.5 Hz) frequency, with action potentials occurring on the depolarized phase, which typically occurs during slow-wave sleep (Steriade et al., 1993). Long-lasting depression occurred instead (Figures 2B–2D; average magnitude, 34.1% ± 9.05% decrease, n = 9) at higher (5 Hz) oscillation frequencies that are generated while performing a working memory task (Lee et al., 2005). Together, these results suggest that somatic inhibition is up- and down-regulated during slow-wave sleep and arousal states.
If long-lasting potentiation and depression occur daily with changes in behavioral state, the modifications should be reversible in either direction. Indeed, we found in five out of seven cells tested that potentiation produced by repetitive firing from −90 mV was reversed by a second episode of repetitive firing from −70 mV that was started 20 min after the first (Figure S4). However, depression was reversed in only one out of six cells tested (Figure S4). Repetitive firing may change how easy it is to modify inhibitory synaptic strength, because the induction mechanism for activity-dependent synaptic modification is often altered by activity (Abraham and Bear, 1996). In particular, once depression is induced, the threshold for the production of potentiation may increase considerably for a certain period of time after induction. It is also possible that the reversal of depression is more difficult to induce if the induction of potentiation is more sensitive to cytosolic wash out during whole-cell recording than induction of depression.
We observed long-lasting potentiation and depression of IPSCs following experimental manipulations that mimic slow-wave sleep and arousal, respectively. This finding suggests that inhibitory synaptic transmission is more effective during slow-wave sleep than arousal. This prediction was tested by comparing somatic IPSCs in slices prepared from rats during arousal and slow-wave sleep states. We used four pairs of littermates for this experiment. One of each littermate pair was anesthetized with urethane at a dose inducing slow-wave sleep-like states in their electroencephalogram (EEG; Friedberg et al., 1999). We began preparing our slices from the rats at a time when the slow wave had dominated the EEG for at least 15 min (lower trace in Figure 3A). On the previous or next day, slices were prepared from littermate controls immediately after they were deeply anesthetized with a high dose of urethane after confirming that the rat had been awake for more than 20 min on the basis of its behavior and EEG pattern (upper trace in Figure 3A).
The mean IPSC amplitude of successful events evoked using minimal stimulation strengths (with failures of synaptic transmission in about half of the stimulation trials) was significantly larger in slices prepared from rats during slow-wave sleep-like states than arousal (Figure 3B) when compared between the two groups of rats in a pair-wise manner (Figure 3C; p < 0.002) or between the two groups as a whole, including all sampled neurons (Figure 3D; p < 0.003). No significant difference (p > 0.05) was detected in either series or input resistance between the two groups (Figures 3E and 3F).
We also compared miniature IPSCs (mIPSCs) recorded from neurons in slices prepared from the two groups of rats, using an additional three pairs of littermates (Figure 3G). The mean mIPSC amplitude was significantly larger in slices prepared from rats during slow-wave sleep-like states than arousal when compared between the two groups of rats in a pair-wise manner (black circles in Figure 3H; p < 0.05) or between the two groups as a whole, including all sampled neurons (black bars in Figure 3I; p < 0.0001). On the other hand, there was no significant difference in the mean mIPSC frequency (black circles and bars in Figures 3J and 3K, p > 0.2) between the two groups. No significant difference (p > 0.3) was detected in either series or input resistance between the two groups (Figures 3L and 3M).
We further examined mIPSCs in slices prepared from rats in natural (not anesthetic-induced) slow-wave sleep or arousal states. Rats were kept in a small transparent box with a small hole in the top. After confirming that the animal had been in natural slow-wave sleep for more than 80% of the 20 min observation period, based on EEG pattern and behavior, vaporized isoflurane was quickly introduced into the box through the hole. Rats were anesthetized within 30 s after isoflurane application and then slices were quickly prepared from them. Slices from aroused animals were prepared in the same way, except that animals were in an aroused state that continued for more than 20 min. As before, a significant difference was found in the amplitude but not the frequency of mIPSCs between arousal and slow-wave sleep groups (red circles and bars in Figures 3H–3K). It is likely that the majority of mIPSCs we detect in layer 5 pyramidal neurons with somatic recordings arise from synapses on the soma and proximal dendrites (Salin and Prince, 1996; Ling and Benardo, 1999). Thus, these observations suggest that somatic inhibition of layer 5 pyramidal neurons undergoes bidirectional modification in accordance with the behavioral state.
Long-lasting changes in inhibitory synaptic strength were similarly produced by applying depolarizing voltage pulses (amplitude 70 mV, duration 10 ms; Figure 4A, top left) at the same frequencies as those used for stimulation with brief current pulses. Stimuli that started from a holding potential of −70 mV induced depression, while stimuli started from −90 mV produced potentiation (Figures 4A and 4C). Thus, we used the better membrane control inherent in this voltage pulse protocol to examine in detail the mechanisms mediating changes in inhibitory somatic synapse strength. Postsynaptic loading of the Ca2+ chelator BAPTA (10 mM) blocked both long-lasting depression and potentiation of somatic IPSCs (Figures 4B and 4C). This indicates that both modifications require postsynaptic Ca2+ increases for induction. We also found that long-lasting depression and potentiation were associated with a decrease and an increase of IPSC conductance, respectively, with no significant changes in the reversal potential of IPSCs (Figure S5).
To test whether these modifications are expressed at the postsynaptic site, we examined responses to GABA applied locally to the soma of recorded cells from a sharp pipette placed very near to the cell soma (Figure 5A). GABA responses were completely abolished by bath application of BMI (20 μM), confirming that GABA responses were mediated by GABAA receptors (Figure 5A). Voltage pulse stimuli started from −70 mV produced long-lasting depression of GABA responses in all cells, whereas stimuli started from −90 mV produced long-lasting potentiation (Figures 5B and 5C). We also found that changes in GABA response were associated with changes in GABA conductance, but not changes in the reversal potential, as was the case for the long-lasting changes in somatic IPSCs (Figure S6). These results strongly suggest that bidirectional modification in inhibitory synaptic strength occurs at the postsynaptic site, although some additional changes could also occur presynaptically. In addition, long-lasting potentiation and depression of GABA responses also occurred in the absence of pharmacological blockade of glutamate receptors (Figure S7), supporting the view that this bidirectional modification of somatic inhibition occurs in the physiological condition.
L- and R-type Ca2+ channels are abundantly present in the soma of neocortical pyramidal neurons (Hell et al., 1993; Yokoyama et al., 1995). R-type Ca2+ channels are activated and inactivated at membrane potentials more hyperpolarized than L-type channels (Zhang et al., 1993; Foehring et al., 2000). Notably, R- but not L-type Ca2+ channels are inactivated around the resting membrane potential. This prompted us to test with specific pharmacological agents whether the direction of synaptic modification depends on the subtype of Ca2+ channel activated. In the presence of nifedipine (20 μM), a selective L-type Ca2+ channel blocker, voltage pulse stimuli that started from −70 mV induced long-lasting potentiation instead of depression, and this potentiation was completely blocked by the addition of 500 nM SNX-482 (Figures 6A and 6C), a toxin that blocks R-type Ca2+ channels selectively (Newcomb et al., 1998). In the presence of SNX-482, voltage pulse stimuli that started from −90 mV induced long-lasting depression instead of potentiation, and the addition of nifedipine abolished this depression (Figures 6B and 6C). These results suggest that the direction of synaptic modification depends on the relative amount of Ca2+ entering through the two different Ca2+ channels.
Calcium imaging of pyramidal neuron soma showed that R-type Ca2+ channel-mediated Ca2+ increases were larger for voltage pulse stimuli started from the depolarizing phase of the membrane oscillation at 0.5 Hz, but not 5 Hz, than those started from −70 mV (Figure S8). In contrast, no difference was found in L-type Ca2+ channel-mediated Ca2+ increases when voltage pulse stimuli were started from the membrane potential of −70 mV or from the depolarizing phase of the membrane oscillation at either frequency (Figure S8). Thus, potentiation likely occurs when the ratio of R-type Ca2+ channel currents to L-type Ca2+ channel currents is relatively high, whereas depression instead occurs in the opposite condition.
Although these results suggest that Ca2+ entering through R- and L-type Ca2+ channels is the main source of intracellular Ca2+ transients responsible for the modification of inhibition, other sources of Ca2+, coming from outside of the neuron or released from internal Ca2+ stores, could also contribute to the production of the synaptic modifications (Cavazzini et al., 2005). We confirmed that Ca2+ entry through NMDA receptors and Ca2+ release from internal Ca2+ stores are not essential for either long-lasting potentiation or depression (Figures S9 and S10). However, we cannot rule out the possibility that voltage-gated Ca2+ channels other than L- and R-type Ca2+ channels play some role in these modifications.
Changes in the property and/or number of somatic GABAA receptors can account for the observed modifications in inhibitory synaptic strength. To discriminate between these possibilities, we measured the variation in GABA response before and after inducing long-lasting changes in synaptic strength. Changes in the coefficient of variation (CV) of GABA response are thought to reflect changes in channel number or open probability but not single-channel conductance at the synapse (Faber and Korn, 1991). The CV increased and decreased significantly in association with long-lasting depression and potentiation, respectively (Figure 7A). These changes in CV are suggestive of a change in GABAA receptor channel number, but could also be explained by a change in open probability.
The number of GABAA receptors at the cell surface is regulated by endocytosis and exocytosis (Barnes, 2000; Kittler and Moss, 2003). Thus, we tested whether interfering with the trafficking of GABAA receptors affects long-lasting changes in inhibitory synaptic strength. Because endocytosis of GABAA receptors requires the interaction of dynamin and amphiphysin (Marsh and McMahon, 1999), we interrupted endocytosis by loading the whole-cell patch pipette with a “P4” peptide (100 μM), which specifically blocks this interaction (Kittler et al., 2000). Although loading the neuron with the P4 peptide did not affect baseline GABA responses (Figure S11), the voltage pulse stimuli that normally produce depression induced potentiation instead (Figures 7B and 7D). When a scrambled P4 peptide (100 −M) was loaded into the cell, depression occurred (Figures 7B and 7D) similarly to the case of loading no peptide (p > 0.2; cf. Figures 5B and 5C). Thus, it is likely that long-lasting depression is caused by the endocytosis of GABAA receptors.
We next tested the effect of botulinum toxin D (BoNT/D), a specific antagonist for vesicular exocytosis (Xu et al., 1998; Schiavo et al., 2000), on potentiation produced by voltage pulse stimuli started from −90 mV. Loading of the toxin (100 nM) did not affect baseline GABA responses (Figure S11). Depression occurred far more frequently than potentiation in BoNT/D-loaded cells, whereas potentiation occurred in heat-inactivated BoNT/D-loaded cells (Figures 7C and 7D) as in the case of control cells (p > 0.9; cf. Figures 5B and 5C). These results suggest that potentiation is ascribed to exocytosis of GABAA receptors. Therefore, GABAA receptor withdrawal and insertion may contribute to long-lasting depression and potentiation, respectively, as was shown for AMPA receptors in NMDA receptor-dependent synaptic plasticity (Malinow and Malenka, 2002; Song and Huganir, 2002; Bredt and Nicoll, 2003; Collingridge et al., 2004), although some additional changes could occur in the property of GABAA receptor channels.
In the present study, we demonstrated a new type of bidirectional synaptic modification in which repetitive firing of cortical pyramidal neurons produces long-lasting potentiation and depression of somatic inhibitory synaptic transmission depending on the firing patterns of these neurons, mimicking those during slow-wave sleep and arousal state, respectively. The direction of modification depends on the relative amount of Ca2+ entry through R- and L-type Ca2+ channels. It is likely that Ca2+ entry through R-type Ca2+ channels initiates insertion of GABAA receptors into synaptic membrane, leading to long-lasting potentiation, while that through L-type Ca2+ channels initiates removal of those receptors from synaptic membrane, leading to long-lasting depression (Figure 8). The dependence of modification polarity on membrane potential state is likely explained by the difference in the voltage dependence and kinetics of inactivation of R- and L-type Ca2+ channels (Zhang et al., 1993; Foehring et al., 2000). Because both Ca2+ channels are commonly present in the soma of cortical pyramidal cells (Hell et al., 1993; Yokoyama et al., 1995), state-dependent modification of somatic inhibition might occur generally in neocortex.
It is generally believed that synaptic plasticity is more useful for information storage and processing if it is bidirectional (Willshaw and Dayan, 1990; Linden, 1994). Many forms of long-term synaptic modifications are indeed bidirectional (Linden, 1994; Bear and Abraham, 1996; Gaiarsa et al., 2002). The induction of enduring potentiation and depression often requires a temporary increase in postsynaptic Ca2+ concentration (Cavazzini et al., 2005). The Ca2+ transient may occur through entry from outside of the cell via ligand-gated receptors or voltage-gated Ca2+ channels, or release from internal Ca2+ stores (Cavazzini et al., 2005). LTP and LTD can both be induced by NMDA receptor activation at many excitatory synapses, such as hippocampal CA1 pyramidal neuron synapses (Bliss and Collingridge, 1993; Malenka and Nicoll, 1993; Bear and Abraham, 1996). It has been proposed that the direction of modification is determined by the amount of Ca2+ elevation (Artola and Singer, 1993; Malenka and Nicoll, 1993; Lisman, 1994). A high level of Ca2+ increase produces LTP, whereas a low level of Ca2+ increase produces LTD. A similar hypothesis was also suggested for modifications at inhibitory synapses between Purkinje neurons and deep cerebellar nucleus neurons, although voltage-gated Ca2+ channels are involved in their induction (Aizenman et al., 1998).
There are some reports suggesting that the direction of modification depends on the source of Ca2+. At inhibitory synapses in layer 5 neurons of developing rat visual cortex, high-frequency stimulation of presynaptic fibers induces LTP that requires IP3 receptor-mediated Ca2+ release from internal Ca2+ stores when NMDA receptors are pharmacologically blocked (Komatsu and Iwakiri, 1993; Komatsu., 1996). When the same stimulation was allowed to activate NMDA receptors, it induced LTD instead (Komatsu and Iwakiri, 1993; Yoshimura et al., 2003). The direction of modification seems to be gated by NMDA receptors in post-synaptic neurons. Insufficient NMDA receptor activation may allow LTP induction, whereas sufficient activation may prevent LTP production and produce LTD instead. Similar LTP and LTD were reported for inhibitory synaptic transmission in CA3 pyramidal neurons in neonatal rats (Mclean et al., 1996).
In the current study, we demonstrated a novel mechanism for determining the sign of synaptic modification that depends on the relative amount of Ca2+ entry through different types of Ca2+ channels, with R- and L-type Ca2+ channels contributing to long-lasting potentiation and depression, respectively. These two types of channels have very different voltage dependence and kinetics, especially for their inactivation (Zhang et al., 1993; Foehring et al., 2000). In dissociated cortical pyramidal cells, R-type Ca2+ channel currents are maximal when voltage steps are applied from hyperpolarized membrane potentials (Foehring et al., 2000), which may explain why potentiation is produced by repetitive firing preferentially from hyperpolarized membrane potentials. Blockade of L-type Ca2+ channels converted depression to potentiation, while blockade of R-type Ca2+ channels converted potentiation to depression. Therefore, it is likely that the ratio of Ca2+ currents through R-type channels relative to L-type channels rather than the absolute value of R- or L-type Ca2+ channel currents is the primary factor in determining the direction of modification. In our previous study using sharp glass microelectrodes (Kurotani et al., 2003), action potentials elicited from a resting membrane potential of about −55 mV induced long-lasting depression, and nifedipine abolished the depression but failed to convert it to potentiation. This suggests that action potentials elicited from that membrane potential could not produce R-type Ca2+ channel currents necessary for the induction of long-lasting potentiation, due to a considerable inactivation of R-type Ca2+ channels (Foehring et al., 2000).
The time course of R-type Ca2+ channel inactivation is voltage dependent, and the inactivation seems much slower around the resting membrane potential than around the peak of action potentials, although the detailed kinetics are not yet well known (Zhang et al., 1993; Jouvenceau et al., 2000; Magistretti et al., 2000). The recovery from inactivation is very slow, such that full recovery requires about 1 s (Zhang et al., 1993; Randall and Tsien, 1997; Nakashima et al., 1998; Sochivko et al., 2003). Thus, the level of inactivation on the depolarizing phase of an oscillation may depend on its frequency. Ca2+ imaging suggested that the inactivation level of R-type Ca2+ channels on the depolarizing phase of a membrane oscillation at 0.5 Hz was significantly lower than that at the steady-state membrane potential of −70 mV. However, this difference in inactivation level was found neither for 5 Hz oscillation in R-type Ca2+ channels nor for either frequency oscillation in L-type Ca2+ channels. Therefore, slow but not fast oscillations may produce long-lasting potentiation.
Depolarizing stimulation produced bidirectional modifications in somatic GABA responses, which were essentially the same as those in somatic IPSCs, suggesting that inhibitory synaptic modifications occur at the postsynaptic site. We tested whether GABAA receptor trafficking is involved in these modifications. It has been reported that endocytosis and exocytosis of GABAA receptors contribute to the regulation of the number of those receptors in postsynaptic membrane (Wan et al., 1997; Nusser et al., 1998; Kittler et al., 2000). The direction of modification was reversed by pharmacological blockade of either endocytosis or exocytosis. If only either exocytosis or endocytosis can occur at one time, depending on the ratio of activation of R- and L-type Ca2+ channels, blockade of either trafficking process would merely abolish modifications but not reverse its direction. Thus, Ca2+ entry through R- and L-type Ca2+ channels may drive exocytotic and endocytotic processes of GABAA receptors, respectively, with each type of channel located in close proximity to the Ca2+ sensors for their specific trafficking process. It is likely that the opposite transport of receptors into and out of the synaptic membrane occurs in parallel, and the direction of modification is determined by the net movement of GABAA receptors. It was suggested that endocytosis and exocytosis regulates the number of synaptic GABAA receptors constitutively on a rather short timescale because the loading of P4 alone quickly increased the amplitude of miniature IPSCs in culture preparations (Kittler et al., 2000). In our slice preparation, neither P4 nor botulinum toxin affected the amplitude of GABA responses in a range of 10 min, suggesting that their constitutive turn-over is not very fast, although postsynaptic Ca2+ increases could move them quickly.
Many recent studies have examined the role of AMPA receptor trafficking during synaptic plasticity, establishing it as a fundamental mechanism for the long-term modification of excitatory synapses (Malinow and Malenka, 2002; Song and Huganir, 2002; Bredt and Nicoll, 2003; Collingridge et al., 2004). In hippocampal CA1 pyramidal neurons, it is well documented that NMDA receptor-dependent LTP and LTD are expressed by the insertion and removal of AMPA receptors into and out of synapse, respectively. Thus, receptor trafficking is likely used for activity-dependent modifications expressed postsynaptically at both excitatory and inhibitory synapses. Although at CA1 excitatory synapses, blockade of exocytosis and endocytosis prevented LTP and LTD, respectively, the direction of modification was not reversed by the blockade of either process alone (Lledo et al., 1998; Lüscher et al., 1999; Man et al., 2000). Thus, in contrast to the present results, at excitatory synapses, either endocytosis or exocytosis can be initiated at a given time by separate Ca2+-sensitive signaling pathways, such as calcium calmodulin-dependent protein kinase and calcineurin, each with different Ca2+ sensitivities (Malenka and Nicoll, 1993; Lisman, 1994).
At the postsynaptic site, the efficacy of inhibitory synaptic transmission can be modified by changing intracellular Cl− concentration in addition to changing the number and/or efficacy of GABAA receptors on the postsynaptic membrane. A recent study using whole-cell gramicidine perforated-patch recording, which keeps the intracellular Cl− concentration intact, unlike conventional whole-cell recordings (Kyrozis and Reichling, 1995), tested the effect of repetitive firing on IPSCs in hippocampal neurons (Fiumelli et al., 2005). Although the repetitive firing used in that study is similar to that we used to induce depression, it produced no change in IPSC conductance but shifted IPSC reversal potentials in a depolarized direction, leading to a reduction in inhibitory synaptic strength. The reversal potential shift requires postsynaptic Ca2+ entry through L-type Ca2+ channels (Fiumelli et al., 2005), like found for long-lasting depression. Although we did not observe any changes in IPSC reversal potential, such a shift could occur in association with a reduction in IPSC conductance, because our experiments used conventional whole-cell recording techniques. However, the reversal potential did not change at all in association with depression, even in our previous study using sharp glass microelectrodes (Kurotani et al., 2003), suggesting that a reversal potential shift, if it exists at all, is small in layer 5 pyramidal neurons, and the primary changes occur in IPSC conductance.
Cortical pyramidal neurons show very different membrane potentials and firing patterns during arousal and slow-wave sleep (Steriade et al., 1993, 2001). Although the functional meaning of this difference is not clear, the present study provides some insight into this issue. Somatic inhibition is likely up- and down-regulated by firing during slow-wave sleep and arousal. Consistent with this, the quantal amplitude of IPSCs in slices prepared from aroused animals was smaller than that in slices prepared from animals during natural slow-wave sleep or in a urethane-induced slow-wave sleep-like state. Thus, behavioral state-dependent changes in pyramidal neuron firing may alter the strength of somatic inhibitory synaptic transmission and hence the mode of signal processing in cortical circuits. This behavioral state-dependent switching of signal processing modes in the cortex brought about by the modulation of somatic inhibitory synaptic strength is a potential new role for synaptic plasticity in brain function. Accumulating evidence suggests that the brain is involved in some kind of memory processing even during sleep (Walker and Stickgold, 2004; Hobson, 2005), and the reduction of output gain might be important in achieving the functional goals of slow-wave sleep.
The modification of somatic inhibition may also contribute to the persistence of a particular behavioral state. Although sensory inputs are reduced at the level of the thalamus during sleep (McCormick and Bal, 1997), they still impinge onto the cortex to some degree (Livingstone and Hubel, 1981; Edeline et al., 2000; Portas et al., 2000). Potentiation of somatic inhibition during slow-wave sleep may further reduce sensory input to the cerebral cortex and contribute to the persistence of sleep states. Because inhibitory connections are likely involved in the generation of slow waves (Steriade et al., 1993), potentiation of inhibition might also play a role in the persistence of the slow oscillations.
Coronal slices (300 μm thick) were prepared from primary visual cortex of 20- to 30-day-old Sprague-Dawley rats, anesthetized with isoflurane, as described previously (Kurotani et al., 2003; Yoshimura et al., 2003). An adult-like sleep-wake cycle is already established at this age in rats (Gramsbergen, 1976). In the experiment comparing the amplitude of IPSCs or mIPSCs in slices prepared from rats during arousal and slow-wave sleep, seven pairs of littermates were used. Two to three days before a slice experiment was conducted, screw electrodes for EEGs were implanted into the right skull (6 mm posterior to bregma, 3 mm lateral to midline and 2 mm posterior to lambda, 3 mm lateral to midline) under anesthesia with pentobarbital (40–50 mg/kg). One of each littermate pair was anesthetized with urethane at a dose (1.5 g/kg) inducing a slow-wave sleep-like state in EEG (Friedberg et al., 1999). Slices were prepared from the rat after confirming that slow wave had dominated EEG for at least 15 min. On the previous or next day, slices were prepared from the other littermate quickly and deeply anesthetized with a high dose of urethane (3.0 g/kg) after confirming that it had been awake for more than 20 min on the basis of both EEG pattern and behavior. We considered the rats to have been aroused behaviorally if they kept their eyes open and showed some voluntary movements. In four additional pairs of littermates with similarly implanted EEG electrodes, each rat was kept alone in a small box (25 × 20 × 15 in cm) made from transparent plates with a small hole on the top (3 × 3 in cm), and EEG was recorded continuously. Inspection through the transparent box allowed us to judge whether rats were asleep or aroused. We considered the rats to have been sleeping if they had their eyes closed and stayed quiet without obvious voluntary movements. Determination of slow-wave sleep and arousal state was based on EEG together with the observed behavior. Vaporized isoflurane was quickly introduced into the box through the hole after slow-wave sleep had dominated for more than 80% of the 20 min observation period (one of each littermate pair) or after the aroused state had continued for more than 20 min (the other littermate). Rats were anesthetized within 30 s after isoflurane application and then slices were quickly prepared from them. The slices were perfused with oxygenated (95% O2 and 5% CO2) artificial cerebrospinal fluid (ACSF) containing (in mM) NaCl 126, KCl 3, NaH2PO4 1.2, MgSO4 1.3, CaCl2 2.4, NaHCO3 26, and glucose 10. The experiments were performed at 27°C–30°C in ACSF containing high concentrations of an N-methyl-D-aspartate (NMDA) receptor antagonist, DL-2-amino-5-phosphonovaleric acid (DL-APV, 100 μM), and a non-NMDA receptor antagonist, 6,7-dinitroquinoxaline-2,3-dione (DNQX, 40 μM), to block excitatory synaptic transmission.
IPSCs and GABA responses were recorded from layer 5 pyramidal neurons using visualized whole-cell patch-clamp methods (Yoshimura et al., 2003). Patch pipettes were filled with an internal solution containing (in mM) KCl 65, K-gluconate 65, NaCl 10, MgSO4 5, HEPES 10, Na-ATP 2, Na-GTP 0.6, K-creatinephosphate 10, and EGTA 0.6 (pH 7.3 with KOH, 305 mOsm, resistance 3–7 MΩ). With this internal solution, IPSC reversal potentials were about −15 mV, and IPSCs were recorded as inward currents under voltage clamp at −70 mV. Throughout the recording session, series resistance was continuously monitored by applying small (2–5 mV) hyperpolarizing voltage pulses unless otherwise mentioned. Cells with a high seal resistance (>1 GΩ) and a series resistance <30 MΩwere selected for analysis. The series resistance was not compensated. If the series resistance changed by more than 10%, the recording was discarded. Somatic IPSCs were evoked by microstimulation of pre-synaptic fibers with a sharp glass pipette filled with normal ACSF, which was placed in close proximity (<10 μm) to the soma of recorded cells, at 0.1 Hz. To evoke GABA responses in the cell soma, we used the same kind of sharp glass pipettes as those used for microstimulation, which were filled with a solution containing (in mM) GABA 0.5–1.0, NaCl 150, and NaOH 10 (pH 11.7). The tip of the electrode was placed close to the soma of the pyramidal neuron, and GABA was ejected by a brief (10–200 ms) constant current pulse (50–500 nA, tip negative) at 30 s intervals. Similar sharp pipettes were also used for iontophoresis of BMI (100 μM in ACSF, 5–50 nA, tip positive). We used a dual microiontophoresis current generator (Model 260, World Precision Instruments, Sarasota, FL) for the local application of GABA and BMI. We detected and analyzed mIPSCs, recorded at −70 mV in the presence of tetrodotoxin (1 μM), using MiniAnalysis software (Synaptosoft, Decatur, GA).
Long-lasting changes in response were induced by one of three protocols, unless otherwise mentioned. (1) To induce long-lasting changes with repetitive firing of postsynaptic neurons, we temporarily changed our whole-cell recording from voltage-clamp mode to current-clamp mode, and then action potentials were elicited with brief depolarizing current pulses (2 ms, 0.8–1.5 nA), each of which produced one action potential, from depolarized (−60 mV), resting (about −70 mV), and hyperpolarized (−90 mV) membrane potentials, which were maintained with DC current injection. The brief pulses were applied at 20 Hz for 5 s, which was repeated 15 times at intervals of 10 s. (2) In the membrane oscillation protocol, sinusoidal current (±0.1–0.2 nA at 0.5 or 5 Hz) was injected at the resting membrane potential (−70 ± 1.0 mV, n = 21) so that the membrane potential oscillated approximately between −90 and −50 mV, eliciting action potentials on the depolarizing phase. If necessary, a slight depolarizing DC current was superimposed on the oscillatory current in order to elicit action potentials. Oscillatory current injection was continued for 10 min at 0.5 Hz, while at 5 Hz it was continued for 3–5 min. (3) In the remainder of the experiments, long-lasting changes in response were induced with depolarizing voltage pulses (amplitude 70 mV, duration 10 ms) started from a membrane potential of −70 or −90 mV in voltage-clamp mode. In this protocol, the time interval and the total number of voltage pulses were the same as in the first protocol using brief depolarizing current pulses in current-clamp mode to elicit action potentials.
Chemical compounds were purchased from the following sources: DL-APV and DNQX from Tocris (Bristol, UK), nifedipine from Research Biochemicals (Natick, MA), and SNX-482 from Peptide Institute Inc. (Osaka, Japan), P4 and scrambled P4 from SIGMA-Genosys (Tokyo, Japan), tetrodotoxin, botulinum toxin D, and urethane from Wako Pure Chemical Industries Ltd. (Osaka, Japan), GABA and BMI from Sigma (St. Louis, Mo), and Isoflurane from Abbott Laboratories (North Chicago, IL). To inactivate botulinum toxin D, it was heated to 100°C for 20 min.
The magnitude of long-lasting changes in response was assessed by measuring the mean IPSC (or GABA) response for the period 30–40 min after repetitive firing or voltage pulse stimulation, compared with the baseline level determined for the 10 min period just before the conditioning stimulation. Data were expressed as mean ± SEM, and Student’s t test or Kolmogorov-Smirnov test (K–S test) was applied to evaluate statistical significance.
This study was supported by grants from the Japanese Ministry of Education, Culture, Science, Sports and Technology to Y.Y. (17500208 and 18021018) and Y.K. (17300101 and 18021017). The authors declare that they have no competing financial interests.
The Supplemental Data for this article can be found online at http://www.neuron.org/cgi/content/full/57/6/905/DC1/.