In primary auditory cortex (AI) neurons, tones typically evoke a brief depolarization, which can lead to spiking, followed by a long-lasting hyperpolarization. The extent to which the hyperpolarization is due to synaptic inhibition has remained unclear. Here we report in vivo whole cell voltage-clamp measurements of tone-evoked excitatory and inhibitory synaptic conductances of AI neurons of the pentobarbital-anesthetized rat. Tones evoke an increase of excitatory synaptic conductance, followed by an increase of inhibitory synaptic conductance. The synaptic conductances can account for the gross time course of the typical membrane potential response. Synaptic excitation and inhibition have the same frequency tuning. As tone intensity increases, the amplitudes of synaptic excitation and inhibition increase, and the latency of synaptic excitation decreases. Our data indicate that the interaction of synaptic excitation and inhibition shapes the time course and frequency tuning of the spike responses of AI neurons.
The vast majority of electrophysiological studies of the primary auditory cortex (AI) have used tones as stimuli. AI spike responses to tones have therefore been well described for some time, and much current research is directed toward its spike responses to more complex sounds. Because of the nonlinearity of AI and the infinite number of sounds, supplementing such research with investigations into the mechanisms underlying the spike responses will help the formulation of a model capable of predicting AI spike responses to arbitrary sounds.
Insight into the network aspects of the mechanisms may be gained by decomposing the synaptic inputs of AI neurons into excitatory and inhibitory components. This is because any particular neuron provides either excitatory or inhibitory synaptic input to all of its postsynaptic neurons, but not both. Furthermore, inhibitory synaptic inputs to AI neurons are primarily cortical in origin, because AI neurons receive few inhibitory synaptic inputs from subcortical stations (Douglas and Martin 1998; Winer 1992).
While there have been many studies about the tone-evoked spike responses of AI neurons, far fewer have examined the synaptic inputs underlying those responses. Previous intracellular work has shown that tones typically evoke a brief depolarization, which can lead to spiking, followed by a long-lasting hyperpolarization (De Ribaupierre et al. 1972; Ojima and Murakami 2002; Volkov and Galazjuk 1991, 1992). This is consistent with synaptic excitation being followed by long-lasting synaptic inhibition. When the tone frequency and intensity are such that the depolarization is small, it is unlikely that intrinsic conductances are opened, and the hyperpolarization must be due mainly to synaptic inhibition. However, when the tone frequency and intensity are such that the depolarization is large and causes a spike, the extent to which the hyperpolarization is due to synaptic inhibition is unclear, because much of the hyperpolarization might be due to intrinsic potassium conductances opening as a result of the large depolarization. We have therefore carried out in vivo whole cell voltage-clamp measurements of the tone-evoked excitatory and inhibitory synaptic conductances of AI neurons of the pentobarbital-anesthetized rat.
We have previously examined the relationship between tone-evoked synaptic inputs and frequency-modulated sweep-evoked spike responses (Zhang et al. 2003). Here we examine the relationship between tone-evoked synaptic inputs and tone-evoked membrane potential responses.
All experimental procedures used in this study were approved under UCSF Animal Care Facility protocols. Experiments were carried out in a sound-attenuating chamber. Female Sprague-Dawley rats about 3 mo old, weighing 280-300 g, were used. Each rat was anesthetized by intraperitoneal injection of sodium pentobarbital (50-80 mg/kg), with the dose adjusted to make the rat areflexic. The rat was maintained in an areflexic state for the rest of the experiment by further intraperitoneal injections of sodium pentobarbital (20-60 mg/kg) when necessary. The rat was placed on a heating pad, and its temperature was maintained at ∼37°C. Prior to any skin incision, bupivacaine was injected subcutaneously at the incision site. A tracheotomy was performed to secure the airway and to reduce spurious respiratory noise. (The animal was not artificially respirated.) The head was held fixed by a custom-made device that clamped it at both orbits and the palate, leaving the ears unobstructed. A cisternal drain was performed. The right auditory cortex was exposed by retracting the skin and muscle overlying it, followed by a craniotomy and a durotomy. The cortical surface was kept moist with normal saline. The location of AI was determined by coarse mapping of multiunit spike responses at 500-600 μm below the pial surface with a parlyene-coated tungsten electrode.
Cell-attached and whole cell recordings
A silver wire, one end of which was coated with silver chloride, served as the reference electrode against which potentials were measured; its chlorided end was inserted between the skull and the dura. The reference electrode was assigned a potential of 0 mV. The potential of the cerebrospinal fluid was assumed to be uniform and equal to that of the reference electrode.
Patch pipettes with resistances of 7 MΩ were made. Pipettes contained a potassium based solution that consisted of (in mM) 135 K-gluconate, 5 NaCl, 5 MgATP, 0.3 GTP, 10 HEPES, and 10 phosphocreatine (2Na), pH 7.3. For whole cell voltage-clamp recordings, pipettes contained a cesium-based solution that consisted of (in mM) 125 Cs-gluconate, 5 TEA-Cl, 4 MgATP. 0.3 GTP, 10 phosphocreatine, 10 HEPES, 0.5 EGTA, 3.5 QX-314, and 2 CsCl, pH 7.2. The pia was broken by slowly lowering and raising the jagged tip of a broken pipette in and out of the cortex. An unbroken pipette was lowered into the cortex, with the pressure in the pipette greater than atmospheric. Dimpling of the cortical surface was not visually detectable. The cortical depth of the pipette tip was estimated according to the distance it had traveled. The cortex was covered with 4% agarose in normal saline. Under voltage clamp, an oscillating potential was set up at the pipette tip; the oscillating potential had a time average of −50 mV; its period was much faster than the breathing rate. The resulting current oscillation was measured. When the amplitude of the current oscillation decreased and an even slower oscillation whose period was the breathing rate of the animal became superimposed on the current oscillation, the pipette tip might be touching the cell membrane of a neuron. At this point, the pipette was slowly advanced to further reduce the amplitude of the current oscillation. The pressure in the pipette was suddenly reduced to less than atmospheric and then returned to atmospheric. Often a giga-ohm seal would spontaneously form within 1 min; if not, additional gentle suction sometimes helped. With a giga-ohm seal, or even a mega-ohm seal, the recording is in cell-attached mode. A giga-ohm seal is required to bring the recording into whole cell mode. A pulse of reduced pressure in the pipette would often break the cell membrane and bring the recording into whole cell mode. Cell-attached and whole cell current-clamp recordings used an Axoclamp 2B amplifier in current-clamp mode. Whole cell voltage-clamp recordings used an Axopatch 200B amplifier in voltage-clamp mode; the whole cell capacitance was compensated and the initial series resistance (25-60 MΩ) was compensated to achieve an effective series resistance of 15-30 MΩ. Signals were filtered at 5 kHz and sampled at 10 kHz (Margrie et al. 2002; Metherate and Ashe 1993; Moore and Nelson 1998; Zhu and Connors 1999).
In cell-attached and whole cell current-clamp recordings, stimuli were delivered into the left ear canal by a tube sealed to a calibrated speaker. Stimuli consisted of 675 pure tone pips, each with 1 of 45 frequencies spanning 1.1-31.1 kHz uniformly on a logarithmic frequency scale, and 15 intensities spanning 5-75 dB SPL uniformly, each with a 50-ms duration and 3-ms cosine rising and falling phases. Single tone pips or noise bursts with the same duration, rising and falling phases were also used. The interstimulus interval was 500 ms.
In whole cell voltage-clamp recordings, stimuli were delivered by a calibrated free field speaker directed toward the left ear. Stimuli consisted of 568 pure tone pips, each with one of 71 frequencies spanning 0.5-64 kHz uniformly on a logarithmic frequency scale, and 8 intensities spanning 0-70 dB SPL uniformly, each with a 25-ms duration and 5-ms cosine rising and falling phases. Single tone pips or noise bursts with the same duration, rising and falling phases were also used. The interstimulus interval was 500 ms.
It should be noted that the spike response of an AI neuron to a 50-ms tone with 3-ms rising and falling phases and the spike response to a 25-ms tone with 5-ms rising and falling phases are virtually identical (Heil 1997a,b).
The excitatory synaptic conductance Ge(t) and inhibitory synaptic conductance Gi(t) at time t were derived using (1) where V is the clamping voltage; Ee and Ei are the reversal potentials of the excitatory and inhibitory synaptic conductances, respectively; and I(V,t) is the amplitude of the synaptic current, relative to the resting current at V. The values of Ee and Ei were set by the ionic composition of the pipette solution and the cerebrospinal fluid (Davson and Segal 1996; Johnston and Wu 1995); the value of Ei was based on the permeability of GABAA conductances to Cl−, but it should be noted that they also pass HCO3− (Bormann et al. 1987). Ge(t) and Gi(t) were the two unknowns in Eq. 1 at any particular t. Measurement of I(V,t) at two different values of V yielded a system of two equations that could be solved for Ge(t) and Gi(t) at any particular t (Anderson et al. 2000; Borg-Graham et al. 1998; Hirsch et al. 1998).
Currents into the neuron were assigned a negative value. Ee and Ei were 0 and −70 mV, respectively; in some cases, values of −60 and −80 mV were also used for Ei.
The resting or leak conductance Gr was derived using (2) where Er is the membrane potential, and Ir(V) is the resting current. Gr and Er were the two unknowns in Eq. 2. Measurement of Ir(V) at two different values of V yielded a system of two equations that could be solved for Gr and Er.
Rather than using V in the equations above, a corrected clamping voltage Vc, given by (3) was actually used, where Rs was the effective series resistance.
We did not correct for any junction potentials.
Estimated membrane potential response
The estimated membrane potential response Vest for the voltage-clamp recordings was derived using (4) where Vr is the resting membrane potential. Vr was assumed to be −60 mV, and Gr, Ge(t), and Gi(t) were derived as described above. If only the excitatory synaptic conductance was taken into account, Gi(t) was set to zero.
We define two terms before proceeding: the tonal receptive field (TRF) of a neuron consists of its responses to tones of various frequencies and intensities; the characteristic frequency (CF) of a neuron is the tone frequency at which the intensity needed to evoke a response is lowest.
Sounds were played through a speaker directed to the left ear of a pentobarbital-anesthetized rat. Recordings were carried out in its right AI or the surrounding auditory cortex. Cell-attached or whole cell recordings of tone or noise-evoked responses were obtained from 71 neurons. Full TRFs were obtained for 32 neurons, all of which were in AI. AI, which is defined by a rostro-caudal CF gradient, was located by coarse mapping of multiunit spike responses (Sally and Kelly 1988). The CFs of the spike responses, membrane potential responses, and synaptic conductances of the 32 neurons matched the CFs of multiunit spike responses at nearby locations.
The main aim of our measurements of tone-evoked synaptic conductances was to examine their contribution to the long-lasting hyperpolarization that follows tone-evoked depolarization. However, because the measurements have permitted us to also examine the relationship between the synaptic inputs and several other aspects of tone-evoked spike and membrane potential responses, we first recapitulate the main characteristics of tone-evoked spike and membrane potential responses before reporting the synaptic conductances underlying them.
Cell-attached recordings of tone-evoked spike responses were obtained from three neurons. Full TRFs were obtained for two neurons. In cell-attached recordings, the pipette is attached to the outside of the cell membrane of the fully intact neuron. Cell-attached recordings are equivalent to extracellular single-unit recordings.
The spike responses from one neuron are shown in Fig. 1. Figure 1A shows three responses, each evoked by a 9-kHz, 75-dB tone. Each response consists of just one or two spikes about 15 ms after tone onset. Figure 1B shows the neuron's TRF; each trace shows the response to one tone; the responses are arranged in a grid according to the frequencies and intensities of the tones; tone frequency increases from left to right; tone intensity increases from bottom to top. The neuron's CF is 9 kHz. When tone intensity is low, only tones with frequencies very close to the CF can evoke spiking. As tone intensity increases, the tone frequency range capable of producing spikes also increases, giving the TRF a V-shape. Figure 1C shows that the spike response strength increases then levels off as the intensity of a CF tone increases. Figure 1D shows that the first-spike latency decreases as the intensity of a CF tone decreases.
The neuron of Fig. 1 is representative of the three neurons from which cell-attached recordings were obtained, except that one neuron often responded with bursts of two or three spikes. The onset-only responses, V-shaped TRFs, and monotonic rate-intensity and latency-intensity functions are characteristic of AI neurons in anesthetized animals (Brugge et al. 1969; DeWeese et al. 2003; Erulkar et al. 1956; Heil 1997a,b; Hind 1960; Katsuki et al. 1959; Merzenich et al. 1975; Oonishi and Katsuki 1965; Phillips and Irvine 1981; Phillips and Kelly 1989; Sally and Kelly 1988).
Membrane potential responses
Whole cell current-clamp recordings of tone or noise-evoked membrane potential responses were obtained from 29 neurons. Resting potentials ranged from −54 to −72 mV, with −63 ± 5 (SD) mV. Full TRFs were obtained for 12 neurons. Cortical depths ranged from 400 to 700 μm, which corresponds to layers III–V (Games and Winer 1988).
The membrane potential responses from a neuron that had spike response characteristics like the neuron of Fig. 1 are shown in Fig. 2. The neuron was 400 μm deep. Its resting potential was −68 mV. Figure 2A shows three responses, each evoked by a 20-kHz, 75-dB tone. About 15 ms after tone onset, depolarization begins and leads to a spike about 5 ms later. A return to the resting potential occurs about 30 ms after the start of the depolarization. The brevity of the depolarization is such that the membrane potential is above the −40 mV threshold for spiking for only 10 ms. Thus the brief depolarization is the basis of the onset-only spike response. There is no hyperpolarization after the depolarization. Figure 2B shows the neuron's TRF. Spikes have been truncated to display subthreshold responses more clearly. The neuron's CF is 20 kHz. Both the membrane potential and spike TRFs are V-shaped, but the membrane potential TRF is broader in its frequency extent, and has a tip lower in intensity (and frequency, in this case). Figure 2C shows that the spike response strength increases then levels off as the intensity of a CF tone increases. Figure 2D shows that the first-spike latency decreases as the intensity of a CF tone decreases.
The neuron of Fig. 2 is representative of the 29 neurons from which current-clamp recordings were obtained, except that a few neurons responded with bursts of spikes, and one neuron had nonmonotonic rate-intensity and latency-intensity functions. Across the 29 neurons, the time courses of tone-evoked and noise-evoked membrane potential responses were similar; the depolarization evoked by CF tones or noise at 60 dB began 15 ± 3 ms after tone onset and lasted 53 ± 23 ms. The brief depolarizations, V-shaped membrane potential TRFs, and broader subthreshold TRFs are characteristic of AI neurons in anesthetized animals (De Ribaupierre et al. 1972; Ojima and Murakami 2002; Volkov and Galazjuk 1991, 1992). [Note that membrane potential responses are similar to local field potentials in some respects (Norena and Eggermont 2002).]
We only occasionally observed a long-lasting hyperpolarization following a tone-evoked depolarization, at least in individual membrane potential responses. Our observations, however, are not inconsistent with those of previous studies; if the conductances that were visible as hyperpolarizations in those studies had reversal potentials close to the resting potential, slightly lower resting potentials in our study could have rendered them invisible as hyperpolarizations. Furthermore, those conductances had an observable effect on the membrane potential of some neurons by affecting their spontaneous activity. In the period before tone onset, spontaneous activity usually consisted of spontaneous depolarizations from the resting potential; if tone onset was followed by the opening of a long-lasting conductance with a reversal potential close to the resting potential, this would cause a long-lasting suppression of the spontaneous depolarizations. The suppression would be visible as a long-lasting hyperpolarization of the average membrane potential in neurons with sufficient spontaneous activity and for which there were recordings of the membrane potential responses to a sufficient number of repetitions of the tone (Destexhe et al. 2003; Monier et al. 2003). In 17 neurons, the hyperpolarization of average membrane potential responses evoked by CF tones or noise at 60 dB lasted until 184 ± 52 ms after tone onset. These timings are comparable to those of the long-lasting hyperpolarizations of individual membrane potential responses previously observed (De Ribaupierre et al. 1972; Ojima and Murakami 2002; Volkov and Galazjuk 1991, 1992).
To measure the synaptic conductances underlying the above-described tone-evoked spike and membrane potential responses, whole cell voltage-clamp recordings of tone or noise-evoked synaptic current responses were obtained from a separate population of 39 neurons. Full excitatory synaptic conductance TRFs were obtained for 18 neurons; full inhibitory synaptic conductance TRFs were also obtained for 8 of the 18 neurons. Cortical depths ranged from 400 and 700 μm, which corresponds to layers III–V (Games and Winer 1988).
Voltage clamping suppresses the activation of intrinsic voltage-sensitive conductances and helps ensure that the recorded current is synaptic. With the 18 neurons, the patch pipette contained a cesium based solution with TEA and QX-314, which further prevented the activation of intrinsic conductances. Cesium and TEA block intrinsic potassium conductances; QX-314 blocks intrinsic sodium conductances. Cesium also blocks GABAB, an inhibitory synaptic conductance. The remaining unblocked excitatory synaptic conductances were AMPA and N-methyl-d-aspartate (NMDA); the remaining unblocked inhibitory synaptic conductance was GABAA. (Douglas and Martin 1998; Hille 2001).
The ionic composition of the patch pipette solution was such that the reversal potentials of AMPA receptors and NMDA receptors were both 0 mV and that of GABAA receptors was −70 mV. Thus when a neuron was clamped at −70 mV, the synaptic current would be mainly excitatory (or more precisely, the synaptic current would be mainly from the excitatory synaptic conductances), and the excitatory current inward; when a neuron was clamped at −30 mV, the synaptic current would be a mixture of excitatory and inhibitory currents, with the excitatory current inward, and the inhibitory current outward. From the synaptic currents at −70 and −30 mV, the corresponding excitatory and inhibitory synaptic conductances were derived with the assumption that each of the currents is linearly proportional to the membrane potential. This assumption is only approximately met by the excitatory current, because the NMDA current is voltage-sensitive. The voltage sensitivity of NMDA receptors is such that more than one-half of them can be opened only at membrane potentials greater than −30 mV (Hestrin et al. 1990; Jahr and Stevens 1990a,b). We clamped neurons only at membrane potentials less than −30 mV to suppress the NMDA current, and accordingly avoid, if only partially, the excitatory current nonlinearity.
We present in detail the time courses of tone-evoked synaptic currents and conductances of five of the eight neurons for which full excitatory and inhibitory synaptic conductance TRFs were obtained. We chose the 5 out of the 8 because they had traces that were the cleanest or most representative of the 39 neurons. A–E of the remaining figures will each show data corresponding to a particular 1 of the 5 neurons: the neuron of Fig. 3C will, for example, be the same as the neuron of Fig. 4C.
Tone-evoked synaptic conductances
The first three panels of Fig. 3A show the synaptic currents evoked by an ∼1-kHz, 70-dB tone in the first of these neurons. The neuron was 680 μm deep. The first panel shows the synaptic current evoked by three repetitions of the tone with the neuron clamped at −70 mV; the net inward current is indicative of the time course of synaptic excitation. The second panel shows the synaptic current evoked by three repetitions of the tone with the neuron clamped at −30 mV; the initial net inward current is followed by a net outward current, indicating that synaptic excitation is followed by synaptic inhibition. The third panel shows the average −70- and −30-mV currents in black and gray, respectively. The resting currents have been subtracted. The last panel shows the corresponding excitatory and inhibitory conductances in black and gray, respectively. The excitatory conductance begins ∼16 ms after the tone onset, rises quickly, peaks 10 ms after that, and takes a further 20 ms to fall to one-half its peak. The inhibitory conductance begins later, peaks later, and lasts longer. It begins 21 ms after the tone onset, peaks 15 ms after that, and takes another 70 ms to fall to one-half its peak. The conductances have had their negative portions set to zero. This only removed a presumably artifactual negative peak from the inhibitory conductance that was coincident with the start of the excitatory conductance. The negative peak was probably due to voltage-clamp errors that resulted in the initial rise of the inward current being faster at −30 than at −70 mV, as can be seen by comparing the first and second panels. Voltage-clamp errors would have been most common near the start of the excitatory conductance, because the initial rise of the inward current was faster than all other phases of the recorded currents. The negative peak causes an uncertainty in the start times of the excitatory and inhibitory conductances of 1 ms, the half-width at half-height of the negative peak.
Figure 3, B–E, similarly shows the synaptic currents and excitatory and inhibitory conductances evoked by a tone in the other four neurons. The four tone frequencies and intensities were 1 kHz and 60 dB; 1.4 kHz and 70 dB; 1.2 kHz and 60 dB; and 22 kHz and 40 dB; respectively. The four neurons were 670, 474, 570, and 630 μm deep, respectively. The five cases all show current and conductance time courses that are qualitatively similar, except that in some cases, the start of the inhibitory conductance rather than coming a few milliseconds after the start of the excitatory conductance, appears to be almost coincident with it. The neuron in Fig. 3C was the only other neuron that had a negative peak like that of the neuron of Fig. 3A; its negative peak causes an uncertainty in the start times of its excitatory and inhibitory conductances of 3 ms. There are probably voltage-clamp errors of similar magnitude in all neurons. We therefore estimate the uncertainty in the start times of the excitatory and inhibitory conductances of the neurons of Fig. 3, B, D, and E, to be between 1 and 3 ms.
Table 1 gives the resting and peak amplitudes of each conductance, and the times in its rising and falling phases when it was at 47.5 and 95% of its peak amplitude. Because the respective excitatory and inhibitory reversal potentials of 0 and −70 mV used to derive the conductances of Fig. 3 are approximate, the amplitudes and times obtained with inhibitory reversal potentials of −60 or −80 mV are also shown in Table 1. Assuming different reversal potentials affects the amplitudes of each excitatory and inhibitory conductance and the (peak normalized) falling phase of each excitatory conductance, it hardly affects either the rising phase of each excitatory conductance or the full time course of each inhibitory conductance. The considerable uncertainty in the falling phase of each excitatory conductance may be because the NMDA current, whose nonlinearity we have only approximately taken into account, contributes more significantly to the falling phase of the excitatory current. There is no similar uncertainty in each inhibitory conductance, although each inhibitory conductance overlaps the falling phase of an excitatory conductance; this is probably because the inhibitory conductance is much larger than the excitatory conductance during its falling phase.
Indeed, the most salient feature of the conductances of Fig. 3 is that the inhibitory conductances last much longer than the excitatory. This was generally true. Across 39 neurons, the time courses of tone-evoked and noise-evoked synaptic conductances were similar; the excitatory conductance evoked by CF tones or noise at 60 dB began 12 ± 3 ms after tone onset and lasted 48 ± 35 ms. In 14 neurons, the inhibitory conductance evoked by CF tones or noise at 60 dB lasted until 215 ± 52 ms after tone onset. The inhibitory conductance can therefore account for the long-lasting hyperpolarization or suppression of spontaneous activity lasting until 184 ms after tone onset on average.
A second feature of the conductances of Fig. 3 is that the inhibitory conductances have substantial temporal overlap with the excitatory conductances. This shows that inhibitory conductance also shapes the depolarization. To examine this shaping, we have estimated the membrane potential responses that the conductances would produce.
Figure 4 shows the membrane potential responses estimated with excitatory and inhibitory conductances taken into account in black; while those estimated with the excitatory conductances taken into account, but not the inhibitory, are gray. They were estimated by assuming that the membrane capacitance could be neglected and that the membrane potential responds so quickly to any synaptic conductance change that it is always instantaneously in equilibrium. The resting conductances in Table 1 were used; the resting potential was assumed to be −60 mV; and the reversal potentials of the excitatory and inhibitory conductances were assumed to be 0 and −70 mV, respectively. Comparison of the black and gray traces shows that the temporal overlap of the inhibitory conductance with the excitatory conductance often reduces the amplitude of the depolarization (Fig. 4, A, B, D, and E) and is essential for the brevity of the depolarization, because it always reduces the depolarization's duration closer to the observed average of 53 ms (Fig. 4, A–E).
The long-lasting hyperpolarization visible in some of the estimated membrane potentials (Fig. 4, B, C, and E) shows the ability of the falling phases of the inhibitory conductances to account for the long-lasting hyperpolarizations and suppressions of spontaneous activity. Overall, the estimated membrane potential responses show that the synaptic conductances can account for the gross time course of those typically measured.
TRFs of excitatory and inhibitory synaptic conductances
The TRFs of the currents are shown in Fig. 5; the TRFs of the conductances are shown in Fig. 6. As expected, the TRFs of the excitatory conductances (Fig. 6, column 1) are V-shaped, as spike and membrane potential TRFs typically are. The TRFs of the inhibitory conductances (Fig. 6, column 2) are also V-shaped, with the same frequency tuning as the excitatory conductances. Furthermore, tones of all frequencies and intensities capable of evoking responses cause an increase of excitatory synaptic conductance, followed by an increase of inhibitory synaptic conductance, as was shown in Fig. 3 for selected tones. In fact, the time courses of the conductances are similar for all tones; only their amplitudes change, with the peak amplitude of the excitatory conductance co-varying with the peak amplitude of the inhibitory conductance (Fig. 6). In other words, to first approximation, the TRFs are frequency-time and intensity-time separable.
However, deviations from frequency-time and intensity-time separability are revealed on examination of the TRFs of the peak amplitudes, the time in the rising phase of the excitatory conductance at which it was at 95% of its peak amplitude (“excitatory conductance latency”), and the time in the falling phase of the inhibitory conductance at which it was at 47.5% of its peak amplitude (“inhibitory conductance duration”). These parameters are little affected by the errors discussed above. Although also little affected by these errors, the TRFs of the times at which the inhibitory conductance was at 95% of its peak were not plotted, because the inhibitory peak is broad and near 95% of its peak for a long time. Again, it may be seen that the peak amplitudes of the excitatory and inhibitory conductances co-vary (Fig. 7, columns 1 and 2). The peak amplitudes (Fig. 7, columns 1 and 2) increase as the intensity of a CF tone increases, as the spike response strength is known to do (e.g., Figs. 1C and 2C). The excitatory conductance latency (Fig. 7, column 3) decreases as the intensity of a CF tone increases, as the first-spike latency is known to do (e.g., Figs. 1D and 2D). There are also systematic changes in the excitatory conductance latency and inhibitory conductance duration with tone frequency (Fig. 7, columns 3 and 4). These nonuniformities in the TRFs of the excitatory conductance latency and inhibitory conductance duration are deviations from frequency-time and intensity-time separability (Fig. 7, columns 3 and 4). In some neurons, the inhibitory conductance duration appears to co-vary with the peak amplitudes, with longer durations corresponding to larger amplitudes (Fig. 7, B and D, columns 2 and 4).
We have shown that tones evoke an increase of excitatory synaptic conductance, followed by an increase of inhibitory synaptic conductance, in AI of the pentobarbital-anesthetized rat. The synaptic conductances can account for the gross time course of the typical membrane potential responses. Intrinsic conductances will further shape membrane potential responses by determining, for example, whether the gross time course results in a single spike or a burst of spikes, and will contribute additional long-lasting inhibition (McCormick et al. 1985; Schwindt et al. 1989, 1992).
In this work, we have assumed that neurons are isopotential, and that a pipette at the cell body would be able to hold the cell body and the dendrites at the same potential. If neurons have extensive, nonisopotential dendrites that are passive and if the excitatory and inhibitory inputs due to different tones are evenly distributed over the dendrites, violation of the isopotentiality assumption might result in underestimation of the synaptic conductances (relative to the apparent resting conductance) and the ratio of synaptic inhibition to excitation; a distortion of the fastest portions of the synaptic conductances will also occur; the frequency and intensity tuning of the excitatory and inhibitory inputs will probably be preserved, as well as their gross time courses. However, anatomical data suggests that excitatory and inhibitory input is not uniformly distributed over the dendrites, with most inhibitory inputs to neocortical pyramidal neurons lying close to the cell body; in this case, the ratio of synaptic inhibition to excitation might be overestimated. More complex distributions of synaptic input in the dendrites will lead to other errors. Problems also arise because dendrites are not passive; active dendrites with voltage-sensitive conductances can magnify small fluctuations in the clamping voltage. An estimation of the contribution of such conductances requires knowledge of their location, as well as that of the synaptic input, in the dendrites. If dendritic processing is important, a neuron cannot be modeled as isopotential. In this work, we have made the assumption of isopotentiality twice: once in obtaining the conductances and once in obtaining the estimated membrane potentials. Since the estimated membrane potentials are reasonable, it appears that the two assumptions of isopotentiality are consistent with each other and reasonable. Nonetheless, the deviation of real neurons from isopotentiality must be kept in mind (Häusser et al. 2000; Meunier and Segev 2002; Segev and London 2000).
Similar results have been obtained recently by Wehr and Zador (2003). They showed that tones evoke synaptic excitation. Synaptic inhibition follows shortly after the onset of the synaptic excitation and temporally overlaps it to ensure a brief depolarization. Synaptic inhibition also had the same frequency tuning as synaptic excitation. However, the durations of the synaptic inhibition they observed were shorter than what we observed. This is probably because they anesthetized rats with ketamine, whereas we used pentobarbital, which increases the duration of synaptic inhibition (Nicoll et al. 1975). A second explanation might be that they used young rats whose ages ranged from P17 to P24, whereas we used 3-mo-old adult rats; it is intriguing and plausible that the duration of synaptic inhibition is different between rats of those two age groups, because it is known that they differ in tonotopy and plasticity (Chang and Merzenich 2003; Zhang et al. 2001, 2002). Since Wehr and Zador (2003) recorded in the auditory cortex, but not necessarily AI, another explanation might be that most of their recordings were made in a different part of the auditory cortex. For these reasons, it is unclear whether the long-lasting hyperpolarizations or suppressions of spontaneous activity that have been observed in awake and ketamine-anesthetized animals (De Ribaupierre et al. 1972; Volkov and Galazjuk 1991, 1992) are due to the same synaptic inputs as those in pentobarbital-anesthetized animals (Ojima and Murakami 2002).
Temporal shaping of spike responses by cortical inhibition
Our data indicate that tone-evoked synaptic inhibition contributes to two forms of adaptation, which is a decrease in the ability of a neuron to respond to sound due to previous sound exposure. First, tone-evoked synaptic inhibition, being essential for the brevity of the depolarization, helps ensure that the neuron spikes only in response to the tone onset, but not the remainder of the tone duration (Fig. 4). Second, the long falling phase of the inhibitory conductance, which can account for the hyperpolarization that typically follows the depolarization, reduces the ability of the neuron to spike in response to a second tone (Fig. 4, B, C, and E).
Cortical synaptic inhibition thus helps generate the increasing amount and duration of adaptation present in successive auditory neural stages from the auditory nerve to AI, which may be ultimately manifested in forward masking. Forward masking is a perceptual phenomenon in which a tone causes a reduction in the listener's ability to hear subsequent tones. Forward masking is so robust a perceptual phenomenon that it can be used by the MPEG digital file format to store sounds compactly (Painter and Spanias 2000). The increasing adaptation of the successive stages is important for explaining forward masking, because the adaptation of the auditory nerve is insufficient to explain it (Eggermont 2001; Frisina 2001; Langner 1992; Phillips et al. 2002; Relkin and Turner 1988). Interestingly, the synaptic inhibition we observed had durations of 100-200 ms, similar to the durations of the adaptation observed in the spike responses of AI neurons and of forward masking (Brosch and Schreiner 1997; Calford and Semple 1995; Moore 2003). Adaptation generally permits a system to ignore the constant, less immediately relevant aspects of its surroundings and frees it to better deal with the changing, more immediately relevant aspects (Phillips and Hall 1992; Ulanovsky et al. 2003). Constancy and change, however, are relative concepts, and have meaning only with respect to a given time scale. The build-up of adaptation in successive neural stages may be to provide multiple time scales of adaptation that correspond to the multiple time scales on which aural features occur (Atzori et al. 2001; Fairhall et al. 2001; Kronland-Martinet and Grossmann 1991; Risset 1991). Adaptation can also explain a stream segregation phenomenon in which two tones of different frequencies are perceptually grouped into a single stream when alternated at a low rate but are perceptually divided into two streams when alternated at a high rate (Fishman et al. 2001).
Frequency shaping of spike responses by cortical inhibition
Our data also indicate that tone-evoked synaptic inhibition sharpens the frequency tuning of the spike responses of AI neurons. The temporal overlap of tone-evoked synaptic inhibition with tone-evoked synaptic excitation often reduces the amplitude of the depolarization that would be produced by synaptic excitation alone (Fig. 4, A, B, D, and E). If this is the case for all tone frequencies, tone-evoked synaptic inhibition will result in the range of frequencies capable of causing suprathreshold depolarization being narrower than that resulting from tone-evoked synaptic excitation alone. Previous pharmacological work has shown that cortical synaptic inhibition sharpens the frequency tuning of the spike responses of AI neurons; because all cortical synaptic inhibition was pharmacologically blocked in those studies, whether the sharpening is due to continually present spontaneous synaptic inhibition or to tone-evoked synaptic inhibition had not been determined (Wang et al. 2000, 2002).
Implications for cortical circuitry
Our recordings were made within layers III–V. Layers III and IV are the major thalamo-recepient layers of AI. The tone-evoked synaptic excitation that we measured might therefore be provided by frequency-tuned, monosynaptic excitatory thalamic input onto the cortical neuron. The temporally delayed synaptic inhibition might be provided by the same thalamic input, disynaptically relayed by inhibitory cortical interneurons (Cruikshank et al. 2002). Other possibilities involving recurrent cortical circuitry also exist. As discussed above, the long duration of the synaptic inhibition may be due to the pentobarbital anesthesia we used. It would also be consistent with the long time that GABAB receptors take to close (Connors et al. 1988); however, cesium in our pipette probably blocked most of the GABAB receptors. Another possibility is that the synaptic inhibition is due to inhibitory cortical interneurons that fire at a considerable delay after the end of the stimulus.
Although onset-only responses are the most common type of response in AI of the anesthetized animal, other types of responses have been observed. For example, neurons that fire throughout the duration of the stimulus have been observed in layers V and VI (Volkov and Galazjuk 1991). Such neurons presumably lack synaptic inhibition at the tone frequencies and intensities that evoke spiking. Interestingly, there is evidence to suggest that neurons in layer II also lack long-lasting synaptic inhibition at the tone frequencies and intensities that evoke spiking (Ojima and Murakami 2002). These and other studies (Atzori et al. 2001; Hefti and Smith 2000, 2003; Martinez et al. 2002; Monier et al. 2003) indicate that patterns of tone-evoked synaptic excitation and inhibition other than what we have reported are probably found in AI, particularly outside its major thalamo-recepient layers.
This work was supported by the Hearing Research Institute (M. M. Merzenich and C. E. Schreiner), a Howard Hughes Medical Institute Predoctoral Fellowship (to A. Y. Y. Tan), the John C. and Edward Coleman Fund to C. E. Schreiner and M. M. Merzenich, National Institutes of Health Grants DC-02260 to C. E. Schreiner and NS-10414 to M. M. Merzenich, a National Organization for Hearing Research Foundation research award to L. I. Zhang, and the Sooy Fund to M. M. Merzenich.
We thank C. Atencio, M. Caywood, K. Imaizumi, T. Lauritzen, and K. Miller for discussions.
↵* A.Y.Y. Tan and L. I. Zhang contributed equally to this work.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 by the American Physiological Society