Abstract
The impact of synaptic inhibition depends on the passive and active properties of the neuronal membrane as well as on the characteristics of the underlying synaptic conductances. Here, we evaluated the contributions of these different factors to the IPSPs produced by two kinetically and anatomically distinct inhibitory synapses onto hippocampal CA1 pyramidal neurons: somatic GABAA,fast and dendritic GABAA,slow. Using combined current-clamp and voltage-clamp recordings from neurons in hippocampal brain slices, we found that despite pronounced differences in kinetics and only weak voltage dependence of the underlying synaptic conductances, there were much smaller differences in duration but strong voltage dependence of IPSPs arising from somatic and dendritic synapses. Pharmacologic tests and compartmental modeling showed that these effects were produced by the hyperpolarization-activated cation current, IH, which accelerated IPSP decay over a broad range of membrane potentials and reduced IPSP amplitudes at hyperpolarized potentials, and the persistent sodium current, INaP, which prolonged and amplified IPSPs at depolarized subthreshold potentials. The relative magnitudes of their influences depended on the kinetics of the underlying synaptic conductances: the effect on duration was greater for GABAA,fast and on amplitude was greater for GABAA,slow. Passive and active factors thus influence the impact of synaptic inhibition in a location- and voltage-dependent manner.
- GABAA receptor
- dendrite
- IPSP
- hyperpolarization-activated cation current
- persistent sodium current
- compartmental modeling
Introduction
Studies of synaptic integration in cortical neurons have focused primarily on EPSPs and on how they are shaped by active membrane processes (Lipowsky et al., 1996; Magee, 1999; Hausser et al., 2000). However, inhibitory synaptic transmission also exerts a powerful influence on integrative function. In hippocampal neurons, inhibition regulates excitability in a location-specific manner (Miles et al., 1996; Tsubokawa and Ross, 1996; Tamas et al., 2004; Somogyi and Klausberger, 2005), participates in the generation of theta and gamma oscillations (Soltesz and Deschenes, 1993; Ylinen et al., 1995; Penttonen et al., 1998; Klausberger et al., 2003), and modulates excitatory synaptic plasticity (Davies and Collingridge, 1996; Evans and Viola-McCabe, 1996). Although its importance is recognized, comparatively little is known about the factors that shape IPSPs.
Voltage-clamp studies of inhibition in CA1 pyramidal neurons have shown that synapses located at the soma versus dendrites produce kinetically distinct responses, termed GABAA,fast and GABAA,slow (Pearce, 1993; Banks et al., 1998). Their anatomical and kinetic properties may allow these synapses to serve different functional roles. For example, fast somatic inhibition (τdecay ∼7 ms) is well suited to control action potential generation (Cobb et al., 1995) and facilitate coincidence detection (Pouille and Scanziani, 2001), whereas slow dendritic inhibition (τdecay ∼50 ms) may regulate dendritic electrogenesis, action potential backpropagation, and NMDA receptor activation (Miles et al., 1996; Tsubokawa and Ross, 1996; Kapur et al., 1997).
The physiological impact of these inhibitory synapses may also depend on other active and passive membrane properties. Two particularly relevant active currents in this regard are the hyperpolarization-activated cation current, IH, and the persistent sodium current, INaP (French et al., 1990; Magee, 1998). Both display strong voltage dependence within the appropriate voltage range (Crill, 1996; Pape, 1996). IH is a mixed cation current that in CA1 pyramidal cells is essentially inactive at −57 mV but becomes activated at more hyperpolarized potentials (Spruston and Johnston, 1992; Magee, 1998; Surges et al., 2004). INaP is the noninactivating fraction of the sodium conductance, and its activation is strongly voltage dependent at −57 mV, but it is inactive at −72 mV (French et al., 1990). These currents shape IPSPs and EPSPs in neocortical pyramidal neurons, with IH accelerating IPSP decay and INaP prolonging and amplifying inhibitory potentials in a voltage-dependent manner (van Brederode and Spain, 1995; Stuart, 1999; Williams and Stuart, 2000, 2003). Whether they exert the same influences in CA1 pyramidal neurons and whether this depends on synaptic location or intrinsic kinetics remain unknown.
Here, we address these issues using combined current-clamp and voltage-clamp recordings and compartmental modeling. Although their underlying conductance kinetics are quite distinct and show little voltage dependence, the durations of fast somatic versus slow dendritic IPSPs are relatively less distinct and strongly voltage dependent. IH and INaP produce the voltage dependence by opposing actions: at hyperpolarized potentials, IH accelerates IPSP decay and reduces IPSP amplitude, and at depolarized subthreshold potentials, INaP prolongs and amplifies IPSPs. The effect on duration is greater for GABAA,fast and on amplitude is greater for GABAA,slow. Thus, passive and active factors influence the temporal impact of synaptic inhibition in CA1 pyramidal neurons in a location- and voltage-dependent manner.
Materials and Methods
Preparation of slices.
Male Sprague Dawley rats (21–32 d of age) were anesthetized with 2% isoflurane, decapitated, and the heads were immediately immersed in cold (4°C) saline. The brain was removed, and a midsagittal cut was made to separate the hemispheres. A downward cut was made 2–5 mm from the posterior end of each hemisphere, displaced from the frontal plane ∼15° toward the anterior and ∼35° toward the midline to obtain a transverse hippocampal section. Brain slices 400–500 μm thick were cut using a vibrating microtome (VT 10005, Leica, Bannockburn, IL) and transferred to an incubation chamber, where they were maintained submerged at 32°C in artificial CSF (ACSF) containing the following (in mm): 127 NaCl, 1.21 KH2PO4, 1.87 KCl, 26 NaHCO3, 2.17 CaCl2, 1.44 MgSO4, 10 glucose, saturated with 95% O2/5% CO2, pH 7.4, with 20 μm 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and 40 μm 2-amino-5-phosphonovaleric acid (APV). After 45 min, the temperature was lowered to room temperature (22–24°C), and the slices were maintained under these conditions until they were transferred to a recording chamber perfused with ACSF at 3 ml/min at elevated temperature (32–34°C).
Patch-clamp electrophysiology.
Neurons in the stratum pyramidale (SP) of CA1 were visualized using a video camera (C2400; Hamamatsu, Hamamatsu City, Japan) connected to an upright microscope (BX-50WI; Olympus America, Melville, NY) equipped with an infrared bandpass filter (775 ± 75 nm), a long working-distance water-immersion objective (40×; numerical aperture, 0.8), and differential interference contrast optics. The microscope and recording pipette were positioned using an integrated motorized control system (Luigs and Neumann, Ratingen, Germany).
Whole-cell recordings were obtained using a Multiclamp 700A (Molecular Devices, Union City, CA) patch-clamp amplifier. All data were collected using pClamp 8.0 software (Molecular Devices). Data were filtered at 4 kHz, sampled at 10 kHz (Digidata 1200; Molecular Devices), and stored on a Pentium-based computer. Recording pipettes were fabricated from borosilicate glass (KG-33; 1.7 mm outer diameter, 1.1 mm inner diameter; Garner Glass, Claremont, CA) using a Flaming-Brown two-stage puller (P-87; Sutter Instruments, Novato, CA). Tight-seal whole-cell recordings were obtained using standard techniques (Hamill et al., 1981; Sakmann et al., 1989). Recording pipettes had open-tip resistances of 2–4 MΩ when filled with the recording solution containing the following (in mm): 140 K-gluconate, 10 HEPES, 5 NaCl, 4 MgATP, 10 phosphocreatine, 0.3 GTP, and 5 EGTA, pH 7.2, and 305 mOsm. Access resistances were typically 10–20 MΩ and were then compensated 60–80%. Evoked GABAA IPSCs and IPSPs were isolated by bath application of 20 μm CNQX and 40 μm APV to block AMPA- and NMDA-mediated currents and by bath application of 2.5 μm (2S)-3-([(1s)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl)(phenylmethyl)phosphinic acid (CGP 55845A) to block GABAB-mediated currents. Drugs were allowed >5 min to wash in before data collection. For voltage-clamp experiments, cells were held at −77 mV and stepped to different membrane potentials to characterize evoked IPSC properties. Blockade of IH by 4-(N-ethyl-N-phenylamino)-1,2-dimethyl-6-(methylamino)pyridinum chloride (ZD 7288) was confirmed by monitoring responses to hyperpolarizing current pulses (0.2–0.5 nA). These current pulses resulted in depolarizing sags characteristic of the IH conductance, with an onset of 40–60 ms and a peak to steady state voltage difference of ∼5 mV. A concentration of 20 μm ZD 7288 abolished this sag in all cells tested (n = 5).
Synaptic responses were evoked using bipolar tungsten electrodes and constant-current stimulus isolation units (A-365; World Precision Instruments, Sarasota, FL) with a maximum stimulation rate of 0.08 Hz. GABAA,fast was evoked by stimulating (20–100 μA) SP ∼50 μm from the recording site. GABAA,slow was evoked by stimulating (50–300 μA) at the border between stratum lacunosum-moleculare (SL-M) and stratum radiatum (SR), ∼100 μm deep in the tissue and displaced 50–100 μm mediolaterally from the apical dendrite of the cell being studied. In some cells, this required that stimulating electrodes be repositioned so that isolated GABAA,fast and GABAA,slow responses could be elicited. For investigations of evoked responses, IPSCs were first characterized in voltage-clamp mode, and the amplifier was then switched to current-clamp mode. IPSPs were then evoked using the same stimulus intensity and electrode positions used during the IPSC characterization. “Injected IPSPs” were obtained by passing current waveforms generated from the α function Iinj = Imax × s × [exp(−t/τdecay) − exp(−t/τrise)], where Iinj is the current passed through the recording electrode, Imax is the maximal current, and s is a scaling factor so that Iinj is equal to Imax at its peak. The time constants for the fast and slow waveforms were, respectively, τrise = 0.2 and 3 ms, and τdecay = 6 and 40 ms.
CGP 55845A was a kind gift from Dr. Froestl (CIBA-Giegy, Basel, Switzerland). ZD 7288 was obtained from Tocris Cookson (Bristol, UK). These drugs were prepared directly from their powdered forms and used within 3 d. All other drugs were obtained from Sigma-Aldrich Chemicals (St. Louis, MO) and prepared as stock solutions, frozen, and diluted in ACSF the day of the experiment.
Data analysis.
Data were analyzed on a Pentium-based personal computer using ClampFit (Molecular Devices), Origin (OriginLab, Northampton, MA), and MatLab (MathWorks, Natick, MA). In a minority of cases (<30%), GABAA,slow was fit well with a single decaying exponential component. The remaining events were fit with the sum of two decaying exponential components, in which case the decay was characterized by the weighted time constant τdecay, wt = (A1τ1 + A2τ2)/(A1 + A2), where Ai and τi are the amplitude and time constant of decay, respectively, of the ith component. GABAA,slow responses were used if their IPSCs displayed a 10–90% rise time >3 ms and had a monoexponential or weighted τdecay > 40 ms. GABAA,fast IPSCs decayed biexponentially (Banks et al., 1998) and were characterized as τdecay, wt. IPSP duration is the time from IPSP peak to 1/e of peak amplitude. Statistical comparisons were made using Student’s t tests or two-way ANOVA with Tukey’s test for post hoc means comparisons. Differences were considered significant for α = 0.05. All data are presented as mean ± SD.
Modeling.
Simulations were performed with the compartmental modeling program NEURON (Hines and Carnevale, 1997). The compartmental model was adapted from Poirazi et al. (2003) and run on a personal computer with a Microsoft Windows operating system (Microsoft, Seattle, WA). The cell was a reconstructed CA1 pyramidal cell and included several active membrane mechanisms, with biophysical parameters and spatial distributions based on experimental studies. For most simulations Rm = 45,000 Ω·cm2; however, simulations were also performed using Rm = 65,000 Ω·cm2, and these yield similar results to those reported here. The reversal potential of the leak conductance was set constant at −75 mV, except for simulations in which the membrane potential of the cell was varied (see Fig. 7), in which case the current–balance function was used to set the reversal potential of the leak conductance to obtain the desired membrane voltage. The model also included Hodgkin–Huxley-type sodium and potassium currents, with gNaT,max = 0 to simulate sodium channel inactivation at depolarized potentials and resting Vm of −57 mV without tonic firing. The model included an M-type potassium current, a hyperpolarization-activated current (IH) with an increasing density at more distal sites along the apical dendrite (Magee, 1998), T- and L-type calcium currents, and slow and medium afterhyperpolarization currents. A-type potassium and R-type calcium currents were present in the original model; however, these currents imparted a dramatic voltage dependence to the input resistance in the absence of IH and INaP that was not observed experimentally, so these were omitted for the simulations shown here.
GABAA,fast and GABAA,slow were modeled with the Exp2Syn point process, of the form Is = s × gs[exp(−t/tdecay) − exp(−t/trise)](Vm − Erev), where s is the scale factor, gs is the maximal conductance (5 nS for both GABAA,fast and GABAA,slow), trise was 0.2 and 3 for GABAA,fast and GABAA,slow, respectively, and tdecay was 10 and 40 ms for GABAA,fast and GABAA,slow, respectively, Vm is the membrane potential, and the reversal potential of the synapse, Erev, was −75 mV. GABAA,fast was a single point process at the soma, GABAA,slow was distributed evenly over the entire dendritic tree, with a total conductance summed over all synapses equal to 5 nS.
Experimental effects of ZD 7288 were simulated by removing the IH conductance. For simulations in which IH activation was set constant at steady state, the activation was calculated using the Boltzmann equation with Vm set constant at the baseline membrane potential.
To simulate a persistent sodium conductance, a new mechanism was created, based on experimental parameters determined in entorhinal stellate cells (Magistretti and Alonso, 1999), the activation properties of which are similar to those described in CA1 cells (French et al., 1990). Because inactivation properties of INaP in CA1 pyramidal cells have not been reported, the inactivation properties determined from stellate cells were used. The INaP in stellate cells display an inactivation time constant of several seconds, which is sufficiently large compared with the duration of IPSPs (25–100 ms) to model the inactivation as time independent. The level of inactivation was calculated based on the baseline membrane voltage preceding the simulated IPSPs and was then held constant throughout the remainder of the simulation. INaP activation is very fast, complete within several milliseconds (French et al., 1990; Crill, 1996), so it was modeled as an instantaneous change in conductance.
The persistent sodium membrane mechanism was of the form gNaP = m × h × gNaP,max. Activation gate m = 1/(1 + exp[(V1/2 − Vm)/k]), and inactivation gate h = 1/(1 + exp[−(V1/2 − Vm)/k]), with the following parameters: activation, V1/2 = −52.6 mV, k = −4.6 mV; inactivation, V1/2 = −48.8 mV, k = 10 mV. The maximal conductance, gNaP,max, was set at 10 μS/cm2, which is within the range of maximal conductances observed physiologically (French et al., 1990; Crill, 1996; Hammarstrom and Gage, 1998). If IA was not present, then gNaP,max = 10 μS/cm2 was the maximum conductance that did not result in regenerative sustained depolarizations as Vm approached −57 mV. Including IA allowed gNaP,max to be increased threefold, which is still in the physiological range, and simulations with this conductance level produced results similar to those reported here, excepting a decrease in steady state RN after depolarization in the absence of INaP and IH, which was not observed experimentally with simultaneous TTX and ZD application. INaP was distributed evenly throughout the somatic and dendritic compartments, because evidence indicates the presence of INaP in CA1 dendrites (French et al., 1990; Lipowsky et al., 1996).
Results
Characterization of evoked IPSCs versus IPSPs
Using standard whole-cell patch clamp recording techniques, we studied monosynaptic inhibitory responses in hippocampal CA1 pyramidal neurons. With the amplifier set to voltage-clamp mode, evoked IPSCs (eIPSCs) were obtained at membrane potentials ranging from −115 to −35 mV (Fig. 1Ai,Bi). As reported previously, GABAA,fast synaptic currents produced by stimulating SP near the somatic recording site decayed rapidly, with weighted time constants of ∼10 ms, whereas GABAA,slow currents evoked by stimulating SL-M decayed more slowly, with weighted time constants between ∼35 and 50 ms. For voltage-clamped eIPSCs, GABAA,fast and GABAA,slow τdecay were weakly voltage dependent within the voltage range of −115 to −30 mV, as determined by linear regression (Fig. 1Ci) (p < 0.05). GABAA,fast τdecay increased by e-fold for every 159 mV depolarization, and GABAA,slow increased by e-fold for every 138 mV depolarization, consistent with previous studies which found e-fold increases for 130–200 mV depolarizations for GABAA IPSCs in hippocampal and neocortical neurons (Collingridge et al., 1984; Otis and Mody, 1992; Salin and Prince, 1996).
Characterization of evoked GABAA,fast and GABAA,slow IPSCs and corresponding IPSPs at different voltages. GABAA,fast (Ai) and GABAA,slow (Bi) IPSCs were evoked in voltage-clamp mode. The recording mode was then switched to current-clamp mode, and the corresponding IPSPs (Aii, Bii) were evoked using the same stimuli. Ci, IPSC weighted τdecay at different command voltages for several cells. GABAA,fast (filled symbols) and GABAA,slow (open symbols) IPSC τdecay,wt was voltage dependent at membrane potentials from −115 to −30 mV (p < 0.05, linear regression; depolarization required for e-fold increase of IPSC τdecay: GABAA,fast, 159 mV; GABAA,slow, 138 mV). Cii, IPSP duration at a range of membrane potentials. GABAA,fast and GABAA,slow IPSP duration was voltage dependent (p < 0.01, linear regression; depolarization required for e-fold increase of IPSP duration: GABAA,fast, 25.2 mV; GABAA,slow, 20.0 mV). n = 7 for GABAA,fast and n = 5 for GABAA,slow. Calibration: Ai, Bi, 100 pA, 50 ms; Aii, Bii, 2 mV, 50 ms. SP, Stratum pyramidale; SLM, stratum lacunosum-moleculare.
To examine the properties of the corresponding evoked IPSPs (eIPSPs), we switched the recording mode of the amplifier to current-clamp and elicited monosynaptic responses using the same stimuli (Fig. 1Aii,Bii). Membrane potential was controlled by manually adjusting direct current injection. As expected, SL-M-evoked responses lasted longer than SP-evoked responses (Fig. 1C). However, two other interesting differences were observed between responses in voltage-clamp and current-clamp mode. First, the relative differences in the durations of eIPSPs (Fig. 1Cii) were not nearly as pronounced as the differences in the corresponding eIPSCs (Fig. 1Ci). Second, the degree of voltage dependence of eIPSP durations were sixfold to sevenfold greater than for τdecay of eIPSCs (Fig. 1Cii) (depolarization required for e-fold increase in IPSP duration: GABAA,fast, 25.2 mV; GABAA,slow, 20.0 mV; p < 0.01 for both, linear regression).
Effects of membrane potential on IPSP duration
The substantially greater voltage dependence of IPSPs than IPSCs suggests that in addition to a weak voltage dependence of the underlying synaptic conductance kinetics, other active processes also influence IPSP duration and, in fact, may be primarily responsible for prolonging IPSPs at depolarized potentials. To examine the underlying mechanisms within a physiologically relevant voltage range, in additional experiments, we quantified IPSP properties at membrane potentials of −72 and −57 mV. These two voltages were chosen because −72 mV is near the resting membrane potential, and −57 mV was found to be the most depolarized membrane potential that was consistently still subthreshold. Because ECl = −79 ± 6 mV (determined from GABAA,fast and GABAA,slow IPSC amplitude–voltage relationships; n = 6), eIPSPs were hyperpolarizing at both voltages. Examples of evoked IPSCs and IPSPs at −72 mV are shown in Figure 2, Ai and Bi, and average characteristics at the two voltages are presented in Figure 2Ci. As expected, both GABAA,fast and GABAA,slow eIPSP durations at −57 mV were significantly greater than those at −72 mV (−57:−72 mV duration ratios, GABAA,fast = 1.8 ± 0.2, n = 7; GABAA,slow = 1.6 ± 0.3, n = 8; p < 0.05, ANOVA). Also, the duration of the GABAA,slow eIPSP was significantly greater than that of GABAA,fast (GABAA,slow:GABAA,fast eIPSP duration ratios, −72 mV = 1.9 ± 0.3; −57 mV = 1.6 ± 0.2), but the difference was not as great as for the corresponding eIPSCs (Fig. 2Ci) (GABAA,slow:GABAA,fast eIPSC ratios for duration at 1/e amplitude, 4.4 ± 1.8 at −72 mV, 4.3 ± 0.5 at −57 mV; n = 4–8; p < 0.05, ANOVA). Thus, the characteristic differences observed over a much greater voltage range (Fig. 1) are also present over this more restricted subthreshold range of membrane potentials.
Comparison of evoked and artificial IPSPs. Evoked GABAA,fast and GABAA,slow IPSCs and IPSPs were characterized in voltage-clamp (Ai) and current-clamp (Aii) modes. Bi, Injected somatic current waveforms with time courses that mimicked GABAA,fast and GABAA,slow IPSCs. Bii, Artificial IPSPs generated with the GABAA,fast and GABAA,slow IPSC waveforms. Ci, Summary graph of evoked IPSC durations and the resulting IPSP durations at two voltages. GABAA,slow IPSP duration was significantly greater than GABAA,fast duration at both −72 and −57 mV (**p < 0.01, Student’s t test), but the relative difference was not as great as that observed for corresponding IPSCs. The durations of both GABAA,fast and GABAA,slow IPSPs increased after depolarization (*p < 0.05, Student’s t test). Cii, Artificial GABAA,fast and GABAA,slow IPSP durations differed significantly (**p < 0.01, Student’s t test). In each case, IPSP duration increased with depolarization from −72 mV to −57 mV (*p < 0.05, Student’s t test). Evoked IPSCs at −57 mV: GABAA,fast, n = 7 cells; GABAA,slow, n = 8 cells. Evoked IPSCs at −72 mV: GABAA,fast, n = 5; evoked GABAA,slow, n = 4 cells. Injected IPSPs: n = 5 cells. Traces shown were normalized averages of three to five traces. Error bars indicate SD.
To determine the degree to which active conductances contribute to the voltage dependence of IPSP duration and to examine the underlying mechanisms, we turned to the study of “artificial IPSPs” (aIPSPs). For these experiments, we injected current waveforms through the somatic recording pipette that matched the characteristics of GABAA,fast and GABAA,slow IPSCs (Fig. 2Aii) and examined the resulting voltage transients. As expected, aIPSP amplitudes depended strongly on the kinetics of the injected currents, so that to produce comparable IPSP amplitudes at −72 mV required that the peak amplitude of the GABAA,fast injected waveform be 2.5-fold greater than the peak of the GABAA,slow injected waveform (GABAA,fast, 250 pA peak current injection resulting in 2.8 ± 0.4 mV IPSP amplitude; GABAA,slow, 100 pA peak current injection resulting in −3.3 ± 0.6 mV IPSP amplitude). Using these stimuli, we found that both the fast (τdecay = 6 ms) and slow (τdecay = 40 ms) aIPSPs did exhibit voltage-dependent increases in duration (Fig. 2Bii,Cii) (−57:−72 mV aIPSP duration ratios: GABAA,fast = 2.0 ± 0.4, n = 5; GABAA,slow = 1.5 ± 0.2, n = 5) that were similar to the voltage dependence of evoked IPSPs (Fig. 2Bi,Ci). Also, as for eIPSPs, GABAA,slow aIPSPs were slower than GABAA,fast aIPSPs at both voltages (GABAA,slow:GABAA,fast aIPSP duration ratios: −72 mV = 2.3 ± 0.2; −57 mV = 1.7 ± 0.2, n = 5; p < 0.05; Student’s t test). These results thus confirm the presence of voltage-dependent mechanisms that alter IPSP duration in CA1 pyramidal neurons, operating at both fast and slow time scales. Furthermore, the similarity between the voltage dependence of evoked IPSPs (Fig. 2Ci) and artificial IPSPs (Fig. 2Cii) indicates that the effects of voltage on IPSP duration are primarily caused by voltage-dependent currents rather than an intrinsic voltage sensitivity of receptor kinetics.
Mechanism of voltage dependence of IPSP duration
We considered the identities of the active conductances that are responsible for the voltage dependence of IPSPs, focusing primarily on IH and INaP. To assess their possible contributions to GABAA,fast and GABAA,slow duration and amplitude, we tested the effects of pharmacological blockade of IH by 20 μm ZD 7288 and blockade of INaP by 1 μm TTX, using identical current waveforms at the two voltages (Fig. 3). We used aIPSPs for these studies, because a complete block of INaP by a high concentration of TTX (1 μm) would abolish evoked responses. Also, using aIPSPs allowed us to compare the relative effects of IH and INaP on IPSP amplitudes at different voltages without the confounding influence of a change in driving force.
Effects of IH and INaP on duration and amplitude of artificial IPSPs. A, GABAA,fast aIPSPs were produced by injecting a current waveform matching the time course of the voltage-clamped IPSC, at −57 mV (gray traces) or −72 mV (black traces). Note the larger response amplitude at −57 mV. Application of ZD 7288 (20 μm) blocked the depolarizing voltage sag during hyperpolarizing current injection (data not shown) and prolonged the IPSP, but IPSP duration remained voltage dependent. Subsequent addition of TTX (1 μm) accelerated the IPSP primarily at −57 mV and reduced the voltage sensitivity of duration or amplitude. Bottom, Time course of pharmacological actions, at −57 mV (open circles) and −72 mV (filled circles). B, The same experiments as in A, except that TTX was applied first followed by ZD 7288. TTX alone greatly diminished the voltage dependence, and addition of ZD 7288 prolonged the duration. Scale bars: vertical, 2 mV; horizontal, 50 ms.
The membrane potential was maintained at approximately −72 or −57 mV by manually adjusting the magnitude of injected tonic current, and aIPSPs were generated by injecting current waveforms that mimicked the GABAA,fast and GABAA,slow IPSCs (as illustrated in Fig. 2). IPSP durations at the two voltages were monitored continuously during control periods and after sequential application of ZD 7288 followed by TTX (Fig. 3A), or TTX followed by ZD 7288 (Fig. 3B). As expected, ZD 7288 greatly reduced the voltage-dependent “sag” following hyperpolarizing current injection (n = 5; data not shown). When ZD 7288 was applied in the absence of TTX, it also slowed GABAA,fast aIPSP decay, but aIPSP duration remained voltage dependent (Fig. 3A). Subsequent addition of TTX eliminated the voltage dependence by reducing duration to a greater extent at −57 mV than at −72 mV (Fig. 3A). Similarly, application of TTX alone eliminated the voltage dependence observed in control conditions, and subsequent addition of ZD 7288 prolonged aIPSP duration at both voltages (Fig. 3B).
Figure 4 summarizes the pharmacological effects on voltage dependence of duration (Fig. 4A) and amplitude (Fig. 4B). For both GABAA,fast (τdecay 6 ms) and GABAA,slow (τdecay 40 ms) waveforms, a strong voltage dependence of duration was present under control conditions and in the presence of ZD 7288 (−57:−72 mV duration ratios >1; p < 0.01; Student’s t test) but absent in TTX (p > 0.05; two-way ANOVA). Qualitatively similar results were obtained for pharmacologic sensitivity of aIPSP amplitude, but in this case, ZD 7288 alone did produce a significant effect (p < 0.01; two-way ANOVA). GABAA,fast aIPSP duration was more strongly voltage dependent than GABAA,slow, whereas GABAA,slow aIPSP amplitude showed a greater voltage dependence than GABAA,fast (p < 0.01; two-way ANOVA). Therefore, intrinsic voltage-dependent conductances increased aIPSP durations and amplitudes, with the fast waveform more strongly augmented in duration and the slow waveform more strongly augmented in amplitude.
Summary of the effects of IH and INaP on voltage dependence of artificial IPSPs. A, ZD 7288 alone did not significantly alter the voltage sensitivity of duration, but TTX decreased the voltage sensitivity both in the presence and absence of ZD 7288. Effects of voltage on duration were significantly greater for the fast injected waveform (p < 0.01, two-way ANOVA for all comparisons). B, Under control conditions, amplitudes of artificial IPSPs were greater at −57 mV than at −72 mV. Differences were significantly reduced by ZD 7288 and also by TTX. The effects of voltage on amplitude were significantly greater for the slow injected waveform (p < 0.01, two-way ANOVA for all comparisons). Error bars indicate SD.
To confirm that the above findings derived from studies of aIPSPs also applied to evoked IPSPs, the same experiments illustrated in Figure 3A were conducted using GABAA,fast and GABAA,slow eIPSPs, evoked at 0.05–0.1 Hz by electrical stimulation. So that synaptic responses could still be obtained, a lower concentration of TTX (100 nm) that would permit action potential conduction but still reduce persistent sodium current was used. As observed for aIPSPs, addition of ZD 7288 prolonged eIPSP durations but did not abolish the voltage dependence (Fig. 5A). Subsequent application of TTX shortened IPSP durations and greatly reduced their voltage dependence (Fig. 5B). Figure 5C summarizes these effects on duration of GABAA,fast (n = 5) and GABAA,slow (n = 4) eIPSPs.
Effects of IH and INaP on duration of evoked IPSPs. A, ZD 7288 prolonged IPSP duration, but IPSPs remained voltage dependent. B, In a separate experiment, addition of a low concentration of TTX (100 nm) to saline containing ZD 7288 accelerated IPSP decay at −57 mV and greatly decreased the voltage sensitivity. Traces are normalized averages of three to five responses. C, Summary of effects of ZD 7288 and TTX on the voltage dependence of IPSP duration. ZD 7288 alone did not significantly decrease the voltage dependence (two-way ANOVA), but ZD 7288 plus TTX significantly reduced the voltage dependence (**p < 0.01, two-way ANOVA). As for aIPSPs, the duration of GABAA,fast was significantly more voltage dependent than the duration of GABAA,slow (**p < 0.01, two-way ANOVA). Error bars indicate SD.
We attempted to also confirm the effects of ZD 7288 and TTX on voltage dependence of eIPSP amplitudes by normalizing amplitudes at the two membrane voltages to the magnitude of the driving force of the GABAA synaptic currents. However, a substantially greater variability in amplitudes than durations, possibly resulting from intrinsic variation in the number of terminals releasing transmitter with each stimulus, meant that many more evoked responses must be averaged than we were able while maintaining sufficiently stable recording conditions, and this precluded the successful completion of these experiments. Thus, the effects of IH and INaP on IPSP amplitudes reported here are confined to studies using aIPSPs.
Table 1 summarizes the pharmacological effects on duration and amplitude at −72 and −57 mV. For both eIPSPs and aIPSPs, ZD 7288 prolonged IPSPs to a greater extent at −72 mV than at −57 mV, whereas TTX accelerated IPSPs more strongly at −57 mV. For aIPSPs, the increase in duration produced by ZD 7288 was significantly greater at −72 mV than at −57 mV when TTX was present (p < 0.05; two-way ANOVA) but not when TTX was absent (two-way ANOVA). The most consistent effect of TTX on duration was a decrease at −57 mV, and this reduction was larger at −57 mV than at −72 mV whether ZD was present or absent (p < 0.05; ANOVA). The only consistent effect on aIPSP amplitude was a decrease with TTX application. Thus, ZD 7288 prolonged IPSP duration more robustly at −72 mV, and TTX decreased both duration and amplitude more robustly at −57 mV.
Effects of TTX and ZD 7288 on IPSP duration and amplitude
NEURON modeling: mechanisms of IH and INaP effects
Because a significant fraction of IH conductance is active at the membrane voltages studied here, it is possible that IH introduces a sufficiently large leak conductance that it influences IPSP duration and amplitude by decreasing input resistance and the membrane time constant. The relative influence of IH via effects on passive cellular properties resulting from constitutive activation at −72 mV versus the dynamic shaping of voltage transients is unknown. Because we were unable to separate the relative contributions of active versus passive effects experimentally, we turned to a computational model to address this question.
We used a publicly available NEURON model of a reconstructed CA1 pyramidal cell (Poirazi et al., 2003), which included IH and supplemented this by an additional INaP mechanism (see Materials and Methods for details). This model reproduced the modulation of IPSP duration by both IH at −72 mV (Fig. 6A) (simulated GABAA,fast IPSP durations at −72 mV: +IH, 34.8 ms; −IH, 44.9 ms) and INaP at −57 mV (Fig. 6A,C) (simulated GABAA,fast IPSP durations at −57 mV: +INaP, 88.5 ms; −INaP, 54.9 ms). We used this model to investigate the relative contributions of active shaping versus changes in passive resistance for IH and the relative contributions of IH and INaP over a range of voltages.
Dynamic shaping of IPSPs by IH. A, Using a compartmental model implemented in the NEURON simulation environment, IPSPs produced by introduction of a transient conductance waveform with characteristics of GABAA,fast were compared in the absence of IH (gray trace), in the presence of a time- and voltage-dependent “dynamic” IH (black trace), and in the presence of an added conductance that lacked the time- and voltage-dependent activation and inactivation of IH but had a constant level of IH activation calculated from the baseline Vm (IH,SS; dotted trace). Introducing IH,SS had little impact on IPSP duration, demonstrating that the effects of IH on duration were caused by dynamic properties of the conductance. B, Current densities of the dynamic IH conductance throughout the time course of the IPSP at the soma and in the distal dendrite. C, Total charge flowing through the GABAA conductance (left), compared with the IPSP-induced charge flowing through a dynamic IH (middle) and time-independent IH,SS (right). More than 98% of the current flowing through the dynamic IH was located in the dendrites.
IH: active versus passive shaping of IPSPs
At −72 mV, the additional steady state tonic IH conductance may itself have changed the membrane time constant. To test this, the time dependence of activation was removed, without affecting other properties of IH, such as density, distribution, and reversal potential. This was done by setting the IH conductance constant, based on the activation Boltzmann function, using the baseline membrane voltage of −72 mV. This conductance will hereafter be referred to as IH,SS, to denote that the IH conductance activation was held fixed at a given steady state level. Introducing the time-independent IH,SS had little effect on simulated somatic GABAA,fast IPSP duration compared with normal, time-dependent IH (Fig. 6A). Therefore, most of the effects of IH on IPSP duration were caused by its dynamic time dependence.
The total change in current density flowing through the dynamic IH conductance induced by the IPSP was greater than that of IH,SS and peaked later during the time course of the IPSP (Fig. 6B) (at the soma, IH peaks 33 ms after IPSP onset, IH,SS peaks 10 ms after IPSP onset). Furthermore, the amount of current flowing through the IH conductance changed substantially during the time course of the IPSP in the distal dendrite (Fig. 6B), indicating that dendritic IH may have contributed to the shaping of a somatically generated IPSP.
To confirm that the IH currents induced by the IPSP were substantial enough to impact IPSP duration, the total membrane charge transfer over the entire cell was measured for the IH current. This was determined with a membrane mechanism that integrated the IH current over time in every compartment. The net charge transfer through the IH conductance induced by the IPSP was determined by comparing identical simulations with and without the IPSP. The current flowing through the GABAA conductance was also measured, and the total charge was calculated by integrating this current over time. The charge through the standard dynamic IH conductance induced by the IPSP was 44% of the charge flowing through the GABAA conductance, whereas the IPSP-induced charge through the time-independent IH,SS was only 10% (Fig. 6C) (GABAA charge, 0.54 pC; IH charge, −0.24 pC; IH,ss charge, −0.05 pC). Thus, for IH, a majority of the IPSP-induced charge was a result of dynamic changes in IH activation during the IPSP. Furthermore, 97% of this charge was attributable to dendritic IH conductances. Therefore, in the model, the shaping of the duration of the somatically generated IPSP was dominated by current influx through dynamically activated dendritic channels. Thus, we conclude that the experimentally observed effects of IH on IPSP duration are mostly explained by dynamic shaping of voltage trajectories and not by changes in the passive properties of the cell.
Relative and cooperative effects of IH and INaP
Although IH and INaP impart the same type of voltage sensitivity to IPSPs, they do so by complementary actions: IH shortens IPSPs at more hyperpolarized potentials, whereas INaP prolongs IPSPs at more depolarized potentials. To investigate the relative effects of these two conductances as a function of membrane voltage and to test whether the previously characterized biophysical properties of IH and INaP can explain our experimentally observed voltage dependences of IPSP duration and amplitude, we simulated fast and slow synaptic currents at somatic and dendritic locations and measured the durations and amplitudes of the resulting IPSPs as recorded at the soma.
We used a 5 nS maximal conductance for both types of synapses. This conductance was modeled as a single point process at the soma for GABAA,fast and divided throughout the apical compartments as multiple point processes for GABAA,slow. The resting potential in all compartments was varied between −73 and −55 mV by adjusting the passive leak reversal potential. Analysis of GABAA,fast and GABAA,slow IPSPs revealed a voltage dependence of duration (Fig. 7A) and amplitude (Fig. 7B) in the presence of IH and INaP (Fig. 7, filled symbols). Removal of both IH and INaP greatly reduced and, in fact, reversed this voltage dependence (Fig. 7, open symbols). The pattern of responses thus mimics our experimental observation that there is a greater relative increase in duration than amplitude for GABAA,fast, whereas there is a greater increase in amplitude than duration for GABAA,slow (Fig. 4).
Effects of IH and INaP on simulated GABAA,fast and GABAA,slow IPSP duration and amplitude. A, For both GABAA,fast and GABAA,slow IPSPs, duration was strongly voltage dependent under control conditions (filled symbols). Voltage dependence was reduced and reversed after removal of both IH and INaP (open symbols). B, IPSP amplitude, normalized for driving force, was more strongly voltage dependent for GABAA,slow than GABAA,fast under control conditions (filled symbols). Voltage dependence was reduced and reversed after removal of IH and INaP (open symbols). C, D, Characterization of relative effects of IH and INaP on duration (C) and amplitude (D), expressed as the fractional change induced by removal of each conductance individually. Removal of IH altered duration at depolarized as well as hyperpolarized potentials, suggesting interactions of the IH and INaP currents (arrow). The location of GABAA,slow in the dendrites (solid black trace, compare with conductance with slow kinetics located at the soma, dotted trace) influenced the impact of IH on voltage dependence of duration (C) and of both INaP and IH on voltage dependence of amplitude (D). E, In the absence of INaP, removal of IH slowed IPSP decay more strongly at hyperpolarized potentials (compare with C, −IH). F, In the absence of IH, removal of INaP accelerated IPSP decay more strongly at depolarized potentials. Note that this effect is slightly greater than in the presence of IH (compare with C, −INaP).
The above simulations demonstrate that, together, IH and INaP impart a voltage sensitivity that is similar to that observed experimentally (Fig. 2Ci,Cii). We separated and assessed the individual effects of IH and INaP by simulating IPSPs in the presence and absence of each of these currents. We expressed the effects as the change in duration and amplitude after the removal of either IH or INaP (Fig. 7C,D). Removal of IH increased both GABAA,slow and GABAA,fast durations (Fig. 7C) but had a much greater impact on GABAA,slow than GABAA,fast at hyperpolarized potentials. In contrast, removal of INaP reduced IPSP durations, and the effects were strongest at depolarized potentials and greater for GABAA,fast than GABAA,slow. These results are consistent with the experimental observation that INaP is mostly responsible for imparting the voltage sensitivity to IPSP duration, and that the relative effects are greater for GABAA,fast than GABAA,slow (Figs. 4A, 5).
Surprisingly, for both GABAA,fast and GABAA,slow the effect of IH on duration was almost as large at some depolarized potentials as it was at hyperpolarized potentials (Fig. 7C, arrow). This effect was caused by an interaction between IH and INaP, because in a model lacking INaP, the removal of IH resulted in a monotonic pattern of changes, with greater effects at hyperpolarized potentials (Fig. 7E). Thus, these two conductances had synergistic interactions that allowed IH to influence IPSP duration similarly throughout the entire voltage range. This interaction may account for our experimental observation that application of ZD 7288 slowed IPSPs but did not strongly affect their voltage sensitivity (Figs. 3, 4A, 5).
In contrast to its effects on duration, INaP more strongly affected the amplitude of GABAA,slow than GABAA,fast at all voltages (Fig. 7D), and its effects were greater at more depolarized potentials. Similar to its effect on duration, IH more strongly affected the amplitude of GABAA,slow than GABAA,fast. Thus, the individual effects of both IH and INaP on amplitudes were greater for GABAA,slow than GABAA,fast.
Influence of location versus kinetics of GABAA,slow
To determine whether the dendritic location of GABAA,slow contributed to the differences in voltage-dependent effects of IH and INaP on GABAA,slow versus GABAA,fast IPSPs, we implemented a somatic conductance with GABAA,slow kinetics. In most cases, the effects of INaP and IH on somatic GABAA,slow amplitude and duration were intermediate to the effects of somatic GABAA,fast and dendritic GABAA,slow (Fig. 7C,D). This indicates that both location and underlying kinetics contributed to the differences between effects on GABAA,fast and GABAA,slow. The one exception to this finding was the effect of INaP on duration, in which case synaptic location did not have an influence. This finding is reminiscent of our experimental observation that evoked (dendritic) GABAA,slow and somatically injected GABAA,slow waveforms displayed similar voltage dependencies (Fig. 5A,B). Again, this result suggests that effects of voltage on IPSP duration are primarily mediated by INaP. In the model, INaP exerted a greater effect on duration at depolarized potentials in the absence of IH than in its presence (Fig. 7, compare F and C), indicating that IH is to some extent able to counteract the effects of INaP. As in the presence of IH, removal of INaP in the absence of IH (Fig. 7F) had a markedly greater effect on duration at depolarized than hyperpolarized potentials and thereby imparted a pronounced voltage dependence.
The model predicts that in the absence of TTX, IH should have a relatively modest effect on the voltage dependence of duration (duration at −57/duration at −72 mV decreased from 1.0–0.84 after removal of IH for GABAA,fast, and from 1.3–0.88 for GABAA,slow). For cells in which ZD 7288 was added in the presence of TTX, we did observe significant decreases in voltage dependence of duration for aIPSPs (duration at −57/duration at −72 mV decreased from 1.00 ± 0.07 to 0.83 ± 0.07 with ZD application for GABAA,fast, p < 0.01, paired t test; and from 1.04 ± 0.07 to 0.85 ± 0.08 for GABAA,slow, p < 0.01, paired t test). Application of ZD 7288 in the absence of TTX did not result in a detectable change of voltage dependence (p > 0.05; paired t test), consistent with the “boosting” of IH by INaP at depolarized potentials (Fig. 7C). Thus, the experimentally measured changes in voltage dependence produced by ZD 7288 and TTX support the findings of the model regarding the relative and cooperative effects of IH and INaP.
Discussion
Past studies using voltage-clamp recordings to compare the physiological properties of fast somatic versus slow dendritic inhibitory synapses revealed substantial differences in conductance kinetics, with somatic IPSCs approximately fourfold to sevenfold faster than dendritic IPSCs depending on the experimental temperature (GABAA,fast has a higher Q10 than GABAA,slow) (Banks et al., 1998). Our present results demonstrate that passive and active properties strongly influence the resulting synaptic potentials, serving to reduce the dissimilarity in their duration and to prolong IPSPs at depolarized voltages.
Passive properties minimize differences in IPSP duration
We showed previously that the difference in duration of GABAA,fast versus GABAA,slow IPSCs is not simply the result of space clamp artifact, because there are equivalent differences in the kinetics of the underlying synaptic conductances (Pearce, 1993). The smaller difference under current-clamp conditions, ∼1.5-fold to twofold, is primarily a consequence of the slow (τ ∼20–30 ms) membrane time constant of CA1 pyramidal cells (Spruston et al., 1992). Because τdecay of GABAA,fast (∼6 ms) is faster than the membrane time constant, it is prolonged to a greater fractional extent than GABAA,slow (τdecay ∼40 ms).
Active properties produce voltage-dependent IPSP duration
The decay rate of inhibitory currents was only weakly voltage dependent (Fig. 1) and approximately the same for GABAA,fast and GABAA,slow IPSCs (e-fold change over 138 and 159 mV), similar to previous reports (Collingridge et al., 1984; Otis and Mody, 1992; Jones and Harrison, 1993; Salin and Prince, 1996). In contrast, IPSP duration was strongly voltage dependent (e-fold change over 25 and 20 mV). Whereas passive properties reduced differences in IPSP durations, active conductances produced the voltage dependence. The effect was not small: its magnitude was equivalent to changes brought about by threefold to sixfold changes in the decay of synaptic conductances (Fig. 2).
The combined influences of INaP and IH produced slower and larger IPSPs at more depolarized potentials. These results are similar to the findings that IPSPs in neocortical neurons are prolonged by INaP after depolarization from −80 to −50 mV (Williams and Stuart, 2003) and that IH accelerates IPSP decay at hyperpolarized potentials (van Brederode and Spain, 1995). Our present results extend these findings by showing that: (1) the effects are also present in CA1 neurons; (2) they are substantial even over a smaller (15 mV) voltage range varying from rest to threshold; (3) they are observed with evoked synaptic responses as well as artificially imposed currents or conductances; and (4) influences on physiologically distinct inhibitory synapses are qualitatively similar but differ quantitatively.
Static versus dynamic influences on IPSP duration
Based solely on passive properties, it might be expected that blocking a membrane conductance would increase the input resistance and thereby increase amplitude and prolong IPSPs (τ = RC). However, we found that TTX paradoxically reduced IPSP duration and amplitude (Table 1). It was such a paradoxical effect that caused INaP initially to be termed an “anomalous rectifying” current (Hotson et al., 1979). The explanation for amplitude effects is straightforward: a fraction of a tonic INaP conductance is turned off as the IPSP hyperpolarizes the membrane, and this augments the IPSP. The rapid kinetics of INaP activation and deactivation are an essential characteristic allowing rapid changes in channel activity during the IPSP. The effect on duration is less intuitive, but it is clearly reproduced in the computational model (Fig. 7C).
In the case of IH, activation and deactivation kinetics are relatively slower. Thus, we considered how changes in this current contributed to the effect of IH on duration. Compartmental modeling showed that a nonchanging “leak” (a “static” effect) produced only minor changes in IPSPs (Fig. 6A). Thus, the primary effects of IH are through “dynamic” changes of IH activation level during the time course of the IPSP.
Other investigators have found that IH shapes dendritic voltage transients (Magee, 1998; Williams and Stuart, 2003). Interestingly, using compartmental modeling, we found that IH in the dendrites is almost entirely responsible for the observed effects of IH on duration even for IPSPs generated entirely at the soma. Thus, dendritic IH is important for integration throughout the somatodendritic axis in these neurons.
Passive versus active influences on IPSP amplitude
Passive membrane properties influence IPSP amplitudes in two ways. First, for synaptic currents briefer than the membrane time constant, the duration of the IPSC influences the amplitude of the resulting IPSP. Thus, in our experiments with aIPSPs, it was necessary to inject current 2.5-fold larger for fast synapses to achieve the same amplitudes as slow synapses. A second passive influence is that of driving force. Given the reversal potential of −79 mV that we measured, this would cause IPSPs to be approximately three times larger at −57 mV than at −72 mV.
How does the magnitude of the passive effect of voltage on IPSP amplitude compare with that produced by active membrane properties? Using aIPSPs, we characterized the influences of IH and INaP on IPSP amplitudes. As opposed to their greater influence on the duration of GABAA,fast than GABAA,slow, these active conductances caused a greater amplification of GABAA,slow aIPSP amplitude than GABAA,fast (Fig. 4). Still, even for GABAA,slow, INaP and IH together only increased amplitude by 50%. Thus, compared with the 300% increase in amplitude expected from the change in driving force, the contributions of active conductances to amplitude amplification are relatively small.
IPSPs and the functional impact of inhibition
In some circumstances, for example, when inhibition is superimposed on a tonic excitatory influence, the primary influence of inhibition will be via hyperpolarization. Under other circumstances, such as in a resting and hyperpolarized cell, IPSPs may actually have a depolarizing influence that facilitates action potential firing (Gulledge and Stuart, 2003). In either case, the timing of action potentials following IPSPs will be subject to the influences of passive and active membrane properties that we describe here. Because inhibition plays an important role in synchronization of neuronal networks (Soltesz and Deschenes, 1993; Ylinen et al., 1995; Penttonen et al., 1998), the active conductances that influence IPSPs may thereby also influence neuronal synchrony and determine spike timing, for example, during phase precession of place cell activity during locomotion (Harris et al., 2002). Given the critical role of spike timing in determining the strength and direction of changes in synaptic strength (Markram et al., 1997; Bi and Poo, 1998), IH and INaP may also thereby influence synaptic plasticity.
If passive membrane properties minimize differences in the duration of GABAA,fast and GABAA,slow IPSPs, then under what circumstances might the difference in their IPSC kinetics come into play? The durations of the underlying conductances, as well as relative locations of inhibitory and excitatory synapses, determine the temporal window during which inhibition attenuates action potential firing (Vida et al., 2006) or regulates facilitation of dendritic NMDA responses to burst stimulation (Kanter et al., 1996). One scenario in which the more rapid decay of the fast somatic conductance may matter is during nested θ-γ oscillations. Here, bouts of synchronous IPSCs impinge on pyramidal cells at γ frequency (∼40 Hz), and these bouts of somatic inhibition recur at θ frequency (8 Hz). Curiously, it is during the phase of the θ oscillation when somatic-targeting basket cells firing rate is greatest, and γ oscillations are largest, that pyramidal cells fire most strongly in awake behaving animals (Fox et al., 1986; Bragin et al., 1995; Skaggs et al., 1996; Csicsvari et al., 1999). This occurs despite the fact that the interval between IPSCs (∼25 ms) is substantially briefer than the fast IPSP duration (∼50 ms at −60 mV). Thus, just as shunting inhibition facilitates synchronous firing of inhibitory interneuron networks to generate γ oscillations (Vida et al., 2006), it also constrains the timing of action potentials in pyramidal neurons so that firing is strongly phase locked to the γ oscillation (Penttonen et al., 1998b).
The slower time course of dendritic inhibition may analogously serve to shunt local excitatory inputs during θ oscillations. Dendrite-targeting interneurons are most active out of phase with basket cell activity (Klausberger et al., 2003, 2004); i.e., dendrites are most strongly depolarized when the soma is inhibited, and the dendrites are most strongly inhibited as basket cell firing wanes (Kamondi et al., 1998). It has been proposed that this curious “see-saw” timing of inhibitory synaptic input permits encoding and retrieval during different phases of theta activity (Kunec et al., 2005). The slow time course of dendritic inhibition may be an important characteristic that supports this function, and its interplay with active currents that modify the level and duration of hyperpolarization may contribute to its ability to control dendritic polarization and synaptic plasticity.
Whether inhibitory synapses exert their effects via hyperpolarization or shunting, it is apparent that differences between fast somatic and slow dendritic inhibition contribute to differences in their ability to control integrative function and spike timing. In this way, they contribute to the ability of the hippocampus to use coding strategies that rely on spike timing to perform its critical roles in learning and memory and in other aspects of higher cognitive function.
Footnotes
-
We thank Dr. Mathew Jones (University of Wisconsin, Madison) for allowing use of laboratory equipment for some experiments, Dr. Frostl (CIBA-Geigy) for the kind gift of CGP55845A, Dr. Ted Carnevale (Yale University) for providing a neuron tool to integrate current in all cellular compartments, Dr. Nelson Spruston for helpful discussions and suggestions, and Mark Perkins for excellent technical support.
- Correspondence should be addressed to Robert A. Pearce at the above address. Email: rapearce{at}wisc.edu
