Bidirectional changes in synaptic strength are the proposed cellular correlate for information storage in the brain. Plasticity of intrinsic excitability, however, may also be critical for regulating the firing of neurons during mnemonic tasks. We demonstrated previously that the induction long-term potentiation was accompanied by a persistent decrease in CA1 pyramidal neuron excitability (Fan et al., 2005). We show here that induction of long-term depression (LTD) by 3 Hz pairing of back-propagating action potentials with Schaffer collateral EPSPs was accompanied by an overall increase in CA1 neuronal excitability. This increase was observed as an increase in the number of action potentials elicited by somatic current injection and was caused by an increase in neuronal input resistance. After LTD, voltage sag during hyperpolarizing current injections and subthreshold resonance frequency were decreased. All changes were blocked by ZD7288 (4-ethylphenylamino-1,2-dimethyl-6-methylaminopyrimidinium chloride), suggesting that a physiological loss of dendritic h-channels was responsible for the increase in excitability. Furthermore, block of group 1 metabotropic glutamate receptors (mGluRs) or protein kinase C prevented the increase in excitability, whereas the group 1 mGluR agonist DHPG [(RS)-3,5-dihydroxyphenylglycine] mimicked the effects. We conclude that 3 Hz synaptic stimulation downregulates Ih via activation of group 1 mGluRs and subsequent stimulation of protein kinase C. We propose these changes as part of a homeostatic and bidirectional control mechanism for intrinsic excitability during learning.
Information processing in the brain requires neurons and neuronal networks to continuously undergo changes in their input–output properties. Synaptic efficacy is a well studied mechanism for regulating neuronal output, and increases and decreases in synaptic strength are believed to be the cellular substrates for learning and memory (Bliss and Collingridge, 1993; Bear, 1996). The ability to regulate excitability within a modifiable range is vital for maintaining the computational power of the neuron. Although many studies have focused on morphological, biochemical and physiological changes that occur within or near synapses, recent studies examined changes in voltage-gated channels (Wang et al., 2003; Frick et al., 2004; Fan et al., 2005; Xu et al., 2005; Kim et al., 2007). The resulting influence on the active and passive membrane properties is called plasticity of intrinsic excitability (Frick and Johnston, 2005). The regulation of voltage-dependent ion channels by activity may represent a homeostatic mechanism for maintaining neuronal excitability within physiological limits (Abbott and Nelson, 2000; Turrigiano and Nelson, 2000). Because of its role in the regulation of membrane excitability, the hyperpolarization activated, nonselective cation channel (Ih) is a suitable candidate.
H-channels, first characterized in cardiac tissue, are widely distributed in the CNS (DiFransesco, 1993; Pape, 1996). Ih is a mixed-cation conductance with slow kinetics (CA1 pyramidal neurons, τact = 20 ms; τdeact = 18 ms) (Magee, 1998). Ih contributes to synchronized network oscillations, intrinsic oscillations, and pacemaker activity in the thalamus, hippocampus, and neocortex (Hutcheon et al., 1996; Maccaferri and McBain, 1996; Lüthi et al., 1998). Because a significant amount is active a rest, Ih contributes to the resting membrane potential (Magee, 1998; Aponte et al., 2006). Furthermore, with its unique activation by hyperpolarization, Ih opposes changes in membrane potential and strongly influences synaptic integration (Magee, 1998; Magee, 1999). Although the h-channel has a relatively small single channel conductance compared with many other channels (Magee and Johnston, 1995; Chen and Johnston, 2004; Kole et al., 2006), the high density of h-channels in the dendrites, which increases with distance from the soma (Magee, 1998), contributes significantly to the total membrane conductance of the neuron. Subtle modifications in the physiology of h-channels can therefore produce significant changes in intrinsic excitability.
We used a long-term depression (LTD) protocol that consisted of pairing back-propagating action potentials and Schaffer collateral EPSPs at 3 Hz (Christie et al., 1996). In addition to producing LTD of EPSPs, 3 Hz pairing also resulted in a long-lasting increase in CA1 pyramidal neuron excitability. This increase was caused by a downregulation of Ih, because it was accompanied by changes in multiple electrophysiological measurements sensitive to Ih and was blocked by the Ih blocker ZD7288 (4-ethylphenylamino-1,2-dimethyl-6-methylaminopyrimidinium chloride). Furthermore, the decrease in Ih was blocked by group 1 mGluR antagonists and inhibitors of protein kinase C and was mimicked by application of a group 1 mGluR agonist. We suggest that activity-dependent increases in intrinsic excitability mediated by decreases in Ih act in a homeostatic manner to prevent the loss of neuronal output during hippocampal long-term depression.
Materials and Methods
Acute hippocampal slices.
All experiments were conducted in accordance with the University of Texas at Austin Institutional Animal Care and Use Committee. Hippocampal slices (350 μm) were prepared from 4- to 6-week-old male Sprague-Dawley rats using standard techniques (Magee and Johnston, 1997). Briefly, animals were anesthetized using a lethal dose of ketamine and xylazine. Once deeply anesthetized, animals were perfused intracardially with ice-cold modified ACSF containing (in mm) 210 sucrose, 2.5 KCl, 1.2 NaH2PO4, 25 NaHCO3, 0.5 CaCl2, 7.0 MgCl2, and 7.0 dextrose bubbled with 95%O2/5%CO2. The brain was removed and bisected along the midline. To promote an orientation favoring dendritic projection in a plane parallel to the surface of the slice, an additional cut was made on the dorsal surface at a 30° angle lateral to the midline. The brain was mounted and sliced using a microtome (Vibratome, St. Louis, MO). Slices were placed in a holding chamber filled with ACSF containing (mm) 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 2 CaCl2, 2 MgCl2, and 25 dextrose (see below) warmed to 35°C for 20 min and then placed at room temperature for <6 h until needed for recording.
For all recordings, the control ACSF solution contained (mm) 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 2 CaCl2, 1 MgCl2, and 25 dextrose and was bubbled continuously with 95%O2/5% CO2 at 31–33°C. Slices were placed individually as needed into a submerged recording chamber continuously perfused with control extracellular saline. Slices were viewed with a Zeiss (Oberkochen, Germany) Axioskop using infrared video microscopy and differential interference contrast (DIC) optics. In those experiments examining LTD of EPSPs, GABAA-mediated inhibition was blocked by inclusion of 10 μm bicuculline and 10 μm picrotoxin in the extracellular saline. In these cases, area CA3 was removed before placing the slice in the recording chamber to prevent spontaneous epileptiform-like discharge. Patch pipettes were pulled from borosilicate glass and had a resistance of 4–8 MΩ when filled with the internal recording solution containing (in mm) 120 potassium methanesulfonate or potassium methylsulfate, 20 KCl, 10 HEPES, 4 NaCl, 0.2 EGTA, 4 MgATP, 0.3 TrisGTP, and 14 phosphocreatine, pH 7.3 (with KOH).
All whole-cell recordings were made from the soma or along the first 50 μm of the proximal apical dendrite of CA1 pyramidal neurons using a Dagan (Minneapolis, MN) BVC-700 in current-clamp mode. Series resistance was monitored throughout the recording and experiments in which the series resistance exceeded 30 MΩ were discarded. EPSPs of 4–6 mV were elicited using tungsten bipolar stimulating electrodes (180–220 μA for 0.1 ms) placed near (<20 μm) the main apical dendrite ∼150–180 μm from the soma in stratum radiatum. Back-propagating action potentials (b-APs) were elicited by current injection into the soma (1–2 nA for 2 ms) or by antidromic extracellular stimulation. LTD was induced by pairing EPSPs and b-APs at 3 Hz for 5 min (Christie et al., 1996).
Data acquisition and analysis.
Data were sampled at 10 kHz, filtered at 3 kHz and digitized by an ITC-18 interface connected to computer running custom software written in IgorPro (Wavemetrics, Lake Oswego, OR). All data analyses were performed with custom written software in IgorPro (Wavemetrics). EPSPs were quantified by measuring the initial slope (linear fit over 1–2 ms). The EPSP slope provides a better measure of synaptic transmission than EPSP amplitude because of the potential contribution of voltage-dependent conductances to the amplitude of the EPSP. Input resistance (RN) was determined by the slope of the linear regression line through the V–I plot (fit through the linear range of a plot of the amplitude of the steady-state voltage against the corresponding current injection from a family of 500–750 ms current steps). Input–output curves were constructed by plotting the numbers of action potentials against the amplitude of the current injections (150–300 pA, for 500 ms in 50 pA intervals). Action potential threshold was defined as the voltage where the first derivative of the voltage (dV/dt) first exceeds 10 mV/ms (Fan et al., 2005). Estimates of the membrane time constant (τm) were made by fitting a single exponentials to the same family of voltage responses used for input resistance measurements and calculating the average τm. Exponentials were fit to relaxation of the membrane voltage over a 25 ms window beginning 2 ms after the end of the current injection.
For measurements of temporal summation, simulated EPSPs (αEPSPs) were used to remove any contribution of presynaptic mechanisms. αEPSPs were simulated by the injection of current using the following function: I = Imax (t/α) e−αt. Imax and α were adjusted to produce EPSP-like waveforms with a peak amplitude of 5 mV and a time-to-peak of 10 ms similar to evoked EPSPs. Temporal summation ratio was measured as the amplitude of the fifth αEPSP relative to first in a train of 5 αEPSPs [(αEPSP5 − αEPSP1)/αEPSP1].
Voltage sag was measured as the percentage change between the maximum and steady state voltage change during hyperpolarizing current injections [(Vmax − Vss)/Vmax] × 100. The stimulus used for characterizing the impedance amplitude profile (ZAP) was a sinusoidal current of constant amplitude (50 or 100 pA), with its frequency linearly spanning 0–15 Hz in 15 s. The magnitude of the ratio of the Fourier transform of the voltage response to the Fourier transform of the current stimulus formed the impedance amplitude profile. The frequency at which the impedance amplitude reached its maximum was the resonance frequency.
All data are expressed as mean ± SEM. Statistical comparisons were made using ANOVA (one-way, two-way or repeated measures design as appropriate) followed by Tukey–Kramer multiple-comparisons post hoc test or Student's t test (paired or unpaired as appropriate) with InStat software. Linear fits and correlations were made using either Igor (Wavemetrics) or InStat (GraphPad, San Diego, CA). Data were considered statistically significant if p < 0.05.
Drugs and solutions.
All drugs were made from stock solutions in water, equivalent NaOH, or DMSO (final concentration of DMSO ≤0.1%) accordingly. ZD7288, d, l-APV, (+)-MK-801 maleate, CNQX, bicuculline, picrotoxin, (2S)-2-amino-2-[(1S,2S)-2-carboxycycloprop-1-yl]-3-(xanth-9-yl) propanoic acid (LY341495), (RS)-3,5-dihydroxyphenylglycine (DHPG), and GF109203X were obtained from Tocris Cookson (Bristol, UK). When 20 mm K4-BAPTA (Sigma, St. Louis, MO) was including in the recording pipette, potassium methylsulfate or methanesulfonate was reduced to 40 mm and the osmolality adjusted with sucrose.
LTD is accompanied by increased CA1 pyramidal neuron excitability
Long-term depression of CA3-CA1 synaptic transmission was induced by delivering a train of b-APs paired with synaptic stimulation at 3 Hz for 5 min (900 total b-AP–EPSP pairs; b-AP preceded EPSP by 5 ms) (Christie et al., 1996). All measurements were made at 31–33°C and the holding current was adjusted to keep the membrane potential at −70 mV. The membrane potential was allowed to vary during the pairing protocol. Three hertz pairing resulted in a significant decrease in EPSP slope that persisted for the duration of recordings (Fig. 1A). In the absence of the induction protocol, there was no significant change in EPSP slope. The 3 Hz pairing was accompanied by increased CA1 pyramidal neuron excitability measured as the number action potentials elicited by a given current injection (Fig. 1B,C). The number of action potentials elicited by current injection in the soma increased at all injection amplitudes after LTD induction (Fig. 1C). The underlying cause for this increase in APs per current injection was an increase in CA1 pyramidal neuron input resistance. Input resistance (RN) increased by 87 ± 10% after 3 Hz pairing (baseline, 54 ± 3 MΩ; 3 Hz, 99 ± 5MΩ; n = 20; p < 0.001) (Fig. 1D). We also found that the membrane time constant (τm) was significantly increased after 3 Hz pairing (baseline, 21 ± 2 ms; 3 Hz pairing, 28 ± 2 ms; p < 0.005) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). There was a small but significant increase in input resistance (RN) in the absence of the pairing protocol (Fig. 1D–F). A previous report on the effects of internal anions on intrinsic membrane properties suggests that methylsulfate based internal recording solutions can increase RN during whole-cell recording (Kaczorowski et al., 2007). Although we observed this as well, the effect of dialysis was significantly smaller than the increase in RN after 3 Hz pairing (Fig. 1D–F). The increase in AP firing was not likely caused by alterations in sodium channel activation as there was no significant change in action potential threshold after 3 Hz pairing (baseline, −50.7 ± 0.6 mV; 3 Hz, −52.9 ± 1.5 mV; n = 5, p = 0.2). CA1 pyramidal neuron input resistance increased gradually after LTD induction reaching an apparent plateau ∼30 min after 3 Hz pairing (Fig. 1E). A plot of input resistance versus EPSP slope, measured for 30 min after 3 Hz pairing, revealed a linear correlation (r = −0.85; n = 13; p < 0.001) between the decrease in synaptic strength and the increase in input resistance (Fig. 1F). We described a similar correlation in CA1 pyramidal cells between the decrease in input resistance and magnitude of the increase in EPSP slope after LTP induction (Fan et al., 2005).
Downregulation of Ih underlies increased CA1 pyramidal neuron excitability
An increase in Ih is responsible for the decrease in input resistance that accompanies LTP (Fan et al., 2005). Accordingly, we hypothesized that a decrease in Ih would explain our observation of increased input resistance after 3 Hz pairing. Furthermore, 3 Hz pairing resulted in a hyperpolarization of the resting membrane potential (baseline, −69 ± 2 mV; 3 Hz, −74 ± 2 mV; n = 10; p < 0.01), consistent with a decrease in Ih (Gasparini and DiFrancesco, 1997). To explore the role of Ih in the increased excitability associated with LTD, the 3 Hz pairing paradigm was repeated in the presence of the Ih blocker ZD7288 (20 μm). Because there is a significant amount of Ih active at rest in CA1 pyramidal cells, Ih contributes to the resting input resistance (Magee, 1998). As expected, block of Ih with ZD7288 resulted in a 180% increase in CA1 pyramidal neuron input resistance (initial, 77 ± 15 MΩ; 30 min, 211 ± 35 MΩ; n = 4; p < 0.05). Although we observed no significant effect on EPSP slope by ZD7288 (data not shown) (cf. Chevaleyre and Castillo, 2002), we nonetheless sought to remove any presynaptic influences by including ZD7288 in the postsynaptic recording electrode. Block of Ih by postsynaptic application of ZD7288 did not significantly affect the decrease in EPSP slope after 3 Hz pairing (LTD in control, 40 ± 4%; LTD in ZD7288, 55 ± 10%; p > 0.05) (Fig. 2A). However, there was no further increase in RN after 3 Hz pairing (Fig. 2B,C). If 3 Hz pairing increases RN by downregulating Ih, then the total increase in RN induced by 3 Hz pairing followed by ZD7288 application should be the same as the total increase in RN when ZD7288 is applied to naive CA1 pyramidal cells. In a separate set of experiments, ZD7288 was applied extracellularly 30 min after the 3 Hz pairing induced increase in input resistance (Fig. 2D). After 3 Hz pairing, RN significantly increased by 43 ± 8MΩ (p < 0.05, n = 4). Application of ZD7288 further increased RN, for a total increase of 133 ± 36 MΩ. Application of ZD7288 to naive slices increased RN by 126 ± 26 MΩ (n = 4) and was not significantly different from the ZD induced increase after 3 Hz pairing (Fig. 2D). These results suggest that 3 Hz pairing results in the physiological loss of a significant fraction of CA1 pyramidal neuron Ih.
A loss of Ih should result in longer EPSP decay times and increases in temporal summation (Magee, 1999). We previously showed activity-dependent increases in Ih after LTP were accompanied by decreases in EPSP decay time and temporal summation (Fan et al., 2005). After 3 Hz pairing, EPSP decay time increased significantly. The increase in decay time was blocked by ZD7288 and absent when there was no induction protocol (Fig. 3A). One potential problem of these experiments is that a decrease in EPSP amplitude alone should increase decay time as smaller EPSPs will deactivate Ih less than larger EPSPs. Simply increasing the stimulation intensity to return EPSP amplitude to control levels, however, could present additional problems, because this could recruit new synapses that did not undergo plasticity. To address this issue, we stimulated two nonoverlapping pathways that made synapses onto the same CA1 pyramidal neuron to compare the change in EPSP slope and decay after 3 Hz pairing. 3 Hz pairing induced a significant decrease EPSP slope in the test pathway compared with the control pathway (34 ± 2% vs 97 ± 3%) (Fig. 3B,C). As expected, there was a significant increase in input resistance after 3 Hz pairing (43 ± 11 MΩ; 3 Hz, 84 ± 19 MΩ; p < 0.05). Accordingly, there was a linear correlation between the increase in input resistance and the decrease in EPSP slope of the test pathway but not the control pathway (Fig. 3D). Interestingly, the 10–90% decay time of the both the test and control EPSPs was significantly greater after 3 Hz pairing (Fig. 3E). Because the decay time was longer for control EPSPs despite no change in amplitude, we conclude that the increase in decay of depressed EPSPs is caused by the activity-dependent decrease in Ih as well as the decrease in EPSP amplitude.
To assess temporal summation, a train of five current pulses (modeled by an α function, αEPSPs) was used to mimic EPSPs. The use of αEPSPs removes any potential presynaptic components that may occur during repetitive stimulation (Brager et al., 2002). Five αEPSPs were injected into the soma at 5, 10, 20, and 50 Hz and summation was expressed as the percentage increase in the fifth EPSP relative to the first. Temporal summation was significantly increased during the 10 and 20 Hz trains, but not the 5 or 50 Hz trains after 3 Hz pairing (Fig. 3F). These intermediate frequencies are most sensitive to changes in Ih (Poolos et al., 2002).
These results suggest that the increase in input resistance after 3 Hz pairing was caused by a decrease in Ih. As an additional test of this hypothesis, two additional electrophysiological parameters sensitive to changes in Ih were examined: membrane potential sag (voltage sag) and resonance frequency (fR). With its activation by hyperpolarization and deactivation by depolarization, Ih opposes changes in membrane potential. This is observed as the characteristic voltage sag after a depolarizing or hyperpolarizing change from rest (Poolos et al., 2002). This same characteristic allows Ih to act as a resonator conductance and changes in resonance frequency can reflect changes in Ih (Hutcheon and Yarom, 2000) (Narayanan and Johnston, 2007). Pharmacological block of Ih with ZD7288 resulted in increased input resistance, complete loss of voltage sag and a decrease in the resonance frequency (baseline, 3.8 ± 0.24 Hz; ZD7288, 1.2 ± 0.12 Hz; p < 0.05). If the observed increase in input resistance is caused by the physiological loss of Ih, a decrease in sag and resonance frequency after 3 Hz pairing would be expected.
Voltage sag was measured from a family of hyperpolarizing current injections from the resting membrane potential measured before and after 3 Hz pairing. Because Ih is voltage-sensitive, sag was compared between two different current steps that had a similar peak voltage deflection (Fig. 4A). After 3 Hz pairing, voltage sag significantly decreased from 24 ± 2% to 9 ± 1% (Fig. 4B). Resonance frequency (fR) was determined by recording the voltage response during injection of the ZAP current stimulus (Fig. 4C). The maximum of the impedance amplitude profile (see Materials and Methods) occurs at the resonant frequency of the cell (Fig. 4D). The fR of CA1 pyramidal neurons was significantly decreased from 3.8 ± 0.4 Hz to 2.4 ± 0.5 Hz (n = 9; p < 0.05) after 3 Hz pairing (Fig. 4E). Neither voltage sag nor fR changed significantly in the absence of 3 Hz pairing (Fig. 4B,E). As for the change in input resistance, the decrease in sag and resonance frequency after 3 Hz pairing was less than that observed after application of ZD7288. In combination, these results suggest that 3 Hz pairing results in a physiological loss of a fraction of the total Ih.
Although these data support the hypothesis that a decrease in Ih underlies the increase in input resistance, we cannot at this point rule out the loss of a voltage-independent conductance. If the increase in input resistance was caused by a change in a voltage independent conductance (i.e., Ileak), then there should be no voltage dependence to the increase in RN after 3 Hz pairing. CA1 pyramidal neuron input resistance was measured at various membrane potentials between −60 and −90 mV before and after 3 Hz pairing (Fig. 5A). The increase in input resistance was greater at the hyperpolarized membrane potentials. A fit of the normalized increase in RN as a function of membrane potential had a slope of −0.22 and was significantly different from a slope of zero (p < 0.005; n = 5) (Fig. 5B). These results support the hypothesis that a change in a voltage-dependent conductance underlies the increase in input resistance after 3 Hz pairing. Although voltage sag during a hyperpolarizing injection can be used to estimate Ih, the rebound potential after the offset of the current can also be used. Other channels such as the inwardly rectifying potassium channel (IKIR) will contribute to the steady-state voltage response during hyperpolarizing current injections. Because the deactivation of IKIR, caused by the voltage-dependent block by Mg2+ ions, is fast (8–9 ms at 15°C) (Matsuda et al., 1987) compared with the deactivation of Ih, the rebound potential amplitude should be indicative of the amount of Ih activated at a given membrane potential. Given that long (500 ms) current injections were used to measure RN and sag, the steady-state voltage can be treated as the new holding potential. We therefore plotted the rebound potential amplitude as a function of the steady-state voltage (Fig. 5C). The slope of the line fit to the data were decreased after 3 Hz pairing (baseline, −0.31 ± 0.007; 3 Hz, −0.15 ± 0.006; n = 16) (Fig. 5D). In the absence of 3 Hz pairing, there was no significant change in rebound potential amplitude as a function of membrane potential (baseline, −0.31 ± 0.005; 30 min, −0.27 ± 0.008; n = 8). These results suggest that after 3 Hz pairing, Ih is reduced because the h-conductance (gh) is smaller.
The reduction in gh after 3 Hz pairing could be caused by a hyperpolarizing shift in the activation of Ih (shift in V1/2), a decrease in the maximum conductance (gmax) or a combination of both. A plot of resonance frequency (fR) as a function of holding potential should produce a bell-shaped relationship that is well described by the first derivative of the Boltzmann function (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Shifts in the voltage dependence of this relationship can be used to make estimates of changes in gmax and V1/2 of Ih (Narayanan and Johnston, 2007). A change in the voltage where the maximum resonance frequency occurs (VB) reflects a shift in V1/2; changes in maximum resonance frequency reflect a difference in gmax (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). The fR of CA1 pyramidal neurons (n = 5) was measured at six membrane potentials between −60 and −90 mV before and after 3 Hz pairing. As the membrane potential was hyperpolarized, fR increased because of an increase in the proportion of active h-channels (Fig. 5E). Fitting the group data with the derivative of the Boltzmann revealed that the maximum resonance frequency was decreased from 6.6 to 5.3 Hz, consistent with a decrease in gmax (Fig. 5F). Although, the voltage where the maximum resonance frequency occurred was shifted toward hyperpolarized potentials (baseline, −80.4 ± 0.5 mV; 3 Hz pairing, −83.6 ± 2 mV; n = 5), this difference was within sampling error (fR was measured in 5 mV intervals). These data support the hypothesis that a decrease in Ih occurs after 3 Hz pairing. Furthermore, these data suggest that a decrease in gmax underlies the reduction in Ih although we cannot rule out changes in V1/2 (see Discussion). These results together with the sensitivity to ZD7288, the increase in input resistance, hyperpolarization of resting Vm, and increase in temporal summation and EPSP decay support the hypothesis that the increase in CA1 pyramidal neuron excitability after 3 Hz pairing is caused by the activity-dependent downregulation of Ih.
Increased excitability requires mGluR activation
Hippocampal LTD requires the activation of either NMDA receptors (NMDARs) (Dudek and Bear, 1992; Mulkey and Malenka, 1992; Bliss and Collingridge, 1993; Christie et al., 1996; Oliet et al., 1997; Malenka and Bear, 2004) and/or metabotropic glutamate receptors (mGluRs) (Bolshakov and Siegelbaum, 1994; Selig et al., 1995; Oliet et al., 1997; Huber et al., 2000; Normann et al., 2000; Rammes et al., 2003). We previously described a role for NMDA receptors in the plasticity of h-channels during LTP (Fan et al., 2005). H-channels have not been shown previously to undergo either NMDAR- or mGluR-dependent modulation during LTD.
Decreased CA1 pyramidal neuron excitability after LTP induction required Ca2+ influx via NMDA receptor activation (Fan et al., 2005). The increase in CA1 pyramidal neuron excitability that accompanied LTD is similarly Ca2+-dependent because the inclusion of 20 mm BAPTA in the recording pipette blocked the increase in input resistance (Fig. 6A,B). The activation of both NMDA receptors and mGluRs can occur during LTD inducing protocols (Oliet et al., 1997; Huber et al., 2002). To determine whether the activation of these receptors is required for the activity-dependent downregulation of Ih, we repeated the 3 Hz pairing in the presence of either 50 μm d, l-APV and 10 μm (+)-MK-801, to block NMDA receptors, or 100 μm LY341495 to block mGluR receptors. These antagonists are known to block NMDA- and mGluR-dependent LTD respectively (Oliet et al., 1997; Huber et al., 2000, 2002). Bath application of either APV/MK-801 or LY341495 reduced LTD of EPSP slope (Fig. 6C). This suggests that the activation of both NMDA and mGluRs occurs during the 3Hz pairing-induction of LTD. APV/MK-801 had no effect on the increase in input resistance (Fig. 6D–F) or on the decrease in voltage sag (baseline, 23 ± 2%; 3 Hz, 15 ± 2%; n = 8, p < 0.05) after 3 Hz pairing. In contrast, LY341495 significantly reduced the increase in input resistance after 3 Hz pairing (Fig. 6D–F) and the decrease in voltage sag (baseline, 22 ± 2%; 3 Hz, 19 ± 2%; n = 5, p > 0.05). The increase in τm was also blocked by LY341495 but not APV/MK-801 (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). These results suggest that activation of mGluRs during 3 Hz pairing is necessary for the decrease in Ih.
As a test of this hypothesis, 3 Hz pairing was repeated with fast glutamatergic transmission blocked by including CNQX (10 μm) and d, l-APV/(+)-MK-801 (50 μm/10 μm) in the extracellular saline (Fig. 7A). Under these conditions, with all ionotropic receptors blocked, presumably only mGluR receptors are activated during the pairing protocol. In the presence of CNQX/APV/MK-801, 3 Hz pairing resulted in a 66 ± 10% increase in input resistance, similar to the increase seen in the absence of CNQX/APV/MK-801 (Fig. 7B). The change in input resistance in the presence of LY3431495 was not significantly different from that seen in the absence of 3 Hz pairing. There was a decrease in fR after 3 Hz pairing, observed as a leftward shift of the impedance amplitude profile in both control saline and in the presence of ionotropic glutamate antagonists (Fig. 7C,D). This was not present when mGluRs were blocked by LY341495 or without 3 Hz pairing (Fig. 7C,D). Because CA1 pyramidal neurons do not express group 2 mGluRs (Luján et al., 1997) and group 3 mGluRs are predominantly presynaptic (Ohishi et al., 1995a,b), we hypothesize that the increase in CA1 pyramidal neuron excitability, via a decrease in Ih, requires the activation of group 1 mGluRs.
As a direct test of this hypothesis, we asked whether the selective group 1 mGluR agonist DHPG could reliably reproduce the changes observed with 3 Hz pairing. Bath application of DHPG has been used previously in the study of hippocampal mGluR-dependent LTD (Huber et al., 2000). In CA1 pyramidal neurons, application of DHPG is known to increase excitability and is accompanied by depolarization and action potential discharge (Gereau and Conn, 1995). We demonstrated previously that increases in neuronal activity increase Ih via NMDA receptor activation (Fan et al., 2005). We found that 30 min after a 10 min application of DHPG, there was no significant change in RN. If however, DHPG was applied in the presence of TTX, to block action potentials, or APV, to block NMDA receptors, a significant increase in RN was observed (Fig. 8A). We hypothesized that in control saline, the decrease in Ih by mGluR activation was opposed by AP/NMDA-dependent increases in Ih. To remove this potentially confounding process, DHPG application was performed with NMDA receptors blocked by APV/MK-801. After a 5 min baseline period, DHPG (100 μm) was applied to the slice for 10 min. During DHPG application, CA1 pyramidal neurons depolarized, fired a series of action potentials, and settled at a new resting membrane potential near −40 mV (Fig. 8B). After wash out of DHPG, the membrane potential slowly returned to near the baseline value. EPSP slope was significantly decreased 30 min after DHPG washout (baseline, 1.2 ± 0.16 mV/ms; 30 min, 0.8 ± 0.14 mV/ms; p < 0.005, n = 5) and there was a leftward shift of the impedance amplitude profile (Fig. 8C). In agreement with our hypothesis, input resistance was increased (baseline, 54 ± 4 MΩ; DHPG, 106 ± 7 MΩ), sag was decreased (baseline, 19 ± 2%; DHPG, 11 ± 2%; p < 0.05), and resonance frequency was decreased (baseline, 3.2 ± 0.5 Hz; DHPG, 1.6 ± 0.3 Hz; p < 0.05) 30 min after DHPG application (Fig. 8D). The slope of the linear fit of rebound amplitude as a function of membrane potential was also reduced (baseline, −0.2 ± 0.004; 30 min after DHPG, −0.06 ± 0.006; n = 5) (Fig. 8E). These data further support the hypothesis that activation of group 1 mGluRs increase excitability by decreasing Ih.
Previously, we demonstrated that the decrease in input resistance observed with LTP did not require evoked EPSPs, but in fact, only required b-APs elicited in a theta burst pattern (Fan et al., 2005). Long trains of EPSPs are sufficient to activate mGluRs and induce synaptic plasticity in the hippocampus (Oliet et al., 1997; Huber et al., 2000). We therefore asked whether the downregulation of Ih required the pairing of b-APs and synaptic stimulation. When the stimulation protocol consisted only of b-APs at 3 Hz, there was no significant decrease in EPSP slope. Although synaptic stimulation alone at 3 Hz produced a small depression of EPSP slope, it was significantly less than the LTD caused by 3 Hz pairing (pairing, 62 ± 6%; synaptic only, 24 ± 14%; p < 0.05) (Fig. 9A). More interestingly, whereas 3 Hz b-APs had no significant effect on input resistance, 3 Hz synaptic stimulation produced an increase in input resistance that was not significantly different from 3 Hz pairing (Fig. 9B).
PKC activation is required for downregulation of Ih
Group 1 metabotropic glutamate receptors are coupled to phosphoinositide hydrolysis (Abe et al., 1992; Aramori and Nakanishi, 1992). In the dentate gyrus and area CA1, mGluR-dependent LTD requires the activation of protein kinase C (PKC) (Oliet et al., 1997; Wang et al., 1998). Furthermore, the activation of PKC by neurotensin receptors can inhibit Ih in the substantia nigra (Cathala and Paupardin-Tritsch, 1997). We therefore tested the hypothesis that PKC activation is necessary for the downregulation of Ih. Bath application of the PKC inhibitor GF109203X (10 μm) blocked the 3 Hz pairing induced LTD (control, 38 ± 5% vs 97 ± 12%, p < 0.001) (Fig. 9C). Block of PKC by GF109203X also prevented the leftward shift of the impedance amplitude profile after 3 Hz pairing (Fig. 9D). Block of PKC by GF109203X significantly reduced the increase in input resistance, decrease in voltage sag, and decrease in resonance frequency measured 30 min after 3 Hz pairing (Fig. 9E).
Changes in neuronal output from a given input are thought to form the cellular basis of learning and memory (Bear, 1996). We demonstrated previously that a long-lasting decrease in excitability, accompanying certain forms of LTP, was caused by the activity-dependent upregulation of Ih (Fan et al., 2005). In this study, we show that there is an activity-dependent increase in neuronal excitability that accompanies 3 Hz stimulation-induced LTD. Similar to the decrease in excitability after LTP (Fan et al., 2005), the increase in excitability occurs gradually over the course of several minutes. This increase in excitability is dependent on the activation of group 1 metabotropic glutamate receptors and protein kinase C. Furthermore, changes in electrophysiological characteristics indicative of Ih (e.g., input resistance, voltage sag, temporal summation, and resonance frequency), coupled with sensitivity to the h-channel blocker ZD7288, strongly support the conclusion that a long-lasting downregulation of Ih underlies this increase in excitability.
Early demonstrations of increases in excitability after LTD focused on changes in EPSP–spike (E/S) coupling. Bernard and Wheal (1995) reported an increase in E/S coupling accompanying LTD. Although both the extracellularly recorded EPSP (fEPSP) and population spike (PS) are reduced after LTD, the relationship between fEPSP slope and PS amplitude is increased. Because all of our experiments used intracellular recordings, our observations were localized to a single CA1 neuron and cannot address any changes in the population of neurons. Two potential mechanisms explaining increased E/S coupling are decreased inhibitory drive and/or a change in action potential threshold (Wilson, 1981; Abraham et al., 1987; Taube and Schwartzkroin, 1988; Chavez-Noriega et al., 1989; Hess and Gustafsson, 1990). Because changes in excitability occurred after LTD with inhibition blocked by bicuculline and picrotoxin and we did not observe a significant change in action potential threshold, we suggest that the decrease in Ih during LTD may be a separate phenomenon.
The h-channel blocker ZD7288 has reported presynaptic effects on synaptic transmission. Chevaleyre and Castillo (2002) reported that bath application of ZD7288 (1–50 μm) reduced Schaffer collateral EPSP amplitude and increased paired-pulse facilitation, consistent with a decrease in presynaptic release probability. However, this same report also demonstrated that bath application of 50 μm ZD7288 (more than twice the concentration used in this study) did not significantly reduce EPSC amplitude over a 15 min application although completely blocked postsynaptic Ih. Furthermore, the time required to observe a significant effect of ZD7288 on release probability was longer than most of the recordings in this study (>60 min).
Activity-dependent plasticity has traditionally been associated with increases and decreases in synaptic strength (Bliss and Collingridge, 1993; Bear and Malenka, 1994). Previously, activity-dependent changes in voltage-dependent channels have been shown to alter intrinsic excitability (Turrigiano et al., 1994; Desai et al., 1999; Golowasch et al., 1999; Frick et al., 2004; Fan et al., 2005). There are several reports of changes in intrinsic excitability associated with LTP from cortex, cerebellum, and hippocampus (Aizenman and Linden, 2000; Wang et al., 2003; Cudmore and Turrigiano, 2004; Frick et al., 2004; Fan et al., 2005; Xu et al., 2005). Our results demonstrate the first report of activity-dependent changes in intrinsic excitability accompanying LTD.
Prolonged low-frequency synaptic stimulation or asynchronous pairing of EPSPs and action potentials in the hippocampus leads to both NMDAR and mGluR activation (Oliet et al., 1997; Normann et al., 2000). Our results suggest that although both receptors can contribute to the depression of EPSPs, only mGluR activation is necessary for activity-dependent decreases in Ih.
Previous reports of mGluR activation on CA1 pyramidal neuron excitability examined the changes in intrinsic properties during agonist application (Desai and Conn, 1991; Gereau and Conn, 1995). In contrast, we report persistent effects up to 30 min after either 3 Hz stimulation or DHPG application. Metabotropic glutamate receptor agonists can increase excitability by inhibiting several intrinsic potassium conductances including IAHP and Ileak (Charpak et al., 1990; Desai and Conn, 1991; Gereau and Conn, 1995). Although we cannot exclude the possibility that the spike afterhyperpolarization is reduced, changes in IAHP cannot explain our results because IAHP does not contribute to any of the electrophysiological parameters altered after 3 Hz stimulation. Our results are also not explained by a persistent decrease in Ileak because the increase in input resistance after 3 Hz pairing was voltage dependent (Fig. 5B).
We report a decrease in maximum resonance frequency (fR) without a significant change in the voltage where that maximum occurs (VB) (Fig. 4F). A hyperpolarizing shift in V1/2 would shift the fR versus Vm curve to the left and increase its peak amplitude, without an overall reduction in fR values across Vm. A decrease in gmax would reduce the peak fR without shifting VB (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Given that there is a decrease in the maximum fR, not an increase, that the shift in VB is within sampling limits (5 mV), and that there is an overall reduction in fR across voltages, we suggest that decreases in the maximum conductance and not shifts in V1/2 underlie the activity-dependent downregulation of Ih. However, additional experiments are needed to rigorously test this hypothesis.
Protein kinase C is activated during the induction of mGluR-mediated LTD in the hippocampus and dentate gyrus (Oliet et al., 1997; Wang et al., 1998). We found that block of PKC activation prevented the downregulation of Ih and the subsequent increase in excitability. Activation of group 1 mGluRs is coupled to phosphoinositide hydrolysis and the generation of diacylglycerol and inositide-1,4,5-trisphosphate (IP3), resulting in the activation of protein kinase C and IP3-sensitive Ca2+ stores respectively (Abe et al., 1992; Aramori and Nakanishi, 1992). Because the increase in excitability by 3 Hz pairing was completely blocked by intracellular BAPTA but not by NMDA antagonists, we suggest that Ca2+ release from stores may be involved in the downregulation of Ih.
Impact on neuronal function
We previously showed that an enhancement of Ih counterbalances the increase in synaptic strength during LTP to normalize neuronal firing (Fan et al., 2005). This would prevent saturation of neuronal output caused by potentiation of synaptic transmission (Abraham et al., 2001). For modulation of Ih to be an effective homeostatic mechanism, it is essential that Ih regulate neuronal output in the opposite direction as well. Without a mechanism to maintain action potential firing, uncontrolled LTD could lead to a complete loss of neuronal output. Therefore, a homeostatic mechanism to prevent this hypoexcitability should exist. If increases in Ih serve as the regulator of neuronal firing during potentiated synaptic transmission, then decreases in Ih may serve a similar function during depressed synaptic transmission.
In this study, we demonstrated that 3 Hz pairing-induced LTD was accompanied by increased neuronal firing caused by increased input resistance. A reduction in Ih would thus make naive synapses of the cell more effective at generating action potentials. One possibility is that the overall result of LTD induction is not just to reduce the contribution of the stimulated synapses, but also to increase the signal from the nonstimulated pathways. One consequence of Ih downregulation is an increase in EPSP decay time and increased in temporal summation (Magee, 1998). This would prolong the depolarization produced by an EPSP and increase the likelihood of action potential generation caused by the integration of multiple inputs. We found that the decay time of nondepressed EPSPs was significantly increased by 20% similar to depressed EPSPs (Fig. 3). Under these conditions, depressed inputs would have a reduced role in AP generation whereas nondepressed inputs would have a greater role. From our previously published results on theta-burst pairing (TBP) LTP, the opposite can be hypothesized: TBP-LTP selects for the potentiated synapses. In TBP-LTP a subset of synapses becomes potentiated whereas the entirety of the synaptic inputs become less effective because of a global increase in Ih (Narayanan and Johnston, 2007). Under these conditions, only the potentiated synapses would be more effective at generating action potentials. By modulating dendritic Ih in concert with synaptic strength, a neuron cannot only increase or decrease the likelihood of action potential generation, but also select which sets of inputs contribute to that AP generation.
In conclusion, we have shown that activity-dependent bidirectional plasticity is not restricted to synapses. We demonstrated that decreases in synaptic strength are counterbalanced by concomitant increases in intrinsic excitability. We demonstrated previously that increases in synaptic strength were counterbalanced by decreases in intrinsic excitability (Fan et al., 2005). This bidirectional regulation of intrinsic excitability is mediated by the activity-dependent increase or decrease in a single voltage-gated conductance, Ih.
This work was supported by National Institutes of Health Grants MH48432, MH44754, and NS37444 (D.J.). We thank R. Narayanan for technical assistance with the analyses of resonance frequency and the neuron simulations, R. A. Chitwood and J. A. Rosenkranz for helpful discussions, and members of the Center for Learning and Memory for critical review of this manuscript.
- Correspondence should be addressed to Dr. Darrin H. Brager, Center for Learning and Memory, University of Texas at Austin, 1 University Station, C7000, Austin, TX 78712-0805.