I_H current generates the afterhyperpolarisation following activation of subthreshold cortical synaptic inputs to striatal cholinergic interneurons

Abstract

Pauses in the tonic firing of striatal cholinergic interneurons emerge during reward-related learning and are triggered by neutral cues which develop behavioural significance. In a previous in vivo study we have proposed that these pauses in firing may be due to intrinsically generated afterhyperpolarisations (AHPs) evoked by excitatory synaptic inputs, including those below the threshold for action potential firing. The aim of this study was to investigate the mechanism of the AHPs using a brain slice preparation which preserved both cerebral hemispheres. Augmenting cortically evoked postsynaptic potentials (PSPs) by repetitive stimulation of cortical afferents evoked AHPs that were unaffected by blocking either GABA_A receptors with bicuculline, or GABA_B receptors with saclofen or CGP55845. Apamin (a blocker of small conductance Ca²⁺-activated K⁺ channels) had minimal effects, while chelation of intracellular Ca²⁺ with BAPTA reduced the AHP by about 30%. In contrast, blocking hyperpolarisation and cyclic nucleotide activated (HCN) cation current (I_H) with ZD7288 or Cs⁺ diminished the size of the AHPs by 60% and reduced the proportion of episodes that contained this hyperpolarisation. The reversal potential (−20 mV) and voltage dependence of the AHPs were consistent with the hypothesis that a transient deactivation of I_H caused most of the AHP at hyperpolarised potentials, while the slow AHP-type Ca²⁺-activated K⁺ channels increasingly contributed at more depolarised membrane potentials. Subthreshold somatic current injections yielded similar AHPs with a median duration of ∼700 ms that were not affected by firing of a single action potential. These results indicate that transient deactivation of HCN channels evokes pauses in tonic firing of cholinergic interneurons, an event likely to be elicited by augmentation of afferent synaptic inputs during learning.

Introduction

Tonically active neurons (TANs) in the striatum, believed to be cholinergic interneurons, pause their firing activity in response to rewarding or aversive events, and sensory cues that predict these events (Aosaki et al. 1994a; Blazquez et al. 2002; Ravel et al. 2003; Morris et al. 2004; Joshua et al. 2008). In response to sensory stimulation, a pause in firing is initially present in a minority of neurons, but more neurons respond during conditioning, until it is present in a majority of TANs throughout the dorsal striatum in both brain hemispheres (Aosaki et al. 1994b). TANs pause firing when dopaminergic neurons show a phasic increase in activity, providing a period of reduced acetylcholine ‘tone’ during which dopamine release is enhanced and synaptic weights on striatal projection neurons are modifiable (Morris et al. 2004; Rice & Cragg, 2004; Wang et al. 2006; Pakhotin & Bracci, 2007). Elucidating the mechanisms underlying the pause response is therefore a fundamental step towards understanding the processes of reinforcement learning and goal-directed behaviour (Graybiel et al. 1994; Apicella, 2007).

Investigations conducted in striatal cholinergic interneurons have led to the proposal of a number of candidate mechanisms. Firstly, GABA-mediated inhibition via interneurons or collaterals of spiny projection neurons has been proposed (Watanabe & Kimura, 1998), supported by the recent demonstration of recurrent inhibition of cholinergic interneurons. Sullivan et al. (2008) found that nicotinic activation can induce polysynaptic GABA_A-mediated PSPs in cholinergic interneurons. These nicotinic synapses exhibit strong short-term depression in response to repetitive activation (Sullivan et al. 2008), and it remains to be determined if this mechanism contributes to the pause response in vivo. Secondly, since the pause response of TANs is abolished following dopamine depletion (Aosaki et al. 1994a), and activation of D2 dopamine receptors reduces N-type calcium as well as sodium currents in cholinergic interneurons (Yan et al. 1997; Maurice et al. 2004), it has been suggested that phasic dopamine release itself triggers the pause in firing. However, dopamine application has a net depolarising effect on cholinergic interneurons (Aosaki et al. 1998), and direct injection of dopamine antagonists into the striatum does not consistently inhibit the pause response in TANs (Watanabe & Kimura, 1998). Similarly, pause responses are still observed when phasic dopamine responses are absent (Joshua et al. 2008).

Previously, we demonstrated in vivo in a small sample of striatal cholinergic interneurons that a plasticity protocol which potentiated cortical inputs was associated with the appearance of an afterhyperpolarisation and an appropriately timed transient suppression of tonic firing (Reynolds et al. 2004). This raises a third possibility, that a hyperpolarisation driven by converging excitatory inputs may underlie the pause response. Contralateral responses could be evoked at short latency (Reynolds & Wickens, 2004), and hence bilaterally projecting corticostriatal neurons may play a role in mediating and synchronising pause responses between the two hemispheres. These attributes suggest that corticostriatal inputs are responsible for, or, more likely, co-operate with, the more numerous thalamic inputs in the generation of conditioned pauses (Lapper & Bolam, 1992; Matsumoto et al. 2001).

A potential mechanism intrinsic to the cholinergic interneuron that may contribute to the pause response following an excitatory input involves Ca²⁺-activated potassium currents with medium and slow activation kinetics (Goldberg & Wilson, 2005; Wilson & Goldberg, 2006; Deng et al. 2007). The generation of an AHP in response to depolarising events has also been attributed to deactivation of the hyperpolarisation and cyclic nucleotide activated (HCN) cation current (I_H) in lobster olfactory receptor neurons (Corotto & Michel, 1998) and hippocampal pyramidal neurons (Takigawa & Alzheimer, 2003; Otmakhova & Lisman, 2004). The presence of this current is one of the defining characteristics of cholinergic interneurons (Kawaguchi, 1992; Kawaguchi et al. 1995; Tepper & Bolam, 2004). I_H contributes to intrinsic membrane oscillations (Bennett et al. 2000), is regulated by dopamine (Deng et al. 2007) and, when activated at hyperpolarised levels of cell membrane potential, shortens the AHP following a somatic depolarisation (Wilson, 2005; Deng et al. 2007). The aim of the present study was to characterise the AHP induced by an excitatory synaptic event. Here we report that the AHP following a cortically evoked postsynaptic potential (PSP) is due to a transient deactivation of I_H and critically depends on the preceding levels of cortical synaptic excitation.

Methods

Ethical approval

All procedures involving live animals were approved by the University of Otago Animal Ethics Committee in accordance with the New Zealand Animal Welfare Act 1999. In addition, experiments fully comply, in all aspects, with UK regulations and policies for animal experimentation, as clearly stated in a recent article in the Journal of Physiology (Drummond, 2009).

Slice preparation

Acute brain slices were prepared from 85 P14–P24 male Wistar rats following decapitation under deep pentobarbital anaesthesia (100 mg kg⁻¹ intraperitoneal). Brains were perfused transcardially with ice-cold dissection solution (in mm: 225 sucrose, 10 glucose, 2.5 KCl, 7 MgCl₂, 0.5 CaCl₂, 28 NaHCO₃, 1 NaH₂PO₄; bubbled with 95% O₂ and 5% CO₂). Slices (300 to 400 μm) were cut using a vibrating microtome (VT1000S, Leica, Nussloch, Germany) at an oblique angle of ∼30 deg to the horizontal plane, in order to maximally preserve corticostriatal connections (Fig. 1B). Slices were transferred to artificial cerebrospinal fluid (ACSF) consisting of (mm): 125 NaCl, 2 MgSO₄, 2 CaCl₂, 2.5 KCl, 10 glucose, 26 NaHCO₃, 1 NaH₂PO₄; bubbled with 95% O₂ and 5% CO₂. After cutting, slices were kept for the initial 45 min at 35°C, and then at room temperature for a minimum recovery period of 1 h.

Figure 1. Characteristics and site of afferent stimulation of striatal cholinergic interneurons.

A, electrophysiological characteristics of a striatal cholinergic interneuron (inset) include spike accommodation on positive current injection, followed by a pronounced sAHP. Other features include a depolarising ‘sag’ during negative current injection (double arrow) and a rebound depolarisation after the end of the current pulse (asterisk) which sometimes triggers a spike. B, brain slices were sectioned at an oblique angle as indicated by the dashed line (left, sagittal view). Both hemispheres were left intact to stimulate either contralateral or ipsilateral cortical inputs (right, oblique transverse view). Electrical stimulation and recording locations are indicated. Cx: cerebral cortex; R: recording pipette.

Electrophysiological recording

For recording, slices were transferred to a temperature controlled recording chamber (TC-2, Bioscience Tools, San Diego, CA, USA) perfused with oxygenated ACSF (2 ml min⁻¹, 33°C). Striatal cholinergic interneurons were visualised (Fig. 1A) using infrared-differential interference contrast optics (BX51WI, Olympus Optical, Tokyo, Japan) and an IR-1000 CCD camera (DAGE-MTI, Michigan City, IN, USA). Recording pipettes were prepared from borosilicate glass capillaries (1.5 mm outer diameter, 0.86 mm inner diameter; Harvard Apparatus, Edenbridge, UK) using a horizontal pipette puller (P87; Sutter Instruments, Novato, CA, USA). Pipettes had a resistance of 5–7 MΩ when filled with a solution containing (mm): 132 potassium gluconate, 6 KCl, 6 NaCl, 2 Na₂ATP, 0.4 Na₂GTP, 2 MgCl₂ and 10 Hepes (pH 7.4, 290–300 mosmol l⁻¹). The modified (high calcium buffering) pipette solution contained (mm): 112 potassium gluconate, 12 KCl, 5 Na₄BAPTA, 2 Na₂ATP, 0.4 Na₂GTP, 2 MgCl₂ and 10 Hepes (pH 7.4). Current- and voltage-clamp recordings were made in the whole-cell configuration using a MultiClamp 700B amplifier (Molecular Devices, Union City, CA, USA), with series resistance compensation of 12–17 MΩ. Recordings in which the series resistance changed by more than 25% during the recording period were discarded. The signals were lowpass filtered at 4 kHz and digitised at 20 kHz (1322A Digidata and pCLAMP9 acquisition software; Molecular Devices). In hybrid-clamp experiments, the switch between current- and voltage-clamp modes was controlled by a digital pulse via the Digidata interface. The switching command was specified by the voltage setting in the MultiClamp with external commands disabled. Switching to voltage-clamp was triggered 45–200 ms following the peak of the PSP (during the decaying phase of the PSP and before the onset of the AHP) and lasted for 250 ms, before returning to the current-clamp mode. Four to six episodes were acquired for each voltage setting, alternating between PSP and ‘no stimulus’ (control) conditions.

Electrical stimulation and drug application

Postsynaptic potentials were evoked in cholinergic interneurons by electrical stimulation (0.1 Hz) of the cerebral cortex or white matter in the contralateral or ipsilateral hemisphere relative to the recording site in the dorsolateral striatum (Fig. 1B). Biphasic stimuli (0.1 ms duration, up to 1 mA) controlled by a Master-8 pulse generator (A.M.P.I, Jerusalem, Israel) were delivered by a stimulus isolator (A13–75, Coulbourn Instruments, Allentown, PA, USA) via a bipolar stereotrode (∼0.2 MΩ; MicroProbe, Gaithersburg, MD, USA). Unless otherwise stated, all drugs were obtained from Tocris Bioscience (Bristol, UK), dissolved in ACSF on the day of the experiment or diluted from a 1000× stock solution stored at −20°C (apamin and CGP55845), and bath applied. The GABA_A receptor antagonist (−)-bicuculline methochloride (BIC), the GABA_B receptor antagonists saclofen hydrochloride or (2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl](phenylmethyl)phosphonic acid (CGP55845), the SK channel antagonist apamin (Sigma, St Louis, MO, USA), the ryanodine receptor agonist caffeine (Sigma), and the I_H blockers 4-ethylphenylamino-1,2-dimethyl-6-methylaminopyrimidinium chloride (ZD7288) or CsCl were used to examine the relationship between depolarising PSPs and AHPs after 5–10 min wash-in or wash-out periods. The AMPA/kainate receptor antagonist 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX) and the NMDA glutamate receptor antagonist d(−)-2-amino-5-phosphonopentanoic acid (AP5) were used to assess the relative contribution to the PSPs of currents evoked by activation of the respective ion channels.

Data analysis

Resting membrane potential (V_m), peak amplitude responses and PSP latencies were measured in single or averaged traces using Clampfit 9 (Molecular Devices) or Axograph X (AxoGraph Scientific, Sydney, Australia). Cell input resistance (R_IN) was determined from the regression slope of the peak membrane potential in response to 500 ms long step current pulses (−100 to +50 pA). The ‘sag’ potential associated with I_H activation in response to a hyperpolarising step current was defined as the difference between the membrane potential of the peak and the tail of the potential. The magnitude of the slow AHP was measured following membrane depolarisations (25–300 pA, 500 ms), and averaged across neurons according to the number of action potentials fired during the preceding current step. Statistical analysis was performed using Stata 10 (StataCorp LP, College Station, TX, USA) or MATLAB (The MathWorks, Natick, MA, USA) software. Group comparisons were made using Student's t test or ANOVA with P < 0.05 considered significant. A mixed model was used to analyse natural log-transformed means for treatment effects with random effects assigned to individual neurons. Wilcoxon's rank sum test was used for comparing data which were not normally distributed. Values are expressed as means ± standard deviation (s.d.) unless stated otherwise.

A procedure was developed to measure corresponding PSP and AHP areas in single episodes, using MATLAB. Axon binary files were imported using a function written by Dr Michele Giugliano (http://www.mathworks.de/matlabcentral/fileexchange/). Each trace was lowpass filtered (20 Hz; 3rd order Butterworth), down sampled to 1 kHz and high pass filtered (0.2 Hz; 3rd order Butterworth) to exclude slow trends from single episodes. The mean reference potential was computed from three periods in each episode (typically 5–7 s duration): (i) a period immediately before the stimulation; (ii) 2 s after the stimulus; and (iii) after intracellular current injection. Episodes were excluded from analysis if the baseline was unstable, as indicated by the s.d. of the reference potential >0.25 mV, or if an action potential was evoked. Areas under the PSP (mV ms) were measured from the start of the stimulus to the time at which the signal returned to the reference potential. At that point, the AHP was considered to commence. AHP areas were computed in the same manner from the above defined time point to the time at which the signal crossed the reference level again (see Fig. 3). Both PSP and AHP potentials were calculated for a minimum of 100 ms to avoid an early termination of the process due to noise.

Figure 3. Stochastic AHP behaviour following small depolarising PSPs.

A, the AHP (asterisk) was assessed in single episodes and its area (light shading), and that of the preceding PSP (dark shading) measured in single filtered traces. The AHP noise level for each neuron was defined empirically as the 95-percentile of the areas of spontaneous hyperpolarising potentials (arrows). B, stimulation of cortical inputs at 0.1 Hz evoked depolarising PSPs and measurable AHPs were observed only in a small percentage of episodes for most neurons (median 13%, n= 41). C, correlation of the size of the PSP area with the proportion of AHP episodes in the same data set.

For each neuron, the areas of randomly occurring hyperpolarising potentials were computed in episodes without PSPs, or in episodes with PSPs after the signal had returned to baseline. The 95-percentile of a large number (n > 400) of these area measurements defined the noise level (spontaneous membrane potential fluctuations), and was used as the significance threshold for experimentally evoked AHPs. For group comparisons, data from individual episodes were either binned according to the magnitude of the PSP area (bin size = 200 mV.ms), or averaged according to the stimulus number used for each neuron. The percentage of episodes exceeding the AHP significance threshold was calculated for each neuron in bins that contained more than three data points. For evaluation of drug effects on the PSP, the median area of matched stimulus number pairs for each neuron was used and expressed as percentage change of control PSP area. Due to the variability in the magnitudes of the PSP areas obtained across neurons, the natural logarithm of the PSP and AHP areas was used for the regression analysis with a mixed model. For statistical analysis of treatment effects using a t test, normalised AHP areas were derived for each neuron from the ratio of the median AHP and PSP areas.

The hybrid-clamp current traces were lowpass filtered (200 Hz; 3rd order Butterworth) before averaging, and the ‘no stimulus’ control was subtracted from the stimulated average. The average of this difference current was calculated for each voltage setting and neuron over a 150 ms period, following an 80 ms delay after the switch to the voltage clamp mode. Exponential curves were fitted to the difference currents in MATLAB by minimising the sum of squares of errors between the data and a single exponential function Ae^−λt for varying parameters A and λ.

Results

Striatal cholinergic interneurons were visually identified in oblique slices by their relatively large cell body and prominent dendrites (Fig. 1A). Their identity was confirmed by characteristic electrophysiological properties observed in whole-cell current-clamp recordings (Kawaguchi, 1992), including a pronounced slow AHP (sAHP) following a long-lasting (>50 ms) suprathreshold positive current injection, a marked depolarising ‘sag’ during negative current injection (indicative of I_H current), and a rebound depolarisation after cessation of the current pulse (Fig. 1A). In total, recordings were made from 164 cholinergic interneurons (V_m−60 ± 6 mV; R_IN 191 ± 58 MΩ). Potential rundown of voltage- or calcium-activated channels was assessed in prolonged recordings from a subset of neurons that were not exposed to any drugs (n= 14). R_IN, the I_H-associated ‘sag potential’ and the peak amplitude of the sAHP were not significantly different between the beginning and end of the recording period (25–75 min) (Fig. S1 in Supplemental Material; paired t test (R_IN) and two-way ANOVA (I_H-sag and sAHP peak potentials), P > 0.05). This indicates that calcium- and hyperpolarisation-activated currents remained intact in our experimental model over the course of recording.

In order to elicit PSPs that are primarily due to activation of corticostriatal afferents, stimulating electrodes were placed either in the deeper layers of the cerebral cortex or in the underlying white matter (Fig. 1B). In a subset of experiments, crossed corticostriatal fibres were activated selectively by placing the stimulating electrode in the corpus callosum of the contralateral hemisphere. Contralateral stimulation should minimise the antidromic activation of axonal collaterals of thalamostriatal neurons which also project to the ipsilateral cortex (Deschenes et al. 1996) and of ipsilaterally projecting corticostriatal neurons contributing to the pyramidal tract (Landry et al. 1984; Wilson, 1986). Depolarising PSPs evoked from both contralateral (Fig. 2A) and ipsilateral (Fig. 2B) cortical stimulation sites with single pulses were similar in size and relatively small in amplitude (contralateral corpus callosum: 1.7 ± 0.9 mV, n= 17; ipsilateral: 1.2 ± 0.5 mV, n= 24; P= 0.068, two sample t test).

Figure 2. Both NMDA and AMPA/kainate receptors contributed to PSPs in cholinergic interneurons in response to cortical stimulation.

A, representative example of a depolarising PSP elicited by stimulation in the contralateral cerebral hemisphere and its dependence on NMDA (blocked by AP5) and AMPA/kainate (blocked by CNQX) glutamate receptors. B, depolarising PSPs elicited by stimulation in the ipsilateral cerebral hemisphere (averages of 6 traces recorded at a 10 s interval) and corresponding PSP areas recorded over 1 min intervals, illustrating the effect of application of AP5 and CNQX. Representative traces at the time points a, b, c and d are illustrated. C and D, means (±s.e.m.) of normalised PSP areas from 13 (C) and 5 (D) neurons plotted at 1 min intervals. PSP areas were normalised to the mean value of the control PSP areas before drug application. Blockade of either NMDA receptors with AP5 or AMPA/kainate receptors with CNQX significantly reduced the area of the postsynaptic response (one-sample t tests, P < 0.001).

The contribution of glutamatergic receptors to the PSPs was determined using the NMDA antagonist AP5 (50 μm) and the AMPA/kainate antagonist CNQX (10 μm). While most of the response was attributed to activation of AMPA/kainate receptors, a significant contribution was due to NMDA receptors, for both ipsilateral and contralaterally evoked PSPs (Fig. 2A and B). Analysis of the postsynaptic responses to ipsilateral cortical stimulation revealed that blockade of NMDA receptors with AP5 reversibly reduced the area under the PSP to 73 ± 6% of controls (n= 13, P < 0.001, one sample t test, Fig. 2C). In comparison, CNQX dramatically reduced the area of the PSP to 15 ± 6% of controls (n= 5, P < 0.001, one sample t test, Fig. 2D). Together, these data show that corticostriatal stimulation produces glutamatergic PSPs in striatal cholinergic interneurons, and that a significant proportion of the depolarisation is mediated by NMDA channels even for relatively small PSPs.

We have previously observed in a small sample of cholinergic interneurons recorded in vivo that cortical stimulation can produce subthreshold depolarising PSPs which may be followed by small AHPs (Reynolds et al. 2004). In the present study, we wished to measure reliably these AHPs in single episodes, and to determine in a large sample of neurons the consistency with which AHPs are elicited by subthreshold PSPs. We therefore developed a procedure for quantifying the areas under PSPs and any ensuing AHPs (see Methods and Fig. 3A). Only hyperpolarisations with an area that exceeded the 95-percentile noise level of spontaneous hyperpolarisations were considered to represent synaptically evoked AHPs. Using this criterion, measurable AHPs were observed only in a small percentage of episodes during recording from individual neurons (median 13%, n= 41, Fig. 3B). The median percentage of episodes with an AHP in the contralateral corpus callosum (17%, n= 17) and ipsilateral (12%, n= 24) stimulation groups did not differ significantly (P= 0.47, Wilcoxon's rank sum test). A higher proportion of AHP episodes was observed in neurons exhibiting larger PSP areas (Fig. 3C), suggesting that the stochastic appearance of the AHP in particular episodes was linked to the strength of the preceding PSP. Because there were no apparent differences in the nature of the PSP and AHP evoked from stimulation in the contralateral and ipsilateral hemispheres, an ipsilateral stimulation site was used mainly in subsequent experiments and the data of both groups were pooled for analysis.

In addition to excitatory synaptic inputs from the cortex and thalamus (Lapper & Bolam, 1992; Dimova et al. 1993; Thomas et al. 2000), cholinergic interneurons receive inhibitory inputs that may originate from spiny projection neurons or GABAergic interneurons of the striatum (DiFiglia, 1987; Martone et al. 1992; Reynolds & Wickens, 2004). As PSPs evoked from single stimuli were relatively small, we attempted to increase the strength of the synaptic input using short trains of stimuli (Fig. 4A). Repetitive stimulation (2–10 pulses; 100 Hz) increased the magnitude of the PSP within the range obtained in vivo (Reynolds et al. 2004) and triggered an AHP in all neurons examined (n= 44). The magnitude of the AHP was directly proportional to the magnitude of the preceding PSP (Fig. 4B). To determine if fast inhibitory PSPs contributed to AHPs under these circumstances, GABA_A receptors were blocked with BIC (30 μm) in separate experiments. Repetitive stimulation in the presence of BIC produced very similar results (Fig. 4C). The log-transformed mean data were best fitted by parallel lines for each group, using a mixed model regression analysis. This confirmed the direct relationship between PSP and AHP magnitudes, and showed that the AHP was not reduced in the presence of BIC (Fig. 4D; see Fig. S2 in Supplemental Material for time course example). The AHP areas were identical in both groups when normalised to the preceding PSP area (Table 1). In addition, the proportion of episodes in which AHPs were observed following PSPs in response to stimulus trains of increasing pulse number increased at a similar rate in the presence or absence of BIC (Fig. 4E). These observations indicate that GABA_A receptors were not activated in our synaptic stimulation protocol and are unlikely to contribute to the AHPs.

Figure 4. Augmenting the depolarising PSP produced an increase in the size of the AHP that was not mediated by inhibitory GABAergic synapses.

A, representative example illustrating how repetitive stimulation increased the magnitude of the PSP and associated AHP when the number of pulses in the afferent stimulus train were increased. B and C, single episode data demonstrating the linear relationship between AHP and PSP areas in individual neurons in the absence (B) or presence (C) of the GABA_A receptor antagonist bicuculline (30 μm). Open circles represent mean areas according to the number of stimuli applied in a train. Linear regression fits of the AHP to PSP areas were performed on single episode data of individual neurons (P < 0.001). The 95-percentile levels for noise in the AHP area measurement are indicated (dotted line). Insets show representative examples of averaged traces for each neuron (B: 3 stimuli, C: 5 stimuli, scale bars: 5 mV and 200 ms). D, group analysis using a mixed model regression of the log-transformed AHP and PSP areas, with random effects assigned to individual neurons, and performed on mean areas shows that bicuculline had no effect on the AHP (P= 0.91). E, BIC did not alter the mean proportion of episodes exhibiting an AHP for a given PSP area (two-way ANOVA, P= 0.16). F–H, blockade of GABA_B receptors with either saclofen (SCF, n= 3) or CGP55845 (CGP, n= 4 neurons) similarly had no effect on the AHP. F, example showing the linear relationship between mean PSP and AHP areas from a single neuron before (control) and during drug application. Average traces for each condition are shown above for 5 stimuli (vertical lines indicate stimulation times; scale bars: 2 mV and 200 ms). G and H, group analysis of the log-transformed means (G, mixed model regression, P= 0.071), and the proportion of AHP episodes (H, two-way ANOVA, P > 0.56) showed that GABA_B receptors did not contribute to the AHP.

Table 1.

AHP areas normalised to preceding area of depolarisation (mean ±s.d.) for each experimental group

	n	Control	Drug	P*
BIC		0.58 ± 0.31	0.57 ± 0.24	0.87
		(n= 20)	(n= 24)
SCF/CGP5584‡	7	0.52 ± 0.10	0.51 ± 0.17	0.91
AP5	6	0.56 ± 0.13	0.65 ± 0.25	0.27
Apamin	6	0.42 ± 0.25	0.66 ± 0.32	0.093
Caffeine	5	0.41 ± 0.16	0.70 ± 0.53	0.19
BAPTA†	5	0.67 ± 0.33	0.41 ± 0.16	0.057
ZD7299	8	0.83 ± 0.43	0.33 ± 0.23	0.003
Cs+ (PSP)	7	0.52 ± 0.21	0.21 ± 0.17	0.025
Cs+ (ICI)§	6	0.54 ± 0.11	0.20 ± 0.10	0.001

The control groups were not different (one-way ANOVA, P= 0.16).

^*P-value of a two-sample t test (BIC) or paired t test (all other groups).

^‡Pooled data from saclofen (SCF, n= 3) and CGP55845 (n= 4) treated neurons.

^†BAPTA control data: first 8 min of stimulation; drug data: 9–16 min of stimulation.

^§Depolarisations induced by intracellular current injection (ICI).

Striatal cholinergic interneurons also contain metabotropic GABA_B receptors (Yung et al. 1999) that produce a membrane hyperpolarisation when activated (Pisani et al. 1999). In addition to being independent of GABA_A receptor activation, the AHPs which followed subthreshold PSPs were not blocked by GABA_B receptor antagonists. The relationship between PSP and ensuing AHP areas, augmented by increasing the number of pulses in a stimulus train, was unchanged in the presence of 1 μm CGP55845 (n= 4; Fig. 4F) or 50 μm saclofen (n= 3). Statistical analysis of both the log-transformed and PSP area-normalised data confirmed that the AHP magnitude was not changed before and after GABA_B antagonists (Fig. 4G, Table 1). In addition, as successively larger PSPs were evoked there was a similar proportion of episodes that contained significant AHPs whether in the presence or absence of GABA_B receptor antagonists (Fig. 4H). Together, these data show that GABA receptor mechanisms are unlikely to be involved in the AHPs induced in cholinergic interneurons following subthreshold cortically evoked PSPs.

Calcium influx via NMDA receptors may trigger small-conductance (SK-type) or slow-decaying (sAHP-type) Ca²⁺-activated K⁺ currents underlying medium-duration and long-lasting AHPs, respectively (Shah & Haylett, 2002; Ngo-Anh et al. 2005). Since NMDA receptor currents contributed significantly to PSPs elicited in the present study (Fig. 2) and both types of Ca²⁺-activated K⁺ currents are functional in cholinergic interneurons (Goldberg & Wilson, 2005), we tested the contribution of each of these conductances to AHPs associated with subthreshold cortically evoked PSPs. Firstly, blocking NMDA receptors with AP5 (50 μm) on average reduced the area of PSPs evoked by stimulus trains to 83 ± 23% of controls (n= 6; see Fig. S3A in Supplemental Material for time course example). However, the relationship between AHP and PSP areas (Fig. 5A and B; Fig. S3A; Table 1) and the proportion of episodes that contained significant AHPs (Fig. 5C) were unaffected by blocking NMDA receptors. Thus calcium influx via the NMDA receptor is not essential by itself for triggering the AHPs that follow cortically evoked PSPs.

Figure 5. The AHP following subthreshold activation of cortical inputs was not diminished by blockade of NMDA receptors with AP5 (A–C), block of SK calcium-activated potassium channels with apamin (D–F), or depletion of intracellular calcium stores with caffeine (G–I), but was reduced after buffering intracellular Ca²⁺ with BAPTA (J–L).

Correlations of PSP magnitude with the associated AHP before and after drug application are shown for each experimental group, representing single neuron examples (A, D, G and J). Data points are mean areas of episodes grouped by the number of stimuli in a train. Average traces for each condition are shown above for 4 stimuli (scale bars: 5 mV and 200 ms in A; 2 mV and 200 ms in D, G and J). Group analyses of the log-transformed means suggest that apamin and caffeine treatments enhanced, and intracellular BAPTA diminished, the AHP (mixed model regression, P= 0.46 in B, ***P < 0.001 in E, H and K). C, F, I and L, mean proportion of AHP episodes calculated for increasing PSP areas before and after drug application, and following drug washout. None of the treatments had any effect on the proportion of AHP episodes (two-way ANOVA; P= 0.89 in C, 0.15 in F, 0.81 in I, and 0.30 in L).

Secondly, bath application of 100 nm apamin, which selectively blocks SK-type Ca²⁺-activated K⁺ currents (I_SK), had no effect on the PSP areas associated with different stimulus train durations (95 ± 22% of controls, n= 6) and the corresponding AHPs on first inspection appeared unchanged (Fig. 5D). However, the group analysis of the log-transformed means showed that the AHP areas were increased significantly in the presence of apamin (mixed model regression, P < 0.001; Fig. 5D and E). This trend was also apparent when AHP areas were normalised to the preceding PSP area, although it did not reach significance in the paired t test analysis (Table 1). The proportion of episodes with an AHP was increased for relatively small PSPs, but there was no significant difference over the full range (two-way ANOVA, P= 0.15; Fig. 5F). This suggests that activation of apamin-sensitive SK channels has a shunting effect on the PSP but I_SK does not contribute by itself to the AHP evoked by subthreshold synaptic stimulation.

Thirdly, to assess whether sAHP-type Ca²⁺-activated K⁺ currents (I_sAHP) contribute to the synaptically evoked AHPs, endoplasmic reticulum calcium stores were depleted by stimulating ryanodine receptors with 10 mm caffeine, or intracellular Ca²⁺ was chelated with 5 mm BAPTA. These treatments have been shown to substantially reduce the I_sAHP in cholinergic interneurons (Goldberg & Wilson, 2005; Wilson & Goldberg, 2006). In our hands, 10 mm caffeine reduced the tail current of I_sAHP induced by a prolonged somatic depolarisation of 1 s to 0 mV to 62% of controls (Fig. S4 in Supplemental Material). Interestingly, long-duration caffeine treatment increased the magnitude of the PSP area associated with different stimulus strengths (1–11 pulses per train, see Fig. S3B for time course) by 166 ± 93% (n= 5), and significantly increased the AHP areas at similar levels of control PSP magnitudes (Fig. 5G and H; mixed model regression, P < 0.001). Again, this trend was not significant when PSP area-normalised AHP areas were compared in a paired t test analysis before and after the drug (Table 1).

Since I_sAHP can be activated directly by Ca²⁺ influx through voltage-dependent calcium channels, we performed synaptic stimulation with 5 mm BAPTA included in the pipette solution to assess the effect of chelating intracellular Ca²⁺. PSP and AHP areas were measured over repeated cycles of synaptic stimulation of increasing strength and the data binned into 8 min blocks (Fig. S5 in Supplemental Material). BAPTA perfusion decreased AHP areas at similar levels of PSP magnitudes during the 9–16 min period, compared to the initial 8 min (Fig. 5J and K; mixed model regression, P < 0.001, n= 5). The decrease in AHP areas normalised to preceding PSP area over the same time period reached borderline significance (Table 1; paired t test, P= 0.057). While PSP and AHP areas continued to gradually decrease beyond this time as a result of intracellular Ca²⁺ chelation, normalised AHP areas did not further decline and stabilised at 66 ± 20% (Fig. S5). Control measurements at the beginning and end of the BAPTA recording period showed that I_sAHP was decreased to 9 ± 6% of the initial level, while input resistance and the hyperpolarisation-induced sag potential remained largely unchanged (Fig. S6 in Supplemental Material). The proportion of episodes with an AHP for different PSP areas was not changed by either the caffeine or the BAPTA treatment (Fig. 5I and L). Together, these results suggest that I_sAHP contributed only partially to the synaptically induced AHPs.

We next examined if the AHP evoked by subthreshold synaptic stimulation was mediated by the temporary inactivation of hyperpolarisation and cyclic nucleotide-activated (HCN) channels that mediate I_H. We noted that AHPs were triggered efficiently over a range of membrane potentials consistent with deactivation of I_H following a small-magnitude depolarising event (Takigawa & Alzheimer, 2003; Otmakhova & Lisman, 2004). Application of the I_H channel blocker ZD7288 (30 – 50 μm) hyperpolarised the resting membrane potential of cholinergic interneurons by −9.3 ± 3.5 mV, indicating that the I_H inward current was active at rest. A depolarising holding current was therefore applied to compensate for the shift in the resting membrane potential in the presence of ZD7288. Consistent with previous reports that I_H activation has a shunting effect on PSPs (Berger et al. 2001; Angelo et al. 2007), ZD7288 application increased the average magnitude of the PSP area associated with different stimulus strengths (1–11 pulses per train) to 167 ± 63% of controls (n= 8). At the same time AHP areas were strongly reduced in all neurons over the range of stimulus strengths (Fig. 6A and B; mixed model regression, P < 0.001). Similarly, AHP areas normalised to preceding PSP area were significantly reduced in the presence of ZD7288 (Table 1; paired t test, P < 0.01). In addition, there was a strong reduction in the proportion of episodes with an AHP in the presence of the blocker (two-way ANOVA, P < 0.001), even for PSPs of large magnitudes (Fig. 6C). This effect did not reverse on washout of the drug, in agreement with previous findings that the ZD7288 block of HCN channels is irreversible (Harris & Constanti, 1995; Gasparini & DiFrancesco, 1997).

Figure 6. Blockade of I_H with ZD7288 (A–C) or Cs⁺ (D–F) greatly diminished the AHP evoked by subthreshold activation of subthreshold cortical inputs.

Correlations of PSP magnitude with the associated AHP before and after drug application are shown for each experimental group, representing single neuron examples (A and D). Data points are mean areas of individual episodes grouped by the number of stimuli in a train. Average traces for each condition are shown above (scale bars: 1 mV and 200 ms in A, 6 stimuli; 2 mV and 200 ms in D, 4 stimuli). B and E, group analysis of the log-transformed means (mixed model regression, ***P < 0.001). C and F, the proportion of AHP episodes for increasing PSP areas was substantially reduced following ZD7288 (note the irreversible nature of ZD7288 drug action) or CsCl application to the bath solution (two-way ANOVA, P < 0.001). Cs⁺ blockade of I_H channels was dose dependent and fully reversible upon washout (Bonferroni's post hoc correction, *P < 0.05, ***P < 0.001).

Corticostriatal stimulation in the presence of 1–2 mm Cs⁺, an alternative blocker of I_H, produced similar results to those obtained with ZD7288 (Fig. 6D and E). While PSP areas induced by differing stimulus strengths (1–13 pulses per train) increased to 112 ± 42% of controls (n= 7), AHPs were greatly reduced in the presence of Cs⁺ (mixed model regression, P < 0.001). This result was confirmed by the paired t test analysis of the normalised AHP areas (Table 1; P= 0.025). Blocking I_H with 1 and 2 mm Cs⁺ revealed a dose-dependent reduction in the proportion of episodes that contained an AHP (two-way ANOVA, P < 0.001), an effect which unlike ZD7288 was fully reversed on washout (Fig. 6F). Hence the results obtained by blocking HCN channels either with ZD7288 or Cs⁺ suggest that I_H deactivation significantly contributes to the synaptically induced AHP.

HCN channels generate a mixed cation current with a characteristic reversal potential near −20 mV (Stieber et al. 2005). In order to further characterise the underlying AHP conductance activated in response to a subthreshold membrane depolarisation, we applied a hybrid-clamp protocol to measure the kinetics and determine the reversal potential of the underlying AHP current (Fig. 7). Corticostriatal fibres were stimulated with pulse trains to evoke large subthreshold PSPs (mean amplitude 6.9 ± 4.1 mV, resting membrane potential −66 ± 6 mV, n= 5), but instead of recording the ensuing AHP the acquisition was switched to a voltage-clamp protocol during the decaying phase of the PSP (Fig. 7A). By subtracting the averaged current responses of ‘no-stimulus’ controls from the PSP current response over a range of holding potentials from −100 to −35 mV, a slowly activating current became apparent (Fig. 7C). The average value for each difference current when plotted against voltage steps reversed at −20 mV (Fig. 7D). Although the ‘difference’ currents were small, they were expected to fall in this range as a 6 pA current can be predicted from Ohm's law for an AHP amplitude of 1.2 mV and an R_IN of 200 MΩ. Time constants of single exponential fits to the difference current data ranged from 17 to 156 ms. When plotted (Fig. 7E), these time constants revealed a profile that resembled activation (for hyperpolarising voltage steps) and deactivation (for depolarising steps) time constants of HCN currents. Thus, the estimated reversal potential and voltage dependence of the apparent AHP current were consistent with our hypothesis that I_H deactivation significantly contributes to the synaptically induced AHP. We therefore conclude that the net current difference measured during the period when the AHP develops is at least partly due to the transient deactivation of I_H in response to the preceding membrane depolarisation.

Figure 7. The synaptically evoked I_AHP reversal potential and voltage dependence match electrophysiological properties of I_H.

A, graphical representation of the hybrid-clamp protocol showing voltage traces before, during and after a transient voltage-clamp to −85 mV. Top trace: voltage-clamp applied during the decay phase of a PSP evoked by cortical stimulation; bottom trace: voltage-clamp applied in the non-stimulated control condition. B, averaged current responses of a single cholinergic interneuron to a range of voltage settings (−100 to −35 mV) during the voltage-clamp period. C, difference (grey) between the PSP and control current responses shown in B and single exponential fits (black) for each voltage step. Zero current levels are indicated by the horizontal dotted lines. The mean current for each voltage step was calculated over a 150 ms period indicated by the vertical dashed lines. D, reversal potential analysis of the relationship between the magnitude of the difference currents and the membrane potential determined (n= 5). E, time constants of single exponential fits to the difference currents for each step potential (mean ±s.e.m., n= 5). The mean resting membrane potential of the 5 neurons (−66 ± 6 mV) is indicated by the vertical dashed line.

In cortical pyramidal neurons I_H channel density in dendrites increases with distance from the soma (Williams & Stuart, 2000; Berger et al. 2001). The somatodendritic distribution of I_H channels in cholinergic interneurons is unknown but it is possible that I_H channels are selectively coupled to glutamatergic inputs of cortical synapses located preferentially at distal dendrites (Thomas et al. 2000). Therefore we tested if subthreshold membrane depolarisations induced by somatic current injections are also followed by an AHP and if this is mediated by deactivation of I_H. Membrane depolarisations induced by brief current pulses of varying amplitude applied to the soma via the patch-pipette (Fig. 8A) generated reliable AHPs, with the magnitude directly proportional to the degree of the preceding membrane depolarisation (n= 26; Fig. 8B). This relationship was highly reproducible and did not differ from the synaptic stimulation group (Fig. 8C; mixed model regression analysis of log-transformed data, P= 0.41), suggesting that somatodendritic HCN channel distribution in cholinergic interneurons is relatively uniform, similar to cerebellar Purkinje cells (Angelo et al. 2007). Furthermore, the hidden ‘sag potential’ obtained by subtracting ZD7288 average waveforms from controls in response to either synaptic stimulation, or somatic current injection, was found to follow a similar time course in both conditions (Fig. S7 in Supplemental Material). As in the synaptic stimulation group, the AHP was greatly reduced in current injection-induced stimulation trials following I_H blockade with 2 mm Cs⁺ (Fig. 8D; mixed model regression, P < 0.001, n= 6). This was confirmed by the paired t test analysis of the normalised AHP areas (Table 1; P < 0.01). Therefore we conclude that both somatic and dendritic depolarisations are equally coupled to I_H.

Figure 8. Somatic current injection induced a Cs⁺-sensitive AHP that was not significantly affected by the firing of a single action potential.

A, representative example illustrating the current-clamp command to generate subthreshold membrane depolarisations similar to those resulting from synaptic stimulation. B, single episode data demonstrating the linear relationship between AHP and the current-induced depolarisation areas in individual neurons (linear regression, P < 0.001). C, group analysis of the log-transformed means shows that both the synaptic stimulation and current injection methods were similarly effective in evoking AHPs (P= 0.41). D, block of I_H with Cs⁺ significantly reduced the intracellular current-induced AHP as demonstrated by the group analysis of the log-transformed means (mixed model regression, ***P < 0.001). E, averaged traces of a representative neuron (*step current of 100 pA, 50 ms) showing the effect of suprathreshold stimulation on the AHP using intracellular current steps set to a level that induced action potential discharge in a proportion of trials. F–I, group analysis of the effects of action potential discharge on AHP area (F and H) and AHP duration (G and I). Vectors represent the direction of change in individual neurons where the origin represents the mean of subthreshold and the arrowhead suprathreshold trials (n= 10). On average (grand means shown as black vectors) there was a tendency towards an increase in AHP area (F and H; +paired t test, P= 0.087) but not in AHP duration (G and I; paired t test, P= 0.29) following suprathreshold current injection.

A single action potential frequently precedes the in vivo pause response without a marked increase in firing probability (Kimura et al. 1984; Aosaki et al. 1995; Apicella, 2002). Since the medium-duration AHPs associated with single action potentials in cholinergic interneurons have relatively fast time constants when compared to I_H (Bennett et al. 2000), we also tested to what extent a suprathreshold stimulus alters the AHP. Intracellular current injections set to the threshold for triggering a single action potential in only a proportion of episodes allowed for an assessment of the action potential contribution to the AHP (Fig. 8E). Action potentials added an early component to the AHP that resembled activation of I_SK but, on average, the AHP area (Fig. 8F and H; paired t test, P= 0.087, n= 10), as well as AHP duration (Fig. 8E, G and I; paired t test, P= 0.29), was not affected by the presence or absence of the action potential.

A similar relationship between depolarisation area and AHP duration was apparent using synaptic stimulation. The AHP duration measured in cholinergic interneurons (n= 20) in response to subthreshold stimulation of cortical inputs with single or multiple pulses increased at a near linear rate from 276 ± 42 ms following small PSPs to 796 ± 116 ms at PSP areas of 1200–1600 mV.ms. At comparable PSP areas (1313 ± 710 mV.ms), the occurrence of a single action potential in a subset of neurons did not significantly change the AHP duration (782 ± 117 ms, n= 4). Hence the duration of the AHP is directly proportional to the preceding membrane depolarisation and deactivation of I_H and is little influenced by evoked action potentials.

Discussion

We aimed to characterise the conditions under which a cortically evoked depolarisation below the threshold for action potentials leads to the generation of AHPs in striatal cholinergic interneurons. We found that AHPs were triggered by relatively small PSPs elicited by stimulation of ipsilateral or contralateral corticostriatal fibres. However, when evoked by single stimuli, AHPs were present only in a small proportion of all episodes suggesting that the underlying conductance was activated stochastically by depolarising synaptic inputs. In contrast, when stimulus trains were used, the AHPs were evoked more reliably and their magnitude was directly proportional to the size of the preceding depolarisation. The temporal and voltage-dependent characteristics of the AHPs and their sensitivity to antagonists suggest that the AHPs were mainly mediated by transient deactivation of HCN (the channels underlying I_H) and to a lesser degree by the Ca²⁺-dependent I_sAHP. Thus, an AHP resulting from deactivation of an I_H conductance that is active at rest can be engaged in cholinergic interneurons, but becomes effective in evoking sufficient hyperpolarisation only if preceded by excitatory synaptic inputs of sufficient magnitude.

Cortical inputs to striatal cholinergic interneurons

Corticostriatal connectivity is limited in in vitro preparations and, to our knowledge, this is the first demonstration of functional corticostriatal connections across hemispheres in an acute slice preparation. Stimulating in the contralateral hemisphere minimises the unwanted effects of local current spread to the recording site, thus avoiding the direct release of neurotransmitters such as dopamine which might modulate the excitability of cholinergic interneurons (Bennett & Wilson, 1998; Maurice et al. 2004; Deng et al. 2007). Using contralateral stimulation in a subset of experiments also allowed us to activate crossed corticostriatal fibres in isolation from both ipsilateral pyramidal tract fibres and projections from the posterior intralaminar thalamic nuclei that innervate the striatum via collateral projections (Deschenes et al. 1996; Zheng & Wilson, 2002). Thus we were able to show that the postsynaptic NMDA receptor current component described previously (Kawaguchi, 1992), but evoked by stimulation within the striatum, is present in corticostriatal synapses of cholinergic interneurons. A detailed analysis of synaptic response parameters will be required to determine if crossed corticostriatal synapses differ in any way from those of pyramidal tract fibres. Importantly, however, PSPs evoked from contra- and ipsilateral stimulation sites were followed by AHPs that persisted in the presence of antagonists of both GABA_A and GABA_B receptors. Therefore, the hyperpolarisation was not due to direct activation of GABAergic fibres in the striatum or to disynaptic inhibitory PSPs, as previously reported following ipsilateral stimulation of the corpus callosum (Suzuki et al. 2001). We conclude that the observed hyperpolarisation represented an intrinsically generated AHP following excitatory corticostriatal PSPs.

Role of calcium-activated potassium currents in AHP mechanism

One channel class known to generate an AHP in response to a membrane depolarisation are the Ca²⁺-activated potassium channels. These channels are functionally expressed in cholinergic interneurons (Bennett et al. 2000; Goldberg & Wilson, 2005; Wilson & Goldberg, 2006). A Ca²⁺-activated potassium current is thought to generate the prominent slow AHP (I_sAHP) in response to a suprathreshold membrane depolarisation in cholinergic interneurons (Fig. 1; see also Wilson & Goldberg, 2006) and other cell types (Sah & Faber, 2002; Vogalis et al. 2003). I_sAHP is characterised by slow Ca²⁺-dependent and voltage-insensitive activation (time to peak of several hundreds of ms), and a decay time constant of 1–2 s (Sah & Faber, 2002). However, a dependence of I_sAHP on membrane potential (V_half=−43 mV) is apparent in striatal cholinergic interneurons (Wilson & Goldberg, 2006), due to the selective coupling of I_sAHP activation to Ca²⁺ influx through L-type channels and calcium release from intracellular stores (Goldberg & Wilson, 2005). The fact that buffering of intracellular Ca²⁺ with BAPTA diminished the AHP by about one-third of its magnitude, while apamin blockade of I_SK had no effect, suggests that I_sAHP contributes to the AHP following a subthreshold membrane depolarisation. However, it is possible that BAPTA also affected the subthreshold regulation of HCN channels. For instance, the activity of signalling enzymes such as diacylglycerol kinase that facilitate HCN channel gating by regulating the formation of acidic lipids (Fogle et al. 2007) is highly dependent on intracellular calcium levels (Jiang et al. 2000). The AHP reversal potential of −20 mV argues against a major Ca²⁺-dependent potassium current component as this outward current would be expected to reverse near −90 mV and increase linearly at successively more depolarised levels of the membrane potential (Storm, 1989). Thus, I_H is likely to account for most of the AHP at hyperpolarised membrane potentials and I_sAHP increasingly contributes at more depolarised membrane potentials at which L-type calcium channels open (Wilson & Goldberg, 2006).

The increase in PSP and AHP magnitudes in response to caffeine exposure and apamin blockade of I_SK are intriguing. It is possible that some of the effects of caffeine were mediated by its action as an antagonist of adenosine A₁ and A_2A receptors (Fredholm et al. 1999). These receptors are expressed in a majority of cholinergic interneurons and endogenous adenosine activity in striatal brain slices has been demonstrated (Preston et al. 2000). However, it is also possible that stimulation of corticostriatal synapses normally activates SK channels in dendrites of cholinergic interneurons, thereby shunting the cortically evoked PSPs, and caffeine or apamin acts to reduce this shunt. Apamin-sensitive SK channels located within dendrites of dopaminergic neurons can be activated by calcium released from caffeine sensitive intracellular stores (Fiorillo & Williams, 1998). In addition, NMDA receptors and SK channels are colocalised in dendrites of CA1 pyramidal hippocampal neurons (Ngo-Anh et al. 2005) and in the lateral amygdala (Faber et al. 2005). It has been shown that SK channel activation via calcium influx from NMDA receptors has a shunting effect on excitatory PSPs, and apamin blockade of I_SK augments glutamatergic PSPs (Faber et al. 2005; Ngo-Anh et al. 2005; Lin et al. 2008). Thus the increase in the area of PSPs induced by corticostriatal afferent stimulation seen as a result of caffeine-induced calcium store depletion, as well as the increase in AHP areas by apamin, may both be attributed to reduced SK channel activation in dendrites of cholinergic interneurons.

SK channels are activated by calcium influx during action potentials and underlie the medium AHP component that modulates the firing frequency of cholinergic interneurons (Bennett et al. 2000; Goldberg & Wilson, 2005). Activation and deactivation time constants for I_SK typically vary from 5 to 15 and 50 to 200 ms, respectively (Sah & Faber, 2002; Vogalis et al. 2003). In comparison, I_H deactivation time constants obtained from voltage-clamp experiments in cholinergic interneurons are >200 ms (Deng et al. 2007). Thus the AHP due to I_H deactivation should outlast I_SK following an action potential, which was confirmed in our study. Using intracellular current injection to induce membrane depolarisation, an early AHP component was present when action potentials were triggered, but the late component and the total duration of the AHP remained unchanged. Thus, if an action potential is triggered consistently from an excitatory PSP the following interspike interval should be longer than the average interspike interval during spontaneous spiking. This is consistent with descriptions of the TAN pause response in monkeys where a single action potential frequently precedes the pause in firing activity in response to a cued stimulus (Kimura et al. 1984; Aosaki et al. 1995; Apicella, 2002).

Role of HCN channel in AHP mechanism

HCN channels open in response to a membrane hyperpolarisation and carry a mixed cation current (I_H) underlying the characteristic depolarising ‘sag’ response in current-clamp recordings (Fig. 1). HCN channels activate and deactivate slowly in response to hyperpolarising and depolarising shifts of the membrane potential, producing a depolarising and hyperpolarising current response, respectively. Notably, an AHP due to a transient deactivation of HCN channels may be easily confused with the AHP associated with a Ca²⁺-dependent potassium current as both processes have a similar time course (Storm, 1989; Williams & Stuart, 2000; Sah & Faber, 2002; Vogalis et al. 2003; Angelo et al. 2007).

The results presented here strongly argue that HCN channels indeed deactivate in response to subthreshold membrane depolarisation and substantially contribute to the generation of AHPs following an activation of glutamatergic corticostriatal synapses. Neither apamin, an antagonist of Ca²⁺-activated SK-type potassium channels, nor targeting sAHP-type Ca²⁺-activated potassium channels by depletion of endoplasmic reticulum calcium stores with caffeine, produced a measurable reduction of the synaptically induced AHP. This is in contrast to the effects of blocking I_H with either Cs⁺ or ZD7288, which almost completely abolished the AHP response to depolarising PSPs. A previous study has shown that ZD7288 and Cs⁺ at the concentrations used here blocks 91 and 80% of I_H current in cholinergic interneurons, respectively (Deng et al. 2007).

The mechanism by which the voltage-dependent deactivation of I_H in response to a membrane depolarisation generates an AHP can be compared to the antagonising action of I_H on the temporal summation of PSPs (Williams & Stuart, 2000; Berger et al. 2001; Angelo et al. 2007). As I_H increasingly deactivates with successive PSPs during a stimulus train, a (hidden) membrane hyperpolarisation develops that antagonises the summation of PSPs and becomes apparent only on cessation of the stimulation. At this time point, membrane hyperpolarisation activates I_H current and causes a slow return to the resting membrane potential in the absence of the depolarising input (see Supplemental Fig. S7). An AHP generated in this way appears indistinguishable from an AHP resulting from a Ca²⁺-activated potassium current, though the underlying ionic mechanisms are completely different. The observed hyperpolarisation and increase in PSP area magnitudes following I_H blockade are consistent with this mechanism.

Our conclusion that a transient deactivation of I_H current is the main mechanism underlying AHPs evoked by subthreshold membrane depolarisations is supported by the apparent reversal potential of I_AHP, determined by switching to a voltage-clamp protocol during the decay phase of the PSP. Such hybrid-clamp protocols have been used previously to measure slow conductance changes in response to a physiological event such as an action potential (Pennefather et al. 1985; Storm, 1989; Dietrich et al. 2002). We applied a similar technique to measure conductance changes that follow depolarising PSPs. The time constants of the resulting ‘difference’ currents between stimulated and non-stimulated episodes were shorter by a factor of about five in comparison to those reported for I_H activation and deactivation time constants in rat cholinergic neurons (Deng et al. 2007), although followed a similar voltage dependence. The discrepancy is due to the fact that we measured the effect of synaptic stimulation on the difference in HCN channel availability for activation/deactivation in response to a voltage step, rather than measuring the time course of the conductance change directly. Kinetic constants of voltage-dependent channels (including HCN) are typically measured directly from voltage steps that follow a defined holding potential (Santoro et al. 2000; Williams & Stuart, 2000; Angelo et al. 2007; Deng et al. 2007). This was not possible in the current set-up involving synaptic stimulation.

Further support for a significant contribution of I_H to the AHP stems from an analysis of the time course of I_H contribution to the PSP in relation to known and modelled distributions of I_H conductances. The effect of somatodendritic I_H distribution and input location has been modelled in detail by Angelo et al. (2007). These authors demonstrate that, depending on I_H channel distribution, differences in response magnitudes develop locally at the site of stimulation. However, cable properties of dendrites impose differential filtering on the PSP compared to the relatively slower AHP, with the result that PSPs of distal inputs strongly attenuate compared to proximal inputs while the AHP is less affected. The model shows that the propagated hidden ‘sag potential’ measured at the soma in response to distal compared to proximal current injection is significantly larger only if HCN channels are concentrated distally or when HCN channel density increases linearly towards the distal dendrites. This effect is relatively weak if HCN channels are uniformly distributed in the somatodendritic domain. We were not able to detect any significant difference in AHP areas between somatic current injection and synaptic stimulation (reported in the literature to target distal dendrites). As shown in Fig. S7, the time course of the ‘hidden sag’ potential in response to both synaptic or somatic stimulation is relatively similar. This suggests that I_H channels are distributed relatively uniformly in the somatodendritic domain of cholinergic interneurons. However, to examine this in detail one would need to use identical stimulus protocols at distal and proximal sites in order to compare both conditions quantitatively. Measuring I_H directly in membrane patches along the somatodendritic axis would be the preferred method to assess HCN channel distribution in cholinergic interneurons.

Functional significance

Our data overall suggest that any mechanism which potentiates the size of excitatory synaptic responses in striatal cholinergic interneurons could engage a level of depolarisation necessary for inducing a subsequent AHP. During reward-related learning, it is likely that such potentiation occurs via a dopamine D5 receptor-dependent mechanism (Suzuki et al. 2001; Bonsi et al. 2004). If sufficient to evoke bursts of action potentials, PSPs potentiated in this manner would engage the I_sAHP (Reynolds et al. 2004; Deng et al. 2007), resulting in a prolonged period of reduced probability of action potential firing. Alternatively, PSPs that remain subthreshold or fire just a single action potential after potentiation, would evoke the I_H-dependent AHP and a pause in tonic firing (Reynolds et al. 2004) of a duration more consistent with firing pauses observed in behaving animals. Such a mechanism might underlie the appearance of new stimulus-elicited pauses which develop in TANs in behaving animals during reinforcement learning.

Glossary

Abbreviations

AHP

afterhyperpolarisation

AP5

d(-)-2-amino-5-phosphonopentanoic acid

BIC

(–)-bicuculline methochloride

CGP55845

(2S)-3-[[(1S)-1-(3,4-dichlorophenyl)ethyl]amino-2-hydroxypropyl](phenylmethyl) phosphonic acid

HCN

hyperpolarisation and cyclic nucleotide activated

I_H

hyperpolarisation and cyclic nucleotide activated cation current

PSP

postsynaptic potential

sAHP

slow afterhyperpolarisation

SK channel

small-conductance Ca²⁺-activated K⁺ channel

TAN

tonically active neuron

ZD7288

4-ethylphenylamino-1, 2-dimethyl-6-methylaminopyrimidinium chloride

Author contributions

Study concept and design: J.N.J.R., M.J.O. and J.L. Acquisition of data: M.J.O. Analysis and interpretation of data: M.J.O., J.M.S., J.N.J.R. and J.L. Drafting of the manuscript: M.J.O. and J.N.J.R. Critical revision of the manuscript: J.L., J.M.S., J.N.J.R. and D.E.O. Study supervision: J.N.J.R. and D.E.O. All authors approved the final version of the manuscript. All experiments were performed in the Reynolds In Vitro Laboratory, Department of Anatomy and Structural Biology, University of Otago, Dunedin, New Zealand.

Supplemental material

Supplementary Figure S1. Assessment of potentially occurring rundown of current and voltage responses in cholinergic interneurons (n = 14) during the whole-cell patch-clamp recording period (25–75 min). Measurements were made at the start and end of the recording period.

A: Mean I-V plots to assess the change of the membrane potential in response to 500 ms current injections over the course of the recording period. R_IN, measured from −100 to +50 pA, was not significantly changed (Start: 163 ± 45 M., End: 155 ± 59 M.; paired t-test, p = 0.58).

B: The hyperpolarisation-activated current (I_H) was assessed from the ‘sag’ in the membrane potential in response to 500 ms hyperpolarising current injections (difference between peak and tail potentials). There was no significant change in the I_H-associated sag potential over the recording period (two-way ANOVA, p = 0.23).

C: The peak amplitude of the slow AHP (sAHP) was measured following 500 ms current injections of increasing amplitude and averaged according to the number of action potentials fired during the current step. There was no significant change in the sAHP peak potential over the recording period (two-way ANOVA, p = 0.18).

Error bars represent the standard error of the mean (SEM).

Supplementary Figure S2. GABA_A receptor blockade with Bicuculline tended to increase both PSP and AHP areas.

Single episode time course analysis of the effect of Bicuculline (30 μM) on PSP and AHP areas in a cholinergic neuron using repetitive stimulation and increasing number of pulses in a stimulus train. Both the PSP and AHP areas were increased after GABA_A receptors were blocked by bath application of Bicuculline, indicating that inhibitory currents mediated by these channels did not contribute to the observed AHPs. Normalised AHP areas represent the ratio of the AHP and PSP areas.

Supplementary Figure S3. NMDA receptor blockade with AP5 decreased, and caffeine bath application increased, both PSP and AHP areas.

Single episode time course analysis of the effect of AP5 (A, 50 μM) and caffeine (B, 10 mM) on PSP and AHP areas in a cholinergic neuron using repetitive stimulation and increasing number of pulses in a stimulus train. A) Both the PSP and AHP areas were decreased after NMDA receptors were blocked by bath application of AP5 for an extended period of time and this effect was fully reversed following a further 8 min washout period with ACSF. B) PSP and AHP areas were both increased after intracellular calcium stores were depleted for 10 min with caffeine and this effect was fully reversed following a further 10 min washout period with ACSF. Normalised AHP areas represent the ratio of the AHP and PSP areas.

Supplementary Figure S4. Effect of caffeine on the slow AHP current (I_sAHP) determined in voltage-clamp.

A: Representative current averages of a single neuron showing the outward current induced at a command potential of −45 mV in response to a preceding voltage step to 0 mV for 1000 ms, from a holding potential of −60 mV, in the absence or presence of 10 mM caffeine.

B: The calcium-activated I_sAHP was reduced to 62 ± 14 % of the control tail currents, measured at the indicated time in A, following depletion of intracellular calcium stores with 10 mM caffeine (paired t-test, *** p < 0.001, n = 7). The peak current was reduced to 77 ± 18 % of controls (paired t-test, * p < 0.05). The difference between reductions in peak and tail currents is likely due to the fact that I_SK also contributes to the peak current.

Supplementary Figure S5. AHPs were reduced on chelation of intracellular Ca²⁺ with 5 mM BAPTA.

A) Single episode time course analysis of the effect of BAPTA on PSP and AHP areas in a cholinergic neuron using repetitive stimulation and increasing number of pulses in a stimulus train. Both PSP and AHP areas were reduced on chelation of intracellular Ca²⁺. B) Normalised AHP areas over 8 min periods of 5 cholinergic neurons, derived from the ratio of the median AHP and PSP areas for two consecutive rounds of repetitive stimulation. On average, the AHP decreased to 66 % of the initial magnitudes (0 – 8 min) and this effect remained relatively stable over the remaining recording period. Error bars represent the standard error around the mean.

Supplementary Figure S6. Intracellular Ca²⁺ chelation with 5 mM BAPTA decimated the slow AHP.

The slow AHP was assessed at the start (BAPTA1) and end (BAPTA2) of the whole-cell recording period (20 – 35 min). A: Representative current averages of a single neuron, showing the underlying outward current of the AHP, induced at a command potential of −50 mV in response to a preceding voltage step to 0 mV for 500 ms, from a holding potential of −60 mV. B: BAPTA reduced the mean I_sAHP peak current to 9 ± 6 % from initial levels (n = 5; *** p < 0.001, paired t-test). This effect was less pronounced in current-clamp mode (C), most likely due to the shorter-lived AHP mediated by I_H. On average, BAPTA reduced the AHP peak potential in response to positive current injections (100 – 500 pA) to 60 ± 7% of initial levels (inset, n = 5; ** p < 0.01, two-way ANOVA). The I_H-associated hyperpolarisation-induced sag potential remained largely unchanged. R_IN, measured from −100 to +50 pA, was not significantly changed (Start: 185 ± 54 M., End: 198 ± 73 Ω; paired t-test, p = 0.55).

Supplementary Figure S7. A hidden “sag potential” (red traces) was revealed by subtraction of ZD7288 average waveforms from controls.

Average waveforms of synaptic responses to 6 stimuli (100 Hz) before and after application of 30 μM ZD7288 to the bath (left panel) and in response to somatic current injections (90 pA, 10 ms; right panel) from a single neuron are shown. Episodes were filtered before averaging as described in Materials and Methods for the PSP and AHP area analysis. The time course of the “sag potential” overall was similar in both stimulation conditions.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors