Cholinergic mechanisms of high-frequency stimulation in entopeduncular nucleus

Abstract

Chronic, high-frequency (>100 Hz) electrical stimulation, known as deep brain stimulation (DBS), of the internal segment of the globus pallidus (GPi) is a highly effective therapy for Parkinson's disease (PD) and dystonia. Despite some understanding of how it works acutely in PD models, there remain questions about its mechanisms of action. Several hypotheses have been proposed, such as depolarization blockade, activation of inhibitory synapses, depletion of neurotransmitters, and/or disruption/alteration of network oscillations. In this study we investigated the cellular mechanisms of high-frequency stimulation (HFS) in entopeduncular nucleus (EP; rat equivalent of GPi) neurons using whole cell patch-clamp recordings. We found that HFS applied inside the EP nucleus induced a prolonged afterdepolarization that was dependent on stimulation frequency, pulse duration, and current amplitude. The high frequencies (>100 Hz) and pulse widths (>0.15 ms) used clinically for dystonia DBS could reliably induce these afterdepolarizations, which persisted under blockade of ionotropic glutamate (kynurenic acid, 2 mM), GABA_A (picrotoxin, 50 μM), GABA_B (CGP 55845, 1 μM), and acetylcholine nicotinic receptors (DHβE, 2 μM). However, this effect was blocked by atropine (2 μM; nonselective muscarinic antagonist) or tetrodotoxin (0.5 μM). Finally, the muscarinic-dependent afterdepolarizations were sensitive to Ca²⁺-sensitive nonspecific cationic (CAN) channel blockade. Hence, these data suggest that muscarinic receptor activation during HFS can lead to feedforward excitation through the opening of CAN channels. This study for the first time describes a cholinergic mechanism of HFS in EP neurons and provides new insight into the underlying mechanisms of DBS.

chronic, high-frequency (>100 Hz) electrical stimulation, known as deep brain stimulation (DBS), through surgically implanted electrodes in the thalamus and basal ganglia has revolutionized the treatment of movement disorders (Kringelbach et al. 2007). Thalamic DBS alleviates tremor (Rehncrona et al. 2003), whereas subthalamic nucleus (STN) and globus pallidus interna (GPi) DBS relieve Parkinson's disease (PD) (Follett et al. 2010). DBS targeting the GPi is also effective for multiple types of dystonia (Kiss et al. 2007; Vidailhet et al. 2005). How DBS works to achieve clinical benefit remains under investigation.

There are several general hypotheses proposed to explain the mechanism(s) of action of DBS: depolarization blockade (Beurrier et al. 2001), synaptic inhibition (Dostrovsky et al. 2000), inhibition of the soma and activation of efferent axons (McIntyre and Grill 2002), and stimulation-induced disruption of pathological network activity (Montgomery and Baker 2000). Our previous work suggested that DBS rapidly depleted glutamatergic synaptic afferents to the thalamus, where glutamate is the predominant neurotransmitter (Anderson et al. 2004, 2006; Iremonger et al. 2006). Clinically, DBS suppresses tremor immediately within the same time scale as these changes in slice experiments (Anderson et al. 2006). DBS applied in the GPi, however, has a slower time scale of clinical benefit (Reese et al. 2011), suggesting that other mechanisms may also be involved.

Cholinergics are ubiquitous in the basal ganglia, play a neuromodulatory role, and have been largely neglected by those studying DBS mechanisms. GPi receives significant cholinergic inputs from brain stem pedunculopontine tegmental nucleus (PPN) in rodents (Woolf and Butcher 1986) and primates (Lavoie and Parent 1994). To date, no previous study has reported or investigated cholinergic mechanisms of DBS.

In rodents, the entopeduncular nucleus (EP) is analogous to the GPi in primates and is located within the fibers of the internal capsule (Nakanishi et al. 1990). It, along with the substantia nigra pars reticulata, is the major output nucleus of the basal ganglia, placing it in a key location where it could influence the entire basal ganglia thalamocortical circuit (McCracken and Kiss 2014). Previous studies on EP applied high-frequency stimulation (HFS) in the adjacent internal capsule and suggested that HFS depressed neuronal activity in EP via elevation of extracellular K⁺ levels (Shin et al. 2007), mediated by enhancement of hyperpolarization-activated cyclic nucleotide gated (HCN) channels (Shin and Carlen 2008). Because DBS is applied clinically inside the GPi, not the internal capsule, and the time frame of clinical improvement is not immediate for dystonia (Reese et al. 2011), we postulated that DBS may involve other longer lasting mechanisms. Using whole cell patch clamp, we tested the hypothesis that HFS applied within EP will activate cholinergic fibers and that these will modulate neural responses to HFS in EP neurons. HFS induced prolonged afterdepolarizations (ADP) in EP neurons, and these were dependent on activation of muscarinic receptors.

MATERIAL AND METHODS

Animals and brain slice preparation.

Forty-nine Sprague-Dawley rats (Charles River, Toronto, Canada) of either sex (postnatal days 11–20) were used for this study. All the experiments were performed according to protocols approved by the University of Calgary Animal Care Committee in accordance with the guidelines established by the Canadian Council on Animal Care. Under deep isoflurane anesthesia, rats were decapitated and the brains quickly removed and immersed in ice-cold carbogen-bubbled (95% O₂-5% CO₂) slicing solution (in mM: 207 sucrose, 2.5 KCl, 26 NaHCO₃, 1.0 CaCl₂, 2.0 MgCl₂, 1.2 NaH₂PO₄, and 10 d-glucose; pH 7.4). Coronal slices (300 μm) were cut with a vibratome (VT1000S; Leica) and transferred to a holding chamber that contained carbogen-bubbled artificial cerebrospinal fluid (aCSF) solution (in mM: 126 NaCl, 2.5 KCl, 26 NaHCO₃, 2.5 CaCl₂, 1.5 MgCl₂, 1.25 NaH₂PO₄, and 10 d-glucose; pH 7.4). EP slices were incubated at 32–33°C for 60 min and then kept at room temperature.

Electrophysiology.

For whole cell recording, a slice was placed in a recording chamber and continually superfused with 32–33 °C carbogen-bubbled aCSF solution at a rate of 2–3 ml/min. Amplified current or voltage signals were obtained from a borosilicate glass electrode (1.5-mm outer diameter; tip resistance 4–6 MΩ; King Precision Glass, Claremont, CA) filled with intracellular solution containing (in mM) 108 K-gluconate, 8 Na-gluconate, 8 KCl, 2 MgCl₂, 1 EGTA, 10 HEPES, 4 K-ATP, and 0.35 Na-GTP (pH 7.2 corrected with KOH, osmolarity 286 mosM). The EP was visually recognized within the internal capsule and individual neurons identified with the use of an upright microscope (Olympus BX51W; Tokyo, Japan) fitted with differential interference contrast optics.

All recordings were obtained using a Digidata 1440A acquisition system connected to a Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA). The recordings were low-pass filtered at 1 kHz, digitized at 10 kHz online with Clampex 10.3, and analyzed offline using Clampfit 10.3 (Molecular Devices). The data were accepted for offline analysis only if changes in access resistance were <20%. No series resistance compensation was performed. The liquid junction potential was 14 mV, and reported voltage data were not corrected.

Resting membrane potential and input resistance were measured immediately after stable whole cell configuration was established. Constant 30-pA current steps, 1 s in duration, were injected to obtain a current-voltage (I-V) curve to measure input resistance and the action potential (AP). AP threshold was calculated as the membrane voltage potential at which point the slope of the first action potential was >10 V/s. The slope of the I-V plot represented the input resistance. For neurons displaying spontaneous spikes, the resting membrane potential was defined as the average membrane potential halfway between the most repolarized potential after an AP and the threshold potential (Shin and Carlen 2008). For neurons without spontaneous spikes, the resting membrane potential was defined as the average membrane potential over 2 min after stable whole cell recording was established.

EP neurons were identified by the following electrophysiological properties (Nakanishi et al. 1990): anomalous rectification at hyperpolarizing potentials, rebound spikes at the offset of membrane hyperpolarization, and high, nonadapting repetitive firing at depolarizing potentials.

Electrical stimulation and “blanking” operation.

A bipolar concentric electrode (NEX-100; Rhodes Medical Instruments, Woodland Hills, CA) was placed within EP. Stimulation was delivered with a constant-current stimulus isolator (A365; World Precision Instruments, Sarasota, FL) connected to a stimulus generator (model 2100; A-M Systems, Sequim, WA).

The stimulating electrode in EP was positioned such that single pulses could reliably evoke postsynaptic currents (PSCs) under voltage clamp (clamped at −60 mV). To minimize the HFS trains applied while determining stimulation thresholds, we started HFS using a current that produced 30% of a maximum PSC. Stimulation trains consisted of 60-, 150-, or 400-μs pulse-width square-wave monophasic pulses. We tested HFS frequencies of 10 or 125 Hz and train durations of 10 s. For each pulse width, current was increased in 50-μA increments until an ADP response was successfully induced. If response to the previous HFS train did not return to baseline, data were not included in analysis.

HFS produces an artifact that can obscure membrane responses. As described previously (Anderson et al. 2004), to eliminate the artifact during HFS, “blanking” pulses (0.1–1.0 ms) were triggered with each stimulation pulse (model 2100; A-M Systems). These voltage pulses initiated a blanking operation of the Axopatch 200B amplifier, which prevented the membrane voltage from updating during the blanking pulse, thus significantly reducing the stimulus artifacts (Fig. 1A).

Fig. 1.

Schematic representation of whole cell patch-clamp recording setup in entopeduncular nucleus (EP) slice. Recordings were made from EP neurons, and high-frequency stimulation (HFS) was delivered within the EP (A). EP neurons did not show spike adaption when currents up to 110 pA were injected into the cell (B). Typically, these neurons also showed rectification at hyperpolarizing potentials and rebound spikes at the offset of membrane hyperpolarization. The rectification sag (voltage differences between the 2 arrowheads) was plotted as a function of the membrane potential (C). Time and voltage scales of insets in B are the same as that in C.

Chemicals.

Chemicals were diluted in aCSF from aliquots and applied by continuous bath perfusion to the slices. Some recordings were performed during full blockade of ionotropic glutamatergic and GABAergic neurotransmission with a cocktail consisting of either kynurenic acid (2 mM) and picrotoxin (100 μM) or a mixture of 6,7-dinitroquinoxaline-2,3-dione (DNQX; 10 μM), dl-2-amino-5-phosphonopentanoic acid (AP5; 50 μM), and picrotoxin (100 μM). Tetrodotoxin (TTX; 1 μM), atropine (4 μM), oxotremorine M (Oxo; 10 μM; muscarinic receptor agonist), and dihydro-β-erythroidine hydrobromide (DHβE; 2 μM; competitive nicotinic acetylcholine receptor antagonist) were applied from stock solution made in double-distilled water. Flufenamic acid (FA; 50 μM; nonspecific inhibitor of Ca²⁺-activated nonspecific cationic channels) and CGP 55845 (1 μM; selective GABA_B receptor antagonist), were applied from stock solutions made in dimethyl sulfoxide (DMSO). The final concentration of DMSO in aCSF was ≤0.1%, which showed no effect on neuronal activities in control experiments. After drug application of TTX, atropine, or FA, slices were washed with aCSF for 10–40 min until recovery of ADP responses was obtained. Picrotoxin, DNQX, TTX, and FA were purchased from Tocris, and all other chemicals were purchased from Sigma.

Data analysis.

ADP amplitude was measured as the voltage difference between resting membrane potential and peak ADP membrane potential. ADP duration was measured as the time difference between the offset point of electrical stimulation and the time point when the membrane potential dropped to resting. To evaluate the effect of different agonists and blockers on HFS, we compared the ADP responses before and at least 5 min after application of the drugs.

All group data are means ± SE. Data were analyzed using one-way ANOVA, repeated-measures ANOVA (RM-ANOVA), or Student's paired t-test comparing firing and depolarization induced by HFS applied under aCSF vs. bath application of drugs. The RM-ANOVA was followed by post hoc Tukey's HSD (honestly significant difference) test for pairwise comparison if necessary.

RESULTS

Properties of EP neurons.

Electrophysiological properties were recorded from 102 EP neurons, 73 of which had spontaneous spikes at resting membrane potential. The input resistance was 369.6 ± 15.6 MΩ, and resting membrane potential was −53.2 ± 0.8 mV. Spontaneously active EP neurons had more depolarized resting membrane potentials (−51.5 ± 0.9 mV) than nonspontaneously active cells (−59.2 ± 1.8 mV, P < 0.01). When 1-s current steps were injected into the recorded neurons, the responses resembled those described by Nakanishi et al. (1990). Recorded neurons did not show spike adaption at depolarizing potentials, and the maximum spike rate induced by a 110-pA (1 s) current step was 30.0 ± 1.8 Hz (Fig. 1B). At hyperpolarizing potentials, the neurons displayed anomalous rectification and a strong rebound activation at the offset of membrane hyperpolarization (Fig. 1C, insets). The rectification sag was measured as the membrane potential difference between the maximum potential at the rectification sag and the potential at the offset of the hyperpolarization. This sag increased with more hyperpolarized current injection (Fig. 1C) and resembled that described in previous studies (Nakanishi et al. 1990; Shin et al. 2007).

HFS in EP induced ADP responses.

After stable whole cell current-clamp recording was obtained for at least 5 min, HFS was applied with neurons at resting membrane potential (n = 71 neurons, 55 with and 16 without spontaneous activity). The stimulation electrode was placed 229.6 ± 9.9 μm from 54 recorded cells (in which this distance was measured). A 10-s train of HFS induced responses after the offset of stimulation in all neurons: ADP were either short lasting (ADP-S) or prolonged (ADP-Pro). Neurons with ADP-S responses showed depolarization right after the end of HFS, which was accompanied by increased spiking (Fig. 2A, top). The membrane potential of ADP-S neurons gradually returned to resting membrane potential after the ADP reached a peak. Neurons with ADP-Pro responses showed an initial ADP-S at the offset of HFS and then gradually depolarized again (Fig. 2A, bottom). Spontaneous spikes increased during this prolonged depolarization. Typically, the ADP-S responses lasted less than a minute, whereas the ADP-Pro response duration ranged from several minutes to as long as the neurons were recorded. Neurons with (55/71) and without (16/71) spontaneous spikes at resting membrane potential both produced ADP-S (spontaneous, 25/55; nonspontaneous, 6/16) and ADP-Pro (spontaneous, 30/55; nonspontaneous, 10/16) responses to HFS. There were no differences in resting membrane potential between ADP-S and ADP-Pro responding neurons of either spontaneous (−51.4 ± 1.2 vs. −51.8 ± 1.3 mV) or nonspontaneous firing type (−62.3 ± 3.8 vs. −57.9 ± 1.6 mV).

Fig. 2.

Afterdepolarization (ADP) induced by HFS in EP neurons at resting membrane potential. According to ADP duration, the responses were classified as short (ADP-S) or prolonged (ADP-Pro) (A). High-frequency (125 Hz) but not low-frequency (10 Hz) stimulation evoked ADP in the same neuron (B). The pulse width, duration, and intensity of the HFS were 400 μs, 10 s, and 300 μA, respectively. Longer pulse widths were more effective than shorter pulse widths in evoking ADPs. The threshold amplitude required to evoke an ADP was significantly lower (C), whereas the corresponding amplitude (D) and duration (E) of evoked ADP were higher, at 400-μs pulse width than those obtained with 60- and 150-μs pulse widths. Gray circles indicate individual neuronal responses; bar graphs show means ± SE. *P < 0.05; **P < 0.01, RM-ANOVA.

As discussed in the following sections, the effects of drugs on both ADP-S and ADP-Pro responses were similar, indicating that both responses share the same cellular mechanisms. Here we focus on the neurons displaying ADP-S responses.

ADP responses were related to frequency and pulse width of HFS applied.

High-frequency DBS is required to achieve clinical benefit. Benefit for dystonia usually requires longer pulse widths (Krauss et al. 2004). Thus we investigated whether the ADP responses were related to stimulus parameters of frequency and pulse width. During and after a 10-s train of 10-Hz HFS, the resting membrane potential and spontaneous spike frequency did not show significant changes (Fig. 2B, top; n = 4). High-frequency (125 Hz) stimulation resulted in dramatic changes in membrane potential and spike frequency, as well as an ADP-S response, i.e., depolarization at the offset of HFS in which the spike frequency also increased (Fig. 2B, bottom). Spike frequency and membrane potential returned to baseline within 1 min.

Next, we tested the effects of different pulse widths on ADP-S responses. Three pulse widths were used to determine the minimal current required to induce an ADP-S (n = 8): short (60 μm), medium (150 μm), or long (400 μm). The amplitude required to elicit ADP-S using long pulse width (181.3 ± 29.8 μA, P < 0.05, RM-ANOVA with Tukey's post hoc comparisons) was significantly lower than that required using short (759.4 ± 150.5 μA) and medium pulse widths (387.5 ± 77.6 μA; Fig. 2C). The amplitude and duration of the ADP-S responses at long pulse width (10.8 ± 1.9 mV, P < 0.01; 14.3 ± 2.2 s, P < 0.05, RM-ANOVA with Tukey's post hoc comparisons) were significantly higher and longer than those obtained using short (4.6 ± 0.8 mV, 9.2 ± 1.1 s) and medium pulse widths (6.2 ± 1.4 mV, 11.2 ± 2.1 s; Fig. 2, D and E). A 10-s HFS train with long pulse width induced a consistent and reliable ADP; therefore, to investigate the cellular mechanisms of the ADP in further experiments, we used only high-frequency stimulation (125 Hz) with long pulse width (400 μs).

ADP responses are dependent on synaptic transmission.

In one group of experiments, we blocked ionotropic glutamate receptors (2 mM kynurenic acid), GABA_A (50 μM picrotoxin), GABA_B (1 μM CGP 55845), and nicotinic acetylcholine receptors (2 μM DHβE) and applied HFS to our slice. Of four neurons tested, HFS induced ADP-S in one cell, and three cells showed ADP-Pro responses (Fig. 3A).

Fig. 3.

HFS-induced ADP depends on synaptic transmission. Blockade of glutamatergic, GABAergic, and cholinergic nicotinic transmission did not eliminate ADP (A). Tetrodotoxin (TTX) abolished the HFS-induced ADP (B). Individual cell responses (gray circles) and group data (bar graphs; means ± SE) show that both the amplitude (C) and duration (D) of the ADP response were significantly decreased in the presence of 0.5 μM TTX. *P < 0.05, paired t-test. ACSF, artificial cerebrospinal fluid; DHβE, dihydro-β-erythroidine hydrobromide.

To rule out the possibility that the ADP responses were induced via direct activation of recorded EP neurons, we blocked all synaptic transmission with TTX (0.5 μM). Figure 3B shows an example of the ADP-Pro response being abolished after 5-min application of TTX. Both the amplitude and duration of ADP significantly decreased, from 8.4 ± 3.3 to 1.7 ± 1.0 mV and from 7.1 ± 1.3 to 2.3 ± 1.4 s, respectively (Fig. 3, C and D; n = 4).

In another set of experiments we moved the stimulating electrode around the EP neuron and delivered single pulses until a spike was initiated, indicating direct antidromic activation of the recorded neuron. In this case HFS failed to induce ADP responses from these neurons at a stimulus intensity of 300 μA above the threshold shown in Fig. 2C (n = 2; data not shown).

ADP responses are dependent on activation of muscarinic receptor.

Because of the time course of ADP responses and the existence of a cholinergic projection from brain stem to EP (Clarke et al. 1997), we hypothesized that cholinergic synapses may be involved. ADP responses in EP neurons significantly decreased after 4 μM atropine application in aCSF (Fig. 4A). The average ADP amplitude was reduced from 13.2 ± 2.5 mV in aCSF to 4.5 ± 1.8 mV (Fig. 4B), whereas the ADP duration was diminished from 32.4 ± 4.6 s in aCSF to 12.4 ± 4.5 s under muscarinic blockade (Fig. 4C; n = 7; P < 0.01, paired t-test). No ADP response was obtained in three of seven cells (42.9%), and in the remaining four cells ADP response was almost eliminated after atropine blockade. In another group of experiments, cholinergic involvement in generating the ADP responses was tested by applying the muscarinic receptor agonist Oxo (10 μM) and a 10-s current injection into the cell (n = 8). In aCSF, a 10-s, 100-pA current simulating synaptic responses of HFS did not change the membrane potential (Fig. 4D, top). However, when the muscarinic receptor was activated by Oxo in the bath, the same stimulation induced dramatic depolarization of the membrane potential, leading to ADP response (Fig. 4D, bottom). These data together suggest that HFS-induced ADP responses were highly dependent on the activation of muscarinic receptors.

Fig. 4.

HFS-induced ADP was dependent on activation of muscarinic acetylcholine receptors. Blockade of muscarinic receptors decreased the HFS-induced ADP response (A). Group data (means ± SE) show that both the amplitude (B) and duration (C) of the ADP were significantly reduced after bath application of 4 μM atropine. **P < 0.01, paired t-test. The muscarinic agonist oxotremorine (Oxo), together with a 10-s, 100-pA current injection into the recorded cell (line below voltage trace), replicated the ADP responses seen with HFS (D). KA, kynurenic acid; Picro., picrotoxin.

ADP requires activation of Ca²⁺-sensitive nonspecific cationic channels.

Activation of Ca²⁺-sensitive nonspecific cationic (CAN) channels through muscarinic receptors is thought to underlie the ADP and persistent activity in entorhinal cortex (Egorov et al. 2002) and primary visual and motor cortex (Rahman and Berger 2011). Hence, we applied the general CAN channel blocker FA (50 μM) during HFS. In one cell (Fig. 5A), the amplitude and duration of the ADP response induced by HFS disappeared entirely in aCSF plus FA. The average ADP amplitude was diminished from 13.9 ± 1.8 mV in aCSF to 4.1 ± 1.1 mV with FA (Fig. 5B), and ADP duration decreased from 15.0 ± 2.8 s in aCSF to 2.9 ± 0.8 s with FA (Fig. 5C; P < 0.01, paired t-test). In two of eight cells (25%), ADP was entirely eliminated after blockade of CAN channels. These data suggest that CAN channels are involved in HFS-induced ADP responses.

Fig. 5.

HFS-induced ADP was dependent on activation of Ca²⁺-sensitive nonspecific cationic (CAN) channels. Blockade of CAN channels with 50 μM flufenamic acid (FA) prevented ADP (A). Group data (means ± SE) show that both the amplitude (B) and duration (C) of the ADP were significantly reduced after bath application of FA. **P < 0.01, paired t-test.

DISCUSSION

DBS effects depend on the physiological properties of the target neurons, the stimulus parameters (frequency, pulse width, and duration of DBS) applied, and the anatomic connectivity of the target nucleus (Kringelbach et al. 2007). Not only can DBS inhibit or stimulate neural activity in the target nucleus (Vitek 2002), but it also can alter downstream connected nuclei and cortical regions (de Hemptinne et al. 2013; McCracken and Kiss 2014). Most previous studies have focused on how DBS alters GABA or glutamate transmission (Chiken and Nambu 2013; Lee et al. 2004) or membrane properties (Beurrier et al. 2001; Shin et al. 2007). Despite most of the basal ganglia being innervated by cholinergic fibers from the PPN, the effects of DBS on such fibers have not been reported. In the present study we found that HFS exerted poststimulation effects, an ADP dependent on frequency and current density applied, and mediated by muscarinic receptors.

Post-HFS responses were either short (ADP-S) or prolonged (ADP-Pro), based on the duration of the ADP. Both ADP-S and ADP-Pro responses were equally prevented by TTX, atropine and FA, indicating that they shared the same mechanisms. In fact, ADP-S may be a subthreshold form of ADP-Pro, because ADPs are usually considered a subthreshold form of long-lasting depolarizations (Egorov et al. 2002). TTX, atropine, and FA virtually eliminated the ADP, supporting the involvement of cholinergic synaptic transmission and CAN channels. The excitatory effects of acetylcholine are probably mediated via either blockade of a G protein-coupled inwardly rectifying K⁺ channel (Krnjevic 2004) or activation of the phospholipase C pathway, which activates CAN channels (Schaefer et al. 2000). CAN channels are permeable to Na⁺ and Ca²⁺, and the increased intracellular Ca²⁺ could in turn amplify the CAN channel responses and lead to long-lasting depolarization (Rahman and Berger 2011). The intracellular Ca²⁺ is linked to a Ca²⁺-activated K⁺ conductance, which may return the membrane potential from depolarized to resting. We recorded all cells at resting membrane potential to test the effects of HFS in physiological conditions. We suspect that the ADP-S response occurred from the Ca²⁺-activated outward K⁺ conductance overcoming the CAN channel-activated inward conductance in some cells, whereas in other cells it did not and resulted in the ADP-Pro response.

Our previous work in thalamus indicated that the primary target of DBS is the axon and its synaptic terminals, with HFS activating and depleting glutamatergic terminals of neurotransmitter (Anderson et al. 2006; Iremonger et al. 2006). In EP, HFS may also activate the robust cholinergic afferents from PPN (Woolf and Butcher 1986). This is comparable to previous studies (Egorov et al. 2002) and current findings (Fig. 4D) in which activation of muscarinic receptors with agonists (carbachol or Oxo) paired with extracellular stimulation could induce ADP and persistent activities.

There have been very few studies investigating the effects of HFS on EP neurons. Shin et al. (2007) reported that HFS induced an elevation of extracellular K⁺ level, resulting in reduction in EP neural activity. This occurred through the enhancement of HCN channels (Shin and Carlen 2008). Our study differed in two respects. First, Shin et al. (2007) placed their stimulating electrode in the internal capsule, whereas we placed ours in the EP cellular mass. HFS of the internal capsule is expected to activate a large number of axons and result in a massive efflux of intracellular K⁺ at the same time. The increased K⁺ conductance associated with abundant neuronal activity and restricted extracellular space would lead to accumulation of extracellular K⁺, which amplifies the accumulation of extracellular K⁺ (Nicholson 1980). Also, Shin et al. applied 150-Hz, 100–300 μA, and 60-μs pulse-width HFS to their EP slices. We applied a variety of pulse widths, frequencies, and amplitudes. Both long and short pulses could induce the ADP responses, but longer pulses (400 μs) had the lowest thresholds. Although our results are similar to those reported by Shin et al. (2007) with respect to depolarization post-HFS, the differences may be due to the distance and orientation of the EP neurons to the current applied, including voltage gradient, stimulation parameters, and the properties of synaptic transmission on the target neurons (Kringelbach et al. 2007).

The EP receives major inputs from striatum, GP, and STN in the basal ganglia and sends major output to the thalamus (Ogura and Kita 2002). The striatal axons projecting to EP contain GABA, dynorphin, and substance P (Penny et al. 1986; Reiner et al. 1999), and the STN neurons projecting to EP contain glutamate (Nakanishi et al. 1991). Dynorphin may control the activity of EP neurons by regulating striatal terminal GABA release and also by hyperpolarizing the postsynaptic membrane (Ogura and Kita 2002). It is possible that neurotransmitters other than acetylcholine also play a role in generating ADP responses during EP HFS, but our data suggest that the majority of this effect is cholinergic mediated.

Advantages and limitations of the model.

The whole cell patch-clamp technique has both advantages and disadvantages when we try to relate it to effects of DBS in humans. A major advantage is that it allows dissection of the precise neurons, pathways, and synaptic effects of stimulation. Synaptic terminals of afferent inputs are maintained; EP neurons have characteristic biophysical properties, allowing us to precisely know the type of neuron we are recording; and neurotransmitters across species are conserved (Kita 2001; Nakanishi et al. 1991; Ogura and Kita 2002). Also, the slice preparation allows precise placement of the stimulation electrode. Only in a slice are we able to control the cellular milieu completely, without confounding features of anesthetics and muscle relaxants. Finally, brain slices provide an unparalleled stability for intracellular recording and good yield for experiments so as to minimize the number of animals used. However, whether HFS affects neurons in human GPi the same way is not known. Previously, in thalamus, we demonstrated similar results in rodent slice experiments and patients undergoing DBS surgery (Anderson et al. 2006); therefore, we can test the hypotheses generated by slice work in humans. We are pursuing similar experiments in humans undergoing DBS surgery to further investigate the hypotheses generated by these data in EP slice (Luo et al. 2015).

Clinical implications of HFS-induced ADP in EP.

Motor symptoms of PD and dystonia are related to abnormal firing patterns and oscillations in basal ganglia (Starr et al. 2005). DBS is applied to GPi for PD and dystonia (Kringelbach et al. 2007). Because DBS produces similar therapeutic effects to a lesion or blockade of the target nucleus, a common mechanism of DBS is thought to be destruction or normalization of abnormal firing (Kringelbach et al. 2007). In this study we identified another mechanism that modulates neuronal excitability, potentially contributing to the effects of DBS on neurons.

The benefits of DBS for PD are thought to involve an altering of abnormal basal ganglia oscillations in the beta-frequency band (Brown et al. 2001). Recent in vivo work suggests that EP DBS induces both synchronizing and desynchronizing effects by modulating local as well as distant neural activities and altering oscillations (de Hemptinne et al. 2013; McCracken and Kiss 2014). Similarly, in EP slice, HFS modulated responses of local EP neurons via activation of cholinergic fibers. In contrast to the immediate effectiveness of DBS for PD, the clinical benefits for dystonia are gradual (Krauss et al. 2004). This suggests that different mechanisms may be at play in DBS for PD and dystonia. Cholinergic modulation of EP neurons could contribute to rapid disruption of local abnormal synaptic transmission, as well as providing longer lasting effects on pathological network activities.

Dopamine (DA) reduction in PD is accompanied by a rise in cholinergic signaling in striatum (DeBoer et al. 1996). This is thought to be involved in PD symptoms and has motivated the use of anticholinergic drugs acting on muscarinic acetylcholine receptors (mAChR) in treatment (Pisani et al. 2007; Riley et al. 1989). Anticholinergics are also the mainstay of therapy for dystonia (Jankovic 2013). This appears paradoxical if one assumes that the cholinergic ADP responses are critical to the benefit of HFS. However, muscarinic receptor subtypes are likely responsible for different effects in different nuclei. The anticholinergics approved for treating PD and dystonia clinically are nonselective and therefore limited by central and peripheral adverse effects (Langmead et al. 2008). There are five subtypes of mAChRs widely distributed in the central nervous system, divided into two groups on the basis of signal transduction: M1 group (M1, M3, M5) and M2 group (M2, M4) (Langmead et al. 2008). These receptors have been best studied in the striatum (Goldberg et al. 2012): presynaptic mAChRs (M2 group) are largely inhibitory and act as inhibitory autoreceptors with M4 receptors predominating. Postsynaptic mAChRs can be either inhibitory (M2 group) or excitatory (M1 group) (Wess et al. 2003). Other than in striatum and substantia nigra (Weiner et al. 1990), cholinergic receptor subtypes have not been characterized in the GPi/EP or other basal ganglia nuclei. A better understanding of the role of cholinergic modulation in the basal ganglia circuit may allow development of more specific treatment strategies for movement disorders.

Dystonia is characterized by sustained muscle contractions that result in twisting and abnormal postures (Tarsy and Simon 2006). The pathophysiology is not clear, but abnormal somatotopic organization is thought to develop in the thalamo-cortico-basal ganglia loop (Albanese and Lalli 2012). How dystonia is improved gradually by GPi DBS may be related to a gradual change in abnormal sensorimotor cortical maps (Chen and Hallett 1998). Whereas GPi DBS can directly inhibit thalamic neurons (Anderson et al. 2003), it may also activate cholinergics antidromic terminals in GPi, which would have a more gradual effect on downstream structures. DBS may antidromically activate the PPN through cholinergic afferents, and neurons in PPN densely innervate thalamus, motor areas of brain stem, and spinal cord (Garcia-Rill 1991). Activation of PPN paired with a stimulus to the sensory system induces auditory plasticity in both thalamus (Luo et al. 2011) and cortex (Luo and Yan 2013). Therefore, GPi DBS may drive both the GPi GABAergic and PPN cholinergic inputs to thalamus at the same time. Because of the modulatory nature of cholinergic action, it may take weeks or months to induce thalamic and eventually cortical plasticity through the GPi action on the motor system.

Direct involvement of cholinergics is further supported by PPN being a promising target for DBS in patients with PD (Kringelbach et al. 2007). PPN is closely related to movement, because animal studies have shown that PPN stimulation increases movement and inhibition of PPN decreases movement (Brudzynski et al. 1986; Mogenson and Wu 1988). In parkinsonian nonhuman primates, disinhibition of PPN by pharmacological agents (Nandi et al. 2002), as well as driving PPN activity by electrical stimulation (Jenkinson et al. 2004), increases motor activity. In humans, PPN-DBS produces 45% amelioration of the motor Unified Parkinson's Disease Rating Scale subscale score, whereas combined DBS of PPN and STN together improves activities of daily living (Stefani et al. 2007) and gait (Ferraye et al. 2010).

Finally, DBS is also being experimentally applied for Alzheimer dementia (Laxton et al. 2010) and memory enhancement (Suthana and Fried 2014). Some have suggested that the mechanism for such memory improvement is neurogenesis in hippocampus (Stone et al. 2011). Our data on cholinergic mechanisms may provide alternative or additional possibilities.

In conclusion, our results demonstrate that muscarinic receptor activation by HFS can generate poststimulation effects, which are mediated by feedforward excitation through the opening of CAN channels. Therefore, cholinergic afferents play a significant role in the effects of HFS on EP neurons and provide new insight into possible mechanisms of clinical DBS in humans.

GRANTS

Funding was provided by the Alberta Heritage Foundation for Medical Research (AHFMR) and the Canadian Institutes of Health Research. Z.H. Kiss is a Clinical Scholar for AHFMR.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

F.L. and Z.H.K. conception and design of research; F.L. performed experiments; F.L. analyzed data; F.L. interpreted results of experiments; F.L. prepared figures; F.L. drafted manuscript; F.L. and Z.H.K. edited and revised manuscript; F.L. and Z.H.K. approved final version of manuscript.