Logo of jnPublished ArticleArchivesSubscriptionsSubmissionsContact UsJournal of NeurophysiologyAmerican Physiological Society
J Neurophysiol. Aug 2010; 104(2): 911–921.
Published online May 26, 2010. doi:  10.1152/jn.00103.2010
PMCID: PMC2934941

Deep Brain Stimulation Alleviates Parkinsonian Bradykinesia by Regularizing Pallidal Activity


Deep brain stimulation (DBS) of the basal ganglia can alleviate the motor symptoms of Parkinson's disease although the therapeutic mechanisms are unclear. We hypothesize that DBS relieves symptoms by minimizing pathologically disordered neuronal activity in the basal ganglia. In human participants with parkinsonism and clinically effective deep brain leads, regular (i.e., periodic) high-frequency stimulation was replaced with irregular (i.e., aperiodic) stimulation at the same mean frequency (130 Hz). Bradykinesia, a symptomatic slowness of movement, was quantified via an objective finger tapping protocol in the absence and presence of regular and irregular DBS. Regular DBS relieved bradykinesia more effectively than irregular DBS. A computational model of the relevant neural structures revealed that output from the globus pallidus internus was more disordered and thalamic neurons made more transmission errors in the parkinsonian condition compared with the healthy condition. Clinically therapeutic, regular DBS reduced firing pattern disorder in the computational basal ganglia and minimized model thalamic transmission errors, consistent with symptom alleviation by clinical DBS. However, nontherapeutic, irregular DBS neither reduced disorder in the computational basal ganglia nor lowered model thalamic transmission errors. Thus we show that clinically useful DBS alleviates motor symptoms by regularizing basal ganglia activity and thereby improving thalamic relay fidelity. This work demonstrates that high-frequency stimulation alone is insufficient to alleviate motor symptoms: DBS must be highly regular. Descriptive models of pathophysiology that ignore the fine temporal resolution of neuronal spiking in favor of average neural activity cannot explain the mechanisms of DBS-induced symptom alleviation.


High-frequency stimulation of the subthalamic nucleus or the internal segment of the globus pallidus is an effective treatment for persons with Parkinson's disease (PD) whose symptoms have become medically unmanageable. High-frequency deep brain stimulation (DBS) can alleviate tremor, bradykinesia, and rigidity and can enable a reduction in medication dose, thereby reducing dyskinesias. The therapeutic benefits of high-frequency DBS resemble those resulting from surgical lesions in the same locations. However, DBS has the advantages that it is adjustable (the electrode geometry and stimulation parameters are programmable) and reversible (stimulation can be turned off and the electrodes can be removed), and DBS can be implanted bilaterally, whereas bilateral lesions are often associated with unacceptable side effects (Okun and Vitek 2004). Although DBS has been implanted in over 40,000 patients, the mechanisms of action remain unclear.

Based on similarities in clinical outcomes, DBS was thought to share a physiological effect with a lesion: silencing or suppressing the neural activity in the stimulated tissue. Some experimental results support that DBS inhibits the neurons surrounding the electrode (Beurrier et al. 2001; Boraud et al. 1996; Filali et al. 2004). However, axons entering or leaving the stimulated nucleus may also fire synchronously with DBS, propagating increased activity to both afferent and efferent locations (Gradinaru et al. 2009; McIntyre et al. 2004; Li et al. 2007). Downstream changes in concentrations of neurotransmitters and related biochemicals (Stefani et al. 2005; Windels et al. 2003) and electrophysiological activity (Anderson et al. 2003; Degos et al. 2005; Hashimoto et al. 2003; Hershey et al. 2003; Phillips et al. 2006) support the claim that DBS excites axons entering and exiting the site of stimulation. Thus while DBS and lesion may yield similar outcomes, the intuitive hypothesis that they both alleviate symptoms by silencing neural activity appears to be incorrect.

If neuronal firing rates were directly responsible for symptom severity, the emerging view that DBS increases widespread axonal activity could not be reconciled with the cessation of neural activity following surgical lesion. However, symptom severity may be less related to rates of neuronal activity than to patterns of neuronal activity as suggested in: rodents (Degos et al. 2005), non-human primates (Bar-Gad et al. 2004; Hashimoto et al. 2003; Meissner et al. 2005; Wichmann and DeLong 2003), humans with PD (Magnin et al. 2000; Tang et al. 2005), and computational models (Guo et al. 2008). Surgical lesions may work not by eliminating all activity in the lesioned nucleus per se but rather by eliminating specifically the firing patterns associated with symptoms. Similarly, DBS may work—independent of acknowledged changes in firing rate—by replacing pathological firing patterns with innocuous, DBS-induced regularity.

Supporting this regularity assertion, high-frequency stimulation overrode pathological firing patterns in computer models of DBS (Grill et al. 2004; Rubin and Terman 2004; Terman et al. 2002). Tremor reduction by thalamic DBS with different frequencies, amplitudes (Kuncel et al. 2007), or patterns (Birdno et al. 2007) of stimulation was strongly correlated with the regularization of neuron firing patterns. Furthermore, high-frequency DBS that alleviated motor symptoms in a nonhuman primate model of parkinsonism also lowered the firing pattern entropy of neurons throughout the basal ganglia thalamic network (Dorval et al. 2008; Hashimoto et al. 2003). We hypothesize that the abolishment of pathological neuronal activity is the mechanism by which DBS alleviates the motor symptoms of PD.

We propose this specific test of causality: DBS that reduces the variability of synaptic inhibition from globus pallidus to thalamus will reduce bradykinesia in persons with PD, while DBS of the same amplitude and average frequency that does not reduce synaptic variability will not alleviate symptoms. In particular, masking the disease-induced pathological activity with regular (periodic) DBS-induced activity will alleviate parkinsonian symptoms, while masking the disease-induced pathological activity with irregular (aperiodic) DBS-induced activity will not alleviate symptoms. Indeed in this study, we found that periodic DBS patterns, which regularized activity (i.e., lowered neuronal firing pattern entropy) in a computational model of the basal ganglia thalamic circuit (Rubin and Terman 2004), alleviated bradykinesia in human participants with PD. Conversely, aperiodic DBS patterns of the same amplitude and average frequency, which did not regularize neuronal activity in the computational model, did not alleviate bradykinesia in human participants. Thus we provide causal evidence for a mechanism of DBS: regularizing neuronal activity and thereby synaptic release in the basal ganglia thalamic network alleviates the motor symptoms of PD.


The efficacy of high-frequency DBS with increasing degrees of temporal variability was measured in both a computational model of the basal ganglia thalamic circuit and in human participants with PD and existing DBS electrodes under awake and behaving conditions during battery replacement surgery. These methods describe the types of DBS trains used in both cases, the computational model construction and simulation, the analysis of the computational data, and the human protocol and data analysis.

Temporally irregular DBS

Four classes of DBS trains were constructed and denoted by their degree of variability (Fig. 1, A and B). One class had no variability consisting of periodic pulses at 130 Hz, identical to the DBS provided by the pulse generators used clinically (Kinetra or Soletra, Medtronic, Minneapolis, MN). The other three classes were constructed as memoryless point processes where the time from one pulse to the next was a random variable found by drawing a random sample from a gamma distribution of instantaneous frequencies, fi. The gamma distributions all had a mean of 130 Hz, with SDs of 13, 39, or 78 Hz for the 10, 30, or 60% variability classes, respectively. Probability density functions of the instantaneous frequency, defined in terms of the gamma function (Γ), were described by: p(fi) = [fiκ-1 e-fi]/[θκ Γ(κ)] for fi > 0 where the shape parameters (κ = 100 {1, 1/9, 1/36}) and scale parameters (θ = 13/10 {1, 9, 36}) were set to yield a mean of 130 Hz and variabilities of 10, 30, and 60%, respectively.

Fig. 1.
Experimental paradigm. Four classes of Deep brain stimulation (DBS) were presented to human participants with Parkinson's disease and to a computational model. A: example pattern rastergrams of the DBS pulse timings. B: probability densities of the 4 ...

Computational model experiments

All computational experiments and analysis were performed on an x86 personal computer running the Ubuntu distribution (//www.ubuntu.com, Canonical) of the GNU/Linux operating system (Free Software Foundation, Boston, MA). All models and analysis code are available from the authors on request.

The computational model was modified slightly from existing models of the basal ganglia thalamic network (Rubin and Terman 2004; Terman et al. 2002). Sixteen point neurons in each of the four regions—subthalamic nucleus (STN), globus pallidus externus (GPe), globus pallidus internus (GPi), and pallidal receiving thalamic cells (TC)—comprised the model. Terman and colleagues (2002) categorized three anatomically distinct connection schemes for the model, from which our connection scheme is somewhere between their random and structured-sparse networks. In particular, our connections were structured and sparse, the scheme preferred by others (Feng et al. 2007; Rubin and Terman 2004) but included asymmetry present only in their random network (Terman et al. 2002). We also introduced heterogeneity into the network by varying the striatal drive to GPe neurons (see following text). The network asymmetry and heterogeneity broadened the apparent domain of chaotic behavior, and kept the network from highly regular, perfectly entrained activity in the presence of the intense periodicity supplied by regular DBS of the STN cells.

The membrane potential (V) of neurons in basal ganglia was described by: cm dV/dt = IappINaIKICaITIahpILIsyn; where cm is the membrane capacitance and each I variable denotes a current source: INa, IK, ICa, IT, Iahp, and IL are the sodium, potassium, calcium, T-type calcium, afterhyperpolarizing, and leak currents, respectively. The membrane potential of TCs was described by cm dV/dt = IsmcINaIKITILIsyn. The ionic currents and gating variables, with parameter values, are detailed in the supplementary material.1 The applied (Iapp, Ismc) and synaptic (Isyn) currents are discussed in the following text.


The neurons were connected in a sparse, structured fashion by repeating, with periodic boundary conditions, the input scheme illustrating all inputs to the fourth neuron of each region in Fig. 1C: each kth neuron in GPe and GPi received excitation from STN neurons k − 1 and k and inhibition from GPe neurons k + 1 and k + 2; each kth STN neuron received inhibition from GPe neurons k and k + 1; and each TC neuron received inhibition from GPi neurons k − 1 and k. Every synaptic connection was described by a differential equation dz/dt = α(1 – z)z – βz, where z = {1 + exp[(V − θz)/σz]}−1. From each synaptic gating variable z, the synaptic current was found as: Isyn = Gsyn z (VVsyn). Parameter values for the rate constants (α and β), shape parameters (θz and σz), maximal conductance (Gsyn) and reversal potential (Vsyn) are defined for all synapses in the supplementary material. With the neuronal parameters and synaptic connections established, the tonic drive currents to the basal ganglia neurons were varied to yield firing rates and patterns as consistent as possible with published electrophysiological results from in vivo models of parkinsonism.

The STN neurons were presented with a tonic applied current, unchanged from Rubin and Terman (2004); plus a stimulatory current during DBS, and only STN neurons received DBS. The patterns of stimulatory current pulses are described in Temporally irregular DBS, and each monophasic pulse was depolarizing with an amplitude of 300 pA/μm2 and a duration of 300 μs. Without DBS, the STN neurons exhibited mean firing rates of 16.4 ± 1.4 (SD) Hz in the parkinsonian state. Although those rates were lower than published values in humans, difficulties isolating STN neurons may lead to a reported firing rate bias toward higher frequencies (Magnin et al. 2000). More importantly however, the model firing rates in the parkinsonian state were 5–10 Hz higher than in their healthy-normal condition and in the range of STN firing rates of some reported 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) treated non-human primates [e.g., 26 ± 15 Hz (Bergman et al. 1994)], although slightly below others [e.g., 36.1 ± 10.7 Hz (Soares et al. 2004)].

The average 20 pA/μm2 depolarizing current applied to the GPe neurons in control conditions was reduced to 5 pA/μm2 in the parkinsonian state to simulate increased inhibition from striatum. These values were picked on the basis that they yielded interspike interval distributions—encompassing firing rate, burstiness and irregularity—similar to those recorded from non-human primates (Dorval et al. 2008). GPe firing rates fell from 77.0 ± 15.6 Hz in control to 32.2 ± 7.1 Hz in the parkinsonian state within the ranges of neurons recorded from GPe in the non-human primate before (65.1 ± 23.6) and after (45.9 ± 18.9 Hz) MPTP exposure (Soares et al. 2004). While the parkinsonism-induced rate changes were greater in the model (44.7 ± 12.1 Hz) than in the non-human primate (19.2 ± 21.4), adjusting striatal drive to match rate more closely, eliminated the bursting and irregular firing patterns seen in vivo. Slight heterogeneity was introduced to the network by varying the striatal drive to GPe neurons: one-half of the GPe neurons (cells: 1, 3, 5,…, 15) received the average current; one-quarter of GPe neurons (cells: 2, 8, 10, 14) received 2 pA/μm2 above that average; and the final one-quarter (cells: 4, 6, 12, 16) received 2 pA/μm2 below that average. The original model (Rubin and Terman 2004; Terman et al. 2002) included parkinsonian-state modifications to the GPe to GPe synapses that were not incorporated here.

The 21 pA/μm2 depolarizing current applied to the GPi neurons in control conditions was reduced to 12 pA/μm2 in the parkinsonian state. This current was intentionally inhibitory in the parkinsonian relative to the control condition because decreasing striatal inhibition to GPi led to physiologically unrealistic hyperpathological bursts of tens to hundreds of spikes followed by brief pauses in each neuron. Firing rates in model GPi cells in the healthy (82.6 ± 3.2 Hz) and parkinsonian (70.1 ± 4.1 Hz) states compared favorably with the ranges recorded from non-human primates before (59.7 ± 16.8 or 65.1 ± 20.8 Hz) and after (59.9 ± 26.9 or 80.6 ± 19.6 Hz) MPTP exposure (Bergman et al. 1994; Soares et al. 2004) and to human participants with PD (89.3 ± 11.2 Hz) (Tang et al. 2005).

Sensory-motor input to the thalamus (Ismc) was modeled as a gamma distributed pulse train, similar to the trains used for irregular DBS stimulation. Each instantaneous frequency (i.e., the reciprocal of the time between pulses) was drawn from a gamma distribution with shape parameter κ = 25 and scale parameter θ = 2/5, which had an average of 10 ± 2 Hz. Each depolarizing monophasic sensory-motor input pulse had an amplitude of 2 pA/μm2 and a duration of 5 ms.


The computational model was used to quantify the effects of parkinsonism and temporally regular and irregular DBS on neuronal activity. Changes to the bias current of neurons in the globus pallidus shifted the model between healthy and parkinsonian conditions. The DBS and sensory-motor input pulse trains were generated in Octave, the GNU numerical computing language (//www.gnu.org/software/octave). Simulations were run in XPP-AUT (//www.math.pitt.edu/~bard/xpp/xpp.html), the nonlinear differential equation simulation package, with the fourth order Runge-Kutta solver using a maximum time step of 50 μs. The first 1.0 s of simulation time was ignored to allow for initial conditions to settle. Simulations were run for 15 s epochs, and summary results were calculated from 150 s of simulation time (Figs. 2, ,3,3, and and66).

Fig. 2.
Changes in model neuron activity. Patterns of neuronal firing activity varied across conditions and basal ganglia. A: rastergrams depicting representative firing patterns of 3 connected computational neurons, in the identified basal ganglia (labels left ...
Fig. 3.
Thalamic errors increase with DBS irregularity. Thalamic errors were identified when a model neuron failed to spike or spiked more than once to a single excitatory input or spiked in the absence of an input. Example membrane potential traces (left) from ...
Fig. 6.
Changes in synaptic input to TCs. Model GPi cells provide inhibitory synaptic input to TCs. A and B: the average synaptic conductance experienced by a TC, in a window ranging from 100 ms before to 50 ms after the onset (- - -) of each excitatory current ...

Computational model analysis

All state variables and time varying signals generated in the computational simulations were exported to binary files and loaded into Octave. Membrane potential traces were converted to spike trains, where each spike time corresponded to a moment at which the membrane potential crossed −20 mV with a positive slope. Interspike intervals (ISIs) were sorted into histogram bins of equal size in logarithmic time at 50 bins per ISI decade. For each cell, a histogram was constructed from 150 s of simulation time and normalized to yield a probability distribution of logarithmic ISIs. Distributions were averaged across all cells (Fig. 2, B–D). Firing pattern entropy (H) was estimated from the logarithmic ISI distributions to decouple changes in entropy from changes in firing rate (Dorval 2008). The first order entropy estimate was found for each neuron, H = –Σ Pk log2 Pk,where the sum is taken over all ISI bins and Pk is the probability associated with the kth bin. Reported entropy values for each region are the mean and SD across the 16 cells in that region (Fig. 2E). Higher order estimates and different entropy estimation techniques all yielded qualitatively similar results.

Model TC error rates (Guo et al. 2008; Rubin and Terman 2004) were assessed by categorizing responses to the sensory-motor input current pulses into four types of events. A correct event was recorded when a TC fired a single spike within the response window, which began with a sensory-motor pulse and lasted 25 ms. Three types of errors were recorded: a miss event when a TC did not spike in a response window; a burst event when a TC spiked more than once in a response window; and an extraneous event when a TC spike was not in any response window. Error rates were calculated as the number of error events per 15 s epoch, divided by the total number of events per epoch, normalized by 15 to yield errors per second. Reported values (Fig. 3) are the average and SDs of those error rates across all 16 TCs for 10 such epochs.

The synaptic conductance (Gsyn z) for each GPi to TC synapse was examined from 100 ms before to 50 ms after the initiation of each sensory-motor input pulse (Fig. 6, A and B). Individual traces of the same condition (i.e., PD, PD with 130 Hz, etc.) and outcome event (i.e., correct, miss, burst, or extraneous) were combined to yield the average time varying synaptic conductance for each event-condition (Guo et al. 2008). Autocorrelations of each 15 s epoch spike train were found from the waveform generated by convolving each spike train with a normalized Gaussian function with a SD of 1.0 ms. Autocorrelations were averaged across 10 epochs for each neuron, and each reported value is the mean across all 16 cells for that region (Fig. 6C). In addition, the average and SD of each individual trace was calculated for each event, and those values were averaged together to yield overall conductance means and SDs (Fig. 6, D and E).

Measurements in human participants with PD

The human participants protocol was approved by the Institutional Review Board of Duke University, and all participants provided written informed consent. The battery of the implantable pulse generator (IPG) needs replacement every 1–5 years following implantation, depending on the parameters used for chronic stimulation. Participants were chosen for the study if they had PD with bradykinesia that was treated effectively with DBS therapy through one or more existing leads and they were returning to the operating room for battery replacement surgery. Nine patients consented to participate in the study.


Participants entered the operating room when possible having forgone dopaminergic therapy for ≥12 h (6/9 subjects). Participants did not receive presurgical medications for analgesia or sedation. Individuals were resting on their backs but fully awake. Local anesthetic was applied above the IPG, and a small incision opened the subcutaneous pocket housing the IPG. The IPG was removed and disconnected from the extension cable extending to the deep brain leads.

Signal generation was performed by custom software in LabView (National Instruments, Austin TX) on an x86 laptop computer (Latitude D810, Dell, Round Rock, TX) running Window XP (Microsoft, Redmond, WA). Signals were trains of biphasic voltage pulses distributed in time as described in Temporally irregular DBS. Stimulation was presented on the electrical contacts used clinically (4/9 subjects) or a clinically unused contact on the DBS electrode was set as the current return (5/9 subjects) in participants whose clinical settings included the IPG case as a current return; there were no significant differences in the responses of those two groups. Typically, the stimulation amplitude was set to the clinically programmed values (6/9 subjects). The amplitude in one participant (4) was reduced from 3.5 to 3.0 V to minimize transient side effects associated with stimulation onset, and the amplitude in the two other participants was increased to compensate for changing the return electrode from the IPG case to a DBS lead: 4.0 V instead of 2.5 V (8) and 3.5 V instead of 3.1 V (6). The primary pulse duration was set to the clinically used value for each individual. The secondary pulses were 1/10 the amplitude of, 10 times the duration of, and immediately following the primary pulses, similar to the waveforms generated by the IPG (Table 1).

After the computer generated the condition-specific signal, a multifunction data acquisition card (PXI-6052e, National Instruments, Austin, TX) converted that digital signal into an analog output. That output signal was passed through an optical isolator (BP-Isolator, Frederic Haer, Bowdoin, ME) and into a custom passive switch box that was configured uniquely for each participant to tie the signal and return paths to the appropriate leads. The switch box, which housed an extra high-pass filter to remove any constant charge flow, fed the signal into a sterilized custom cable designed to interface with the DBS lead extension cable (Medtronic, Minneapolis, MN). This custom cable was strung into the sterile field and connected to the lead extension cable.

The hand contralateral to the stimulated brain hemisphere was placed on a two-button computer mouse on a flat surface. In each trial, participants were prompted to click alternately the two buttons with their index and middle finger as rapidly as possible. The laptop recorded time stamps for each press and release action of the mouse (Fig. 4A). Each trial lasted for ~2 min: ~100 s of rest followed by 20 or 30 s of prompted mouse button clicking. Each participant began with two to five baseline trials without DBS. Subsequently, the four classes of DBS were presented in random order for 4 min each with 4 min of no stimulation between DBS trains (Fig. 4B). Thus each participant performed two 20 or 30 s clicking trials during each DBS and each no stimulation epoch. After completing the experimental protocol, the extension lead was disconnected from the custom stimulation system and the generator replacement surgery continued. Although many participants could identify when some DBS was being presented, participants were blinded to the stimulation conditions and none expressed any discrimination between the types of DBS.

Fig. 4.
Human protocol. Human participants were tasked with tapping alternately left and right computer mouse buttons with their index and middle fingers, during presentation of the 4 patterns of DBS. A: representative examples of finger tapping from 1 participant ...


Finger tap rate and tap variability are correlated with the symptom severities measured clinically (Tavares et al. 2005).2 The tap rate for each trial was found as the number of button depressions divided by the trial duration. To pool data, each trial of a given participant was shifted by subtracting the mean tap rate of the baseline conditions for that participant. Post hoc analysis revealed that data from the first of the two trials in each condition were more correlated with data from the previous trial than from the subsequent trial. In other words, the symptomatic effects of the DBS condition became evident with a time course >2 min, consistent with previous studies of the time course of effects of DBS (Lopiano et al. 2003; Temperli et al. 2003). To minimize the effects of this nonstationarity, data from the first trial in each condition were not included in this analysis. Summary results are plotted as the mean ± SE rates across participants (Fig. 4C).

Tap variabilities were also found for each condition. Because the first trials are not included in this analysis, all data presented were collected ~210–240 s into their respective conditions. Although this 210 s delay approaches the time constant of DBS onset effects, it is too short for symptoms to return to baseline following DBS offset. Thus the finger tap variabilities of interest are not the absolute variabilities, but the change in variability from the previous condition. These measures will naturally underestimate the absolute changes, but because condition order was randomized across subjects, these measures will not introduce any biases for on or off DBS conditions. The durations of all button tap depressions {Tdur} and the intervals between button taps {Tint} were found to range over two orders of magnitude for most participants. For example, the tap durations of the example participant (Fig. 4B, left) ranged from substantially less than 1/10 of a second to substantially more than 10 s. To accommodate this change of scales, the base 10 logarithm was taken of each tap duration {log10 Tdur} and interval {log10 Tint} (e.g., Fig. 4B, right). The SDs of these logarithmic transforms of tap duration (σdur = STD{log10 Tdur}) and interval (σint = STD{log10 Tint}) were found for each finger of each participant in each trial. To accommodate slow nonstationarities—due to changes in alertness, effort, or enduring effects from the previous DBS condition—the SD from each condition was subtracted from that of the subsequent condition to yield shifts in the variability of duration (Δσdurk +1 = σdurk +1 − σdurk) and interval (Δσintk +1 = σintk +1 − σintk).

Statistical analysis was performed on those normally distributed variability shifts (i.e., {Δσdur} and {Δσint}) pooled together by condition across all participants. Because all measures are changes from the previous condition, the absolute severity of the baseline bradykinesia has minimal effect on the statistics; whether subjects were on (3/9) or off (6/9) dopaminergic medication did not affect subsequent results. The reported values were found as the inverse base 10 logarithm of the variability shifts, to use the more intuitive units of variability gains, ξdur = log10-1 Δσdur and ξint = log10-1 Δσint, but are presented on logarithmically graduated axes to highlight their underlying logarithmic nature (Fig. 5). A bootstrap procedure was used to estimate the median and confidence intervals of the mean. The mean of a population of random samples drawn with replacement from the equally numerous population of variability shifts was found 10,000 times for each condition and sorted to yield confidence intervals (Figs. 5 and and77).

Fig. 5.
Regular DBS reduces tapping variability. The variability of finger tap times was quantified for each participant in each condition, as tap duration variability (ξdur) or tap interval variability (ξint). See Measurements in human participants ...
Fig. 7.
Expected improvements in UPDRSIII induced by different DBS trains. The logarithm of the coefficient of variation of the finger tap durations correlates with changes in the clinically used UPDRSIII motor scores (Tavares et al. 2005). Across nine participants, ...


We measured the effects of temporally regular and temporally irregular high-frequency DBS on motor symptoms in persons with PD and on neuronal activity in a computational model of the basal ganglia thalamic circuit. Four patterns of DBS were constructed to have the same average instantaneous frequency, but differing degrees of regularity. A temporally regular pattern consisted of identically spaced pulses at 130 Hz. For the other three temporally irregular patterns, the instantaneous frequencies were drawn from a gamma distribution with a mean of 130 Hz and a SD equal to 10, 30, or 60% of the mean (Fig. 1, A or B).

Error rates in model TCs track DBS irregularity

A computational model of the basal ganglia thalamic network (Fig. 1C) containing 64 point neurons with voltage-dependent conductances was used to quantify the effects of temporally regular and irregular DBS on neuronal activity. Changes to the bias current of neurons in the globus pallidus shifted the model between healthy and parkinsonian states. In the parkinsonian state, four patterns of DBS were used to activate neurons in the STN.

Changing the model from the healthy to the parkinsonian state increased the firing rate of neurons in the STN. Regular DBS (130 Hz) further increased the average firing rate but also regularized the inter-spike intervals (ISIs, Fig. 2B). Increasingly irregular DBS had the same effect on firing rate as regular DBS but decreased the ISI regularity, broadening the ISI distributions (Fig. 2B).

In contrast to disease-induced rate changes in STN, neuronal firing rates in the GPe dramatically decreased from the healthy to the parkinsonian condition (Fig. 2C). Similar to changes observed in STN, however, the firing rates of GPe neurons increased in response to all patterns of DBS. During regular DBS, GPe firing patterns became more regular, exhibiting fewer bursts and fewer long pauses. Also similar to STN, the ISIs in GPe became increasingly irregular with increasingly irregular DBS (Fig. 2C).

Shifting from the healthy to the parkinsonian state did not have a large effect on the average firing rate of neurons in the GPi but substantially altered the firing patterns. The ISI distribution shifted from primarily unimodal centered around 15 ms in the healthy condition to bimodal in the parkinsonian condition with a broad intra-burst mode centered around 8 ms and a broad inter-burst mode centered around 25 ms (Fig. 2D). All patterns of DBS increased the firing rate of GPi model neurons but had different effects on the patterns of neuronal activity. Regular DBS converted the two broad modes into three relatively narrow modes, representing the GPi neurons firing phase-locked to the DBS pulses in ratios of 1:1, 2:3, or 1:2 (Fig. 2D). Irregular DBS generated broad ISI distributions more similar to the parkinsonian state than to the regular DBS state.

In summary, firing rate changes between the healthy and parkinsonian states were different across all three regions: increasing, decreasing, and nearly constant in STN, GPE, and GPi, respectively. The rate changes generated by DBS amplified those of parkinsonism in STN, counter-acted them in GPe, and simply increased the rate in GPi, regardless of DBS regularity. However, irregular DBS produced ISI distributions that were qualitatively similar to the parkinsonian state, whereas regular DBS produced ISI distributions qualitatively similar to the healthy state. To quantify these similarities, the firing pattern entropy was calculated from each ISI distribution. In all regions, firing pattern entropy increased between the healthy and parkinsonian conditions (Fig. 2E). Regular DBS decreased firing pattern entropy to below the levels observed in the healthy condition. Increasingly irregular DBS increased the firing pattern entropy, which returned to or exceeded the unstimulated parkinsonian level in response to the most irregular pattern, 130 ± 60%.

The fidelity of thalamic transmission was quantified to estimate the functional impact of different patterns of DBS. TCs, in addition to receiving inhibitory synaptic input from GPi cells, received excitatory current pulses that represented motor information passing through the thalamus. Responses were counted as correct when a TC spiked once in response to one excitatory input, and responses were counted as errors when a TC did not spike in response to an input (miss), spiked more than once in response to an input (burst), or spiked in the absence of an input. While errors were rare in the healthy condition (Fig. 3), they were commonplace in the parkinsonian condition (DBS off). Regular DBS reduced the error rate to almost the healthy level. Increasingly irregular DBS increased the thalamic error rate, and with 130 ± 60% DBS, errors were more common than with no stimulation. Thus thalamic errors were not corrected by high-frequency DBS unless the pulse train was highly regular.

Human bradykinesia tracks DBS irregularity

We tested the same four patterns of DBS in human participants to determine whether high-frequency DBS also had to be highly regular for therapeutic effectiveness. The implanted pulse generator used for clinical DBS can only generate regular pulse patterns. Therefore we connected an external pulse generator to the DBS brain leads of participants with PD whose motor symptoms were managed effectively by DBS therapy during battery replacement surgery. Four patterns of DBS were delivered through their previously implanted DBS leads for 4 min epochs using a randomized block design. At regular intervals during and between epochs, bradykinesia was quantified by instructing participants to tap alternately the two buttons of a computer mouse as quickly as possible for 20- or 30-s trials.

Example trials depict the raw tap durations collected from three different stimulation conditions in one participant (Fig. 4A). Data summaries for that participant in all conditions are presented in the order in which the data were collected (Fig. 4B, left). Tap durations for each condition were approximately log-normally distributed (Fig. 4B, right). Changes in tap rate, from the patient-specific mean baseline, were found for each trial, and data from all participants were combined by condition (n = 42, 54, 18, 18, 16, and 18 for baseline, recovery, 130 Hz, 130 ± 10%, 130 ± 30%, and 130 ± 60%, respectively). Omnibus van der Waerden testing (data not normally distributed) yielded a significant effect of therapy condition (P = 0.041) when the stimulation off conditions (i.e., baseline and recovery) were combined into a single group.

The eight pairwise comparisons between the two hypothesized asymptomatic conditions (130 Hz and 130 ± 10%), and the other four conditions were evaluated with the Mann-Whitney U test. Five hypotheses were significant (α = 0.05, uncorrected): 130 Hz increased the tap rate over baseline (P = 0.0024), recovery (P = 0.012), 130 ± 30% (P = 0.038), and 130 ± 60% (P = 0.044); and 130 ± 10% increased the tap rate over baseline (P = 0.031). In other words, regular 130 Hz DBS increased the finger tap rate over the four hypothesized symptomatic conditions, and nearly regular 130 ± 10% DBS increased finger tap rate over baseline. Employing the Holm-Bonferroni method to account for multiple comparisons with eight hypotheses and the same statistical criterion (m = 8, α = 0.05), one hypothesis was highly significant: 130 Hz increased tap rate from baseline.

Changes in the tap duration variability (the SD of the logarithms of finger tap durations) and the tap interval variability (the SD of the logarithms of the intervals between finger taps) were calculated for each trial (Fig. 5), and data from all participants were combined by condition. Analyses of variance showed that changes in variability of both tap duration (P = 0.0080) and tap interval (P = 0.028) were dependent on stimulation condition. Regular and nearly regular (130 ± 10%) DBS decreased tap duration variability and tap interval variability relative to the DBS off and the highly irregular DBS conditions, while neither of the highly irregular conditions changed either the tap duration or tap interval variability relative to DBS off.

Specifically, the data support three of the six hypotheses regarding tap duration: that the two hypothesized asymptomatic conditions decreased tap duration variability shifts with respect to the other three conditions. DBS off increased duration variability from 130 Hz (P = 0.00074) and 130 ± 10% (P = 0.0063); and 130 ± 60% increased duration variability from 130 Hz (P = 0.033). After correcting for multiple comparisons (m = 6, α = 0.05), the first two hypotheses remained highly significant: DBS off increased duration variability over 130 Hz and 130 ± 10%. Additionally, the data support all six hypotheses regarding tap intervals: that the two hypothesized asymptomatic conditions decreased tap interval variability shifts with respect to the other three conditions. Regular DBS decreased interval variability from DBS off (P = 0.00050), 130 ± 30% (P = 0.012), and 130 ± 60% (P = 0.0043); and 130 ± 10% decreased interval variability from DBS off (P = 0.0081), 130 ± 30% (P = 0.040), and 130 ± 60% (P = 0.016). After correcting for multiple comparisons (m = 6, α = 0.05), all six hypotheses remained highly significant.

Pallidal variability determines error rate in model TCs

To examine the cause of irregular-DBS-induced bradykinesia in the human participants and thalamic errors in the model, the inhibitory synaptic conductances from model GPi cells to TCs were measured between 100 ms before and 50 ms after each excitatory input. Thalamic miss errors were associated with a prolonged (100 to 10 ms before the input pulse) weaker-than-average inhibitory conductance followed by increased inhibition coincident with the excitatory input (Fig. 6A). Thalamic burst errors were associated with a prolonged (100 to 10 ms before the input pulse) stronger-than-average inhibitory conductance followed by decreased inhibition coincident with the excitatory input (Fig. 6B). Averaged across all similar events, the amplitudes of these changes in inhibitory conductance were larger in both the unstimulated parkinsonian and irregular DBS states than for either the healthy or regular DBS states.

While the inhibitory conductance profile around error events explained why a TC may have made a particular type of error, it did not explain why different conditions yielded different error rates. To explore the statistical behavior of the inhibitory synaptic inputs, the GPi neuron spike-time autocorrelations were calculated (Fig. 6C). Autocorrelations in the healthy condition were similar to those in the parkinsonian condition and during irregular DBS: small rapidly decaying oscillations that reflect primarily the neuronal refractory period. However, autocorrelations during regular DBS differed dramatically, exhibiting long-lasting oscillations reflecting the DBS periodicity. Thus thalamic error rates could not be predicted from GPi spike-time autocorrelations.

The mean and SD of the inhibitory synaptic conductance were calculated for each thalamic neuron in response to each excitatory input. While the mean conductance across all conditions was roughly equivalent (Fig. 6D), the SD of the inhibitory conductance differed dramatically across conditions (E). Highly variable conductances yielded high thalamic error rates, while minimally variable conductances yielded low thalamic error rates. Additionally, the condition specific ranges of the SDs were so small that essentially no cases overlapped between the high error rate (i.e., symptomatic) and low error rate (i.e., asymptomatic) conditions.

In summary, GPi cell autocorrelations were poor predictors of thalamic errors. The swings in inhibitory conductance that lead to misses and bursts were relatively independent of the average synaptic input, which varied widely within all conditions but not systematically across conditions. Rather the temporal profile and variability of the synaptic conductance, while consistent within each condition, changed dramatically across conditions and were correlated strongly with the thalamic error rate.


DBS of the STN or GPi is an effective therapy for the motor symptoms of PD, but the mechanisms of symptom alleviation are not fully understood. Work over the past decades has suggested that symptom progression in PD is the behavioral manifestation of increasingly pathological neuronal activity in the basal ganglia. In addition to changes in neuronal firing rates, changes in the firing patterns of neurons and neural assemblies accompany the onset of parkinsonism in nonhuman primates (Bergman et al. 1994; Legéndy and Salcman 1985; Wichmann and DeLong 2003; Wichmann and Soares 2006) and persons with PD (Lenz et al. 1994; Magnin et al. 2000; Tang et al. 2005). In this study, we provide evidence for causality of the relationship between neuronal firing pattern variability and symptom severity by showing that for a fixed frequency of DBS, reducing the output variability of basal ganglia neurons is necessary for symptom alleviation in human participants with PD.

Irregular stimulation and symptom severity

We hypothesized that masking the parkinsonism-related pathological activity with regular DBS-induced activity would alleviate the motor symptoms of human participants with PD and that masking the pathological activity with irregular DBS-induced activity would fail to alleviate the same motor symptoms. We tested these hypotheses by substituting clinically effective temporally regular DBS with temporally irregular DBS during previously scheduled surgery to replace the depleted pulse generator. Highly irregular DBS, even when delivered at an effective average frequency, did not improve bradykinesia. However, regular DBS at the same frequency lead to faster (Fig. 4C) and more regular (Fig. 5) motor control.

In a related study, Tavares and colleagues (2005) showed that the logarithms of both the mean and the coefficient of variation of the button hold duration were correlated with the Unified Parkinson's Disease Rating Scale, Section III subscore (UPDRSIII). Because the logarithm of the coefficient of variation of tap duration had been mapped to UPDRSIII scores directly and was the mostly highly correlated measure found (Tavares et al. 2005), we calculated that measure for the nine participants in our study. Multiplying those values by the reported correlation coefficient (R = 0.66) and scaling by the gain (80 UPDRSIII points per log unit), we calculated predicted UPDRSIII score shifts from the unstimulated case (DBS Off) for all DBS conditions (Fig. 7). The estimated improvements in UPDRSIII scores diminished with increasingly irregular DBS.

These findings extend previous work highlighting the effects of DBS pattern regularity, as opposed to DBS rate, on motor symptom severity. In particular, rapidly cycling DBS on and off, thereby creating DBS patterns that were less regular on short time scales, alleviated motor symptoms less effectively than regular DBS (Montgomery 2005). Also in a population of human participants with heterogeneous tremor disorders, regular DBS alleviated tremor, whereas irregular DBS did not (Birdno et al. 2008). Thus independent of the type of irregular DBS or the particular neurological disorders, symptoms may be maximally alleviated when stimulation is maximally regular.

Irregular neuronal activity and thalamic errors

The symptoms of parkinsonism are accompanied by a transition to irregular and burst-like firing patterns in rodents (Degos et al. 2005), non-human primates (Bar-Gad et al. 2004; Hashimoto et al. 2003; Meissner et al. 2005; Wichmann and DeLong 2003) and human participants with PD (Brown et al. 2004). We used a computational model of the basal ganglia thalamic network, which captured this transition, to quantify the effects of different temporal patterns of DBS on neuronal firing patterns. The computational network did not exhibit the highly regular bursting described in previous implementations of the model but did yield burst-like events mixed with nonbursting irregular activity (Fig. 2), similar to that recorded from basal ganglia output neurons in parkinsonian models in rodents (Degos et al. 2005; Shi et al. 2006), non-human primates (Bergman et al. 1994; Dorval et al. 2008; Hashimoto et al. 2003; Wichmann and DeLong 2003) and humans with PD (Tang et al. 2005).

Because all activity within the model basal ganglia was funneled to the thalamus via the GPi, we focused on the spiking and synaptic activity of GPi neurons (Fig. 6). The average synaptic outputs of GPi neurons to TCs were not changed substantially among control, parkinsonian, and any DBS conditions. However, the variability of these outputs varied markedly between conditions. The importance of firing patterns revealed by this model reinforces experimental studies that found average changes in GPi firing rates to be negligible (Bergman et al. 1994) or overshadowed by large variability across cells in any given condition (Hashimoto et al. 2003; Soares et al. 2004).

The parallel between symptom alleviation in human participants and DBS-induced firing pattern changes in the model is consistent with previous computational findings (Rubin and Terman 2004), symptom-correlated changes in bursting activity in the 6-OHDA-rat (Shi et al. 2006) and firing pattern entropy changes in the MPTP-primate (Dorval et al. 2008) in response to DBS. In a computational model similar to the one used here, thalamic neurons presented with parkinsonian firing patterns exhibited reduced fidelity, whereas model neurons presented with regular DBS-induced firing patterns exhibited improved fidelity (Guo et al. 2008). Similarly, in a preliminary study on healthy non-human primates, replacing normal firing patterns with firing patterns recorded from animals in the parkinsonian state induced parkinsonian motor symptoms (Ma Y and Wichmann T, Society for Neuroscience Abstracts 2004).


Collectively, these data suggest that irregular activity is somehow responsible for parkinsonian symptoms. Furthermore, effective therapy must alleviate those symptoms either by adding some new activity or by removing the existing irregular activity. Surgical history indicates that lesion and DBS are similarly effective, implicating the removal of irregular activity as a mechanism of symptom alleviation. Therapeutic lesion removes the pathological activity by eliminating the neural substrate in which it resides; therapeutic DBS merely masks the pathological activity with periodic pulse trains. Indeed, DBS that masks the activity with aperiodic pulse trains does not alleviate symptoms. However, any means of eliminating irregular activity in GPi should reduce information processing errors by thalamic neurons and alleviate symptoms. Indeed Tass and colleagues have proposed one such alternative approach—using well timed stimulus bursts to desynchronize pathological network activity—and shown that it can be effective in computational models of PD (Hauptmann et al. 2005; Tass 2003) and in vitro models of pathological synchronization (Tass et al. 2009).

The present results demonstrate that for a fixed frequency of DBS, the pattern of stimulation has dramatic effects on parkinsonian bradykinesia, highlighting the importance of neuronal firing patterns independent of firing rate. While regular DBS regularized firing patterns, reduced synaptic variability and increased thalamic fidelity, irregular stimulation at the same average frequency simply replaced the pathological disease-induced patterns with pathological stimulation-induced patterns. Irregular DBS was unable to reduce the rate of errors made by TCs in the model and was unable to alleviate bradykinesia in human participants with PD. Four potential mechanisms have been offered to explain symptom relief by DBS (Garcia et al. 2005; Grill and McIntyre 2001): depolarization blockade of neurons around the electrode by activation or inactivation of voltage-gated currents (Beurrier et al. 2001), depression or failure of synaptic transmission resulting in blockade of output from the stimulated neurons (Anderson et al. 2006), synaptic inhibition of the neurons around the electrode by activation of local inhibitory (GABAergic) axon terminals (Filali et al. 2004), and alteration or “jamming” of pathological patterns of activity (Rubin and Terman 2004; Vitek 2002). The data presented here strongly support the fourth potential mechanism of regularization of pathological patterns of activity in the basal ganglia (Birdno and Grill 2008) and argue against the first three potential mechanisms as it is not clear how such mechanisms would depend on the fine temporal structure of the stimulation train.

Understanding the time scales that divide DBS patterning from DBS rate is important to this work and to our conclusion that fine temporal structure matters. During irregular DBS, the time between any two DBS pulses often deviated beyond the therapeutic range. However, the average rate over many consecutive pulses rarely left the therapeutic range. As an example, the time between any two pulses in the 30% irregular DBS case was highly variable, ranging from <2.5 to >25 ms (Fig. 1). In contrast to this pattern variability, the DBS rate was fairly constant. In particular, the probability that 1 s elapsed during which DBS was outside of the clinically accepted therapeutic range (i.e., DBS rate was <100 Hz), was <0.001. In other words, for every 1,000 s of 30% irregular DBS, only 1 s was outside of the therapeutic range comprising <100 pulses. And yet even though the DBS rate was in the therapeutic range 99.9% of the time, symptoms were not alleviated. Thus while DBS rate may play a therapeutic role, we have shown that high-frequency alone is insufficient: DBS patterning is a critical component of symptom alleviation.

Recent work suggests that effective DBS of the STN may be mediated via activation of afferents projecting from layer V of motor cortex (Gradinaru et al. 2009; Li et al. 2007), a possibility that our computational model does not address. However, when participants with STN electrodes were stimulated with irregular DBS, their bradykinetic symptoms were not alleviated. Thus whether the electrophysiological mechanisms of DBS therapy are orthodromic through the basal ganglia thalamic cortical loop, or antidromic directly to motor cortex, regularizing neuronal activity is a necessity for symptom alleviation. Incorporating our model, we provide causal evidence that DBS of the STN alleviates the symptoms of PD by regularizing the firing patterns of basal ganglia neurons, enabling motor thalamic neurons to process what residual streams of information remain in the parkinsonian state.

Applying this understanding to other disorders, electrical stimulation may be beneficial to disorders accompanied by aberrant firing patterns within identified brain regions (Llinás et al. 1999). However, each brain region will require a different minimum stimulation frequency to mask the pathological pattern of neuronal activity (Grill et al. 2004). Regular DBS eliminates pathological firing patterns but does not reintroduce the healthy firing patterns present in the absence of disease. The high-frequency tonic firing imposed on the basal ganglia by DBS masks the information that those neurons would otherwise convey, eliminating all normal output messages to thalamus. While motor symptoms were better treated by regular DBS than irregular DBS at the same high frequency, there is no reason to believe that regular DBS trains are optimal (Feng et al. 2007). The most desirable interventions would eliminate the pathological firing patterns associated with PD without imposing new firing patterns, thereby allowing the healthy normal firing patterns to reemerge.


This work supported by the National Institute of Neurological Disorders and Stroke Grants K25-NS-053544 to A. D. Dorval and R01-NS-040894 to W. M. Grill.


No conflicts of interest, financial or otherwise, are declared by the author(s).

Supplementary Material

[Supplemental Data]


1The online version of this article contains supplemental data.

2If the coefficient of variation and the mean both increase then their product, the SD, will increase more robustly. Additionally, because the variability of tap duration scales with the mean (Fig. 4B, right), the distribution of tap durations (and intervals) are log-normally distributed. Thus we used the standard deviation of the logarithmic transform of the tap durations and intervals.


  • Anderson ME, Postupna N, Ruffo M. Effects of high-frequency stimulation in the internal globus pallidus on the activity of thalamic neurons in the awake monkey. J Neurophysiol 89: 1150–1160, 2003. [PubMed]
  • Anderson TR, Hu B, Iremonger K, Kiss ZHT. Selective attenuation of afferent synaptic transmission as a mechanism of thalamic deep brain stimulation-induced tremor arrest. J Neurosci 26: 841–850, 2006. [PubMed]
  • Bar-Gad I, Elias S, Vaadia E, Bergman H. Complex locking rather than complete cessation of neuronal activity in the globus pallidus of a 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-treated primate in response to pallidal microstimulation. J Neurosci 24: 7410–7419, 2004. [PubMed]
  • Bergman H, Wichmann T, Karmon B, DeLong MR. The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of parkinsonism. J Neurophysiol 72: 507–520, 1994. [PubMed]
  • Beurrier C, Bioulac B, Audin J, Hammond C. High-frequency stimulation produces a transient blockade of voltage-gated currents in subthalamic neurons. J Neurophysiol 85: 1351–1356, 2001. [PubMed]
  • Birdno MJ, Cooper SE, Rezai AR, Grill WM. Pulse-to-pulse changes in the frequency of deep brain stimulation affect tremor and modeled neuronal activity. J Neurophysiol 98: 1675–1684, 2007. [PubMed]
  • Birdno MJ, Grill WM. Mechanisms of deep brain stimulation in movement disorders as revealed by changes in stimulus frequency. Neurotherapeutics 5: 14–25, 2008. [PMC free article] [PubMed]
  • Birdno MJ, Kuncel AM, Dorval AD, Turner DA, Grill WM. Tremor varies as a function of the temporal regularity of deep brain stimulation. Neuroreport 19: 599–602, 2008. [PMC free article] [PubMed]
  • Boraud T, Bezard E, Bioulac B, Gross C. High frequency stimulation of the internal globus pallidus (GPi) simultaneously improves parkinsonian symptoms and reduces the firing frequency of GPi neurons in the MPTP-treated monkey. Neurosci Lett 215: 17–20, 1996. [PubMed]
  • Brown P, Mazzone P, Oliviero A, Altibrandi MG, Pilato F, Tonali PA, Di Lazzaro V. Effects of stimulation of the subthalamic area on oscillatory pallidal activity in Parkinson's disease. Exp Neurol 188: 480–490, 2004. [PubMed]
  • Degos B, Deniau J, Thierry A, Glowinski J, Pezard L, Maurice N. Neuroleptic-induced catalepsy: electrophysiological mechanisms of functional recovery induced by high-frequency stimulation of the subthalamic nucleus. J Neurosci 25: 7687–7696, 2005. [PubMed]
  • Dorval AD. Probability distributions of the logarithm of inter-spike intervals yield accurate entropy estimates from small datasets. J Neurosci Methods 173: 129–139, 2008. [PMC free article] [PubMed]
  • Dorval AD, Russo GS, Hashimoto T, Xu W, Grill WM, Vitek JL. Deep brain stimulation reduces neuronal entropy in the MPTP-primate model of Parkinson's disease. J Neurophysiol 100: 2807–2818, 2008. [PMC free article] [PubMed]
  • Feng X, Shea-Brown E, Greenwald B, Kosut R, Rabitz H. Optimal deep brain stimulation of the subthalamic nucleus–a computational study. J Comput Neurosci 23: 265–282, 2007. [PubMed]
  • Filali M, Hutchison WD, Palter VN, Lozano AM, Dostrovsky JO. Stimulation-induced inhibition of neuronal firing in human subthalamic nucleus. Exp Brain Res 156: 274–281, 2004. [PubMed]
  • Garcia L, D'Alessandro G, Bioulac B, Hammond C. High-frequency stimulation in Parkinson's disease: more or less? Trends Neurosci 28: 209–216, 2005. [PubMed]
  • Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K. Optical deconstruction of parkinsonian neural circuitry. Science 324: 354–359, 2009. [PubMed]
  • Grill WM, McIntyre CC. Extracellular excitation of central neurons: implications for the mechanisms of deep brain stimulation. Thalamus Relat Syst 1: 269–277, 2001
  • Grill WM, Snyder AN, Miocinovic S. Deep brain stimulation creates an informational lesion of the stimulated nucleus. Neuroreport 15: 1137–1140, 2004. [PubMed]
  • Guo Y, Rubin JE, McIntyre CC, Vitek JL, Terman D. Thalamocortical relay fidelity varies across subthalamic nucleus deep brain stimulation protocols in a data-driven computational model. J Neurophysiol 99: 1477–1492, 2008. [PubMed]
  • Hashimoto T, Elder CM, Okun MS, Patrick SK, Vitek JL. Stimulation of the subthalamic nucleus changes the firing pattern of pallidal neurons. J Neurosci 23: 1916–1923, 2003. [PubMed]
  • Hauptmann C, Popovych O, Tass PA. Effectively desynchronizing deep brain stimulation based on a coordinated delayed feedback stimulation via several sites: a computational study. Biol Cybern 93: 463–470, 2005. [PubMed]
  • Hershey T, Revilla FJ, Wernle AR, McGee-Minnich L, Antenor JV, Videen TO, Dowling JL, Mink JW, Perlmutter JS. Cortical and subcortical blood flow effects of subthalamic nucleus stimulation in PD. Neurology 61: 816–821, 2003. [PubMed]
  • Kuncel AM, Cooper SE, Wolgamuth BR, Grill WM. Amplitude- and frequency-dependent changes in neuronal regularity parallel changes in tremor with thalamic deep brain stimulation. IEEE Trans Neural Syst Rehabil Eng 15: 190–197, 2007. [PubMed]
  • Legéndy CR, Salcman M. Bursts and recurrences of bursts in the spike trains of spontaneously active striate cortex neurons. J Neurophysiol 53: 926–939, 1985. [PubMed]
  • Lenz FA, Kwan HC, Martin RL, Tasker RR, Dostrovsky JO, Lenz YE. Single unit analysis of the human ventral thalamic nuclear group. Tremor-related activity in functionally identified cells. Brain 117: 531–543, 1994. [PubMed]
  • Li S, Arbuthnott GW, Jutras MJ, Goldberg JA, Jaeger D. Resonant antidromic cortical circuit activation as a consequence of high-frequency subthalamic deep-brain stimulation. J Neurophysiol 98: 3525–3537, 2007. [PubMed]
  • Llinás RR, Ribary U, Jeanmonod D, Kronberg E, Mitra PP. Thalamocortical dysrhythmia: a neurological and neuropsychiatric syndrome characterized by magnetoencephalography. Proc Natl Acad Sci USA 96: 15222–15227, 1999. [PMC free article] [PubMed]
  • Lopiano L, Torre E, Benedetti F, Bergamasco B, Perozzo P, Pollo A, Rizzone M, Tavella A, Lanotte M. Temporal changes in movement time during the switch of the stimulators in Parkinson's disease patients treated by subthalamic nucleus stimulation. Eur Neurol 50: 94–99, 2003. [PubMed]
  • Magnin M, Morel A, Jeanmonod D. Single-unit analysis of the pallidum, thalamus and subthalamic nucleus in parkinsonian patients. Neuroscience 96: 549–564, 2000. [PubMed]
  • McIntyre CC, Grill WM, Sherman DL, Thakor NV. Cellular effects of deep brain stimulation: model-based analysis of activation and inhibition. J Neurophysiol 91: 1457–1469, 2004. [PubMed]
  • Meissner W, Leblois A, Hansel D, Bioulac B, Gross CE, Benazzouz A, Boraud T. Subthalamic high frequency stimulation resets subthalamic firing and reduces abnormal oscillations. Brain 128: 2372–2382, 2005. [PubMed]
  • Montgomery EB. Effect of subthalamic nucleus stimulation patterns on motor performance in Parkinson's disease. Parkinsonism Relat. Disord 11: 167–171, 2005. [PubMed]
  • Okun MS, Vitek JL. Lesion therapy for Parkinson's disease and other movement disorders: update and controversies. Mov Disord 19: 375–389, 2004. [PubMed]
  • Phillips MD, Baker KB, Lowe MJ, Tkach JA, Cooper SE, Kopell BH, Rezai AR. Parkinson disease: pattern of functional MR imaging activation during deep brain stimulation of subthalamic nucleus–initial experience. Radiology 239: 209–216, 2006. [PubMed]
  • Rubin JE, Terman D. High frequency stimulation of the subthalamic nucleus eliminates pathological thalamic rhythmicity in a computational model. J Comput Neurosci 16: 211–235, 2004. [PubMed]
  • Shi L, Luo F, Woodward DJ, Chang J. Basal ganglia neural responses during behaviorally effective deep brain stimulation of the subthalamic nucleus in rats performing a treadmill locomotion test. Synapse 59: 445–457, 2006. [PubMed]
  • Soares J, Kliem MA, Betarbet R, Greenamyre JT, Yamamoto B, Wichmann T. Role of external pallidal segment in primate parkinsonism: comparison of the effects of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-induced parkinsonism and lesions of the external pallidal segment. J Neurosci 24: 6417–6426, 2004. [PubMed]
  • Stefani A, Fedele E, Galati S, Pepicelli O, Frasca S, Pierantozzi M, Peppe A, Brusa L, Orlacchio A, Hainsworth AH, Gattoni G, Stanzione P, Bernardi G, Raiteri M, Mazzone P. Subthalamic stimulation activates internal pallidus: evidence from cGMP microdialysis in PD patients. Ann Neurol 57: 448–452, 2005. [PubMed]
  • Tang JKH, Moro E, Lozano AM, Lang AE, Hutchison WD, Mahant N, Dostrovsky JO. Firing rates of pallidal neurons are similar in Huntington's and Parkinson's disease patients. Exp Brain Res 166: 230–236, 2005. [PubMed]
  • Tass PA. A model of desynchronizing deep brain stimulation with a demand-controlled coordinated reset of neural subpopulations. Biol Cybern 89: 81–88, 2003. [PubMed]
  • Tass PA, Silchenko AN, Hauptmann C, Barnikol UB, Speckmann E. Long-lasting desynchronization in rat hippocampal slice induced by coordinated reset stimulation. Phys Rev E Stat Nonlin Soft Matter Phys 80:011902, 2009. [PubMed]
  • Tavares ALT, Jefferis GSXE, Koop M, Hill BC, Hastie T, Heit G, Bronte-Stewart HM. Quantitative measurements of alternating finger tapping in Parkinson's disease correlate with UPDRS motor disability and reveal the improvement in fine motor control from medication and deep brain stimulation. Mov Disord 20: 1286–1298, 2005. [PubMed]
  • Temperli P, Ghika J, Villemure J, Burkhard PR, Bogousslavsky J, Vingerhoets FJG. How do parkinsonian signs return after discontinuation of subthalamic DBS? Neurology 60: 78–81, 2003. [PubMed]
  • Terman D, Rubin JE, Yew AC, Wilson CJ. Activity patterns in a model for the subthalamopallidal network of the basal ganglia. J Neurosci 22: 2963–2976, 2002. [PubMed]
  • Vitek JL. Mechanisms of deep brain stimulation: excitation or inhibition. Mov Disord 17Suppl 3: S69–72, 2002. [PubMed]
  • Wichmann T, DeLong MR. Pathophysiology of Parkinson's disease: the MPTP primate model of the human disorder. Ann NY Acad Sci 991: 199–213, 2003. [PubMed]
  • Wichmann T, Soares J. Neuronal firing before and after burst discharges in the monkey basal ganglia is predictably patterned in the normal state and altered in parkinsonism. J Neurophysiol 95: 2120–2133, 2006. [PubMed]
  • Windels F, Bruet N, Poupard A, Feuerstein C, Bertrand A, Savasta M. Influence of the frequency parameter on extracellular glutamate and gamma-aminobutyric acid in substantia nigra and globus pallidus during electrical stimulation of subthalamic nucleus in rats. J Neurosci Res 72: 259–267, 2003. [PubMed]

Articles from Journal of Neurophysiology are provided here courtesy of American Physiological Society

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...