Overview of motor control in the CNS

There are many competing theories of motor control; however, most assume that at some level a motor plan exists to provide an overall strategy for the performance of a movement (13; 71). The motor plan consists of sequences of learned commands, called motor programs, which have been learned during the performance of previous actions. Each motor program is in turn comprised of a sequence or cascade of smaller learned subroutines of muscle activities, each triggering the next subprogram in sequence. Motor programs are likely formulated with their appropriate timing before the beginning of movement so that the entire sequence of movements can be performed in the absence of sensory feedback. The skill in the performance of any movement, therefore, depends very much on how well these components fit together. When an overall program fails, the brain must resort to recomposing the intended movement from simple sub-movements.

Organization of motor control. The neural substrates of motor control are distributed throughout the neuraxis. They include the sensorimotor cortex, limbic, association, and paralimbic cortices; the basal ganglia; the cerebellum; the brainstem premotor and motor nuclei, and the motor-neurons and interneurons of the brainstem and spinal cord. Feedback information from peripheral receptors, as well as inputs from the basal ganglia and cerebellum, project to the thalamus and cerebral cortex, where it is incorporated into the motor circuits. The neural processing within these structures is organized hierarchically and in parallel with additional lateral interactions between motor circuits (22).

Hierarchical organization. Motor control as classically conceived consists of a hierarchy of command. The highest level of organization resides in the limbic, paralimbic, and association cortices and their subcortical connections. These circuits gain access to the motor system primarily via connections to the anterior cingulate cortex, which in turn provides the drive to respond to behaviorally relevant stimuli and to enable the association cortices to convert needs into goals (220).

The regions of association cortex at the highest level of the motor hierarchy are the dorsal lateral prefrontal cortex, which provides executive control of goal-directed behavior, and the posterior parietal cortex, which integrates visual, auditory, and somatosensory information to provide an interface between the sensory cortex and the frontal motor cortical areas. The activity of the posterior parietal cortex is also integral to the formation of the intention to move, which is the earliest motor plan of the movement, specifying both the type of movement and its goal (08).

A middle level of motor control converts the motor plans previously generated in the association cortex into motor programs, providing a specific plan for how the movement is to be performed. This middle level of control is hypothesized to reside in the sensory motor cortex, the motor loops of the basal ganglia, the cerebellum, and the motor nuclei of the brainstem. These areas, together with the primary motor cortex, then project to brainstem and the ventral horn of the spinal cord.

At the lowest level of the control hierarchy, motor neurons and interneurons integrate information provided by the multiple descending and segmental influences and translate the instructions for movement into motor commands to muscles. Segmental reflexes arising from muscle spindle receptors and Golgi tendon organ receptors also regulate the activity of these neurons. The actual state of the limb (its position, force, speed, and stiffness) serves as an external feedback that is monitored at various levels of the central nervous system.

Parallel processing and lateral interactions: the cerebral cortex, basal ganglia, and cerebellar circuits. Cortical projections engage both the basal ganglia circuits and the cerebellum to help control motor programs. The cortical input to the basal ganglia is predominately ipsilateral and is topographically and functionally organized (described below). The outputs from the basal ganglia are subdivided into motor, oculomotor, associative, and limbic domains, each connecting via different portions of the thalamus in partially segregated, parallel, and closed-loop interactions with the cortical areas from which they originally received their input. These loops also interact laterally with each other via open-loop connections (06).

Collaterals from pyramidal tract neurons in motor cortex also continually inform the cerebellum of cortical events by sending the pontine nuclei a “reference copy” of any signal that has been sent to the spinal cord or brainstem to initiate movement. This cortico-ponto-cerebellar pathway represents the major input into the contralateral cerebellum. The cerebellum, via excitatory output from the deep cerebellar nuclei, then projects back to motor relay areas in the contralateral hemisphere.

One of the primary roles of the cerebellum is to sequence and coordinate intended movements. The 2 components of this process are reflected in a functional subdivision within the cerebellum. It is theorized that the lateral cerebellum triggers and establishes the timing of the programs that initiate and hold movement by adjusting the activity of the pre-motor and primary motor cortical areas. The more medial portions of the cerebellum adjust and update the movements based on the feedback they receive from spinal descending pathways originating from the sensory motor cortex as well as relay nuclei such as the red nucleus, superior colliculus, medial pontomedullary reticular formation, and vestibular nuclei. The cerebellar output provides feedback to the motor cortex, continuously furnishing it with a provisional estimate of the position and movement of the limb before real sensory information from the periphery is available to it and even before the movement has occurred (18).

Finally, horizontal cortico-cortical inputs from the adjacent supplementary motor area and premotor cortex to area M1 further direct centrally programmed activity. The supplementary motor area and premotor cortex also receive direct information from the prefrontal cortex and posterior parietal association cortex.

Functional anatomy and physiology of basal ganglia function

Complex interconnected circuits link the basal ganglia with the motor cortex, thalamus, and cerebellum. Because of the primary role of the basal ganglia in many human movement disorders, this article places special emphasis on these structures. Our understanding of their functional anatomy and physiology continues to evolve, but as it is currently understood has been well described by many investigators (124; 107; 74; 74). Only the most basic aspects will be reviewed here.

Functional architecture of the striatum. Virtually the entire cortical mantle projects to the striatum in a topographic manner. These inputs are mostly excitatory and glutamatergic. The majority of neurons within the striatum itself, however, are the GABAergic medium spiny neurons that project and inhibit the output nuclei of the basal ganglia: the globus pallidus internus (GPi), globus pallidus externus (GPe), and the substantia nigra pars reticulata (SNpr). The large dendritic fields of the medium spiny neurons (303) allows them to receive simultaneous input from distributed but functionally related areas of cortex (251); conversely, single cortical regions project to multiple striatal zones (251; 98; 106). In this way, the striatum provides by its very structure an anatomical framework to both map and integrate cortical information within the broader framework of the parallel circuits running through it (106).

Medium spiny neurons receive other inputs in addition to the primary projections from the cortex. These include excitatory glutamatergic inputs from the amygdala and hippocampus, inhibitory GABAergic input from small interneurons (31), aspiny cholinergic interneurons, a somewhat larger input from the serotonergic nerve terminals projecting from the median raphe, and an extensive input from dopamine-containing nerve terminals projecting from neurons in the substantia nigra pars compacta (SNpc) (157) synapsing onto the necks of spines of the dendrites in a “triadic” arrangement with the glutamatergic inputs from the neocortex, amygdala, and hippocampus, which synapse onto the heads of the spines (31). Whereas the topographically arranged glutamatergic cortical afferents probably subserve more specialized functions through parallel circuits, the dopaminergic striatal projections arise from only a small population of midbrain neurons, innervating multiple medium spiny neurons. This organization suggests that they convey some form of general signal to the striatal neurons, perhaps modulating the functional strength of the glutamatergic synapses (125).

The circuit model of basal ganglia function: direct and indirect pathways. One of the earliest and most basic models of basal ganglia function is based on Albin and colleagues’ 2-circuit model of “direct” and “indirect” pathways within the basal ganglia (04). Although this influential model has formed the cornerstone of our understanding of basal ganglia function for several decades and promoted the development of new therapies such as functional neurosurgeries, many of its essential elements remain unproven. Over the last decade, a number of modifications have also been made to the basic model to account for more recent findings; however, the overall concept of circuits running through the basal ganglia to release desired behaviors and inhibit potentially unwanted ones remains unchanged (74). According to this updated model, the major output of the basal ganglia is a tonic one that arises from the globus pallidus internus (GPi) and substantia nigra pars reticulata (SNpr). It projects to and continuously inhibits the ventral motor and intralaminar nuclei of the thalami, which, in turn, project to and regulate motor pattern generators in the cerebral cortex.

Direct pathway. When a particular cortical motor pattern generator initiates a movement, the cortex sends simultaneous signals via glutamatergic projections to the basal ganglia to release the intended movement and also to suppress unintended movements. The release of the intended movement within the basal ganglia occurs primarily via GABAergic inhibitory projections on what is called the “direct” pathway running from striatum to GPi/SNpr. Decreased activity in the GPi/SNpr allows selective release of thalamocortical circuits controlling the relevant motor pattern generators in the cortex. Importantly, separate populations of striatal neurons are felt to subserve each of the pathways. Striatal interneurons subserving the direct pathway contain dynorphin and predominantly express dopamine D1 receptors; stimulation with dopamine, therefore, enhances the release of intended movements.

Indirect pathway. Simultaneously with the first signal to the striatum, the cortex sends secondary signals to inhibit surrounding or competing motor pattern generators. The major pathways for this activity are the indirect and the hyperdirect pathways. The indirect pathway runs from the striatum to suppress the GPe and, thus, dampen its tonic inhibition of the subthalamic nucleus (04), leading to stimulation of the GPi, which inhibits the thalamic-cortical loop. Evidence from human studies points towards this pathway being involved in modulation of movement kinematics (205). Striatal interneurons subserving the indirect pathway contain enkephalin and express dopamine D2 receptors predominantly. Activation of these D2 receptors suppresses the indirect pathway, allowing the release of motor patterns. This is the traditional explanation as to why D2 blockade is useful in treating a variety of hyperkinetic movement disorders, including chorea and tics: blockade of D2 receptors removes dopaminergic suppression of this pathway, thereby allowing it to inhibit background movements that might otherwise interfere with intended movements.

The globus pallidus externus (GPe) is an essential component of the indirect pathway, but by nature of its location and projections it is also a major integrator nucleus. Many of its connections are reciprocal ones, and it is, therefore, in a position to provide feedback inhibition to neurons in the striatum and subthalamic nucleus as well as feed forward inhibition to neurons in the GPi and SNr (04; 05; 157; 216; 31; 125).

Hyperdirect pathway. This important pathway runs from the frontal cortex to stimulate the subthalamic nucleus, which, via relatively divergent glutamatergic projections, then excites GPi and SNpr to suppress the thalamus and inhibit thalamocortical circuits (216). Recent studies have provided neurophysiological evidence of this pathway in humans (Kelly et al 2018; 193). Regarding the STN evidence from its local field potentials recorded in patients with Parkinson disease undergoing deep brain stimulation suggests that upper and lower limb movements are encoded in this nucleus before movement execution, and this can be predicted by machine learning algorithms (144). The hyperdirect pathway appears to play an important role in the cessation of movements (297) and may mediate the cognitive aspects of careful motor preparation (205) and other cognitive processes (Gharemani et al 2018; Kelley et al 2018; 205).

Numerous other feedback loops also contribute to the final output of the basal ganglia. The frontal cortex in particular likely modulates basal ganglia output independently of the striatum by providing direct inputs to the thalamus, pedunculopontine nucleus, and superior colliculus as well as the midbrain dopaminergic neurons that, in turn, project back to the basal ganglia. Prefrontal cortex, precentral and postcentral gyri, and superior parietal lobule project directly to substantia nigra (47). Connections to the subthalamic nucleus include direct input from the motor and prefrontal cortices, the intralaminar nucleus of the thalamus, and the pedunculopontine nucleus. The subthalamic nucleus has reciprocal connections with the GPe and output to the GPi, the striatum, the pedunculopontine nucleus, and the SNr (300).

Adding to the complexity of our understanding of the basal ganglia circuit, a study in normal rats showed that M1 stimulation triggered not a simple monophasic response but multiple excitatory and inhibitory responses in each basal ganglia structure (147).

Regarding basal ganglia cerebellum interactions, studies in animal models (Cebus monkeys) showed a more central interaction than previously thought, pointing toward disynaptic feedback and feedforward connections between the cerebellum and the basal ganglia (33). These studies showed that the dentate nucleus projects to the thalamus, which in turn projects to the GPe. In addition, the cerebellar cortex receives projections from the pontine nucleus to which subthalamic nucleus neurons project. These projections are somatotopically organized and originate from motor and non-motor regions of both the dentate nucleus and the subthalamic nucleus. Furthermore, MRI tractography studies in humans found extensive connections between basal ganglia, specifically between the STN and the cerebellar cortex, between the dentate nucleus and GPi, and between the dentate nucleus and substantia nigra (188). Evidence has also emerged correlating the loss of Purkinje cells in neocerebellum and motor symptoms in Huntington disease (260). These findings open the door to new hypotheses regarding the interaction between the cerebellum and the basal ganglia in motor (50) and non-motor disorders. For motor disorders, these connections provide the anatomical substrate for the interactions between the cerebellum and the basal ganglia that could help to understand the pathogenesis of abnormal movements such as tremor in Parkinson disease, essential tremor, or dystonia.

The center surround model of basal ganglia function. The center surround model of basal ganglia function represents a further elaboration of the basic schema of direct and indirect pathways in the standard model. Here the circuit properties of the standard model remain unchanged, but the structural organization of these pathways is further elaborated. Hallett first proposed that one of the fundamental roles of the basal ganglia was to balance cortical excitation and inhibition and that the direct and indirect pathways might, therefore, act in a "center-surround" manner to focus motor commands (109). In the first formal elaboration of this model, Mink proposed that the release of intended movements in the direct pathway from cortex to striatum occurs in a focused area of striatal cells, which produces a “center” of decreased activity in the GPi/SNpr that selectively disinhibits thalamocortical circuits to release the relevant motor pattern generators back in the cortex. Simultaneously, unwanted motor programs are inhibited by the hyperdirect pathway from the cortex to the subthalamic nucleus, which, in turn, projects broadly to stimulate the relevant portions of the GPi/SNpr to provide a “surround” of inhibited cortical motor activity, which may be mediated by M1 cortical interneurons (276). The result of any action, therefore, is a powerful release of the intended action, coded in the center of the relevant output neurons of the basal ganglia, and a broader, less forceful suppression of unintended actions, coded in the surround (191).

Striasome and matrix in basal ganglia function. Graybiel and colleagues described functional and neurochemical sub-compartmentalization of the basal ganglia that further elaborates the basic circuit model of basal ganglia function (106). With appropriate neurochemical staining, cells of the striatum can be differentiated on the basis of the neurotransmitter elements they predominantly express, for example opioid receptors, catecholamines, or acetylcholinesterase. Staining for acetylcholinesterase in the striatum delineates patches of lightly stained regions designated “striosomes” amidst a “matrix” of heavily stained striatum. The classical direct and indirect pathways run within the matrix whereas other circuits, primarily connecting limbic areas to the SNpc, run in the striosomal compartment. This striosomal circuit is proposed to provide direct frontal cortical modulation of the nigrostriatal tract.

Debate continues regarding the degree to which various circuits passing through the basal ganglia are segregated. Many investigators contend that the main circuits remain separate under normal conditions, as physiologic levels of dopamine should inhibit cross-connectivity between parallel circuits (24). However, anatomical studies provide evidence that seems to support a theory of parallel convergence. Striatal spiny neurons are known to give rise to extensive axon collaterals, which supports the theory that the direct and indirect pathways are synaptically linked at the level of the striatum (27).

Neural oscillation models of basal ganglia function. One of the limitations of the classic rate model of movement disorders is that it relies on a highly simplified conception of basal ganglia circuitry that does not factor in the numerous interconnections among structures in the basal ganglia that may play a role in the pathogenesis of movement disorders. Most importantly, however, the basic reliance on circuit properties, such as firing rate alone to explain the manifestations of movement disorders, is likely inadequate. The oscillation model of brain function is another emerging concept in neuroscience that attempts to address these limitations (269; 69; 83). The most familiar forms of neuronal oscillation are between the cortex and thalamus and are recorded routinely on EEGs. Over the past decade there has been an increasing interest in similar oscillations involving other structures of the brain, including basal ganglia, limbic system, and cerebellum. The synchronous oscillations of neuronal firing in these structures appear to vary in a predictable way with activity and specific goal-directed behavior as well as with rest and in sleep. There has, therefore, been increasing interest in the potential that these oscillations themselves could encode information and participate actively in brain function, and some authors have proposed that oscillations of ensembles of neurons represent the critical “middle ground” linking single neuron activity to behavior (45).

Oscillations of neural activity in ensemble of neurons could encode information in a variety of ways. For example, simultaneous oscillations of membrane potentials of participating neurons could predictably bias the probability of voltage-gated channels opening at a given point in time, thus, providing a “window of opportunity” for information to arrive at a group of neurons and to trigger a response. Another possible role for neuronal oscillations is described in the “binding-by-gamma” hypothesis, which postulates that oscillations of a particular frequency may bind together neuronal activity in distant areas of the brain to subserve a given function at a given time. This would allow components of a circuit to work together in a closely timed, synchronized manner, even when synaptic connections between the components were relatively weak (55).

Increasingly, the oscillation model of basal ganglia function has been applied to the analysis of movement disorders, fueled largely by postoperative local field potential and intraoperative microelectrode recordings from patients undergoing deep brain stimulation. These studies have revealed that abnormal synchronized oscillatory discharges of the GPi and STN, in phase with cortical activity, are a common feature of both hypokinetic disorders, such as Parkinson disease, and hyperkinetic disorders such as chorea, ballismus, and dystonia. One explanation for the clinical manifestation of synchronous discharges in diverse movement disorders is that the abnormal timing, pattern, or synchronization of discharges from the GPi may introduce noise into the cortical motor areas, regardless of whether its tonic output is increased or decreased. This may explain why both hypo- and hyperkinetic movement disorders improve with high frequency stimulation of the STN or GPi. Regardless of the specific pathogenesis or features of the underlying aberrant oscillator activity, these procedures may indiscriminately interrupt the aberrant STN/GPe loop to reduce the amount of noise in cortical motor areas (41; 90) or to allow for proper encoding and processing of motor programs as deep brain stimulation disrupts the “holding” effect of beta band on movement processing (144). Further evidence for the potential anti-kinetic role of high alpha and low beta oscillations (10 to 15 Hz) comes from the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) primate model of Parkinson disease (195). Simultaneous recordings from GPe, GPi, and STN showed that the highly synchronous and coherent oscillations in the 10 to 15 Hz range were attenuated, and the coherence between STN and GPi was lost during a prokinetic intervention such as STN stimulation. In healthy subjects, training with precise timed neurofeedback to suppress cortical beta bursts resulted in improved reaction time, suggesting that there is a link between cortical beta burst and movement initiation (116). In addition, cortical microeletrodes placed in the motor cortex of nonhuman primates delivering stimulation at precise timing and phase were able to specifically modulate peak of beta oscillations. Decrease in beta peak worsened whereas increase in beta peak improved motor performance (222).

Another model to complement the oscillation and the rate model of movement disorders emerges from animal literature, where changes in the millisecond-scale timing patterns of actions potentials (spike timing) evokes changes in how a specific movement is performed (262). This emerging spike timing dependent motor control model may complement and challenge current theories.

Basal ganglia dysfunction in movement disorders

By convention, movement disorders are divided on the basis of their clinical phenomenology into hypokinetic and hyperkinetic forms. The hypokinetic disorders include primarily the various forms of parkinsonism, both typical and atypical. The hyperkinetic movement disorders are a far more diverse group of disorders, both in their phenomenology and pathophysiology. They include dystonia, chorea and hemiballismus, tics, stereotypies, myoclonus, and tremor. However, the phenomenological classification on which the basic division between hypo- and hyperkinetic disorders rests provides little insight into their underlying pathophysiology, as the co-occurrence of a hypokinetic disorder (eg, parkinsonism) with hyperkinetic features (eg, tremor or focal dystonia) attest.

Basal ganglia dysfunction in hypokinetic movement disorders

Parkinson disease. The prototypical hypokinetic movement disorder is Parkinson disease. Its cardinal features are bradykinesia, rigidity, postural instability, and varying degrees of resting tremor. The selective degeneration of dopaminergic neurons in the substantia nigra pars compacta (SNc) is one of the primary pathological findings in both humans with Parkinson disease and in primate models of the disorder (300). Although all components of the basal ganglia sustain marked loss of dopaminergic input in Parkinson disease, the putamen is typically the most severely affected (124). The putamen is the sensorimotor portion of the striatum, and early disease is, therefore, primarily associated with motor and occasionally sensory dysfunction (137). As Parkinson disease progresses, more widespread neuronal loss occurs in the basal ganglia (04), affecting its associative and limbic circuits, in addition to loss of neurons in nondopaminergic systems in brainstem, thalamus, and cortex, which likely contributes to the variable disturbance of cognition, mood, sleep, and autonomic functions that is associated with the disorder (300; 162). Multimodal functional imaging studies provide further evidence of a widespread network dysfunction involving nigrostriatal structures and a sensorimotor network (bilateral precentral gyrus, supplementary motor area, and inferior parietal cortex) affecting patients with Parkinson disease (243).

Dopaminergic neurons of the SNc project to the striatum, as well as to the GPi, GPe, STN, and SNr, and the dopaminergic deficit produced by the degeneration of these dopaminergic neurons, therefore, disrupts processing within the entire basal ganglia network (04). Much of the detailed understanding of changes and motor circuit activity in Parkinson disease were originally based on the MPTP primate model of the disease. Metabolic imaging and electrophysiologic studies of the MPTP model demonstrated that neuronal discharge rates are increased in the STN, GPi, and SNr but decreased in the GPe. These findings prompted the development of the rate model for Parkinson disease in which decreased stimulation of D1 receptors--which predominate on the interneurons of the direct pathway--results in failure of inhibition of the GPi/SNr, which is itself inhibitory on the thalamus causing failure of release of intended motor programs in thalamo-cortical circuits. Conversely, decreased stimulation of D2 receptors that predominate in the indirect pathway leads to overactivation of striatal interneurons in the indirect pathway, thus, inhibiting the GPe and releasing the STN from its tonic inhibition. Projections from the STN stimulate the GPi/SNr and so increase their GABAergic output, which inhibits thalamo-cortical neurons and impairs motor function (04; 61).

In keeping with the standard rate model of basal ganglia function, microelectrode recordings from GPi neurons in patients with Parkinson disease reveal that the D1/D2 receptor agonist apomorphine, administered in doses sufficient to transition patients from the “off” to the “on” state, produces a striking decrease in firing rates of GPi neurons, and when this firing rate reaches a nadir, contralateral dyskinesias may emerge. As the effects of apomorphine wear off, the neurons then return to their pathologically elevated baseline firing rates (174; 130). The standard model provides an explanation for the increased activity that has been consistently observed in the subthalamic nucleus and GPi in Parkinson disease, as well as the marked improvement in parkinsonian symptoms with lesioning or stimulation of either the subthalamic nucleus or GPi (301); however, some investigators have also suggested that additional mechanisms could underlie subthalamic nucleus hyperactivity in Parkinson disease. For example the STN and SNc are reciprocally innervated, and degeneration of dopaminergic neurons in the SNc could, therefore, influence STN activity more directly than via the alterations in the indirect and direct pathways.

Paradoxes of the standard rate model of Parkinson disease. A number of observations from deep brain stimulation and lesion studies in patients with Parkinson disease have been difficult to reconcile with the standard rate model of Parkinson disease. For example, according to the rate model, lesions to the GPi/SNr would be expected to disinhibit the thalamus and produce dyskinesia. However, medial pallidotomy reduces levodopa-induced dyskinesias (17). The standard model also proposes that GPe activity should be decreased in Parkinson disease, yet an extensive literature indicates GABAergic activity of the GPe in parkinsonism is quantitatively similar to that observed under normal conditions. Moreover, biochemical analysis on brains of both parkinsonian humans and MTPT primates has demonstrated normal glutamic acid decarboxylase mRNA expression in GPe neurons, suggesting a normal level of GABAergic transmission between GPe neurons and the STN (121). These inconsistencies have suggested that other features of basal ganglia discharge, particularly altered patterns of discharges or frequencies of oscillations may play a more important role than absolute discharge rates. In the rodent 6OHDA model of Parkinson disease, motor cortex stimulation led to a net decrease in the firing of the GPi compared to control animals. Thus, the authors hypothesized that abnormal response to the motor cortex stimulation might underlie some of the features of Parkinson disease (147). Adding to the complexity of the rate model of Parkinson disease, it has been shown in a mouse model with dopaminergic deficit (sepiapterin reductase knock out mice) that activation of the globus pallidus led to inhibition followed by rebound excitation of the thalamus, and this led to Parkinson disease-like motor symptoms of akinesia and tremor (145). This rebound firing in the thalamus showed characteristics of both tonic and burst firing of actio potentials. It was modulated by the intensity of the globus pallidus inhibition and showed a strong impact on motor behavior when a sufficient number of neurons were stimulated.

Pedunculopontine area in Parkinson disease. Over the last decade, a growing literature has highlighted the importance of structures outside the basic basal ganglia circuits in motor control in both health and disease states. This is particularly true in Parkinson disease in relation to an area in the upper brainstem called the pedunculopontine area or nucleus. This structure is being investigated in several centers as a potential target for deep brain stimulation in both typical and atypical parkinsonisms. The pedunculopontine nucleus is the principal site of the mesencephalic locomotor region and plays a crucial role in gait and sleep in primate models of healthy and parkinsonian animals (141). Radiofrequency lesioning of the pedunculopontine nucleus region in the normal macaque produces akinesia (14), and selective destruction of pedunculopontine nucleus neurons produces an akinetic state and freezing (154). Conversely, injection of a GABA antagonist into the pedunculopontine nucleus of MPTP-treated monkeys has been shown to alleviate akinesia (201).

A number of investigators have postulated that the akinesia of Parkinson disease may arise from the suppression of the pedunculopontine nucleus by overactive GPi inhibitory efferents projecting to it (198; 160). In support of this view, deep brain stimulation of the posteroventral GPi, which projects primarily to the pedunculopontine nucleus, provides partial relief of akinesia in patients with Parkinson disease (198). According to this concept, overactive pallidal and nigral inhibitory inputs to the pedunculopontine nucleus may decrease pedunculopontine nucleus excitatory input to the substantia nigra and, thus, aggravate striatal dopamine deficiency. However, postmortem studies in patients with Parkinson disease have found neuronal loss in the pedunculopontine nucleus itself (317); this prompted speculation that the neuronal loss may cause dopamine-resistant problems, including gait disorders, postural instability, and sleep disturbances, more directly. Preliminary studies suggest that low frequency pedunculopontine nucleus stimulation results in modest improvements in the motor subscale of the UPDRS with particular benefit on rigidity, bradykinesia, gait, and postural stability (266; 142; 93; 196). The use of lower frequency stimulation may be related to different oscillation properties in the pedunculopontine nucleus compared to the basal ganglia (283). Local field potentials (LFP) studies with STN stimulation and recording from the pedunculopontine nucleus region in Parkinson disease suggest a polysynaptic functional connection between the STN and the contralateral pedunculopontine nucleus region (202).

The oscillation model of basal ganglia function applied to Parkinson disease. The oscillation model of basal ganglia function provides another perspective on motor deficits in Parkinson disease. The ability to record intraoperatively and immediately postoperatively from patients undergoing deep brain stimulation surgery has made this a rapidly developing field. Several reviews of this active area of research have been published (41; 295), and only the most salient findings are discussed here.

One of the hallmarks of Parkinson disease is an increased rate and degree of synchronization of oscillatory discharges in the GPi and STN--a feature also reported in hyperkinetic disorders such as chorea, ballismus, and dystonia (41). A consistent finding is reduction in the power spectra of the beta frequency band (13 to 30 Hz) measured from LFP in the STN prior to and during voluntary movements. Recordings from striatum in normal primates demonstrated similar findings (69), and one interpretation is that these beta band oscillations reflect normal physiologic processing for the preparation and execution of voluntary movements. Furthermore STN stimulation at lower frequencies in the beta and alpha range has been shown to reduce the speed of movement in parkinsonian patients off medication compared to the same patients off medication and off stimulation (57; 91). On the contrary, STN stimulation at beta frequency (20 Hz) in parkinsonian patients on medication was shown to slightly increase the speed of movement (150). The findings suggest that levodopa may modify the interaction between the beta oscillations and movement execution.

In Parkinson disease, pathologically increased synchronization of beta band oscillation in the STN correlates clinically with both bradykinesia and rigidity but not with tremor. Treatment with levodopa suppresses this excessive synchronization around the beta band in patients with Parkinson disease in a manner that correlates with the patient’s clinical improvement in bradykinesia and rigidity (but not in tremor) (151; 152). Transient deep brain stimulation has a similar effect on beta oscillations and, likewise, improves clinical signs (114; 104). Furthermore, Little and colleagues used the amplitudes of beta band peaks in STN LFP as a feedback signal to control adaptive deep brain stimulation, and they found greater improvement in parkinsonian motor signs than when using continuous deep brain stimulation (164; 280). Interestingly, oscillations in the STN beta band are also suppressed by ipsilateral stepping (95). Using a device capable of long-term recording of local field potentials, it was found that episodes of freezing of gait were related to increased entropy in the alpha band (8 to 12.9 Hz) (271). During walking without freezing, patients with Parkinson disease who had freezing of gait showed lower beta band power and higher beta entropy compared to nonfreezers (Syrkyn-Nicolau et al 2017). An interesting study found that at onset and during freezing of gait there is decoupling of low frequency band (4 Hz to 13 Hz) between cortex and STN, whereas this coupling is present in effective walking (232). Recording from surface EEG and from either STN or GPi in parkinsonian patients undergoing deep brain stimulation, a study found that during an isometric force exertion task, beta power in the basal ganglia covaried with force rate (movement effort) and that dopamine-related suppression of cortico-basal ganglia beta coupling is linked to faster force adjustments (97). Regarding the underlying physiology of beta oscillations, higher rates of STN beta bursts during movement may relate to bradykinesia in Parkinson disease (165). In patients with Parkinson disease off medications, slow movements were time-locked to multiple beta bursts, and these were coupled across different brain regions (Thinkhauser et al 2020). Moreover, in patients with Parkinson disease, there was a phase locking of background spiking activity of the STN to beta bursts in frontal cortex (49). In a rodent model of Parkinson disease, this phase locking was due to a relationship between the timing of STN and, to a smaller extent, Gpe and striatal cells action potentials to cortical oscillation phases. In this model, the phase-locked network synchronization was initiated with a sudden shift of the preferred angle of phase-locked action potential in subcortical cells and maintained with stable phase-locking between the brain network studied (frontal cortex, Striatum, GPe and STN) during beta bursts (49). There is also evidence that control of lower and upper limb movements involves a different spectrum of beta band, with desynchronization of the higher spectrum of the beta band (24 Hz to 31 Hz) associated to lower limb movements (281). These and other findings provide further support for the oscillation model of the basal ganglia and the potential for practical clinical applications (11).

Whether suppression of excess beta oscillatory activity directly enhances movement is still unclear. One possibility is that the STN and the GPe normally participate in oscillations in phase with the cortex. Some degree of synchronization of neural activity in the basal ganglia-thalamic-cortical circuits occurs normally and is necessary for normal movement. The striatal input to the GPe appears to disrupt this circuit so that it is able to escape from the control of the STN and maintain its usual pattern of tonic activity. Disruption of striatal input, in either a hypo- or hyperkinetic disorder, enhances the reciprocal circuit between the STN and GPe so that the STN is able to entrain the GPe to produce phasic activities driven by the cortex.

The effects of dopamine depletion on oscillatory activity are also an active field of investigation. Although broadly speaking, the effect of dopamine on circuit properties in the basal ganglia is to increase the signal-to-noise ratios in a manner that both facilitates the selection of specific motor programs and reinforces motor learning, it has also been shown to prevent pathological oscillatory activity. In hypodopaminergic states, the STN becomes increasingly sensitive to cortical rhythms, and as a result of this increased excitatory activity, the GPe switches from normal tonic activity to low frequency oscillatory activity that may destabilize the network and increase neuronal synchronization. Consistent with these findings, when patients with Parkinson disease are withdrawn from their medications, they consistently demonstrate prominent oscillations in the basal ganglia beta band. Although these field potentials are generated locally, the excessive synchronization they index is a feature of the whole basal ganglia cortical loop, which couples beta band oscillations identified in the STN with globus pallidus and cerebral cortex (43).

In the pedunculopontine nucleus, beta oscillations appear to be modulated differently by dopaminergic medications and by voluntary movements compared to the basal ganglia. Coupling between the supplementary motor area and the pedunculopontine nucleus in the beta range was observed only in the “on” medication but not the “off” medication state. In addition, during movement preparation in the “on” medication state, beta oscillations in the pedunculopontine nucleus increased in contrast to the reduction observed in the basal ganglia. Beta oscillations may be a prokinetic rather than an antikinetic rhythm in the pedunculopontine nucleus, and this may be related to the use of lower frequencies of deep brain stimulation in the pedunculopontine nucleus compared to the basal ganglia (283).

More detailed evaluations of the effect of dopamine on synchronous oscillation in basal ganglia circuits have taken place in rodent models, which allow more rapid manipulation of dopaminergic states within cortex and striatum. Synchronized activity in corticostriatal circuits increases in low dopaminergic states and decreases in high dopaminergic states (68). These changes may alter the firing properties and signal transduction within the striatum. In low dopaminergic states, the increased synchronization would result in a large proportion of cortical and striatal neurons firing preferentially during a particular phase of the oscillation of the local field potential. Inputs arriving out of phase with the oscillating local field potential would not depolarize target neurons and, therefore, would be “gated.” Conversely, in high dopaminergic states, the lack of synchronization between cortex and striatum would mean that few neurons would entrain to the oscillation of the local field potential, implying that those inputs arriving out of phase with field potential oscillations might more easily result in an action potential in the target neurons of the striatum; however, activation of striatal neurons requires very strong simultaneous input from populations of cortical neurons firing in unison, making the net effect of the unsynchronized local field potentials somewhat difficult to interpret. Some groups recorded from subdural cortical electrodes placed over the sensorimotor cortex together with STN recordings in nontremor dominant parkinsonian patients and found exaggerated coupling between M1 beta-phase and broadband gamma-amplitude in parkinsonian patients (70; 72; 255). They also found that this phase-amplitude coupling was decreased during a motor task compared to rest and that an increase in M1 LFP gamma-power preceded a drop in STN LFP beta-power. In addition, deep brain stimulation reversibly suppressed this exaggerated coupling in M1, and it modulated the shape of the cortical beta oscillations (64). Reversibility was observed 4 minutes after deep brain stimulation was turned off. The authors suggest that excessive phase amplitude coupling of cortical LFP could be used as a feedback signal for closed-loop stimulation that would suppress the excessive coupling to facilitate movement. There has also been an emerging interest in high frequency activity of basal ganglia structures in Parkinson disease. In addition to prominent beta band oscillatory activity in the Off state, the On state in Parkinson disease is characterized by activities in the gamma (60 to 80 Hz) and other high-frequency (300 Hz) bands that are modulated by movement. LFP recorded from the STN in patients with Parkinson disease demonstrated that in the Off state, the amplitudes of the high frequency oscillations (HFOs) are coupled to the phase of the excessive beta activity. In the Off state, the beta coupled HFOs showed little or negative movement related to changes in amplitude. Therefore, the degree of movement-related modulation of the HFOs correlated negatively with rigidity/bradykinesia scores. In contrast, in the On state, the HFOs were not coupled to beta oscillations and displayed marked movement-related amplitude modulation. These findings suggest that nonlinear coupling between different frequencies may not only be a physiological mechanism of movement, but may also be involved in the pathophysiology of parkinsonism (167). Similar findings obtained from STN LFPs support a direct correlation between a high beta peak, high phase amplitude coupling, and a high peak at HFOs and cardinal signs of Parkinson disease (bradykinesia and rigidity) (291). Using simultaneous STN LFP recording and measurement of cortical excitability using transcranial magnetic stimulation in a stop signal task, Wessel and colleagues showed that STN beta oscillations are related to stopping of movements and global motor suppression (297). In addition, specific phase alignments between STN and GPi LFPs correlated with local beta synchrony of both nuclei (48). Levodopa reduced frequency and duration of periods during which both nuclei would be locked to this pathologic oscillation (48). In a previous study performing similar recordings, Tsang and collaborators found individual peak frequencies in the gamma range during the On L-dopa state and during movement (286). These frequencies, which were lower than the usual high-frequency stimulation used for chronic deep brain stimulation, were used to stimulate the STN. Patients showed similar clinical benefit with the individualized gamma frequency and the usual high-frequency stimulation, suggesting that oscillations in the gamma range may be a prokinetic rhythm. In addition, there was increased power and coherence between M1 and STN in the gamma range and a reduction in power between M1 and STN in the beta band during wrist movements compared to the rest (272). Further evidence of the role of oscillations on the gamma range and its cortical interactions as a prokinetic phenomenon is provided by Muthuraman and Fischer and colleagues (96; 200). STN-spike to cortical gamma phase coupling during upper limb movements in patients with Parkinson disease off medications was linked to faster reaction times, was more pronounced over the motor cortex, was frequency- and time-specific, and was not modulated by average firing rate, and there was a systematic phase offset of timing of STN spikes relative to cortical gamma phase between contra- and ipsilateral motor task. These findings suggest that the phase was relevant to the desired movement with the contralateral limb and not the ipsilateral limb.

Basal ganglia dysfunction in hyperkinetic movement disorders. The hyperkinetic movement disorders are among the most phenomenologically varied disorders in neurology. They include chorea, hemiballism and levodopa-induced dyskinesias, dystonia, tics, stereotypies, and tremor.

Chorea, hemiballism, and levodopa-induced dyskinesias. Chorea and ballism as well as levodopa-induced dyskinesias are all characterized by the abnormal release of fragments of normal movements. For many years clinicians distinguished between ballism and chorea, reserving the term “ballism” for large amplitude movements involving proximal muscle groups. However, the 2 disorders share pathophysiology, exacerbating and relieving factors and treatments and often coexist in the same patient. Many researchers now maintain that these movements are 2 extremes of the same spectrum (231). Hemiballism has classically been attributed to lesions of the sensorimotor portion of the subthalamic nucleus (239). According to the classic rate model of basal ganglia function, this would result in reduced excitatory glutamatergic drive from the STN to the GPi. Decreased activity in the GPi would release the motor thalamus from inhibition, which would, in turn, result in increased cortical activation release of movement manifested as hemiballism or chorea. Consistent with this basic model, results from single photon emission CT imaging of patients with hemiballism have demonstrated increased blood flow in the thalamus and decreased flow in the basal ganglia contralateral to the affected limb (146). Likewise, recording from MPTP-lesioned primates with drug-induced dyskinesias have demonstrated reduced activity in the GPi, and similar findings have been shown in patients with Parkinson disease (174). Studies in mice and human subjects found that hyperactivation of D1 receptors in striatal neurons due to multiple alterations in intracellular signaling is a hallmark of levodopa induced dyskinesias (19; 120; 223; 261). However, the pathophysiology is complex (19), and other monoaminergic systems such as cholinergic and serotonergic systems also play a role (44; 32; 215). Moreover, there is evidence for abnormal M1 cortical facilitation in patients with Parkinson disease with levodopa-induced dyskinesias, which may be related to overactive glutamatergic transmission (108), as well as a direct relationship between levodopa-induced dyskinesias and progression of frontal executive dysfunction and development of parkinsonian dementia (307). One hypothesis to explain levodopa-induced dyskinesias of Parkinson disease is that these arise from reduced inhibitory activity in the striatum-GPe projection, which allows augmented activity of the GPe, which in turn over-inhibits the subthalamic nucleus, leading to hypoactivity of the GPi (211). A study that used electrocorticographic recordings in patients with Parkinson disease undergoing STN-deep brain stimulation found a high gamma peak and high gamma cortico-STN coherence during levodopa-induced dyskinesias, suggesting this prokinetic rhythm is related to dyskinesias (270).

Although the classic rate models of hemiballism and chorea have been useful as a starting point for understanding the movement disorders’ pathophysiology, they are not without limitations. With the advent of better imaging technologies, for example, if has become clear that lesions of the putamen and the STN as well as the thalamus can produce chorea and that lesions of the STN are, in fact, the cause of only a minority of cases of hemiballism (231). Hemiballism associated with acute hyperglycemia has also been reported, and MRI of these patients has demonstrated high signal T1 lesions in the putamen, with similar changes sometimes identified in the globus pallidus and caudate (213). Moreover, the classic rate models of the disorder do not explain the effects of additional lesions in the involved circuit. Lesioning of the GPi, for example, should worsen hemiballism; yet pallidotomy effectively treats it. Furthermore, therapeutic lesioning and deep-brain stimulation of the subthalamic nucleus, performed in patients with Parkinson disease, is rarely associated with hemiballism as a complication (231; 287; 187). A lesions network mapping study suggests that hemiballism and chorea are elicited by lesions affecting brain regions functionally connected to the posterior putamen (156).

Many investigators have suggested that alterations in firing patterns encompassing the degree or duration of burst activity, the length of interpotential pauses, and the degree of temporal or neuronal synchronization may determine the clinical manifestation of abnormal movements. According to this view, firing patterns in GPi neurons may signal a code that conveys information to the motor cortical areas, resulting in selection and execution of movement patterns. Abnormal phasic bursting of GPi neurons has been described in chorea or hemiballism (299; 269; 294), and this phenomenon may explain why GPi pallidotomy alleviates rather than exacerbates hemiballism: these procedures would abolish the abnormal firing patterns responsible for excessive movement (231).

Dystonia. Dystonia is usually classified as a hyperkinetic movement disorder on the basis of the excessive involuntary muscle activation that leads to the abnormal, twisting postures. However, the presence of focal dystonia in patients with parkinsonism demonstrates that the hyperkinetic and hypokinetic classification of movement disorders provides little insight into underlying pathogenic mechanisms.

The pathogenesis of dystonias is likely heterogeneous, as many genetic mutations and brain lesions are capable of generating similar presentations (130). Different genetic forms of dystonia (229; 244), as well as primary versus secondary dystonias (148), show distinctive neurophysiological and clinical traits. Primary dystonias often present in childhood without apparent underlying CNS lesions and may be associated with defined genetic mutations. To date, 25 different forms of monogenic dystonias have been identified and classified as DYT1 loci. The relationship between primary childhood onset dystonia and the more common adult-onset focal or segmental dystonias remains unclear (130), but genetic risk factors are also likely to play a role in the pathophysiology of the latter as well. Physiological investigations of dystonia have revealed a number of abnormalities that may reflect the effects of a genetic substrate that confers vulnerability to the condition. Findings from these studies include loss of inhibition in the central nervous system (including loss of surround inhibition, increased plasticity, and additional sensory abnormalities), although which if any of these abnormalities is primary in the disorder remains to be determined (110; 112; 219).

The lack of a suitable primate model for dystonia has limited the investigations of circuit abnormalities underlining the condition. The model predicts that hyperkinetic disorders should be characterized by reduced GPi output leading to disinhibition of thalamocortical neurons and a resultant release of involuntary movements. Intraoperative recordings from the GPi in patients free of systemic sedation have revealed that GPi discharge rates are indeed lower in patients with dystonia than in those with Parkinson disease (294; 314; 263; 306). Starr and colleagues have reported that burst and oscillatory activity in the GPi are broadly similar in the 2 conditions (263) and Tang and colleagues have reported that burst activity is more irregular in cervical dystonia than in Parkinson disease (274). Schrock and colleagues evaluated the role of the STN in dystonia with microelectrode recordings from awake patients undergoing STN deep brain stimulation for dystonia (248). They reported that the mean STN discharge rate is lower in patients with dystonia compared to patients with Parkinson disease, but higher than the published values for subjects without basal ganglia dysfunction, namely patients with essential tremor. Oscillatory activities were present in both disorders, with a higher proportion of units oscillating in the beta range in Parkinson disease. Bursting discharge is a prominent feature of both dystonia and Parkinson disease, whereas sensory receptive fields were expanded in Parkinson disease compared with dystonia. The authors concluded that bursting and oscillatory discharges in basal ganglia output may be transmitted via pathways involving the STN, which would provide a rationale for the STN as a surgical target in dystonia. Studies performed in cervical dystonia and secondary hemidystonia confirmed prominent bilateral beta oscillations in the GPi in dystonia (Tsang et al 2011; 284; 285). However, gamma oscillations increased with movement planning and execution in M1, S1, and GPi only on the side contralateral to the movement. These oscillations likely play a role in self-paced as well as in externally triggered movement. Crowell and colleagues described impaired movement-related beta desynchronization in M1 and S1 in patients with cervical dystonia (70). A metanalysis of 8 studies on resting state oscillations off medication and off stimulation in a total of 127 patients with dystonia and 144 patients with Parkinson disease reported that beta oscillations in GPi were lower and that low frequency oscillations (4 Hz to 12 Hz) in GPi and STN were higher in dystonia than in Parkinson disease. Importantly, this metanalysis highlighted discrepancies in findings and research methodology across studies (226). A probabilistic mapping study suggested the ventroposterior GPi and adjacent subpallidal white matter as major hubs to achieve control of dystonic symptoms by deep brain stimulation (238).

Other findings have called into question the utility of simple rate models of dystonia. For example the rate model cannot explain why pallidotomy should effectively alleviate dystonia as well as chorea (130), when this would be predicted to further disinhibit the thalamocortical circuits contributing to hyperkinesis (182; 42), particularly when spontaneous lesions of the GPi from stroke can produce dystonia. Alternative models for dystonia have, therefore, been proposed, including those postulating changes in discharge patterns of basal ganglia neurons, changes in somatosensory responsiveness within basal ganglia cortical circuits, and alterations in the degree of neuronal synchronization within basal ganglia circuits (299; 293). The finding of upregulation of dopamine D1 receptors in bilateral putamen in writer’s cramp patients, and in the right putamen and caudate nucleus in laryngeal dystonia, suggests that the direct pathway is hyperactive in focal dystonias (258). The authors proposed that the pathophysiology of focal dystonia involves upregulation of D1 receptors, leading to increased excitation of the direct pathway and downregulation of D2 receptor, which results in decreased inhibition of the indirect pathway and weakened nigrostriatal phasic dopamine release, leading to abnormal hyperexcitability of motor thalamo-cortical loops in focal dystonia.

The success of surgical treatments for Parkinson disease has led to successful trials of pallidotomy and deep-brain stimulation for dystonia, specifically targeting the sensory and motor portions of the GPi (129; 184; 292; 52; 288). It is, however, common to see a temporal delay of up to several months between the surgical intervention and maximal benefit, supporting the hypothesis that the emergence of dystonia involves some degree of neuroplasticity. Transcranial magnetic stimulation studies showed that plasticity in the dystonic patient became less than normal at 1 month after the GPi surgery and gradually restored to normal levels at 6 months after surgery (242). Short interval intracortical inhibition (SICI) was reduced in dystonic patients before surgery and was progressively restored to normal levels in parallel with the clinical improvement during the first 6 months after surgery. Because changes in plasticity preceded clinical improvement and restoration of SICI, the authors suggested that changes in plasticity may drive clinical improvement and changes in cortical inhibition.

The center surround model of dystonia: evidence for altered cortical specificity. Several lines of evidence also implicate impaired cortical function in dystonia (131). In general, cortical abnormalities are felt to reflect disorder of basal ganglia and thalamic outflow to sensory motor areas of cortex. According to center-surround model of the basal ganglia (109), dystonia results from incomplete suppression of competing movements due to insufficient surround inhibition of competing motor pattern generators. The deficient surround inhibition could also lead to expansion of the facilitatory center, which would result in “overflow” contraction of adjacent muscles.

Several studies suggest that cortical hyperexcitability and failure of surround inhibition is a feature of dystonia. Transcranial magnetic stimulation studies of excitatory and inhibitory phenomena in motor cortex have demonstrated increased cortical motor excitability (132) and reductions of intracortical inhibition (60; 20; 21) as well as impaired interhemispheric inhibition (204) in patients with focal task-specific dystonia. Various components of the pre-movement EEG potential (Bereitschafts potential) are also reduced in patients with dystonia (23), suggesting that the physiologic mechanism underlying movement preparation may also be altered in the condition.

A number of functional imaging studies likewise suggested dysfunction of premotor and motor cortical networks in generalized dystonia, but results have not been consistent (56). Ibanez and colleagues, in an FDG-PET study of patients with writer's cramp performing a writing task, demonstrated deficient activation of premotor and sensorimotor cortices as well as a decreased correlation of metabolic activity between premotor cortex and putamen compared to controls (131). Carbon and colleagues used a multitracer approach with PET and demonstrated increased metabolism in secondary sensorimotor areas and decrease metabolism in cerebellum, midbrain, rostral pons, and thalamus in both DYT1 and DYT6 manifesting carriers compared to non-manifesting carriers and healthy controls (53; 54). In these patients, diffusion tensor MRI demonstrated microstructural changes in cerebellar pathways that correlated with disease penetrance. The authors speculated that structural abnormalities may be the main intrinsic abnormality underlying observed downstream cortical changes, such as increased sensorimotor metabolism during audiovisual processing compared to control subjects. This is consistent with loss of cerebellar control of sensorimotor plasticity reported in patients with writer’s cramp (128). The role of the cerebellum in the pathophysiology of dystonia is increasingly being recognized (58; 233; 253). Using single cell recording and optogenetics, Chen and associates found that a short latency disynaptic cerebello-thalamo-basal ganglia pathway modulates cortico-striatal plasticity in a reversible manner (58). They also found that the dystonic features in a mouse model of dystonia were reversed when the disynaptic cerebello-striatal connections were interrupted. Their findings demonstrated a well-defined anatomical pathway through which the cerebellum can modulate motor learning when interacting with the basal ganglia, and may underlie the pathophysiology of dystonia. With functional magnetic resonance imaging (fMRI), patients with cervical dystonia showed decreased activation of posterior cerebellar lobules, premotor areas, associative parietal cortex, and visual regions, with reduced connectivity between cerebellum and both bilateral basal ganglia and dorsolateral prefrontal cortex (94). Thus, the cerebellum was included in the dysfunctional network of cervical dystonia. In their review, Prudente and colleagues suggested that dystonic features can be a consequence of irritation of the cerebellar cortex resulting in excessive cerebellar excitability, whereas destructive lesions result in ataxia (233). Hoffland and associates found that functionally inhibiting the cerebellar cortex by continuous theta burst stimulation using repetitive transcranial magnetic stimulation can normalize the blink reflex in patients affected by cervical dystonia (122). This normalization achieved through cerebellar inhibition is suggestive of a reversible role of cerebellar hyperactivation in dystonic patients. A “lesion network mapping” study reported that the occurrence of cervical dystonia was related to lesions in a broad network involving cerebellum, basal ganglia, thalami, and sensorimotor cortex. However, lesions with positive connectivity to the cerebellum and negative connectivity to the somatosensory cortex were specific markers for cervical dystonia (67). Regarding the broader neuronal network involved in the pathophysiology of dystonia, choline acetyltransferase deficiency in the pedunculopontine nucleus of cervical patients with dystonia has been reported, suggesting a cholinergic deficit in this nucleus may be part of the dysfunctional neuronal network in cervical dystonia (185). Functional magnetic resonance imaging (fMRI) studies of focal dystonia demonstrated increased activation of basal ganglia and thalamus in patients with focal dystonia while performing tasks not primarily involving the dystonic musculature (210). Patients with musician’s dystonia demonstrated decreased activation of thalamus and increased activation of the ipsilateral motor areas in hand tapping tasks (138). Levy and Hallett have used MR spectroscopy to demonstrate diminished GABA levels in the sensorimotor cortex and lentiform nuclei contralateral to the affected hand in patients with focal task-specific dystonia. The authors have interpreted these findings as being consistent with reduced intracortical inhibition (161).

Reduced inhibition of brainstem and spinal reflexes have also been described in dystonia. Such reflexes are indirectly modulated by activity in the pallido-thalamo-cortical motor circuits, and 1 interpretation of these findings is that they reflect alterations in descending projections to inhibitory neurons (168). Some authors have also postulated that disordered muscle spindle activity may play a role in dystonia (23).

Sensory abnormalities in dystonia. Sensory inputs are known to influence some forms of focal dystonia, as demonstrated by "sensory tricks" that allow patients to transiently overcome abnormal postures (23) and “reverse sensory tricks” that may aggravate them (12). However, the importance of sensory abnormalities in the genesis of dystonia is still a matter of speculation. Several hypotheses propose a more general derangement in the basal ganglia’s ability to link patterns of proprioceptive input to motor output, which would lead to inappropriate recruitment of muscles during voluntary movement to produce the co-contraction of agonist and antagonist pairs and overflow of movements into other muscles (189; 279). An alternative interpretation of these findings is that both motor and sensory functions are similarly affected in dystonia and that one is not causative of the other (111).

Nonetheless, there is accumulating evidence for disturbances of sensory processing (35) and sensorimotor integration in dystonia (316; 26), particularly in the parietal multimodal multisensory association region with aberrant downstream effects that influences fine motor control (186). Patients with focal hand dystonia have impaired sensory perception with reduced spatial and temporal discrimination (16; 250) that are not somatotopic to the distribution of dystonia (194), and the abnormalities are present even in relatives of patients with dystonia, suggesting that they may be non-manifesting carriers of dystonia genes with incomplete penetrance (212; 36). In part, these sensorimotor deficits can be explained by reduced inhibition in the primary somatosensory cortex (09). Voxel-based morphometric studies in writer’s cramp demonstrate reduced gray matter in the bilateral thalamus and cerebellum, as well as primary sensorimotor cortex contralateral to the affected hand (73). An fMRI study identified focal cortical atrophy to the ventral premotor cortex contralateral to the right dominant hand affected by writer’s cramp (100). This atrophy correlated with symptom durations and with dysfunction in a network specifically used when writing, which includes ventral premotor cortex and inferior parietal cortex. The authors speculated that this original lesion triggers compensatory mechanisms that explain the brain network changes described in this condition. The same group found reduction of GABA-A receptors in vermis of the right cerebellum and in left sensorimotor cortex, which could be a mechanism of loss of sensorimotor inhibition in focal dystonia (99). Restoration of this inhibition may be a way by which alcohol or benzodiazepines may improve dystonic symptoms in some patients (136). Nelson and colleagues, using high-resolution fMRI and surfaced-based mapping techniques to evaluate cortical somatosensory representations of affected digits in patients with task-specific writer’s cramp, have demonstrated that digits directly involved in writing show reduced inter-digit separation, reversals, and overlapping activation, whereas asymptomatic digits preserve their inter-digit separation (203). Sensory training by learning of Braille (310) and motor training to increase individuation of finger movements (311) have been shown to produce mild, transient improvements in motor performance.

The role of overtraining of movements in the genesis of dystonia has also been investigated. The question is particularly relevant for some conditions such as writer’s cramp and musician’s dystonia. Some preliminary evidence has linked overtraining to changes in sensory processing. Byl and colleagues demonstrated with a primate model of focal dystonia that repetitive stereotypic limb movements result in degraded representations of sensory information in the cortex (46). Several studies have demonstrated that in focal hand dystonia structural and functional brain reorganization exists on a spectrum that extends from mild changes present in healthy musicians to more pronounced changes in musician’s dystonia (85; 86; 241). Rosenkranz and colleagues reported that in healthy nonmusicians, proprioceptive input to fingers of the hand results in reduced intracortical inhibition to muscles from the area where the sensory input is directed and increased intracortical inhibition in surrounding areas. In healthy musicians, this pattern is reduced and in musicians with dystonia this inhibition is lost. The authors proposed that the changes observed in healthy musicians are secondary to musical skill learning in order to support high levels of performance. In musician’s dystonia this reorganization may have extended too far to the point where it interferes with motor control rather than assists it (241). Genetic factors also predispose to musician’s dystonia (247).

Tourette syndrome. Tourette syndrome is a heritable neuropsychiatric disorder that presents in childhood with a constellation of motor and non-motor symptoms, but its defining feature is the presence of vocal and motor tics. Tourette syndrome is distinguished from other movement disorders in that it is primarily a disorder of volition that may represent a more general failure of inhibition. Prior to enacting tics, up to 90% of adults (159; 155) and up to 35% of young children (15) report that they feel unpleasant sensations or urges that ultimately drive much of their unusual behavior. Although the trigger for the tic is itself involuntary, the action that results from it is at least under partial voluntary control, possibly involving the dorsal anterior cingulate cortex and associated limbic areas when tics are suppressed successfully (289). The degree of success in volitional inhibition of tics correlated with the degree of reduction in corticospinal excitability (101). In the majority of cases, Tourette syndrome is also associated with behavioral disturbances, such as obsessive-compulsive disorder (240), attention deficit hyperactivity disorder (66), impulse control disorder, and intermittent explosive disorder. All have as their common feature impaired inhibition of unwanted behaviors.

Anatomic localization of Tourette syndrome. Functional imaging studies during the active generation of tics have delineated a probable anatomic substrate for Tourette syndrome, consisting of frontal cortical, paralimbic, and striatal regions of the brain (37; 84; 224; 268; 133; 30). MRI tractography studies found aberrant enhanced connections between subcortical (striatum and thalamus) and cortical structures (M1, S1, supplementary motor area, and parietal cortex) (304). Interestingly, these anomalies correlated with symptom severity and were more prominent in females. A MR spectroscopy study found reductions in striatal concentrations of glutamate that correlated with tic severity and reduction of thalamic glutamate concentration that correlated with premonitory urges (140). However, a similar study with a high field magnet (7 Tesla) did not find this deficit (179). A resting state fMRI study in male Tourette syndrome subjects found that bilateral GPi hypofunction correlated with the severity of symptoms (134). A similar study found that premonitory urges were associated with enhanced connections between the insula and M1 and that the severity of tics was associated with enhanced communication between left thalamus and right M1 (257). Another resting state fMRI study found increased brain activities in the left calcarine sulcus and left cuneus, and it found decreased activities in the cerebellum, left fusiform gyrus, and left insular compared to controls (163). A fMRI study that used dynamic causal modeling of a finger opposition task found a correlation between the strength of connections between subcortical structures (thalamus and putamen) and M1 with tic severity and an inverse correlation between the strength of the connections between premotor cortex and M1 with tics severity, suggesting that the premotor cortex could be attempting to control the aberrant subcortical signal (309). A study that used diffusion tensor imaging and probabilistic tractography found reduced connectivity strength in the right hemisphere but increased global integration representing more efficient communications, which could reflect plastic adaptation to the reduced connections in that hemisphere (246). The aforementioned areas work within interconnected networks to subserve motor planning, behavioral inhibition, motivation, affect, and the detection of threats—all aspects of brain function that to one degree or another are impaired in Tourette syndrome (115; 249; 178).

Although considerable efforts have been directed to elucidate the contribution of the basal ganglia to Tourette syndrome, a substantial body of evidence has also implicated cortical abnormalities in the disorder. The phenomenology of Tourette syndrome would indicate that this might be the case, as tics are under partial voluntary control, are suppressible and, conversely, suggestible—all of which suggests some degree of conscious and, thus, cortical contribution to the phenomenology of the disorder. Evidence from anatomic and functional imaging studies supports this view. Volumetric MRI analyses have revealed significantly larger dorsolateral prefrontal regions in children with Tourette syndrome and significantly smaller ones in adults with the disorder (225) as well as reduced grey matter volumes in anterior cingulate gyrus, sensorimotor areas, and caudate nucleus (197). Studies have demonstrated significant cortical thinning in the front-parietal and somatosensory motor cortices in Tourette syndrome patients relative to controls (92). Studies using voxel-based fractional anisotropy techniques have reported white matter microstructural abnormalities and evidence for abnormal connectivity among components of the frontostriatal-thalamic circuit (180; 277) as well as abnormal interhemispheric connectivity (207). Church and colleagues have used resting-state functional conductivity MRI to evaluate Tourette syndrome patients against a large population of healthy controls and have demonstrated that adolescents with Tourette syndrome have immature functional connectivity in broadly distributed areas, with more profound deviations in frontostriatal areas (62). The authors subsequently demonstrated differences in brain activation in Tourette syndrome patients during tests of task maintenance and adaptive control (63).

Event related PET during the active generation of tics has shown preferential activation of the dorsolateral rostral prefrontal cortex among other cortical and subcortical regions (268); conversely, fMRI during the active suppression of tics has demonstrated increased activation of prefrontal cortex, in addition to relevant areas of thalamus and basal ganglia (224). Transcranial magnetic stimulation studies have showed that cortical inhibition (315; 102) and possibly that long-term potentiation and long-term depression like plasticity (183) are reduced in Tourette syndrome patients. Postmortem analyses of a limited number of Tourette patients have likewise suggested cortical abnormalities (259; 192; 139). A PET study with the ligand [11C]FLB457, a high-affinity D2/D3 receptor antagonist, reported a different pattern of radioligand binding in cortical and subcortical regions. The authors hypothesized that abnormalities of dopaminergic function in these regions may provide a mechanism for the hyperexcitability of thalamocortical circuits in Tourette syndrome (265).

The standard rate model of Tourette syndrome. In his original manuscript published in 1885, Georges Gilles de la Tourette identified no anatomic or pathological causes for the condition and referred scientists interested in the origins of the disease to the field of psychology. For many decades thereafter psychogenic explanations for the disorder dominated thinking within the field, and only with Bockner’s observation in 1959 that neuroleptics effectively improved tics (28) did significant interest develop in organic models of Tourette syndrome.

The most basic model of Tourette syndrome is derived from the classical 2-circuit model of direct and indirect pathways (04). It proposes an imbalance between the direct pathways releasing intended movement and the indirect pathways suppressing involuntary movements, with a net effect of reducing activity of the GPi, which then inappropriately releases thalamocortical motor pattern generators (04). Zhuang and colleagues have provided limited evidence to support this model (312; 313). In 8 patients undergoing unilateral pallidotomy for severe tics, they correlated microelectrode recordings to electromyographic discharges from muscles generating tics. They reported decreased neuronal firing rates and irregular firing pattern in GPi neurons and, in a subgroup of neurons, tic-related alterations in burst activity or pauses in ongoing tonic activity.

Although the standard rate model of Tourette syndrome localizes the origin of motor dysfunction to the basal ganglia, these structures also subserve a variety of cognitive and affective functions through parallel and overlapping pathways, which are likely organized and governed according to the same principals as those for motor function, with signals releasing intended processes and suppressing unintended ones. In this context, it is significant that the 4 major groups of symptoms associated with Tourette syndrome, ie, motor and vocal tics, obsessive compulsive disorder, attention deficit hyperactivity disorder, and impulse dyscontrol, have as their common feature impaired inhibition of unwanted thoughts and behaviors. This observation may reflect a common relationship between the clinical manifestations of Tourette syndrome and the nature of the disruption to the neural circuitry that underlies it.

The center surround model of Tourette syndrome. The previously described center surround model of basal ganglia function has also been applied to provide a more specific framework to understand the motor and behavioral manifestations of Tourette syndrome (191). Mink proposed that tics may be generated by the pathological activation of isolated populations of neurons within surrounding fields in the striatum that project to and inappropriately inhibit basal ganglia output, thus, releasing the relevant thalamocortical circuits to generate unwanted movements, thoughts, and behaviors.

Striasome and matrix in Tourette syndrome. Graybiel and colleagues’ description of functional and neurochemical sub-compartmentalization of striosome and matrix within the striatum (123) has been likewise proposed to have implications for Tourette syndrome (51). The authors reported that within the striosomal compartment increased expression of immediate early genes correlates with the expression of repetitive behaviors seen in mice treated with psychomotor stimulants. Graybiel and colleagues proposed that these repetitive behaviors are a model for human stereotypies, which are voluntary, repetitive movements commonly seen in children with developmental disabilities, as well as in some normal children and adults. The authors have postulated that stereotypies, and possibly by extension, tics, may result from a metabolic imbalance in the activities of the striosome and matrix compartments (190; 245). Subsequently, Glickstein and Schmauss demonstrated in D2 and D3 receptor knockout mice that striasome activation is only an indicator of psychostimulant-induced D2 and D3 receptor coactivation and is not in itself necessary for the stereotypic motor responses seen in psychostimulant-sensitized rodents (105).

Neural oscillations in Tourette syndrome. The oscillator model of brain function provides another perspective on Tourette syndrome. Leckman and colleagues formulated a hypothesis of Tourette syndrome based on oscillatory models of basal ganglia function (158), postulating that abnormal oscillatory activity within corticostriatal circuits produces abnormal discharges in the Gpi, which allows transient release of the thalamus from its tonic inhibition. This in turn leads to ectopic activation of selected groups of cortical pyramidal neurons that generate overt responses in the form of tics or subliminal perceptions in the form of premonitory urges. Pallidotomy and deep-brain stimulation of the GPi as well as the centromedian parafascicular and ventral oralis complex in the thalamus can produce rapid improvements in tics (126; 312; 177; 252; 208; 230) and similar results have been reported for deep brain stimulation of the nucleus accumbens (230). Marceglia and colleagues reported that single unit recordings from the VO complex in Tourette syndrome patients showed a localized pattern of bursting neuronal activity (181). Local field potentials obtained from bilateral pallidum and thalamus showed peaks of activities in the theta (3 Hz to 12 Hz) and beta (13 Hz to 35 Hz) ranges. Prolonged theta bursts in both subcortical regions were associated with greater motor tic severity (206). Evidence supporting or refuting the oscillator hypothesis of Tourette syndrome may emerge from detailed analyses of intraoperative or postoperative local field potential recordings, when the results from Tourette syndrome patients are compared to patients with other motor and nonmotor conditions. Interestingly, recording of intraoperative GPi local field potentials from Tourette syndrome patients found cross-frequency coupling between HFO and the beta band only during tics, but not during voluntary movement or resting (135). Simultaneous recordings from M1 and thalamic centromedian-parafascicular complex found a characteristic thalamic low frequency (1-10 Hz) peak that correlated with cortical desynchronization in beta band (256). These oscillations were concurrent with generalized motor tics lasting more than 5 seconds. As our understanding of Tourette syndrome continues to evolve, additional questions arise. Several reviews summarized the current state of knowledge in this field and have suggested areas for further investigation (02; Albin and 191; 264).

Tremor. The pathophysiology of tremor is generally felt to differ from that of other hyperkinetic movement disorders. Tremor is also a heterogenous entity, and its pathophysiology likely varies with its etiology. Many types of tremor can be distinguished: essential tremor, parkinsonian tremor, enhanced physiologic tremor, dystonic tremor, a midbrain or Holmes tremor, and cerebellar tremor secondary to a wide variety of pathological processes. The physiology of essential tremor and parkinsonian tremor are considered below. An interesting review on this topic has been published (290).

Tremor can arise from 1 of 4 pathophysiologic oscillations: mechanical oscillations, reflex oscillations, feedback-driven oscillations, and central oscillations. Any of these may interact to produce a particular tremor, but one source of oscillation is typically dominant in a given condition (75). Central oscillators produce tremor frequencies that are largely independent of limb mechanics and reflex arc length. However, no source of tremor is completely isolated from the effects of sensory feedback, and body and limb mechanics serve as the final common pathway for all forms of tremor (87).

Essential tremor. Despite the fact that essential tremor is one of the most common movement disorders, its underlying pathophysiology remains speculative. Although the disorder is well known to have a genetic basis, an association with sequence variants of the LINGO1 gene has been reported (267), and the finding has been replicated (273; 275). The LINGO1 gene plays a key role in oligodendrocyte differentiation, maturation, and neuronal myelination. Essential tremor was, for many years, thought to arise from a disturbance of neuronal function, rather than structure, as limited postmortem studies originally failed to demonstrate consistent pathology (76). Louis and colleagues proposed that 2 patterns of pathology are associated with essential tremor. The first, a cerebellar variant, is characterized by loss of Purkinje cells and an increased number of Purkinje cell axonal swellings called “torpedoes.” The authors subsequently reported increased cerebellar heterotopia in essential tremor. In cerebellar heterotopias, Purkinje cell bodies are displaced to the molecular layer, and this occurs in cerebellar degenerative disorders (153). The second pattern is characterized by the presence of Lewy bodies in the brainstem (173; 171), which did not correlate with the presence of rest tremor in essential tremor (169). In the cerebellar variant, increased numbers of torpedoes and Purkinje cell loss are reported to correlate with older age of tremor onset and faster rate of tremor progression (170). Another study found a negative correlation between volume of lobule VIII of the cerebellum and severity of postural tremor (166) and a positive correlation between volume of lobule IV of the cerebellum and severity of resting tremor in Parkinson disease. Further studies are needed to clarify the relationship between these pathological findings and the clinical manifestations of essential tremor. Rajput and colleagues reported different results with decreased numbers of cerebellar Purkinje cells with advancing age, but no differences between patients with essential tremor, Parkinson disease, and normal controls (237). Moreover, 30% of essential tremor patients develop a neurodegenerative parkinsonian syndrome (more frequently Parkinson disease) after years of essential tremor progression (236).

The electrophysiologic properties of essential tremor have been extensively investigated. The tremor is characterized by rhythmic 4 to 12 Hz entrainment of motor unit discharges that forces the affected body part into oscillation (88; 87). The tremor amplitude bears a logarithmic relationship to the intensity and frequency of motor unit entrainment (88).

Essential tremor is thought to be generated centrally by an oscillatory neuronal network influenced by somatosensory feedback loops (76). EMG peak frequencies do not shift with loading in essential tremor (75), and patients with essential tremor exhibit normal mechanical-reflex properties (89). However, an understanding of the neural circuit generating the tremor remains incomplete; intraoperative thalamic burst activity is strongly correlated with forearm EMG signals in essential tremor, suggesting that the thalamus participates in the oscillatory network generating tremor (127), and kinematic and EMG analyses suggest cerebellar involvement in essential tremor as well (39; 76; 149). Regarding neural circuits involved in the genesis of essential tremor, cortical involvement had been controversial because magnetoencephalography failed to detect cortical activity in tremor generation (113). However, other studies using simultaneous EEG-EMG recordings have demonstrated corticomuscular coherences at tremor frequency at least intermittently, suggesting some involvement of sensorimotor cortex in the generation of essential tremor (117; 235; 254). More evidence of a cortico-thalamic-cerebellar network involved in essential tremor has emerged (10). Using fMRI, clinical assessment, and analysis of tremor induced by applying force, the authors found evidence of a more widespread network, including extrastriate visual areas and inferior parietal lobe, which are associated with tremor severity. They found that increased visual feedback can increase entropy of blood level oxygen dependent signal in these areas, which correlated with tremor severity.

Brain connectivity measurements in patients with essential tremor treated with VIM deep brain stimulation found somatotopic segregation of the cerebello-thalamo-cortical network with head versus hand involvement in the tremor and described a location ventrolateral to the thalamus, medial to internal capsule and inferior to the VIM and sensory thalamic nuclei where afferent cerebellar fibers enter to thalamus, as the site of most clinically effective deep brain stimulation to treat essential tremor (01). In contrast, a study on focused ultrasound thalamotomy (34) reported the posterior portion of the VIM nucleus of the thalamus as the best clinical target to treat essential tremor. Further insight from noninvasive stimulation studies exploring the role of the cerebellum in essential tremor can be found in a review (176).

Animal studies have suggested a potential role for the olivocerebellar system as a central component in the pathogenesis of essential tremor (117). The toxin harmaline enhances the inhibition-rebound properties of olivary neurons, resulting in increased rhythmicity and neuronal entrainment throughout the olive. When injected into primates, harmaline induces a fine, action tremor that has the frequency, EMG findings, and drug-response of essential tremor (87). In vivo studies and brainstem slice recordings indicated that the emergence of harmaline-induced tremor in rodents is dependent on the CA(V)3.1 T-type calcium channel in the inferior olive, and knockout mice for the Ca(V)3.1 gene fail to generate tremor in response to harmaline injections (217). A genetically driven defect in harmane metabolism has been postulated as the cause of essential tremor in humans, as a difference in the ratio of blood harmane to its metabolite, harmine, in a large group of essential tremor patients and controls (172). Levels of harmane are highest in familial essential tremor patients, intermediate in sporadic essential tremor cases, and lowest in normal controls.

The harmaline animal model has given rise to the “olivary hypothesis” of essential tremor, which posits that the basis for the disorder is enhanced olivary synchronization that produces a 4 to 12 Hz neuronal rhythmicity. The olivary oscillations are transmitted, and possibly amplified, through the cerebellum, which entrains the thalamus, motor cortex, and brainstem nuclei. The finding of normal excitability of the cerebello-thalamocortical pathway itself in essential tremor is consistent with this hypothesis (227). This widespread entrainment of the motor system by olivary oscillations could explain why lesions in many areas of the nervous system suppress essential tremor. The olivary origin of tremor is also supported by PET studies that have revealed increased olivary glucose metabolism as well as increased blood flow to the cerebellum, red nucleus, and thalamus (65; 29). However, neither essential tremor nor its animal model has been consistently associated with defined morphologic abnormalities. There is, therefore, no pathologic evidence that the harmaline model and by extension the olivary hypothesis truly reflects the pathophysiology of the human disorder (76).

Parkinsonian tremor. The physiologic origin of tremor in Parkinson disease remains unclear. Studies have shown that multiple oscillators are likely involved. Parkinsonian resting tremor has long been known to involve thalamic oscillatory discharges, and tremor cells discharging in time to tremor frequencies are usually found in the ventral intermediate nucleus of the thalamus in patients with Parkinson disease during electrode implantation for deep brain stimulation in this region, which is an effective treatment for severe tremor in Parkinson disease. However, in both animal models and parkinsonian patients, tremor cells have also been reported in the subthalamic nucleus and the GPi. There is evidence that tremor cells are somatotopically organized in the STN and thalamus in Parkinson disease and in the thalamus in essential tremor (221). Many neurons in the dorsal STN exhibit oscillations at Parkinson disease tremor frequencies (296). In addition, significant interneuronal coherence in subthalamic nucleus neurons at the Parkinson disease tremor frequency has been reported. A finding interpreted as indicating that the subthalamic nucleus in Parkinson disease may be part of an extended and strongly coupled tremor network (07). Rodent studies have also shown that loss of inhibitory of striatal GABAergic neurons (fast spike interneurons) induces rest tremor and high coherence between basal ganglia oscillations. The tremors are reduced with pharmacological inhibition of the primary motor cortex, suggesting a role of these striatal interneurons and cortical projection to these interneurons in mediating rest tremor (214). Electrocorticographic measurements in parkinsonian patients undergoing STN-deep brain stimulation showed a distinctive drop in beta power in the sensorimotor cortex and in STN, and a loss of coherence between sensorimotor cortex and STN in the same frequency band during resting tremor (234). Moreover, higher STN-deep brain stimulation amplitude was associated with resting tremor control and a drop in the STN low gamma band power, suggesting that a peak in STN gamma frequency power is associated with parkinsonian tremor (25). Whether ventral intermediate nucleus cells truly represent tremor generators or are part of an unstable oscillating network remains unclear, but stereotactic surgery of these structures also effectively treats parkinsonian tremor, and accordingly parkinsonian tremor has also been attributed to aberrant discharges from the basal ganglia (301). A hypothesis for the generation of parkinsonian tremor is that there is a loss of segregation of signaling leading to abnormal synchronization among structures in a cortical-subcortical loop that involves the subthalamic nucleus, GPi, and thalamus, among other elements. According to this view, stereotactic surgery restores motor function by desynchronizing the striato-pallido-thalamic pathways, allowing a resegregation of signaling (77).

Using magnetoencephalography correlated with limb EMG in patients with Parkinson disease, Timmerman and colleagues demonstrated tremor-related oscillatory activity at tremor frequency and double tremor frequency within a cerebral network consisting of contralateral cortical motor and premotor areas, cingulate, supplementary motor area, diencephalon, secondary somatosensory cortex, posterior parietal cortex, and the contralateral cerebellum (278). This oscillatory activity significantly decreases after administration of levodopa (228). Furthermore, Mure and colleagues used FDG-PET to identify a specific metabolic network of co-varying increases in activity during Parkinson tremor, which consists of the cerebellum/dentate nucleus, primary motor cortex, and, to a lesser degree, the caudate and putamen. Ventral intermediate nucleus (VIM) stimulation reduced the activities of this network (199). Postural tremor in Parkinson disease could be reset by transcranial magnetic stimulation of the cerebellum (209), and resting tremor was entrained by cerebellar transcranial alternating current stimulation (38). These findings, together with animal (33) and human (188) data showing direct connections between the cerebellum and the basal ganglia support a role for the cerebellum as part of the network involved in the generation or transmission of the tremor. Several studies suggested that increased activities of the cerebellar pathways may be a compensatory mechanism in Parkinson disease (Wu and Hallet 2013). Further understanding of the interaction between the cerebellum, thalamus, cortex, and basal ganglia in the genesis of parkinsonian tremor comes from a study that combined functional imaging with EMG recordings in both tremor- and nontremor-dominant parkinsonian patients (119). It was found that tremor amplitude correlated with activities in the cerebello-thalamo (VIM) - cortical (M1) circuit. Tremor-dominant patients had increased functional connectivity between the GPi and putamen and the M1 node of the cerebello-thalamo-cortical circuit. The effectiveness of levodopa in controlling resting tremor is related to direct inhibition of the VIM (79). Cerebral activity timed locked to high amplitude tremor onset and offset was localized to the GPi, GPe, and putamen. Dopamine depletion in the pallidum, but not striatum, correlated with tremor severity. Taking these findings into account, Helmich and colleagues propose a “dimmer-switch” model of parkinsonian resting tremor, where basal ganglia, and more specifically GPi (78), trigger (switch) resting tremor in the cerebello-thalamo-cortical circuit, and this circuit acts as a “light dimmer” modulating tremor intensity (118). Resting tremor amplitude is also by modulated by influences exerted over the cerebello-thalamo-cortical circuit by the ascending arousal system and by the cognitive control network (fronto-parietal cortex, insula, thalamus, and anterior cingulate cortex) (82). The same group also found that patients with dopamine-resistant tremor showed increased cerebellar and reduced somatosensory cortex influences on the cerebellar thalamus whereas patients with dopamine-responsive tremor had reduced cerebellar and increased somatosensory influences on the cerebellar thalamus (81; 308), suggesting that different pathophysiology underlie dopamine-responsive and dopamine-resistant tremors. Further evidence of involvement of the cerebello-thalamo-cortical circuit in rest tremor arises from structural neuroimaging using diffusion tensor MRI (175). Tremor dominant but not nontremor dominant patients with Parkinson disease showed increased axial diffusivity in major white matter tracts including middle cerebellar peduncle, thalamus, internal capsule, and superior corona radiata. A study found that loss of raphe serotonin transporter and putaminal dopamine transporters were contributors to the occurrence of rest tremor. However, the degree of raphe serotonin transporter loss correlated with severity of rest tremor and reduced response to dopamine replacement therapy (218). A study that used multichannel EMG found that 81% of patients with Parkinson disease with postural tremor had reemergent tremor with similar frequency and levodopa response as resting tremor, whereas 19% of patients with Parkinson disease showed pure postural tremor with frequency higher than rest tremor and did not respond to levodopa (80). The authors concluded that reemergent tremor is another phenotypical manifestation of resting tremor, whereas pure postural tremor is mediated by a different neural oscillator.

Clinical applications

After many decades of research, our understanding of the pathophysiology of movement disorders remains limited. However, these efforts have nonetheless improved our understanding of the functional neuroanatomy and neurochemistry of motor control and have helped generate new models of how the brain controls movement. The increased interest in the neurophysiology of movement has led to a renewal of surgical interventions for movement disorders, which have improved quality of life for many individuals with symptoms not adequately controlled with conventional treatments. Intracranial recordings and functional imaging studies from these patients have, in turn, enhanced our knowledge of both normal and abnormal motor control. As ever, the data generated from these studies have raised new questions, highlighted the deficiencies of the existing models, and suggested new avenues for the investigation and treatment of a challenging and fascinating group of disorders.