Current Models of Cortical–Basal Ganglia Networks

Models of the normal and pathologic physiology of the motor system, especially of corticobasal ganglia–thalamic function, have evolved significantly over the last 20 to 30 years to better account for observations such as the similarity between lesions and high-frequency stimulation of certain targets and the abnormalities observed in the parkinsonian subject and in animal models of striatal dopamine depletion. Seminal works from the 1980s and 1990s include the characterization of Albin, DeLong, and colleagues of multiple, parallel, segregated loops transmitted serially from the cortex; through the striatum, pallidum, and thalamus; and back to the cortex.^12–14 Gerfen et al.¹⁵ demonstrated distinct populations of striatal medium spiny neurons that possess mainly subtype-1 dopamine receptors and project directly to the GPi or subtype-2 dopamine receptors and project indirectly to the medial pallidum via the external segment of the globus pallidus (GPe) and STN. These models conceived of pallidal outflow as a tonic, inhibitory brake on the thalamus and helped explain the observations of aberrantly high neuronal discharge frequencies in the STN and GPi in animal models of hypokinetic states such as dopamine depletion¹⁶ and in human parkinsonism. This model of hypokinesia as a result of excessive inhibition of the thalamus by the pallidum was incongruous with some observations, such as the lack of hypokinesia observed with thalamic lesions. Similarly, the therapeutic effects of pallidal, thalamic, or subthalamic lesions on hyperkinesias challenge this classical model, which, in their simplest form, posit that hyperkinetic disorders result from reduced pallidothalamic outflow. Modifications of these models have incorporated other connections in the cortical–basal ganglia network, such as extrathalamic projections of pallidal output, reciprocal GPe-STN interconnections, and direct cortical-STN projections (“hyperdirect” pathway).¹⁷ However, newer models have also significantly shifted the emphasis from abnormal frequency and patterns of neuronal discharge in basal ganglia nuclei during loss of dopaminergic innervation to an emphasis on abnormal network activity, particularly in the form of abnormal neuronal synchrony and network oscillations measured by field potentials. The crux of these newer models, sometimes referred to as “noisy signal” hypotheses,⁸ is that different pathologic conditions may be characterized by abnormal neuronal synchrony and dominance of particular field potential oscillations in particular networks, potentially producing tremor (5 Hz), dystonia (3-10 Hz), or akinesia (20 Hz).^18,19 This shift in models owes much to investigations into the cellular and network substrates of thalamocortical oscillations performed in the 1980s, when Jahnsen and Llinas demonstrated dual behavior in thalamocortical neurons²⁰: When simply depolarized, these cells exhibited a “classical” behavior, emitting a series of spikes and relaying their inputs on to the cortex with high fidelity. When released from hyperpolarization, these cells exhibited a “burst” mode, where afferent information was largely prevented from being relayed to the cortex. The biophysical mechanisms of the bursting were found to result from the interplay between a hyperpolarization-activated cation current (known as I_h) and a transient low-threshold calcium current (known as I_T), which produces a “slow spike” lasting tens of milliseconds when released from hyperpolarization. These produce “pacemaker potentials,” which are rhythmic oscillations of thalamocortical cells in the delta frequency range (3-6 Hz).²⁰ These cellular oscillations are important for understanding oscillating field potentials such as those measured by electroencephalogram, but the cellular oscillations alone are insufficient to generate the delta frequency oscillations measured in field potentials. The relationship between rhythmic bursting at the cellular level and oscillations measured in field potentials is complex, and a number of mechanisms may contribute to synchronization of oscillating and nonoscillating cells. Networked models of separate thalamocortical oscillators that incorporate thalamocortical connectivity and feedback pathways can account for synchronized thalamic burst firing at delta frequency and generate field potentials. Models of these pacemaker potentials interacting in a simple network among the thalamocortical cells, the corticothalamic cells, and the reticular nucleus of the thalamus have been developed to explain how these phenomena generate local field potential (LFP) oscillations in the form of delta waves and spindle waves observed during sleep.²¹

Similar mechanisms for cellular oscillations, rhythmic activity, and synchronization across networks of cells have been proposed for the STN and its interactions with the pallidum. STN neurons have been identified that generate rebound burst activity mediated by hyperpolarization-deinactivated calcium channels,²² and roughly half of STN neurons in rat brain slices switch from a single-spike mode to burst-firing mode when continuously hyperpolarized.²³ Studies of cocultures of STN and GPe neurons demonstrate that this reciprocally interconnected network can generate correlated, rhythmic firing when isolated from the cortex and striatum.²⁴ Models of STN-GPe reciprocal bursting, similar to those described for spindle oscillations in the thalamus, involve synchronous bursting activity in GPe neurons that hyperpolarizes STN neurons sufficiently to generate rebound burst activity, which then drives bursting activity in GPe and perpetuates the rhythm.²⁵ The loss of dopamine, leading to overactivity in the striatal neurons of the indirect pathway would disinhibit GPe neurons, hyperpolarize STN neurons, accentuate burst firing in the STN, and generate pathologic synchrony between groups of neurons. This could result in the generation of oscillations that disturb cortical processing, or it could restrict communication among the cortex, STN, and GPe to certain bandwidths.

Many details of “noisy network” hypotheses remain to be defined, both theoretically and experimentally. By itself, the idea of a noisy network is sufficiently vague to admit all possible mechanisms of DBS and can potentially explain any physiologic observation. A more detailed articulation of such a model, which incorporates physiologic principles derived from the thalamocortical studies described earlier, has several salient features:

In this type of model of subcortical pathophysiology, an intriguing hypothesis regarding the mechanism of high-frequency stimulation is to trigger tonic firing, thus preventing neurons from returning to their hypersynchronized, bursting, oscillatory firing patterns. These models can also offer possible explanations of the diverse etiologies of movement disorders: single-ion channel dysfunction that changes the biophysics of single cells; loss of neuromodulatory transmitters such as dopamine, as seen in idiopathic PD; or altered network connectivity from plastic changes at cortical or subcortical levels, as seen in post-traumatic movement disorders or models of certain dystonias.^26,27 While many details of the cellular and network substrates of normal and pathologic field potential oscillations remain uncertain, some specific rhythms associated with specific symptoms are reasonably well characterized and deserve review.

Beta Oscillations and Parkinsonian Akinesia

Pathologic oscillations in movement disorders have been characterized most extensively in PD. Electrocorticographic and electroencephalographic investigations of motor tasks have long shown a decrease in signal power over motor cortices in the beta frequencies (about 20 Hz) during movement, sometimes referred to as event-related desynchronization.²⁸ Voluntary movement is accompanied by increased power over more focal areas of the motor cortex during movement in higher-frequency bandwidths.^29,30 The inability of the cortex to transition between beta oscillations and these focal higher-frequency oscillations appears to be a feature of parkinsonian akinesia, and the basal ganglia dysfunction in PD appears to be a critical part of this impediment. LFPs recorded from the STN, GPi, and motor cortex in parkinsonian patients demonstrate the presence of increased power in the beta band associated with akinesia.³¹ Not only is this increased power abolished with the administration of dopaminergic agonists or exogenous levodopa (l-dopa),³² but it remains suppressed in the STN immediately following high-frequency stimulation of that structure.³³ Accumulating evidence of this type suggests that suppression of the “beta straightjacket” may be a key to therapeutic efficacy of medications and DBS for the treatment of Parkinsonian akinesia.

Oscillations Associated with Tremor

The two most common types of tremor treated with stereotactic neurosurgery are the resting tremor of PD and the action tremor, most cases of which are diagnosed as essential tremor. Other types of tremor include those associated with lesions of the cerebellothalamic outflow, such as the tremor seen in multiple sclerosis, and “rubral” tremor.

Tremor is one of the cardinal symptoms of PD, consists of a rhythmic (4-8 Hz) activation of mainly distal musculature, and manifests at rest and in some cases while maintaining posture. Identification of pathologic oscillations in the nervous system that correlate with the oscillatory movements of parkinsonian tremor has proved challenging. The question is complicated by the problem of defining a physiologic oscillation that is efferent in character (i.e., drives the tremor) versus one that is afferent (i.e., results from proprioceptive feedback). Physiologic models of tremor have evolved to account for the observations that dorsal rhizotomies had minimal effects on tremor (suggesting central oscillators as drivers) and that Vim DBS is often able to treat parkinsonian tremor but does not affect the other cardinal symptoms of the disease to the same degree. Tremor-frequency single-unit activity (“tremor cells”) in the ventral oral posterior nucleus, Vim, and ventral caudal nucleus thalamus has been described,^34,35 as have cells with tremor-frequency bursting oscillations in the STN³⁶ and globus pallidus.³⁷ Pare et al. hypothesized that rhythms transmitted through inhibitory pallidal outflow to the motor thalamus were transformed into the 3- to 6-Hz tremor frequency in the thalamocortical circuit through the low-threshold calcium spikes induced by transient (T-type) calcium channels.³⁸ Hassler et al. were able to produce contralateral resting tremor with stimulation of the internal segment of the globus pallidus at frequencies of less than 25 Hz,³⁹ and Plaha et al. were able to elicit a 3- to 6-Hz tremor in a previously tremor-free parkinsonian patient by stimulating the caudal ZI at the beta frequency. In the latter study, low-frequency stimulation of the GPi or PPN did not produce tremor, and STN stimulation at low frequencies only produced tremor at high amplitudes.⁴⁰ Because tremor in muscles of different extremities is not usually coherent, it has been concluded that independent central oscillators generate tremor separately in separate extremities.⁴¹ Simultaneous recordings of spatially separated tremor-related cells in the GPi showed that some pairs of cells oscillated in phase and some oscillated out of phase with one another,³⁷ leading the authors to hypothesize that tremor-related activity may be propagated independently by several parallel pathways.

Surgical Targets for Tremor

Since Benabid’s demonstration of efficacy with high-frequency stimulation (more than 100 Hz) of the Vim thalamus for the treatment of parkinsonian tremor, this site has emerged as the preferred target.⁴² While this therapy seems to produce effects that are indistinguishable from a lesion of this target, it remains uncertain what ideal subregion of the thalamus (or possibly extrathalamic structure) provides the best tremor control, whether through stimulation or ablation. This question was posed during the ablative era of functional neurosurgery, highlighted by Laitinen’s survey of 16 neurosurgeons, who were asked to define their choice of target for treatment of a patient with PD who exhibited tremor and rigidity in equal measure. Targets differed by as much as 7 mm and included a wide span of the ventrolateral thalamus; the PSA, including the ZI; and the pallidothalamic fibers of the Forel fields.⁴³ Some of the difficulty in defining the ideal target stems from the multiple parcellation schemes for the motor thalamus,⁴⁴ though most stereotactic neurosurgeons employ Hassler’s terminology, which forms the basis for thalamic parcellation in Schaltenbrand and Wahren’s widely used atlas.⁴⁵ The question of the ideal target is also complicated by questions as to how stimulation differs from ablative surgery. The ideal target for stimulation may be different from the ideal target for ablation. Among the multiple hypotheses that have been put forth regarding the mechanism or mechanisms of DBS, two intriguing ones related to tremor are the observation that adenosine release may be critical to the effect of Vim stimulation⁴⁶ and data that suggest stimulation in the dorsal STN may act by directly modulating pallidothalamic or cerebellothalamic fibers in the immediate vicinity.⁴⁷ Finally, while Vim DBS has been shown to be effective for parkinsonian and essential tremor,^48,49 other tremor syndromes respond only inconsistently or less dramatically to Vim DBS. In addition to controversy surrounding the ideal target for a particular tremor syndrome, there is the question of the ideal target for those patients exhibiting combinations of tremor syndromes. Patients exhibiting symptoms of both PD and essential tremor may be treated with Vim DBS, but other parkinsonian symptoms would not be expected to respond. For these patients, and for patients with tremors that respond inconsistently or less dramatically to Vim DBS, it is important to describe the role of sites such as the STN, which some data suggest it can effectively treat symptoms of essential tremor,⁵⁰ and the PSA.

Posterior Subthalamic Area

The PSA encompasses the ZI and the prelemniscal radiation (RaPRL). This was the target area of subthalamotomies performed frequently from the 1960s to the 1980s. Various terminologies have been used to define approximately the same target area, such as the PSA, ZI, RaPRL, subthalamic area, and posterior subthalamic white matter. In this chapter, we refer to this area as the PSA and discuss in further detail two of its most important components, the ZI and the RaPRL. The PSA is bordered anteriorly by the posterior STN, inferiorly by the dorsal substantia nigra, superiorly by the ventral thalamic nuclei, posteriomedially by the red nucleus, posteriolaterally by the ventrocaudal nucleus, posteriorly by the lateral lemniscus, and laterally by the posterior limb of the internal capsule (Fig. 116-1). As a surgical target, the PSA has been extensively reviewed by Blomstedt et al., who accounted for a total of 95 patients implanted with PSA DBS: 42 of these patients for the diagnosis of PD, 21 for multiple sclerosis, 18 for essential tremor, 7 for torticollis, 2 for dystonic tremor, 2 for post-traumatic tremor, 1 for cerebellar tremor, 1 for Holmes tremor, and 1 for spinocerebellar ataxia.⁵¹ Several reports have suggested that high-frequency stimulation of the PSA offers superior tremor control to stimulation of the Vim. Hamel et al. systematically analyzed the location of active contacts effective in suppressing intention tremor and found that those below the intercommissural plane in the subthalamic region were most effective.⁵²

FIGURE 116-1 T2-weighted MRI demonstrating selected structures based on Schaltenbrand and Wahren’s atlas. A, Axial image corresponding to horizontal slice H.v −3.5. Structures illustrated include the red nucleus (red), STN (orange), Zi (green), and medial and lateral segments of the GPi (purple). B, Sagittal image corresponding to sagittal slice S.I 10.5. Additional illustrated structures include the substantia nigra (yellow), and the RaPRL (blue). (Modified from Schaltenbrand G, Wahren, W: Atlas for Stereotaxy of the Human Brain. Stuttgardt, Thieme, 1977.)

Complications of PSA DBS can include dysarthria (potentially from overly lateral placement of the electrode stimulating the corticobulbar fibers in the internal capsule or overly medial placement of the electrode stimulating the cerebellar fibers of the RaPRL); disequilibrium; lethargy; or worsening symptoms of preexistent depression.⁵³

Analysis of lead locations following intended STN DBS electrode placement demonstrated a correlation between best clinical effect and placement of the electrode dorsal to the STN in the white matter bundles of the ZI and H1 and H2 fields of Forel, outside of the PSA.^54,55

Zona Incerta

The ZI is a collection of gray matter that is continuous ventrally with the reticular nucleus of the thalamus. Functionally, it is divided into four parts: rostral, ventral, dorsal, and caudal. The rostral ZI is separated from the STN by the pallidofugal fibers crossing the internal capsule on the dorsomedial side of the STN. The ZI receives afferent input from the GPI or substantia nigra pars reticulata (SNr), the PPN, and cortical projections. Its gamma-aminobutyric acid–ergic (GABAergic) efferents target the CM/PF, the ventral anterior–ventral lateral nuclei of the thalamus, and the cortex, among others. The caudal or motor component of the ZI extends posterior to the STN and the RaPRL. Forel described it in 1877 as a “region of which nothing certain can be said,” and it remains an understudied region. Although the ZI is not considered part of the basal ganglia, the anatomy and physiology of the two regions overlap considerably. Both receive excitatory cortical input and send inhibitory projections to the thalamus, and both are involved in motor control and attention orientation, among other shared functions.⁵⁶

Recent studies in rodents have demonstrated that this region plays a critical role in the gating of somatosensory input by the brain stem⁵⁷ and the motor cortex.⁵⁸ Like the neighboring reticular nucleus of the thalamus, cholinergic input from the brain stem results in state-dependent input from the ZI to the thalamus, which may contribute to elevated sensory thresholds during sleep. The ZI serves as an important locus for modulating sensory information from diverse sources by gating unwanted sensory input via inhibition of thalamic neurons.⁵⁸ It has been proposed that the efferent GABAergic signal of ZI neurons patterned by cortical activity can play a critical role in synchronizing thalamocortical and brain stem rhythms.⁵⁹ Subsequently, it has been implicated as a generator of tremor in PD and essential tremor.⁴⁰

Given the similar connectivity between the ZI and the basal ganglia, it is perhaps not surprising that DBS achieves many of the same results in the ZI as it does in the basal ganglia. Indeed, high-frequency DBS may have a greater effect in the dorsal ZI than in the STN. In the largest case series published of 35 patients with refractory idiopathic PD, patients were implanted with a total of 64 high-frequency DBS electrodes. Of these, 27 DBS electrodes were implanted into the dorsal ZI, 20 into the dorsomedial STN, and 17 into the STN. This observational study suggested that high-frequency DBS of the ZI is superior to that of the STN in improving contralateral Unified Parkinson’s Disease Rating Scale motor scores. Some patients developed speech and balance disturbances following high-frequency DBS, thought to be due to the proximity of the lead to the RaPRL.⁶⁰ A report by Kitagawa et al. from 2005 targeting the caudal ZI in patients with tremor-dominant PD noted 78.3% improvement in contralateral tremor and even superior improvement in rigidity.⁶¹

Prelemniscal Radiation

The ZI separates posteriorly the dorsal STN from the RaPRL located posterior to the ZI/STN. Its ascending fibers originate from the mesencephalic reticular formation and the cerebellum projecting to the thalamus. Thus, it is thought to play a role in tremor formation and muscle tone in the context of attention. Continuous stimulation of the RaPRL abolishes contralateral tremor with reduction in rigidity by inhibition of local circuits.⁶² With RaPRL DBS, dysarthria can be a complication by cerebellothalamic fiber stimulation and disequilibrium from the stimulation of the ascending mesencephalic reticular formation. Murata et al. targeted the RaPRL in eight patients with essential tremor of the proximal muscles using these coordinates: lateral to the red nucleus (about 10 mm from midline) and 3 to 4 mm posterior to the posterior border of the STN in the axial plane through the largest cut over the STN. Postoperative localization analysis resulted in mean stereotactic coordinates in patients with more than 71% improvement at 7.6 ± 1.2 mm posterior to the midcommissural point, 10.9 ± 0.8 mm lateral from midline, and 3.9 ± 1.7 mm inferior from the AC-PC line.⁶³ This targeting yielded marked improvement in both proximal and distal tremor control.