Neurons in the globus pallidus internus fire continuously. During wakefulness, over eighty times per second. During sleep, the rate drops somewhat, but never stops. This sustained high-frequency inhibitory output projects to the thalamus, which then projects to the cortex. The resting state of this entire pathway is one of suppression.
The striatum projects to the globus pallidus internus. Medium spiny neurons in the striatum are GABAergic. When striatal activity increases, the globus pallidus internus is inhibited, its firing rate decreases, and the thalamus is released from inhibition.
This logic chain was organized into the direct and indirect pathway model in the late 1980s. Albin, Young, and Penney were one group. DeLong was another. The two papers were published a year apart. The model's predictions aligned with the pathophysiology of Parkinson's disease and with the clinical effects of subthalamic nucleus lesioning. Later, the selection of targets for deep brain stimulation also referenced this model.
Classical Basal Ganglia Circuit Model
In primates, the striatum is divided into two parts by the internal capsule. The caudate nucleus follows along the lateral ventricle; the putamen sits laterally. The total number of neurons in the human striatum is around one hundred million, the vast majority being medium spiny neurons. Interneurons account for only a few percent, but they are quite diverse—cholinergic, parvalbumin-expressing, somatostatin-expressing, calretinin-expressing—each with their own electrophysiological characteristics and connectivity patterns.
Cholinergic interneurons are also called TANs, because in behavioral tasks they show a brief pause response to reward-related events. This pause was described by Kimura in the early 1990s. He recorded it using classical conditioning in monkeys. The pause lasts about two hundred milliseconds and occurs after the conditioned stimulus appears.
Aoki in 2015 used transgenic mice to knock out striatal cholinergic neurons, and the mice began exhibiting stereotypic behaviors. Licking, digging, circling. This resembles some manifestations of Huntington's disease. Striatal cholinergic interneurons are indeed reduced in Huntington's patients. But from stereotypic behavior in mice to chorea in humans, the inference must cross many steps.
The classification of direct and indirect pathways is based on dopamine receptor subtypes. D1-positive medium spiny neurons take the direct pathway; D2-positive ones take the indirect pathway. These two types of neurons are intermingled in the striatum with no obvious spatial segregation.
Cui and colleagues in 2013 performed an experiment using calcium imaging to simultaneously record both types of neurons in freely moving mice. When the mouse initiated movement, both types of neurons activated together. Not the pattern of D1 activation with D2 silence. Jin later reported similar results.
Tecuapetla's research went further. Using optogenetics to separately activate both types of neurons, they found both could induce movement. The difference lay in the type and duration of movement. D1 pathway activation made mice initiate new movement sequences; D2 pathway activation maintained ongoing movements.
Pathway Functions Revised
Traditional models suggested D1 (direct) pathway promotes movement while D2 (indirect) pathway inhibits it. Modern optogenetic studies reveal both pathways activate during movement initiation, with D1 driving new sequences and D2 maintaining ongoing actions—a more nuanced picture than the classical dichotomy.
The tremor frequency in Parkinson's disease is between four and six hertz. Neurons in the subthalamic nucleus in Parkinsonian animal models show oscillatory firing at the same frequency. So does the globus pallidus. Whether there is a causal relationship between these pathological oscillations and tremor has been debated for many years.
Deep brain stimulation can suppress tremor. The stimulation frequency is usually set around one hundred thirty hertz. Why does high-frequency stimulation work? Some say it's equivalent to a functional lesion, scrambling the output of the subthalamic nucleus. Some say it simply overrides the pathological oscillation. Others say high-frequency stimulation actually orthodromically activates the subthalamic-pallidal projection.
Clinical Reality
The mechanism remains unclear, but it works clinically. The parameters were found empirically through trial and error.
Hoshi used viral tracers to track projections between the cerebellum and basal ganglia. Injecting from motor cortex, crossing two synapses reaches both the cerebellar dentate nucleus and the subthalamic nucleus. This means the same population of cortical neurons can simultaneously influence both subcortical structures.
Projections in the reverse direction also exist. Dentate nucleus to thalamus, then to striatum. Subthalamic nucleus to pontine nuclei, then to cerebellar cortex.
| Connection | Pathway | Functional Status |
|---|---|---|
| Cortex → Cerebellum | Motor Cortex → Dentate Nucleus | Anatomically confirmed |
| Cortex → Basal Ganglia | Motor Cortex → STN | Anatomically confirmed |
| Cerebellum → Striatum | Dentate → Thalamus → Striatum | Function unknown |
| STN → Cerebellum | STN → Pontine Nuclei → Cerebellar Cortex | Function unknown |
The function of these connections is completely unclear. Anatomy has outpaced physiology.
Deep brain stimulation for obsessive-compulsive disorder received FDA approval in 2009. The target is at the junction of the ventral internal capsule and ventral striatum. Response rate is around sixty percent, defined as a Y-BOCS score reduction exceeding thirty-five percent.
The choice of target has historical reasons. In the 1950s, anterior limb of internal capsule lesioning was performed for treatment-resistant OCD, and it was effective. Later, when lesioning became ethically problematic, it was replaced with reversible stimulation. The target was simply carried over.
In reality, no one truly knows where stimulation should occur. The ventral internal capsule has large numbers of fibers passing through; stimulating it affects multiple pathways simultaneously. The nucleus accumbens is right next door. The extent of current spread depends on stimulation parameters. Different hospitals use different parameters, and electrode positions vary by several millimeters. How much of the outcome variance comes from the patients themselves versus these technical variables cannot be distinguished.
Caudate nucleus atrophy in Huntington's patients is clearly visible on MRI. The area beside the lateral ventricle is indented. The degree of atrophy correlates with disease duration and with CAG repeat number.
But how exactly atrophy occurs, why medium spiny neurons are particularly sensitive to mutant huntingtin protein—the mechanism remains unsettled. The striatum contains various neuron types, yet the ones that die most are the most numerous. Is it because there are many so they die easily? Or is there another reason?
Which is the main factor? Or do all participate, each accounting for a portion? HDAC inhibitors entered clinical trials—failed. BDNF-related strategies—also failed. ASO therapy results recently—also not ideal.
Using the striatum as a target has one problem: it's too large. AAV distribution is uneven. One injection cannot cover the entire structure. Multiple injection points increase surgical risk.