In contrast to β frequencies, γ (31–100 Hz) frequencies in STN were related to normal motor functions. Increased γ frequencies between 60 and 80 Hz in STN were often observed along with improved motor performances when patients were in their ON state or were engaged in voluntary movements.^3,6–8 Moreover, STN γ activities may be related to state of arousal as they disappeared when patients became drowsy.⁷ Taken together, these observations led to the hypothesis that increased STN γ frequencies may represent improved functional state of the BG and its interaction with the cortex in patients with PD. Therefore, γ frequencies may facilitate motor performance during voluntary movements.^3,7,8,11 However, the peak θ, β, or γ frequencies were highly variable from patient to patient.^3,12

Several studies reported that levodopa-induced reduction in β band power in the STN was associated with improvements in bradykinesia and rigidity but not tremor in patients with PD.^13–15 STN DBS at 20 Hz was found to modestly decrease tapping speeds^16,17 but had no effect on motor symptoms¹⁸ whereas at 50 Hz was reported to reduce rigidity and improve tapping speed in a study with 3 subjects¹⁹ but other studies showed no benefit.^17,18,20,21 No previous study examined the effects of DBS at individualized STN peak frequencies, which may be highly relevant to behavioral performances.^10,14 In the present study, we examined the effects of DBS at individualized peak STN frequencies on parkinsonian motor signs. We hypothesized that STN DBS at individualized γ frequencies would improve whereas at θ and β frequencies would worsen PD motor signs compared to no stimulation.

The increased γ oscillations in the BG as patients with PD transitioned from OFF to ON dopaminergic medication states were considered prokinetic because this was associated with concurrent improvements in parkinsonian motor symptoms.^6,7 These dopamine-dependent γ rhythms may also be an indicator of the arousal state as they disappeared in the BG during drowsiness⁷ and ceased to be observed in the ventral thalamus (VT) during deep sleep.²⁵ In the ON state, these dopamine-dependent γ rhythms occurred throughout the BG-thalamo-cortical network as coherences were reported to be observed between the STN, GPi, VT, and the cortex.^7,25,26 These γ rhythms were suggested to be involved in attentional processes in the BG-thalamo-cortical circuit and cortico-cortical interactions to facilitate cognition and motor-related functions.^7,25,27 We demonstrated for the first time the acute clinical benefits of these intrinsic γ frequencies of the BG through STN DBS, which decreased parkinsonian motor signs similar to that produced by well-established HF DBS. Therefore, increased γ frequency in the STN induced by dopaminergic medications is likely a prokinetic rhythm in the human BG.

The increased γ oscillations in the cortico-BG circuit that occurred shortly prior to and during voluntary movements in patients with PD were thought to represent the coding and the transmission of motor information between the cortico-BG network and the muscles during movement executions.^3,28 These γ frequencies were considered to represent normal movement-related activities because similar activities were recorded the GPi in dystonia patients²⁹ and in the motor cortical regions of normal subjects or in patients with epilepsy during voluntary movements.^30,31 However, the precise functions of these movement-related γ oscillations in the cortico-BG circuit remains unclear. Our results indicated that peak γ frequencies in the STN during voluntary movements were effective in decreasing parkinsonian motor signs, consistent with the hypothesis that these γ oscillations may reflect outputs of the BG-thalamo-cortical network representing coding of voluntary movements and promotion of these γ oscillations may thus facilitate movement executions.

Dopamine-dependent γ oscillations in the BG have been suggested to be a carrier rhythm mediating information transfer between the BG and the motor cortical areas during voluntary movements. HF stimulation, empirically effective frequencies for chronic STN DBS, was hypothesized to be the second harmonics of intrinsic BG dopamine-dependent or movement-related γ rhythms and therefore effective in reducing PD motor signs.⁷ Our results indicated that although similar peak frequencies in the γ range were found in 3 patients (patients 1, 2, 6) (table 2), most patients displayed distinct γ peak frequencies in dopamine-dependent and movement-related conditions. Moreover, HF for chronic STN DBS (table 2) was not the second harmonics of the specific dopamine-dependent or movement-related γ peak frequencies we found in the STN. Another hypothesis regarding the therapeutic effects of HF STN DBS involved the disruption of abnormal rhythms in the BG network, in particular the β frequencies.^2,9,10 Whether DBS at peak γ frequencies also suppress excessive β oscillations in the BG needs to be examined. These results suggest that the BG-cortical circuit may have 2 prokinetic γ rhythms with peak frequencies that are variable across patients and they may be effective in reducing parkinsonian motor symptoms.

Previous studies using STN DBS in the γ range at ~50 Hz did not produce consistent clinical benefit and at least 100 Hz is required to produce consistent reductions in parkinsonian motor signs.^18,20,21 These are in contrast to our findings that dopamine-dependent or movement-related peak γ frequencies in the STN between 30 and 90 Hz were as effective as HF in reducing PD motor signs (figure 2, figure e-2). The reason for this difference may be because γ frequency DBS need to be at the intrinsic peak frequency to be effective and the optimal γ frequency varied widely among patients (table 2).^3,12 In the present study, we used the individual peak STN γ frequencies rather than setting the DBS at a fixed γ frequency. However, we need to study γ frequencies other than the individualized peak γ frequencies to determine whether the prokinetic effects are specific to individualized peak γ frequencies. The effects of STN DBS at individualized dopamine-dependent and movement-related peak γ frequencies warrant further investigations to optimize chronic DBS treatments for PD.

Our results showed that applying dopamine-dependent or movement-related peak θ or β frequencies of the STN did not increase PD motor signs whether in the OFF or ON states as measured by mUPDRS. Previous studies of STN DBS at θ and β frequencies did not produce consistent results. Unilateral STN DBS at 5 Hz was found to reduce hand-tapping speed.²⁰ Two studies that measured finger-tapping rate with the repetitive depression of a single key with the index finger found that bilateral STN DBS at 5 and 20 Hz reduced tapping speed by ~12% and ~8% compared with no stimulation but only a subgroup of patients with PD who demonstrated normal baseline tapping rate and stimulation at 10 Hz did not decrease tapping rate compared to no stimulation.^16,17 In contrast, bilateral STN DBS at 10 Hz was found to significantly worsen PD motor symptoms particularly bradykinesia measured by mUPDRS but stimulation at 5 and 20 Hz had no effect.¹⁸ We found that STN DBS at individualized peak θ or β frequencies did not slow hand-tapping speed compared to no stimulation. This may be because we used different methods compared to previous studies. Our hand-tapping test involved more complex arm movements compared to the simple finger tapping.^16,17 We used peak θ or β STN frequencies instead of fixed frequencies and used unilateral STN DBS instead of bilateral stimulations used in previous studies.^16–18 Regardless of differences in methodologies and findings, STN DBS at θ and β frequencies did not produce a consistent increase in parkinsonian motor symptoms. No previous study examined the effect of low-frequency STN DBS in the ON medication state. Our results suggest that STN DBS at peak θ and β frequencies did not block the effects of levodopa (figure 2). Therefore, increased θ or β oscillations in the STN may be indicators^13–15,32 rather than contributors to PD motor symptoms. However, we cannot rule out subtle effects. It is also possible that the single pulse stimulations we used were not sufficient to synchronize the intrinsic θ and β rhythms. A train of pulses may be required³³ but this type of setting was not possible with the pulse generator used. See appendix e-1 for a discussion of the limitations of the study.