Generators of the frequency-following response in the subthalamic nucleus: Implications for non-invasive deep brain stimulation

sound-evoked activity recordings in animals are typically restricted to the IC. Our study is the first to show that the FFR at a high frequency of 333Hz is present in the STN of four patients with DBS electrodes while absent in a fifth patient implanted in the thalamus, hippocampus and other areas. Our finding holds relevance for DBS and auditory-sensorimotor integration in the basal ganglia.


Methods
We recorded five patients (Table S1) with implanted, externalised DBS or depth electrodes at the Ruijin Hospital of Shanghai Jiao Tong University after obtaining informed consent (approved by Ruijin Hospital Ethics Committee).The DBS device was from a local manufacturer (SceneRay).We recorded simultaneous local field potentials (LFP) and EEG (64 electrodes, according to the 10-20 system) using Neuroscan (https://compumedicsneuroscan.com/, 10kHz sampling rate).Patients performed an audiometry screening to ensure sufficient hearing sensitivity (see supplementary material).
The experimental paradigm was a pure-tone 333Hz FFR similar to Ref. [1], with interstimulus interval reduced to 40 ms, no jitter.We recorded 6000 trials of 200 ms in alternating polarity per patient.DBS electrodes were localised with lead-dbs (https://www.lead-dbs.org/).Data was analysed using SPM (https://www.fil.ion.ucl.ac.uk/spm/) and FieldTrip (https://www.fieldtriptoolbox.org/).The data was filtered using a fifth-order high-pass Butterworth filter at 300Hz and a low-pass at 400Hz, bad trials were removed.The DBS electrodes were re-referenced to a bipolar montage.
The power-spectra were calculated using a fast-Fourier transform.Inter-trial coherence (ITC) was estimated by taking the absolute value of the sum of the angle of the Hilbert transform of the signal band-passed at 333Hz±2Hz.Dispersion of the phase was estimated using the circ_var function of the circular statistics toolbox in MATLAB.Significant FFR response compared to background noise was tested using F-statistics (α = 0.5, df = 2,11).Student's t-tests were done for ITC comparison of EEG and LFP.Bonferroni correction was done for number of tests.

Results
All patients showed significant FFR in the STN (12/21 STN electrodes, all p < 0.005).Patient 5 showed a significant response outside the left anterior thalamus (p < 0.0001) and a weak but significant response in the right hippocampus (p < 0.0005).No other electrodes survived correction for multiple comparisons.Fig. 1 shows a summary overview of the results for the phase and amplitude of all patients (also see S1 and S2).

Discussion
As the STN is anatomically closest to the IC (~2cm) compared to other channels (closest is the anterior thalamus at ~4cm), the responses could be due to volume conduction.We do not believe it is the case based on the following arguments: 1) An obtuse shift in phase, close to a phase-reversal was observed in 5 of the 7 STNs.2) Comparing with patient 5, we conclude the FFR response is anatomically specific.The FFR is present in a few non-STN electrodes (albeit with smaller amplitude and ITC).Volume conduction cannot explain why some electrodes show the FFR while others in close proximity do not.3) Finally, the ITC of EEG was significantly higher than the LFP.As LFP is less susceptible to motion and recording artefacts, with pure volume conduction the LFP-ITC would be higher or similar to the EEG-ITC.A reduction in phaselocking could be generated through a functional connection with the IC whereas the EEG is most likely volume conducted directly from IC.These arguments make a case for a locally generated STN-FFR.Note that we cannot fully rule out volume conduction of a far-field, future studies using intraoperative microelectrode recordings are needed.
Auditory perception in the basal ganglia is underexplored, with few  studies investigating direct human neurophysiology [3][4][5].Basal ganglia generators of the FFR may be better revealed with fMRI but studies have focused primarily on cortical activity [6] given the challenges of mapping brainstem nuclei.
To date, there is little to suggest direct connections between the STN and IC or other brainstem auditory nuclei.However, there are several lower brainstem motor centres such as the mesencephalic locomotor region (MLR) that both directly synapse with the STN and auditory nuclei [7].The link between midbrain auditory pathways, lower brainstem motor systems and the STN requires further assessment, and may have therapeutic implications.Evidence of interactions comes from "Paradoxical kinesia" in Parkinson's disease where there is a transient improvement in axial symptoms or increased grip force in response to sudden unexpected auditory stimuli, with the connections between the IC and MLR emerging as candidate pathways that regulates this motor response [8][9][10].
Future research should investigate whether auditory stimuli at common frequencies and waveforms of electrical DBS can induce clinical improvement.Furthermore, this opens new avenues for IC-DBS [10].

Fig. 1 .
Fig. 1. Results of FFR response in the STN of four patients.a) time and frequency domain representation of the auditory stimulus, generated in MATLAB b) time and frequency domain representation of neural response to FFR in an example patient for the channel pair optimally placed within STN as confirmed by lead-dbs c)Summary ITC and power of all invasive (LFP) channels (significant and not-significant), showing power at 333Hz and phase-locking value at 333Hz.d) bar plot for ITC between all invasive (LFP) channels and EEG for each patient.For all patients, the difference was significant at p < 0.00001 e) relative phase difference for each STN patient in the left and right hemisphere.The first channel-pair (01) is arbitrarily set at zero phase and the phases of other channel-pairs are shown relative to the first.Magnitude of the phase vectors indicates ITC for that channel-pair, dispersion of the phase is shown by the red striped lines for each channel pair.Nonoverlapping dispersion intervals indicate significant differences in phase delay between pairs.Solid lines, marked with an asterisk indicate a significant FFR response, using F statistics. f) visualisation of STN DBS electrodes in the STN, using lead-dbs.Lowest contact pairs are placed within the STN.