We tested seven right-handed patients with PD (6 males, 1 female; median age 61 years [range: 67–53 years]; median disease duration 11 years [range: 10–19 years]). All patients were diagnosed according to the UK Parkinson Disease Brain Bank criteria (Hughes et al., 2002) and evaluated with the Unified Parkinson Disease Rating Scale motor part (UPDRS-III). All subjects were right-handed as assessed by a modified Edinburgh handedness inventory. Patients were implanted at the University Hospital of Würzburg between December 2013 and May 2014 with the Activa PC+S^® neurostimulation system (Medtronic, PLC). This system allows therapeutic DBS as well as on-demand LFP recordings from the implanted STN electrodes (Rouse et al., 2011; Stanslaski et al., 2012). The Activa PC+S^® system and the related hardware and software for programming and readout were provided under a request for application agreement by Medtronic, PLC. The company had no impact on study design, patient selection, data analysis, or reporting of the results. All patients had been selected based on established criteria for DBS surgery (Pollak, 2013). Of relevance, none of the subject had cognitive decline or mood disturbances as evaluated using the Parkinson neuropsychometric dementia assessment (PANDA), Mattis Dementia Rating Scale (MDRS), Hamilton Depression Rating Scale (HDRS) and the Non-Motor Symptoms Scale (NMSS).

The surgical procedure has been described elsewhere (Steigerwald et al., 2008). In brief, implantation was performed under local anesthesia using Leksell’s Frame (Elekta, Leksell Stereotaxy System, Stockholm, Sweden). The DBS electrode used was model 3389 (Medtronic, PLC) with four platinum–iridium cylindrical contacts of 1.5 mm each and a contact-to-contact separation of 0.5 mm. Contact 0/8 was the lowermost and contact 3/11 the uppermost (E0–3 refers to right- and E8–11 to the left-hemisphere). The intended coordinates for STN were 12 mm lateral, 2 mm posterior, 4 mm ventral to the mid-commissural point and were adjusted according to individual STN delineation on T2-weighted and SWI images (Magnetom Trio, Siemens Healthcare, Erlangen, Germany) and with intraoperative microelectrode recordings. Micro- and macro-electrode stimulation and intraoperative CT scan also served to confirm targeting. Postoperative scanning (1 mm slice-thickness, CT scan fusion with the pre-operative MRI) confirmed electrode location. Of note, the presence or absence of LFP power in the beta band was not used to determine the placement of the DBS lead (Quinn et al., 2015).

The precise localization within the STN of the active contacts used for chronic stimulation was further confirmed by image fusion of a non-stereotactic postoperative CT with the preoperative planning MRI by means of Optivise^® software under a research agreement with Medtronic, PLC. Correct placement of the DBS electrode was also verified by the clinical response to DBS (meds-off/stim-on) compared to the preoperative improvement of the UPDRS motor score (UPDRS-III) during levodopa challenge (meds-off vs. meds-on; Table 1). Therapeutic response to DBS or levodopa was expressed as percentage of improvement, according to the formula: ((a—b)/a) × 100 (adapted from Isaias et al., 2008) where a = meds-off UPDRS-III score and b = meds-on UPDRS-III at pre-DBS or b = meds-off/stim-on UPDRS-III at the time of the test, 4 months after surgery (post-DBS). The mean percentage of improvement was 65.77% (range: 42.5%–92.72%) due to dopaminergic medication and 68.82% (range: 52.5%–83.63%) due to STN stimulation, thus further supporting correct placement of the DBS electrodes.

Demographic and clinical information for all subjects is listed in Table 1. At the time of this study, all patients were on stable dopaminergic treatment (for at least 2 months) and chronically stimulated for 4 months (at least 1 month with unchanged DBS stimulation parameters). The local institutional review board of the University Hospital Wuerzburg approved the study and all patients gave written informed consent.

SPECT data acquisition, reconstruction (Lapa et al., 2015) and analysis has been described in detail previously (Isaias et al., 2010, 2011). All patients but one (i.e., wue5) were willing to perform a SPECT with FP-CIT to measure dopamine reuptake transporter (DAT) density. SPECTs were performed within 3 months before surgery. Scans were started 180 min after injection of 182.3 ± 3.6 MBq of FP-CIT on a dual-headed integrated SPECT/CT system (Symbia T2; Siemens, Erlangen, Germany) in the meds-on condition. In brief, SPECT data were spatially normalized onto a FP-CIT MNI-based template and volumes of interest (VOI) of caudate nucleus, putamen and striatum (for both hemispheres), as well as a reference region in the occipital cortex, were defined using the automated anatomical labeling (Tzourio-Mazoyer et al., 2002). The non-displaceable binding potential (BP_ND) was then assessed using average regional uptake values from VOI analysis and the occipital cortex as the reference region (Innis et al., 2007). The asymmetry index (AI, expressed as a percent) of whole striatal DAT availability was calculated as the BP_ND difference (Striatum_ipsilateral−Striatum_{contralateral}) relative to the mean value of both striatum.

Striatal DAT binding measurements for each patient were compared with normal values of 15 healthy subjects (4 males, 11 females, age range: 44–68 years; Table S1, Supplementary Material). The striatal dopaminergic loss exceeded 50% bilaterally in all but one subject (i.e., wue2, left striatum: 67.7% and right striatum: 47.5%). As previously reported (Panzacchi et al., 2008), the clinically most affected hand (higher UPDRS-III scores) always corresponded to the striatum with less nigro-striatal dopaminergic innervation. The percentage loss of DAT binding values is listed in Table 1 section B. The most affected side was the right one in two out of the four patients who completed the whole study protocol. Based on molecular imaging and clinical data, we identified the hemisphere with less (STN− and motor cortex, MC−) or more (STN+ and MC+) dopaminergic innervation.

The motor tasks have been extensively described in previous studies (Ghilardi et al., 2000; Perfetti et al., 2010; Isaias et al., 2011; Moisello et al., 2015). Briefly, subjects performed a single, multi-joint, uncorrected movement, as accurate and as fast as possible (Figure S1, Supplementary Material). They moved a cursor with either their dominant (right) or non-dominant (left) hand on a digitizing tablet, straight out-and-back, from a central starting point to one of eight equidistant (4 cm) radially-arranged targets that appeared on a screen. An opaque panel prevented the arm vision. All targets were displayed on a screen as circles (2 cm diameter). Targets were presented in random order every 3 s in three blocks of 16 movements each. Participants were tested in meds-off/stim-off condition (i.e., after overnight withdrawal of all dopaminergic drugs and after pausing DBS for at least 1 h) and were asked to perform the task first with the dominant (right) hand and then with the non-dominant (left) hand, irrespective of the more affected body side.

A neurologist (IUI) supervised the absence of any mirror movement or tremor in the hand not performing the task. The data presented refer only to the four subjects who completed the whole study protocol. Three patients were not able to complete the task in the required amount of time (i.e., 3 s per movement) with the right or left hand, and therefore their data were excluded from the analyses.

For each movement, we measured: onset time (OT, time from target appearance to movement onset), movement time (MT, time from movement onset to reversal), peak velocity (PV) and peak acceleration (PA) and path length (PL, from onset to reversal).

LFPs were recorded with a single bipolar contact configuration for each STN and amplified by 1000. The recording contacts were chosen according to the chronic stimulation setting as a bipolar montage of the two contacts surrounding the stimulation cathode (Devos et al., 2006; Quinn et al., 2015). High density EEG (hdEEG) signals were acquired with a 128-channel EEG Brainamp system (BrainAmp ExG, Brain Product) with sampling frequency at 1000 Hz. LFP and hdEEG recordings were synchronized by means of a common external signal (Figure S2, Supplementary Material), re-sampled at 250 Hz, bandpass filtered in the range 0.5–80 Hz and segmented into 6 s epochs based on the movement onset latencies from −4 s to 2 s after the return-time (RT). RT equals the time point in which the subject came back to the central target (Figure S1, Supplementary Material).

The hdEEG channels affected by bad scalp-electrode were visually identified and replaced with spherical spline interpolation. Trials with sporadic artifacts were excluded by visual inspection. Stereotypical artifacts (e.g., blinks, heartbeat, and muscle tension) were removed by independent component analysis (ICA; Jung et al., 2000; Onton and Makeig, 2006; Onton et al., 2006). For LFPs artifacts management, please refer to the supplementary material (Figure S3). LFP and hdEEG signals were processed and analyzed by means of MatLab-based custom script, EEGLAB (Delorme and Makeig, 2004), Brainstorm (Tadel et al., 2011) and SPM M/EEG Toolbox (Litvak et al., 2011b).

We then calculated the event-related power relative changes (ERD and ERS), normalizing the mean beta power by subtracting and dividing the average power of the whole task interval (from −3 s to 0 s) relative to the RT, multiplied by 100. For efficient spectral estimation of the relatively small number of trials, we used multitaper spectral analysis (Thomson, 1982). We estimated the spectra between −4 s and 2 s relative to the RT of each trial, in overlapping windows of 400 ms with a time resolution of 50 ms. The time-frequency bandwidth was set to 1.5, resulting in two tapers being used. The time-frequency images were then averaged using robust averaging (Wager et al., 2005; Holland and Welsch, 2007; Litvak et al., 2012) and the event-related power changes were obtained by normalizing to the whole trial (−3 s to 0 s) before RT. A representative time-frequency plot showing the cortical and subcortical ERD and ERS is shown in Figure S4 (Supplementary Materials, Methods).

To study cortical beta variations, we assessed the rebound in the beta band (13–35 Hz). The beta-rebound is the largest peak-to-peak difference between the minimum of the ERD during movement and the maximum of the post-movement ERS in beta band calculated, respectively, between −2 s and −1 s, and between 0 s and 1 s after RT.

Topological maps of these parameters showed two main clusters around CCP3h and CCP4h, as described also in previous studies (Alegre et al., 2005; Meziane et al., 2015; Moisello et al., 2015). Accordingly, we defined two region of interest (ROI) of eight electrodes surrounding CCP3h and CCP4h respectively. The selected electrodes predominantly represent activity over the primary sensorimotor cortex, being the supplementary motor area more medially and the premotor cortex more frontally located (Lalo et al., 2008).

To study the functional connectivity between cortical and subcortical structures, we then computed the coherency between the LFP signals of both the STN and the aforementioned cortical areas (Friston, 2011). We adopted the same method used for spectral estimation for the estimation of coherency. In this case, we also performed a robust averaging (Litvak et al., 2012). Finally, we computed the absolute value of the coherency (i.e., the coherence Coh) and the imaginary part of the coherency (iCoh) to isolate the part of coherency possibly affected by volume conduction (Nolte et al., 2004, 2008; Hohlefeld et al., 2013).

Statistical significance of the behavioral performances was assessed by means of a two sample unpaired t-Test with a significance level p < 0.05.

Beta ERD and ERS were calculated computing the mean beta event-related power changes and then the mean beta rebound values for each subject. We assessed significant differences for all the six possible comparisons: −_CONTRA vs. +_CONTRA; −_IPSI vs. +_IPSI; −_CONTRA vs. +_IPSI; −_IPSI vs. +_CONTRA; −_IPSI vs. −_CONTRA; +_IPSI vs. +_CONTRA. We used a permutation test for each time point of the beta event-related power changes (with Bonferroni correction for multiple comparisons) and for the beta rebound values. For each comparison (e.g., STN−_CONTRA vs. STN+_CONTRA), we computed the observed statistic (T_obs) as the difference between the event-related power changes or between the rebound, respectively. For constructing the surrogates (T_surr), we shuffled the trials of each of above listed comparisons (separately), we then recomputed the surrogated event-related power changes and the surrogated beta rebound values. We performed the shuffling 10,000 times with replacement obtaining 10,000 T_surr values for each test. The p value was computed as the Pr{T_surr > T_obs}. The significance level was set at p < 0.05.

Significant regions of Coh and iCoh were determined by statistical comparison to a population of 50 surrogate Coh maps in which any coherence was destroyed. For each pair of channels, the surrogates Coh were generated shuffling the order of trials in one of the two channels. The significance level was set at p < 0.05. Multiple comparisons were corrected with false discovery rate method (Benjamini and Yekutieli, 2001). All the analyses were performed in Matlab.

Movement of the most-affected hand showed a longer MT and lower PV and PA with respect to the less affected hand, though not statistically significant (Table 2). These differences were also present when comparing the movement of the non-dominant vs. dominant hand (Table 2).

The most striking finding of this study was the stronger β-modulation in the STN of the most dopamine-depleted hemisphere (Figure 1A). Compared to STN+, the STN− (i.e., the one in the hemisphere with less striatal dopaminergic innervation) exhibited greater beta-modulation, both stronger beta-reduction and a higher post-movement rebound, during contralateral hand movements (i.e., the clinically most affected side) and, although weaker, also during ipsilateral hand movements (Tables 3, 4). Beta-modulation in STN+ instead did not significantly differ according to the moving side (Figure 1A; Table 4, subject by subject comparisons are shown in Figure S6A).

At a cortical level, MC+ and MC− showed a similar temporal evolution of beta-power, with a steep beta reduction followed by an increase after movement end (Figure 1A). Similar to STN−, the MC− exhibited a more pronounced ERD and ERS for contralateral movements compared to ipsilateral ones. Such a difference was not found in MC+, where movements of both hands evoked similar responses (Table 4, subject by subject comparisons are shown in Figure S6B).

Of note, in one patient with relatively preserved right striatal dopaminergic innervation (<50% loss; i.e., wue2), we showed the smallest beta modulation in the corresponding cortical and subcortical areas (i.e., MC+ and STN+), whereas the patient with the greatest striatal dopaminergic innervation loss ( >70%, bilaterally; i.e., wue9) showed the strongest bilateral beta modulation (Figure 1B).

When grouping the data by handedness, we found a similar time evolution of beta-power between the two MC (i.e., MC^L and MC^R) and of the two STN (i.e., STN^L and STN^R), regardless of the moving hand (Figure 1C).

Cortical-subcortical (i.e., ^CohMC−/STN−, ^CohMC+/STN− and ^CohMC−/STN+, ^CohMC+/STN+) and subcortical (i.e., ^CohSTN−/STN+) coherences are shown in Figure 2. Each patient displayed a distinctive frequency of coherence within the beta range. Cortical-subcortical coherencies diminished during movement execution (i.e., from −2 s to 0 s). In all patients, ^CohMC−/STN− was greater than ^CohMC+/STN− and ^CohMC+/STN+ irrespective of the moving hand. Of note, the patient with the greatest striatal DAT loss (i.e., wue9) showed the most persisting and strongest cortical-subcortical coherences, also in the MC+ hemisphere (Figure 2, Supplementary Figures S7,S8).

We also found a significant subcortical, cross-hemispheric coupling (i.e., ^CohSTN−/STN+), although weaker than the ipsilateral cortical-subcortical ones. This subcortical coherence was not affected by movement (Figure 2) and mirrored the cortical-subcortical coherence, being higher in the patient with the greatest bilateral striatal DAT loss (i.e., wue9; Figure 2, Supplementary Figures S7, S8).