Do human recordings reveal drastic modulations in the discharge of striatal projection neurons in Parkinson’s disease?

1 Department of Medical Neurobiology, Institute of Medical Research Israel Canada (IMRIC), The Hebrew University Hadassah Medical School, Jerusalem 91120, Israel 2 The Edmond and Lily Safra Center for Brain Sciences, The Hebrew University, Jerusalem 91904, Israel 3 Department of Neurosurgery, Hadassah University Hospital, Jerusalem 91120, Israel 4 University of Bordeaux, UMR 5293, IMN, 33000, Bordeaux, France 5 CNRS, UMR 5293, IMN, 33000, Bordeaux, France 6 CHU de Bordeaux, IMN Clinique, 33000 Bordeaux, France


Introduction
Parkinson's disease (PD) and dystonia are two of the most common movement disorders, and have a wide spectrum of etiologies and clinical presentations. To date, the pathophysiology of PD and dystonia is still debated. Unlike in PD, there is no degeneration of the midbrain dopaminergic neurons in dystonia. Nevertheless, an imbalance between the midbrain dopaminergic and striatal cholinergic systems (1-4), as well as cerebellum dysfunction, (5)(6)(7)(8)(9) are present in both PD and dystonia. Moreover, both diseases are traditionally viewed as basal ganglia (BG) disorders (10,11) and deep brain stimulation (DBS) in BG (subthalamic nucleus, STN and internal segment of the globus pallidus, GPi) is an effective invasive treatment for both diseases (12)(13)(14)(15).
The striatum (i.e., the main input structure of the BG network) is the main recipient of midbrain dopaminergic neurons (16) and also receives major cerebellar projections (17,18).
Most (>95%) of the striatal neurons are medium spiny projection neurons (SPNs) that receive afferents from the cortex and the thalamus, and together with the STN, innervate the central (i.e., external segment of the globus pallidus, GPe) and output (i.e., GPi and substantia nigra reticulata, SNr) BG structures (19,20). Therefore, alteration of striatal signaling disrupts normal BG activity and may lead to the manifestation of the motor and non-motor symptoms of PD and dystonia (19)(20)(21).
Dysregulation of BG activity may consist of changes in discharge rate (19,20). Unlike in all other BG structures, extracellular recordings of spiking activity in non-human primates (NHPs) reveal that SPNs have a very low discharge rate (~1-2Hz at rest) and are phasically active (i.e. emit short bursts) around relevant behavioral events (22)(23)(24). However, earlier studies of the BG in the NHP model of PD have mainly focused on the STN, GPe and GPi which are structures with a high frequency tonic discharge (i.e. 25-70Hz at rest) (25)(26)(27).
These studies have reported excessive GPi/SNr inhibitory inputs in PD leading to an increase of BG inhibitory outputs to the thalamus and the frontal cortex motor areas (27).
Conversely, it is assumed that dystonia (and other hyper-kinetic states) are characterized by reduced GPi/SNr activity (28).
More recent studies of the BG in animal (rodent and NHP) models of PD and human patients (undergoing DBS procedures) have focused on changes in discharge patterns and synchronization. Parkinsonism-related β oscillations have been observed in local field potentials (LFPs) recorded in all BG structures, including the striatum (25,(29)(30)(31). Similarly, low frequency (4-12Hz) LFP oscillations have been recorded in the BG network of dystonic patients (32,33). Finally, synchronous β oscillations are commonly observed in the spiking activity of the STN, GPe and GPi of MPTP-treated monkeys (25,34,35) and PD patients (36,37). All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint Nevertheless, direct evidence for abnormal activity of the striatal SPNs is still elusive. Earlier studies in the NHP model of PD reported striking increases (~15-fold increase from the normal discharge rate of ~1-2Hz) in the firing rate of the SPNs subsequent to striatal dopamine depletion and the induction of parkinsonism (38,39). The same research group also reported a high discharge rate of SPNs recorded in PD and dystonic patients (~30Hz and 9Hz, respectively) (39). They also found a significant change in the firing pattern of striatal neurons, with many SPNs exhibiting bursting activity in PD patients and MPTP monkeys as compared to a smaller fraction in patients with dystonia (39). Finally, the SPNs of patients with an essential tremor (ET, considered as non BG disorder) have a very low discharge rate (~2Hz) and no tendency to burst, as reported in normal NHPs (38)(39)(40).
These spectacular changes in the discharge rate and pattern of SPNs in the NHP model of PD run counter results obtained in our research group. We recorded the activity of SPNs and other BG neurons in Vervet monkeys before and after systemic 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP) treatment and the induction of severe parkinsonian symptoms.
Although we found robust changes in discharge properties (rate, pattern and synchronization) of the STN, GPe and BG output structures, we did not observe any difference in the discharge rate (~2-3Hz) or pattern of SPNs in the MPTP-treated monkeys in comparison to the recordings in the same monkeys before MPTP treatment (25,34,41). Extracellular recordings of SPN spiking activity of anesthetized (42)(43)(44) and awake (42,45) rats before and after striatal dopamine depletion by 6-hydroxydopamine (6-OHDA) treatment have revealed a significant, but very slight, increase in the SPN discharge rate. Rodent studies make it possible to differentiate between SPNs expressing D1 and D2 dopamine receptors. A significant imbalance, [but see (46)] in the discharge rate and calcium dynamics of D1 and D2 SPNs was observed in the dopamine-depleted striatum (43,44,47). In particular, D2 SPNs increased their discharge (43,44) and were also prone to being entrained to parkinsonian β oscillations (44). Nevertheless, the absolute increase in the discharge rate even of the D2 SPNs was still modest (from ~0.5 to ~2.8Hz) (44). Moreover, recent studies have reported no significant increase in the low discharge rate of either SPN subpopulation in striatal dopamine-depleted mice (46,48).
To date, only one research group has been able to record striatal spiking activity in patients (39). Accordingly, we examined our human patient data to check whether we could observe similar drastic spiking activity increases in striatum of PD patients. Unlike in animal (rodent and NHP) studies, there is no control condition (healthy recordings). Furthermore, extracellular recording methods do not allow us to discriminate between the spiking activity of the striatal D1 and D2 SPNs. Nevertheless, the current study was conducted under the assumption that if only one population of striatal neurons is strongly affected by the PD state, we should observe a significantly higher discharge rate of SPNs than that reported in normal All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. NHPs and ET patients, and/or distinct clusters of SPN activity in our patients. To this end, we carefully identified single-unit activity from the striatum (putamen) of both PD and dystonic patients undergoing GPi-DBS surgeries and compared their discharge rates and patterns

Results
Database and spike sorting results. In this study, 93 microelectrode trajectories (48 in PD and 45 in dystonia) were used, yielding a total of 934 and 718 microelectrode recording (MER) segments within the striatum of patients suffering from PD and dystonia, respectively.
Applying spike sorting on these MER segments identified 5336 units (3020 in PD and 2316 in dystonia). The MER segments and the sorted units from patients suffering from non-genetic (N = 27) and genetic (N = 18) dystonia were pooled, since no difference was detected between them. All surgeries were carried out while the patients were fully awake (no sedation or anesthesia) and the PD patients were off dopaminergic medication (overnight washout > 12 hours). The DBS target was the ventro-posterior-lateral portion of the GPi for all patients, and trajectory angles were only slightly modified according to patient's anatomy.
All patients provided their written informed consent and the study was approved by the Institutional Review Board of Hadassah Hospital in accordance with the Helsinki Declaration (reference code: 0168-10-HMO).  (39). Similarly, we found that the firing rate of the dystonic patients was significantly lower than the PD firing rate (Mann-Whitney U-test, p < 0.001, Fig.S1, first column). Nevertheless, this significant difference disappeared and the striatal firing rate decreased dramatically when only comparing the firing rate of the well-isolated units (Iso.SC All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
Inspection of the spike train of the well-isolated units revealed that although the units had been graded as high-isolation score units, their spike train could be non-stationary (Fig.S2A).
To assess the stationary of the spike trains, we examined the temporal (linear) evolution of their firing rate and spike amplitude (see Materials and Methods). and Fig.S1 first column). Moreover, the striatal firing rate remained similar between both diseases after this maneuver.
Equally important, we found that mean values of the firing rate (Fig.S1, first column) were systematically higher than the median values (Fig.2, first column), suggesting that whatever the quality of the isolation and the stationarity of the striatal units, the distributions of the firing rate were not normal and were skewed to the right. We therefore used the median (Fig.2) rather than the mean (Fig.S1) to represent the central moment of the distribution of the discharge properties (including the discharge rate) of the SPNs. The median ± MAD (median absolute deviation) of the discharge rate was 3.85 ± 1.18Hz and 2.55 ± 0.81Hz for the well-isolated stationary SPNs in PD and dystonia respectively -i.e. in the same range as reported for the controls in animal studies and with no significant difference between PD and dystonia.
No evidence for bursty patterns in the striatal spiking activity of PD and dystonic patients. To characterize the pattern of the spike trains (i.e., irregular, periodic or bursty) and compare them between PD and dystonic patients, we examined the time interval histograms (TIHs) of the inter-spike intervals (ISIs). Although weak (Fig.2, second column), we found a significant difference in the coefficient of variation (CV) of the ISIs of all the sorted striatal All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint units between PD and dystonic patients (Mann-Whitney U-test, p < 0.001, Fig.S1, second column). Nevertheless, as for the striatal firing rate, this difference vanished when only considering the well-isolated or well-isolated stationary units ( Fig.2 (Fig.S4A). Accordingly, the mean autocorrelograms did not reveal any periodic or bursty firing patterns in the striatal wellisolated stationary units (Fig.S4B). Similarly, the prevalence of bursts was relatively low Therefore, we further calculated the spike-triggered averages of the LFPs (STAs LFPs). The β power in the LFPs recorded around the time of the spikes of all the striatal sortedunits was significantly higher in PD patients than in dystonic patients (Mann-Whitney U-test, p > 0.0 1, All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  Cluster analysis of striatal discharge. To further test whether two distinct SPN subpopulations could be observed in the current study, we performed 2D k-means cluster analysis with k=2, using the discharge rate and the CV of the ISIs of each unit as input parameters (Fig. 3). Using k=2 as predefined numbers of clusters systematically enforces the separation into two distinct SPN subpopulations; however calculation of the silhouette values revealed that the two clusters were not well-separated, whatever the quality of the isolation and the stationarity of the striatal units (Fig. 3 3). Similar 2D or higher-order (3 or 4D) k-means cluster analyses (using the discharge rate, CV of the ISIs, MUA β power or STA-LFP β power as input parameters) also failed to reveal two well-separated clusters.

Discussion
There is no consensus as to the impact of dopamine depletion on the discharge rate and pattern of the SPNs. (25, 38-40, 43, 44, 46). Recordings of the activity of the SPNs are inherently difficult to perform given their very low firing rate and small size, leading to frequent loss and damage of these units (53). These difficulties can distort the reported population averages of SPNs. For these reasons, we used an automated data driven approach. A total of 934 and 718 MER segments from the striatum of PD and dystonic patients were analyzed. The striato-pallidal border was automatically detected using a machine learning algorithm (54). Spike detection was performed offline with fully automatic quantification of the isolation quality and stationarity of the identified units. Unlike previous reports (39), we found no drastic modulation in the SPN discharge rate and pattern (compared to values reported for normal/control animals or ET patients) in PD and dystonic patients.
Low discharge rate and similar levels of striatal activity in PD and dystonia. We found that the firing rate of the SPNs of both the PD and dystonic patients was extremely sensitive to the quality of the isolation of the units. Specifically, we found a negative correlation between striatal firing rate and isolation score ( Fig.1A and B). The negative correlation between the discharge rate of the detected units and their isolation quality revealed that -All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint unlike in STN (55) -the spike detection and sorting algorithm used in this study and/or the physical properties of the striatal neurons tended to erroneously classify noise events as spikes (false positives) rather than missing real spikes (false negatives). Removing the most non-stationary units also reduced the striatal firing rate of SPNs in both PD and dystonic patients (Fig.2, first column), thus further indicating that the inclusion of non-stationary ("injured") units can also erroneously increase the striatal firing rate in PD and dystonic patients. This automatic approach to spike detection, sorting and quality assessment revealed that the median firing rate of the striatal well-isolated stationary units was 3.85 and 2.55Hz in PD and dystonic patients, respectively (Fig.2, first column).
Our results are inconsistent with a recent study by Singh et al. (39) where the SPN discharge rate in PD (30.2 ± 1.2Hz) and dystonic (9.3 ± 0.6Hz) patients increased by ~15-and 5-fold compared to the low striatal discharge rate (2.1 ± 0.1Hz) found in patients suffering from ET (i.e., a disorder without any known BG malfunction). In the current study, we did not have a control condition, nor did we record the striatal activity from ET patients. Therefore, we cannot rule out the possibility of an increase in the firing rate in the striatum of both PD and dystonic patients compared to human controls. Moreover, we did not differentiate between SPNs and striatal interneurons (including the TANs). About 95% of the striatal neurons are SPNs and only ~2% are cholinergic interneurons (presumed TANs) [see e.g., (56)]. We therefore considered our sorted striatal spiking activity to reflect SPN spiking activity and neglected the small fraction of TANs probably included in our sample. In any case, we did not find any significant difference between PD and dystonic patients that might reflect a distinct level of striatal hyperactivity in PD (Figs 2 and S1, first column). These different results may be due to: (i) a difference between automatic and manual striato-pallidal border demarcation, (ii) the use of algorithms for the quantification of the isolation quality and stationarity of the units, and (iii) different recording techniques (microelectrode type and step size) and/or spike detection methods.
Lack of SPN bursty and oscillatory pattern in the SPN spiking activity of PD and dystonic patients. The emergence of bursty and periodic oscillatory patterns after striatal dopamine depletion has been observed in other BG nuclei than the striatum (25,26,34,57,58). Previous studies by our research group reported the emergence of oscillatory spiking activity in the TANs (presumably the striatal cholinergic interneurons) of the MPTP-treated monkey, thus indicating that abnormal oscillatory activity following striatal dopamine depletion did not spare the striatum (25,59). However, we found no significant parkinsonismrelated oscillations in SPN spiking activity (25,34). Here, in line with this finding, we did not observe the emergence of bursting (Figs 2 and S1, second column and Fig.S4) or oscillatory All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020.  (40) showed that most bursty SPNs in the parkinsonian NHP exhibited a D1 receptor response to levodopa, thus suggesting that the emergence of an abnormal firing pattern predominantly affected the D1 SPNs (i.e., the SPNs of the direct pathway). However, when using a single-cell juxtacellular recording-labeling technique in the 6OHDA rodent, Sharott et al. showed that in addition to their excessive firing rate, D2 SPNs (that primarily composed the indirect pathway) displayed aberrant phaselocked burst firing to cortical β oscillations (44). In that study, one might assume that D2 SPNs, rather than D1 SPNs, would exhibit an abnormal firing rate and pattern in the dopamine-depleted striatum. Therefore, technical advances and further studies in PD patients and animal models should be carried out to reach a consensus.
As reported above (25, 29-32, 34, 63), abnormal oscillations of LFP were recorded in the striatum of both PD and dystonic patients (Fig.S6A). However, we found no significant phaselocking between striatal spiking activity and mono-polar LFP β oscillations (Figs 2 and S1 fourth column and Fig.S6B). These results support the view that β oscillatory spiking activity is not materialized in the SPNs and that the BG oscillatory activity observed in both PD and dystonic patients does not resonate through the striatum. BG LFP represents sub-threshold (e.g., synaptic input) activity at best (64,65). Recent studies have demonstrated that monopolar and bipolar BG LFPs may be contaminated by the volume conductance of cortical electroencephalogram (EEG) activity (66,67). Therefore, it is likely that BG LFPs, including the LFPs recorded in the striatum, do not accurately reflect local cellular activity and should be at best interpreted with caution. Exaggerated striatal LFP oscillations in PD and dystonia cannot therefore be regarded as direct evidence for the presence of SPN spiking oscillatory activity.
Is the imbalance in the activity of SPN subpopulations evident in PD and dystonia?
Obviously, striatal activity must be affected in PD and dystonic patients. One claim suggests that hypoactivity of the direct pathway (originating from D1 SPNs) and hyperactivity of the indirect pathway (originating from D2 SPNs) lead to excessive GPi/SNr inhibitory inputs to the thalamus in PD (27) and vice versa in dystonia (28). These aberrant GPi/SNr inhibitory inputs to the thalamus lead to the release of abnormal output commands and result in the emergence of the clinical symptoms. Unlike in rodent studies [see e.g., (44)], extracellular All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint recording methods cannot discriminate between the spiking activity of the striatal D1 and D2 SPNs. However, if striatal dopamine depletion drastically enhances the differences (in discharge rate and pattern) between the two distinct SPN subpopulations [but see (46)], one would expect to see a bimodal distribution of SPN discharge properties after dopamine depletion. Instead, visual inspection of the distributions of the discharge rate (Fig.S3) and the CVs of ISIs (Fig.S4A) revealed long-tailed unimodal distributions in both diseases. Moreover, the cluster analysis using SPN discharge properties failed to identify two well-separated SPN subpopulations.
In conclusion, our current results in patients extend our previous study in the MPTP NHP model of PD (25) and studies in the 6-OHDA rodent model [e.g. (46,48)] and demonstrate that abnormal activity along both the direct and indirect pathways of the BG network is not caused by dramatic changes in SPN spiking activity. Therefore, we propose that there is an abnormal recruitment (e.g., by behavioral events) of D1 and D2 SPNs (which are mostly silent at rest in the healthy condition) in pathological conditions that result in an aberrant net balance of striatal outputs (48). Moreover, it is likely that these moderate changes in SPN discharge are amplified by BG downstream structures, thus leading to the clinical symptoms of PD and possibly of dystonia.

Materials and Methods
Patients and surgery. Patients with PD and dystonia were recruited from the movement disorders clinics at the Hadassah Medical Center in Jerusalem. All patients were scheduled to undergo implantation of DBS electrodes into the GPi and underwent MR imaging, and evaluation for motor and non-motor impairments within the 30 days prior to surgery. Data were collected from 16 PD and 13 dystonic (non-genetic and genetic dystonia) patients.
Patient demographic information appears in table S1. Note that our DBS recordings in ET patients start 10mm above the thalamic target and therefore do not include striatal recordings. All patients met the criteria for DBS and signed a written informed consent for surgery that involved microelectrode recording. This study was authorized and approved by the Institutional Review Board of Hadassah Hospital in accordance with the Helsinki Declaration (reference code: 0168-10-HMO) Surgery was performed using a CRW stereotactic frame (Radionics, Burlington, MA, USA).
BG target coordinates were chosen as a composite of the indirect anterior commissureposterior commissure (AC-PC) atlas-based location and direct (1.5 or 3Tesla) T2 magnetic resonance imaging, using Framelink 4 or 5 software (Medtronic, Minneapolis, USA). All recordings used in this study were made while the patients were fully awake (no sedation or All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
Data acquisition. The data were acquired using two systems: MicroGuide [prior to 2015, previously described (55) and Neuro Omega (from 2015, previously described (54) (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
In order to assess the stationarity of the firing rate and the spike amplitude of the wellisolated units (Iso.Sc ≥ 0.6), the firing rate and the average spike amplitude of each wellisolated unit were Z-normalized over 10 bins (from the first to the last spike of the spike train). Then, the slope value of the linear regression line for the firing rate and the average spike amplitude were calculated and expressed as the z-score/bin. Units with slope values of the linear regression lin. e for firing rate or average spike amplitude greater than or equal to the 70th percentile were defined as non-stationary units.
Discharge pattern assessment of spike train. For each spike train of the well-isolated stationary unit, we calculated the inter-spike intervals (ISI) and generated the ISI histograms of the well-isolated stationary units. In parallel, we also computed the autocorrelograms of the spike train of the well-isolated stationary units, calculated for ±500 ms offset with 10msbins. For each autocorrelogram, values were normalized so that autocorrelation values ranges from 0 to 1.
For burst detection, we applied the Poisson surprise method with the surprise maximization (SM) search algorithm (69) to each spike train of the well-isolated stationary units, using the following parameters: minimal burst length = 3 spikes; threshold surprise value (S) = 10 (i.e., only bursts having S>10, corresponding to an occurrence rate of ~ 0.01 bursts/1000 spikes in a random spike train, were considered to be of interest); burst ISI limit = mean (ISI)/ 2; add limit = 150% of burst ISI limit and inclusion criteria (IC) = 5.
For each spike train of the well-isolated stationary units, the frequency (number of bursts/s) and mean duration (s) of the burst episodes were calculated over their entire recording span.
These two metrics were used to determine the burst prevalence for each unit which was defined as the: burst frequency * mean burst duration. For each unit, the burst prevalence (range: 0-1) represents the probability that the discharge pattern is bursty.
Power spectral density. For the power spectral density (PSD) calculations, the band-pass filtered spiking signal was Z-score normalized to obtain an unbiased estimate (by the electrode impedance, the A/D characteristics of the recording system, or the amplitude of the recorded neuronal activity) of the oscillatory activity (37). The Z-normalized signal was rectified by the "absolute" operator (25,34,37,55,70). The rectified signal follows the envelope of the MUA and therefore enables the detection of burst frequencies below the range of the online band-pass filter. Since the LFP frequency domain was filtered out, the resulting PSD only represented the oscillatory features of the spiking activity. All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint The PSD of each multi-unit site was calculated using Welch's method with a 1-s Hamming window (50% overlap) and a spectral resolution of 1Hz (nfft = 44000 or 48000, sampling frequency = 44 or 48kHz depending on the acquisition system). To evaluate the β power, the baseline values in the 13-30Hz range of each PSD were linearly interpolated [based on the two closest points that flanked the 13-30Hz band, namely the values at 12 and 31Hz (spectral resolution of 1Hz, see above)] and averaged. Then, the β power (i.e., the mean of the observed values between 13 and 30Hz) was compared to the interpolated β power (baseline).
Similarly, we also calculated the PSD of the striatal monopolar (0-300Hz) LFP. To do so, the PSD of each Z-normalized LFP was calculated using Welch's method (see above for the parameters; but nfft = 1375, sampling frequency = 1.375kHz) and without prior rectification by the absolute operator Spike-triggered average of the LFP. To investigate the spike-LFP relationship in the temporal domain, we also calculated the spike-triggered average (STA) of the LFP (25,71) for well-isolated stationary units. In doing so, the LFP was recorded in the vicinity of wellisolated stationary units (i.e., spiking activity and LFP were recorded on the same electrode).
Each Z-normalized LFP was offline band-pass filtered from 4 to 12Hz or from 13 to 30Hz (4pole Butterworth filter, filtfilt Matlab function). For comparison, STAs-LFP were also calculated after randomly shifting the timestamp of each spike of the spike train [i.e., random time (comprised between 0 and 1s) was added to the timestamp of each spike of the spike train] in order to abolish any relationship between LFP and spiking activity (Shift predictor).
The PSD of the STA-LFP was calculated as for the LFP.
Software and Statistics. All the data and statistical analyses were carried out using custommade MATLAB R2016a routines (Mathworks, Natick, MA, USA). Mann-Whitney U-test and Wilcoxon signed rank tests were used for statistical comparisons of paired and unpaired samples, respectively. The criterion for statistical significance was set at P < 0.05 for all statistical tests Data availability. The data are available from the corresponding author upon request. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  k-means cluster analysis with k=2, using the firing rate and the CV of the ISIs of each unit as input parameters. Analysis was performed when considering all (upper panels), only the well-isolated (middle panels) and only the well-isolated stationary (lower panels) units.
Markers (x) represent cluster centroids. Inset: silhouette values (range between -1 and 1) were calculated for each clustering to assess how well-separated the two resulting clusters were.
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. and non-significant differences, respectively (Mann-Whitney U-test).
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint spiking activities recorded in the vicinity of well-isolated stationary units. To interpolate the β power, PSDs were linearly interpolated [based on the two closest points that flanked the 13-30Hz band, namely the values at 12 and 31Hz (spectral resolution of 1Hz)]. Interpolated β power is the mean of the linearly interpolated values between 13 and 30Hz. Total β power is the mean of the observed values between 13 and 30Hz. Error bars represent SEMs. NS: Non significant (Wilcoxon signed rank test and Mann-Whitney U-test).
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  spike-triggered averages (STAs) of LFP. LFP was recorded in the vicinity of well-isolated stationary units (i.e., spiking activity and LFP were recorded on the same electrode) and offline band-pass filtered from 4 to 12Hz (upper panels) or from 13 to 30Hz (lower panels).
For comparison, STAs-LFP were also calculated after randomly shifting the timestamp of All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted March 6, 2020. ; https://doi.org/10.1101/2020.03.04.20030999 doi: medRxiv preprint each spike of the spike train [i.e., random time (comprised between 0 and 1s) was added to the timestamp of each spike of the spike train] in order to abolish any relationship between LFP and spiking activity (Shift predictor). Dashed grey vertical lines indicate the time of the spikes (time = 0). N is the number of STAs-LFP averaged.
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.