Fig 1.
Structural imaging used to target multi-microelectrode DBS to the central lateral thalamus.
A: Schematic of coronal thalamic section (right hemisphere) with tailored DBS electrode placed such that contacts span the dorsal-ventral extent of CL (yellow). We simultaneously stimulated via 16 electrode contacts (200μm spacing between contacts), with the centermost contact (contact 8) shown in red. B: Track reconstruction of DBS sites overlain on the high-resolution structural image of the thalamus (monkey R, right hemisphere, A8). Only the centermost contact location shown (colored circles) for clarity. For each stimulation, 8 contacts above and 7 contacts below the colored circle would also be stimulated. To improve visualization by reducing overlap, positions plotted with up to 1 voxel (.5mm) jitter separately for 10, 50, and 200Hz experiments. Black lines demarcate different thalamic nuclei: central lateral, CL; centromedian, Cm; lateral dorsal, LD; lateral posterior, LP; mediodorsal, MD; parafascicular, Pf; ventral posterior lateral, VPL. Stimulation performed at all frequencies for most sites.
Fig 2.
Experimental paradigm for manipulating consciousness in awake macaques.
Schematized paradigm to reveal stimulation frequency-specific effects of thalamic DBS across multiple time scales. On the far left, a schematized timeline of the paradigm shown shifting between a series of experiments with (Stim) and without (No-Stim) DBS. The first two recording days of a series were considered transitional, and the following days in the same series were considered established. The top half of the figure presents an example day (D5) in the DBS paradigm, comprised of a pair of experimental runs with pseudorandom stimulation frequency assignment. Here, a sample resting-wake block (420s) using 200Hz DBS is featured, with typical on/off periods (S+/S-) of DBS within the block (60s on, 60s off). Above the block, one stimulation with duration of 60s is subdivided into shorter time periods: onset (O, 0-2s), early (E, 2-10s), mid (M, 10-40s), and late (L, 40-60s) periods with respect to the start of DBS. The bottom half of the figure presents an example day (D13) in the established No-Stim portion of the paradigm, with a sample fixation task block featured. Here, the eye-tracker trace indicates the animal spends most of the experiment fixating on the central target (centered, steady) as compared to the upper trace where the animal shows typical eye behavior during resting wakefulness (eyes moving around).
Fig 3.
Example of simulated null distribution permutation test for probability analysis.
A, Example of hypothetical population, size N, consisting of 4 members and yielding observed sample, size n, consisting of the 4 members. The question to address: Is there category-specific variation in the probability of occurrence for different members of the observed samples? B, Equations demonstrating how to represent the question in A as a null hypothesis, and how to simulate samples from the population. C, Distributions (histograms) and fitted normal distributions (curves) showing the proportion of simulations yielding a given probability of member occurrence under the null hypothesis. Red lines indicate where the observed sample falls relative to the null distribution, with matching Z statistic and p-value before correction. D, Category-specific probability of occurrence where error bars indicate ± 1SD of the null distribution. * indicates statistical significance. E, Similar figure to (D), but representing the relative proportion of sample occurrences and the null hypothesis that the member proportions observed are the same as those observed in the population.
Fig 4.
Thalamic DBS triggers periods of behavioral stillness coinciding with low-frequency activity in the EEG, similar to absence epilepsy.
A, Example time course of behavioral and neural signatures of a VPC event initiating after 10Hz stimulation. B,C,D,E, Average behavioral signatures during VPC relative to pre and post conditions (±SE) for the (B) variance in polar eye position from eye tracker, and change in luminance for video data centered on the (C) eyes, (D) mouth, and (E) nose. Luminance changes indicate movement. F, Average EEG power (±SE) for VPC and control stares compared to pre and post conditions. Colored straight lines at top indicate regions of significance for t-tests across frequencies comparing spectra from the different conditions. Vertical black lines indicate the cutoff mark of different frequency bands: δ = (0, 4), θ = (4, 8), α = (8, 15), β = (15, 30), γL = (30, 60), γH = (60, 100). G, Spectrogram of average differences in normalized EEG power (VPC–control stare) aligned to event onset (white line). Frequency is represented across described power bands. Significant cluster outlined in black.
Fig 5.
Measure of neural complexity and integration (Φ*) associated with consciousness selectively decrease during VPC, and integration patterns shift to reflect anesthetized rather than conscious states.
A, Average entropy (H) and B, Φ* (±SE) centered relative to the fixation task for VPC and control stares relative to pre and post event conditions. C, D, Average Φ* (±SE) for stim-off (gray bars) and stim-on (colored bars) epochs at different frequencies excluding VPC for resting wake (C) and fixation (D) controls. E, F, Gaussian kernel-density estimates fitted to the probability distribution for the occurrence of different MIP types of the full system for (E) pre, VPC, control stare (STR), post, (F) resting wake (RW), fixation (FX), isoflurane (Iso), and propofol (Prop) conditions. G, Results for the KS tests comparing the MIP kernel-density estimates across all conditions. Color scales with the strength of the KS stat. * indicates corrected p-value (pc) < .05.
Fig 6.
MIP changes reflect switches in parieto-striatal association indicative of consciousness.
A, B, Probability (connection weight) of each brain area associating with any other on the same side of the MIP for states assumed to be (A) conscious (pre-VPC, post-VPC, resting wake (RW), fixation task (FX) and control stares) or (B) less conscious (VPC, propofol (Prop), or isoflurane (Iso)). C, D, Results (Z statistics ± SD of the null distribution) of the permutation test comparing cortico-striatal MIP associations under the null hypotheses that (C) MIP types occur randomly, or (D) MIP types reflect the same patterns as conscious states (defined by the resting wake and fixation task samples). Both approaches separate more conscious from less conscious states. E, Normalized Φ* (±SE) for VPC and control stares (STR) calculated in 1s sliding windows (.1s steps) and aligned to event onset (black vertical line). Horizontal pink line shows regions of significance for the pairwise t-test across time comparing VPC Φ* at each sample to the maximum. No significant differences were found in the control stare condition. F, Cortico-striatal significance (Z) based on the sliding window approach in E for VPC and stare conditions. Computations used the same approach as in C, but now for each sample in the sliding window.
Fig 7.
Thalamic DBS modulates VPC occurrence over acute and longer time frames.
A, Relative proportion (± SD of the null distribution) of VPC events occurring around the onset (O, within 2 secs of stimulation start), early (E, 2-10s), middle (10-40s), or late (40-60s) phase of the 60s stimulations. B, Probability of VPC (± SD of the null distribution) across different experimental conditions: no stimulation (NS) and stimulations at 10, 50 and 200Hz. Dashed line indicates the average probability of VPC (the null hypothesis). C, D, Causal power by stimulation condition across experimental history (x axis starting when there are at least 10 blocks per condition to ensure robust causal power estimates) for (C) Monkey W and (D) Monkey R. E, Population CP (± SD of the null distribution) under the null hypothesis that CP is equal across conditions and no different from 0. F, Conditional probability (± SD of the null distribution) of VPC occurring with a particular stimulation frequency (10, 50 or 200Hz), given it was preceded by a different stimulation condition, under the null hypothesis that probability of VPC is consistent irrespective of preceding stimulations. G, Correlation (best fit line ± SE of the point estimate) of local VPC probability and stimulation predominance (% maximum, sliding window across 32 recording blocks). Individual data points shown for both animals (R and W) across stimulation and no-stimulation runs during transitional (first two days) and established (third day onwards) phases of an experimental series. H, Local VPC probability (± SE) for all data (All) in stimulation or no-stimulation runs, and further separated into established (Est) and transitional (Trs) phases.
Fig 8.
Neural firing rates decrease during VPC relative to control conditions.
Spike rate differences normalized to the pre-condition (± SE of the point estimate) for different brain areas under VPC, control stare (STR), post, resting wake (RW), and fixation (FX) conditions. Neurons were analyzed in resting wake, fixation, and pre/post-event epochs only for neurons recorded during VPC or STR events (NCL = 60, NCN = 19, NFd = 15, NFs = 38, NLd = 46, NLs = 21).
Fig 9.
VPC is associated with increases in low-frequency power relative to control stares.
A-F Spectrograms of power differences (VPC–control stare) for each frequency band across time aligned to event onset (white vertical line). Results include all recorded VPC and control stares for (A) superficial FEF, (B) superficial LIP, (C) deep FEF, (D) deep LIP, (E) CN, and (F) only the subset of events without thalamic DBS for CL. Significant clusters are outlined in black with numerical labels referenced in Table 1 for statistical details.
Table 1.
Statistical results of cluster analysis for LFP power spectral density time-frequency spectrograms.
Area identity (Tp), N (VPC, control stare), cluster identity, Tsum statistic, and p-value for each significant cluster as identified in Fig 9. Positive Tsum indicates VPC power spectral density > control stare.
Fig 10.
VPC and stare controls are readily decodable on a similar time-frame as changes in consciousness, and represent a clear shift in the relative importance of different frequencies.
A, Decoding accuracy (±SE) across time relative to event onset of VPC and stare controls using 25 frequency (δ, θ, α, β, γ) by brain area (CN, Ld, Ls, Fd, Fs) features. Thin saturated pink line shows period of significance when decoding accuracy is greater than chance (50%). B Mean decrease in accuracy (MDA) attributed to each model feature across time. Higher values indicate higher feature importance. D-F, MDAs further analyzed for the period of time just prior to event onset (thick blue horizontal bar in (A)) for features by (C) area and (D) frequency, and for the period of time just after VPC onset (thick red horizontal bar in (A)) for features by (E) area and (F) frequency. Bar graphs show average MDA (± SD).
Fig 11.
Functional connectivity increases, especially at low frequencies, just before and throughout VPC.
A, Diagram showing anatomical pathways isolated in our study. Superficial and deep cortical layers are represented by the respective light and dark regions of the labeled area. Arrow colors relate to the panel subtitle underlines featured in B-G. B-G, Time-frequency plots showing coherence differences (VPC–control stare) aligned to event onset (white vertical line) for pairs of LFPs within/between brain area(s). Color scales with the strength of the resulting t-stat. Significant clusters are outlined in black with numerical labels referenced in Table 2 for statistical details. Results are shown for the key (B) Thalamo-frontal, (C) Thalamo-parietal, (D) Cortico-striatal, (E) Subcortical, (F) Intracolumnar, and (G) Cortico-cortical feedforward (FF, Ls-Fs) and feedback (FB, Fd-Ld and Fd-Ls) pathways.
Table 2.
Statistical results of cluster analysis for coherence time-frequency spectrograms.
Area identity (Tp), N (VPC, control stare), cluster identity (Clust), Tsum statistic, and p-value for each significant cluster as identified in Fig 11. Positive Tsum indicates VPC coherence > control stare.
Fig 12.
Pathways involving thalamus and striatum dominate the changes in coherence attributed to VPC.
A-E, For each pairwise comparison across all areas, proportion of significant pixels indicating coherence increases/decreases at higher-frequency (β, γL, γH) or increases/decreases at lower-frequency (δ, θ, α) in time windows that (A) predate significant decoding and changes in consciousness (Pre 1, P1), (B) coincide with increased decoding and the start of changes in consciousness (Pre 2, P2), (C) coincide with behavioral changes (event onset, EO), (D) coincide with sustained changes in behavior and consciousness (After 1, A1), and (E) coincide with changed behavior but gradual restoration of consciousness as gauged with Φ* (After 2, A2).
Fig 13.
Neural correlates of VPC involve substantial impairment of communication in thalamo-cortical and cortico-striatal circuits.
A-C, Connection diagrams summarizing key coherence and power patterns within and between different areas (A) just before, (B) at onset, and (C) ongoing during VPC. Arrows show increased (thicker) or decreased (thinner) coherence between specified areas at lower (light blue) or broadband (pink) frequencies. Light gray arrows indicate pathways where coherence changes are relatively small or unchanged. Blue shading in areas shows power increases at low frequencies. Ls and Ld, superficial and deep layers of LIP; Fs and Fd, superficial and deep layers of FEF; CN, caudate nucleus; CL, central lateral thalamus. Diagrams summarize results from Figs 9–11, and values thresholded at .4 in Fig 12.