Optogenetic fMRI reveals therapeutic circuits of subthalamic nucleus deep brain stimulation

Abstract

While deep brain stimulation (DBS) is widely employed for managing motor symptoms in Parkinson’s disease (PD), its exact circuit mechanisms remain controversial. To identify the neural targets affected by therapeutic DBS in PD, we analyzed DBS-evoked whole brain activity in female hemi-parkinsonian rats using functional magnetic resonance imaging (fMRI). We delivered subthalamic nucleus (STN) DBS at various stimulation pulse repetition rates using optogenetics, allowing unbiased examination of cell-type specific STN feedforward neural activity. Unilateral optogenetic STN DBS elicited pulse repetition rate-dependent alterations of blood-oxygenation-level-dependent (BOLD) signals in SNr (substantia nigra pars reticulata), GP (globus pallidus), and CPu (caudate putamen). Notably, this modulation effectively ameliorated pathological circling behavior in animals expressing the kinetically faster Chronos opsin, but not in animals expressing ChR2. Furthermore, mediation analysis revealed that the pulse repetition rate-dependent behavioral rescue was significantly mediated by optogenetic DBS induced activity changes in GP and CPu, but not in SNr. This suggests that the activation of GP and CPu are critically involved in the therapeutic mechanisms of STN DBS.

1. Introduction

Deep brain stimulation (DBS) is a well-established neurosurgical therapy for controlling motor symptoms in humans with Parkinson’s disease (PD) who do not respond to dopaminergic medications or have developed levodopa-induced motor complications. DBS for PD involves implanting stimulating electrodes in subcortical brain areas, including subthalamic nucleus (STN). STN DBS decreases motor disability and enhances quality of life [1] and may delay symptom progression in PD patients [2,3].

Despite the clear clinical benefits of STN DBS in PD, its underlying neural mechanisms remain elusive. One obstacle is that electrical stimulation non-specifically activates both local neurons and axons around the stimulation site, which in turn alters neural activity across widespread areas of the brain [4-8]. Accordingly, several hypotheses have been proposed for the neural pathways involved in the therapeutic effects of STN DBS. The use of optogenetics offers a promising approach to activate specific cell types [9], which facilitates circuit dissection of the therapeutic STN DBS mechanisms in PD. Notably, several pioneering studies using Channelrhodopsin-2 (ChR2) reported that STN DBS can achieve its therapeutic effects by antidromic activation of motor cortical neurons via the hyperdirect pathway [10,11], shifting the focus of DBS research toward this pathway, rather than other STN feedforward circuits. However, the widely used ChR2 suffers from slow kinetics and cannot follow photostimulation above 100 Hz [12,13]. Considering that high pulse repetition rate stimulation (>100 Hz) is a hallmark of therapeutic STN DBS in PD, it is likely that the contributions of circuits downstream of STN have been overlooked. Indeed, a recent study reported robust behavioral therapeutic effects by stimulating STN cell bodies without direct engagement of its motor afferents [14]. Despite this critical finding, it remains unclear how activity changes in different brain regions downstream of STN are related to the behavioral rescue in PD. To decipher the circuit mechanisms governing therapeutic STN DBS, we used functional magnetic resonance imaging (fMRI), which enables mapping whole-brain evoked responses in an unbiased manner. Such brain-wide patterns of changes in activity during stimulation with different parameters represent critical information to reveal the precise neural circuit changes that are necessary for therapeutic STN DBS [15-21].

Here, we combined optogenetic activation of STN neurons and fMRI to map the brain areas that are modulated by stimulating STN cell bodies in hemi-parkinsonian rats. We varied the pulse repetition rate, which is known to modulate the therapeutic effects of DBS [22,23], and we hypothesized that brain areas mediating the effects of STN DBS would exhibit responses differentially tuned to the pulse repetition rate in a manner correlated with the behavioral effects. We used an ultrafast opsin (Chronos) [24] as required to follow faithfully the high pulse repetition rates necessary for effective STN DBS [14]. We found that a range of subcortical brain regions, including substantia nigra pars reticulata (SNr), globus pallidus (GP), and caudate putamen (CPu) were activated by optogenetic STN DBS, and that the activation of GP and CPu mediate the pulse repetition rate-dependent therapeutic effect of STN DBS in PD.

2. Material and methods

2.1. Animals and surgery

All animal care and experimental procedures were approved by the University of North Carolina Institutional Animal Care and Use Committee. Fourteen female Sprague Dawley rats (weight 250–275g, ~90 days old) were used in this study. The animals were housed in a temperature-controlled environment on a 12-h light-dark cycle and provided with ad-libitum food and water. Rats received either rAAV5-CAMKII-Chronos-GFP (n = 8) or AAV-CAMKII-ChR2-GFP (n = 6) injection. There was an unexpected failure of the MRI system during the research, and data collection in five rats was abandoned due to the concerns of headcap integrity and health after the long delay. The remaining nine rats (Chronos group n = 5, ChR2 group n = 4) were used both in methamphetamine-induced circling measurements and functional imaging experiments.

Surgery was conducted under anesthesia with sevoflurane (induced at 7 % and maintained at 3.25 %). The animals were fixed in a stereotaxic frame. The body temperature was maintained at 37 °C by a heated water blanket. Meloxicam (2.0 mg/kg, s.q.) and Bupivacaine (0.2 ml of 0.25 %, s.q.) were administered for analgesia. Dexamethasone (0.5 mg/kg, s.q.) was administered to reduce cerebral edema. Craniotomies were performed over the STN (3.6 mm posterior and 2.6 mm lateral from bregma) and medial forebrain bundle (MFB, 2.0 mm posterior and 2.0 mm lateral from bregma) in the left hemisphere. rAAV5-CAMKII-Chronos-GFP (1.8 × 10¹³ vg/ml) or AAV-CAMKII-ChR2-GFP (4.6 × 10¹² vg/ml, University of North Carolina vector core) was injected into the STN (depth ~7.0 mm from cortical surface; 0.6 μl at a rate of 0.15 μl/min). Six μl of 6-hydroxydopamine (6-OHDA) (2.5 mg/ml dissolved in 0.2 % ascorbic acid in saline) was injected into the MFB (depth 7.5 mm from cortical surface) at a rate of 1.5 μl/min to induce unilateral dopaminergic lesion of the substantia nigra pars compacta (SNc). Desipramine (5 mg/kg) and pargyline (50 mg/kg) were administered intraperitoncally 30 min before the 6-OHDA injection to protect noradrenergic neurons and inhibit monoamine oxidase, respectively. An optical fiber (200 μm core diameter, 0.37 NA, Precision Fiber Products) was implanted into the STN 300 μm dorsal to the virus injection site. The optical fiber was then fixed to the skull by MRI compatible bone screws and dental cement.

2.2. Behavioral testing and optogenetic stimulation

The success of the dopaminergic lesion and effects of DBS were assessed using methamphetamine-induced circling [25] starting four weeks after the surgery. The animals were administered 1.875 mg/kg of methamphetamine (intraperitoneal, 1.25 mg/ml dissolved in saline) and then placed in a dark cylindrical chamber 30 cm in diameter. The locomotor activity of the rat was captured by an infrared camera. Six different conditions were tested in random order: no stimulation, 5Hz, 20Hz, 75Hz, 100Hz or 130Hz stimulations. In each test condition, laser light (473 nm, Shanghai Laser) was applied via an optical fiber. The laser power was 10 mW (calibrated as output power at the end of the fiber) and the pulse width was 1 ms. The duration of stimulation was 10 s with 20 s intervals between epochs of stimulation. Each stimulation condition was repeated six times.

2.3. fMRI data collection

The fMRI scans were conducted one week after the circling tests following the procedures established in several recent studies [15, 26-28]. Briefly, rats were intubated and ventilated under dexmedetomidine (0.025 mg/kg/h, i.p infusion) and 0.5 % isoflurane. Pancuronium bromide (0.5 mg/kg/h, i.p infusion) was administered for paralysis. Heart rate and oxygen saturation level were continuously monitored by a MouseOX Plus system (STARR Life Science Corp.) and maintained at 350–400 bpm and 95 %, respectively. Body temperature was maintained at 37 °C by a circulating water pad. End-tidal CO₂ was monitored by a capnometer (Surgivet, Smith Medical) and stabilized within 2.6–3.2 %. MR images were collected through the UNC Center for Animal MRI (CAMRI) service on a Bruker BioSpec 9.4-T, 30 cm bore system (Bruker BioSpin Corp., Billerica, MA) with Paravision 6.0.1 on an AVANCE II console. A custom-made surface coil with an internal diameter of 1.6 cm placed directly over the head was used as a transceiver [16]. Magnetic field homogeneity was first optimized by global shimming, followed by local second-order shims using a MAPSHIM protocol. For each subject, the BOLD fMRI data were acquired using a 2D multi-slice, single-shot, gradient echo-planar imaging (EPI) sequence: repetition time (TR) = 2000 ms, echo time (TE) = 14 ms, bandwidth = 250 kHz, flip angle = 70°, field of view (FOV) = 28.8 × 28.8 mm, matrix size = 72 × 72, slice number = 32, and slice thickness = 0.4 mm, resulting in an isotropic voxel resolution of 0.4 mm [27,28]. The optical stimulation parameters were the same as in the behavioral test. Five different stimulation frequencies (5Hz, 20Hz, 75Hz, 100Hz, or 130Hz) were delivered in random order. Three repeated scan sessions of 20 s rest period followed by six repetitions of the optical stimulation train, including 10 s of stimulation and 20 s rest period block, were applied for each pulse repetition rate.

2.4. Data analysis

The position of the nose and the base of tail over time were extracted from the video recording of the methamphetamine-induced circling using TopScan analysis software (Clever Systems). The normalized angular velocity and linear speed of the body for each trial were calculated by dividing the angular velocity or linear speed during the 10 s stimulation-on period by the averaged angular velocities or linear speeds during the 20 s pre-stimulation and post-stimulation periods immediately before and after the stimulation-on period. Normalized angular velocity or linear speed for each pulse repetition rate were then computed by averaging across all repetitions.

For MRI data analysis, all EPI data were first corrected for slice timing, followed by motion correction using AFNI [29]. Each 4D dataset was then averaged across time to improve signal-to-noise ratios (SNR) for subsequent skull stripping. The brain mask for each time-averaged EPI image was manually drawn using ITK-snap [30]. The skull-stripped data were then spatially normalized to the BOLD-EPI rat brain template [27], constructed from a BOLD-EPI image processed under the same protocol as the current study, but with Fomblin in the ear canal to restore signal in regions affected by susceptibility artifacts at air-tissue interfaces near the ear canal. Although this does not completely restore the signal intensity in the affected areas, the adjustment reasonably corrects distortions during spatial normalization [27], using ANTs SyN diffeomorphic registration [31] prior to further group-level analysis. Before proceeding to GLM and group-level statistics, the dataset was smoothed using a Gaussian kernel with a full width at half maximum (FWHM) of 0.5 mm to enhance the SNR. The design matrix for the first-level General Linear Model (GLM) analysis was generated using the canonical HRF with the BLOCK4 function via the 3dDeconvolve tool in AFNI. This approach was employed to model the evoked responses from the optogenetic stimulation paradigm, applied in each scan session as outlined in section 2.3. The matrix was subsequently fitted using 3dREMLfit for a Generalized Least Squares time series fit. This fitting process employed Restricted Maximum Likelihood (REML) to correct for autocorrelation with an ARMA(1,1) model. AFNI’s statistical package was then employed to perform group-level statistics using 3dLME. We assessed the effects of pulse repetition rate, the type of opsins used, and the interaction between pulse repetition rate and opsin. In this model, the intercepts and repeated scan trials were treated as random effects. We also designed a contrast matrix and applied it to the estimated linear mixed model for evaluating the effect size of each factor and testing the significant differences between their averages. Specifically, we applied an ANOVA to test for significant differences between the averages of each factor, including the main effects of pulse repetition rate, opsin, and their interaction. For the multiple comparison correction, the spatial autocorrelation function (sACF) was estimated by fitting the mixed model of gaussian and mono-exponential function for the residual of the linear mixed model analysis (using 3dFWHMx). Subsequently, the probability of false positive clusters was estimated from the sACF to identify the minimal size of significant clusters at 5 % of the false positive rate (using 3dClustSim). The clusters larger than the estimated cluster size were determined as significant findings. Additionally, to illustrate succinctly the characteristic hemodynamic responses for each region, we averaged the time-course data of evoked responses into single response curves. This approach enhances the clarity and interpretability of the response patterns among regions and pulse repetition rates. To investigate how optogenetic STN DBS influences pathological circling behavior via brain activations, we constructed a mediation model using linear mixed-effect regression (LMER) [32]. This model included the applied pulse repetition rates, the averaged beta coefficients from first-level GLM results for each ROI, and normalized angular velocity values from the behavioral measurements. We calculated the direct relationship between pulse repetition rate and normalized angular velocity using the LMER model incorporating random intercepts for subjects and trials to accommodate repeated measures. Subsequently, the mediation model estimated the indirect relationship between pulse repetition rate and normalized angular velocity mediated by the beta coefficients of selected ROIs. In this mediation model, pulse repetition rate was the independent variable, normalized angular velocity was the dependent variable, and the BOLD response coefficients from SNr, GP and CPu served as mediators. To assess the robustness of our findings, we conducted 1000 bootstrap tests and employed the Sobel test to evaluate the mediation effect of each significant pathway [33].

2.5. Histology

At the conclusion of the MRI scan, rats were deeply anesthetized with phenytoin sodium (Euthasol) and transcardially perfused with saline followed by 4 % paraformaldehyde. The brains were removed and stored overnight in 4 % paraformaldehyde and transferred to 10 % sucrose (in DI water) for 4 h and then 30 % sucrose solution for 2–3 days until they sank. Brains were cut horizontally at 40 μm thick sections on a freezing microtome and mounted on glass slides for fluorescent imaging. Tyrosine hydroxylase (TH) immunohistochemistry was used to evaluate the degree of degeneration of dopaminergic neurons in the SNc. Briefly, brain slices were rinsed and then incubated in PBST (0.3 % Triton-X in PBS) with 3 % goat serum for 30 min at room temperature. The sections were then incubated in anti-TH antibody (1:1000, AB152, Abcam, CA, USA) at 4 °C for 16 h. After rinsing in PBST 3 times for 10 min each, the sections were incubated in secondary antibody (1:1000, goat anti-rabbit Alexa Fluor 546) for 2 h at room temperature before being rinsed and mounted to slides. Vectashield mounting medium with DAPI stain (Vector laboratories, Item # H-1200) was used to provide a cell body counterstain. To confirm the expression of Chronos-GFP, brain sections were directly mounted. To validate the cell-type specific expression of Chronos-GFP, brain sections were stained for CaMKII. Slices were incubated with primary anti-CaMKII antibody (1:300, Millipore #05-532) with 10 % goat serum and 0.25 % Triton X-100 overnight at 4 °C. Second antibodies (Alexa Fluor 594 goat anti-mouse IgGI) were used to visualize CaMKII. All sections were mounted with DAPI-FluoroMount-G (Southern Biotech) and were imaged with a 20x oil lens with Olympus FV3000RS confocal microscope or Nikon Eclipse Ti2 microscope. The locations of the optical fibers were determined by registering the MR anatomical images to a rat brain atlas [34].

2.6. Statistics

A two-way ANOVA analysis was used to test the significance of pulse repetition rate effect and injected virus type on angular velocity or linear speed in the methamphetamine-induced circling test. Post hoc paired comparisons with Tukey-Kramer method were used to compare the difference between two different pulse repetition rate conditions. All statistic results are represented as mean ± standard deviation unless otherwise indicated.

3. Results

3.1. Histological validation of dopaminergic degeneration, virus expression, and optical fiber implants

To evaluate DBS pulse repetition rate-dependent effects on STN-related circuitry, we conducted a series of optogenetic-fMRI experiments in hemi-parkinsonian rats. For each optogenetic-fMRI experiment, blue light (473 nm) pulse-trains were delivered by fiber optic to the STN to activate either Chronos or ChR2 fused with green fluorescent protein (GFP) expressed by an adeno-associated-virus (AAV) vector. Expression was directed to excitatory neurons in STN under control of the calmodulin kinase IIα (CaMKIIα) promoter. Animals used in behavioral and imaging data collection (Chronos group, n = 5; ChR2 group, n = 4, Fig. 1A) exhibited degeneration of dopaminergic cells in substantia nigra pars compacta (SNc) in the hemisphere receiving 6-hydroxydopamine (6-OHDA) injection (Fig. 1B), as indicated by the clear reductions in tyrosine hydroxylase (TH) immunoreactivity (Fig. 1C). The expression of Chronos or ChR2 in STN was confirmed by GFP (Fig. 1D), and the positions of optical fibers in STN were confirmed by structural MRI (Fig. 1E-F).

Fig. 1. Chronos or ChR2 expression, TH immunofluorescence, and optical fiber implant location in hemi-parkinsonian rats.

(A–B) schematic illustration of virus (A), and 6-OHDA (B) injections. (C) TH immunofluorescence showing 6-OHDA lesion of dopaminergic neurons in the left substantia nigra (SN). (D) Fluorescent micrograph showing expression of Chronos-GFP in the left STN. (E) Schematic illustration of MRI experimental setup. (F) Anatomical MR image showing the optical fiber track in the brain and the location of the tips of the optical fibers in STN across all animals used in this study, marked with distinct symbols: ‘*’ for Chronos and ‘o’ for ChR2 (inset). The optical fiber implantation was targeted 300 μm above the STN viral expression site, allowing illumination of the STN.

3.2. Optogenetic STN DBS reduced pathological motor behavior

The motor dysfunction induced by 6-OHDA lesion was characterized behaviorally by ipsilateral circling (Fig. 2A) following administration of methamphetamine (mean ± SD = 8.85 ± 5.87 turns/min, n = 9; p = 0.0039, Mann-Whitney test compared to zero). Pathological circling behavior was ameliorated by 130 Hz optogenetic STN DBS in the Chronos group (F_{(2, 75)} = 73.4, p < 0.001, one-way repeated measure ANOVA), but not in the ChR2 group (F_{(2, 60)} = 1.52, p = 0.23). The effects of optogenetic DBS on angular velocity were dependent on pulse repetition rate (Fig. 2B-D; two-way ANOVA with significant main effect of the pulse repetition rate (F_{(4, 260)} = 6.77, p < 0.001) and opsin type (F_{(1, 260)} = 78.6, p < 0.001), and a significant interaction between factors (F_{(4, 260)} = 7.96, p < 0.001). Post hoc paired comparisons showed that high pulse repetition rate optogenetic DBS (75, 100 and 130 Hz) applied to animals in the Chronos group reduced ipsilateral circling (all p < 0.01), whereas no changes in circling rate were detected during low-rate optogenetic DBS (5 and 20 Hz) in the Chronos group or during optogenetic DBS at any pulse repetition rate in the ChR2 group (all p > 0.1). Importantly, neither optogenetic pulse repetition rate nor opsin type affected the linear speed of animals’ motion (Fig. 2E; two-way ANOVA with non-significant main effects of pulse repetition rate (F_{(4, 260)} = 1.03, p = 0.39) or opsin type (F_{(1, 260)} = 2.36, p = 0.13), and a non-significant interaction between factors (F_{(4, 260)} = 1.42, p = 0.23). These results are consistent with previous studies of the effects of electrical [35] and optogenetic DBS [14], and established the comparative effectiveness of different rates and different optogenetic constructs prior to fMRI studies of the effects of optogenetic STN stimulation.

Fig. 2. Effects of STN optogenetic DBS at different pulse repetition rates on the angular and linear velocities in methamphetamine-induced circling test.

(A) Schematic illustrations of the pathological (counter-clockwise) circling behavior and non-pathological behaviors (no circling). (B–C) Measured number of turns per minute in Chronos (n = 5) and ChR2 (n = 4) rats. (D) Normalized angular velocity and (E) normalized linear velocity during stimulation in two groups of rats. STN optogenetic DBS pulse repetition rate had a significant effect on angular velocity in Chronos rats but not in ChR2 rats. DBS had no effect on linear velocity in either group of rats. Data are expressed as mean ± sem. **p < 0.01, and ***p < 0.001; n.s., not significant.

3.3. Optogenetics STN DBS activated three nuclei, SNr, GP and CPu, within the basal ganglia

Next, the same cohorts of Chronos and ChR2 rats underwent fMRI. Six randomly ordered paradigms were used (5 different pulse repetition rate conditions plus no stimulation condition, Fig. 3A), resulting in a group of n = 90 epochs for Chronos rats and n = 72 epochs for ChR2 rats per paradigm. Together, these experiments provided a grand total of n = 540 and 432 fMRI epochs for Chronos and ChR2, respectively, for the entire study. A general linear model (GLM) analysis was conducted to evaluate voxel-wise response patterns of unilateral optogenetic STN DBS evoked blood-oxygenation-level-dependent (BOLD) fMRI signal changes in hemi-parkinsonian rats. The group-averaged response maps for each stimulation pulse repetition rate are presented by type of opsin in Fig. 3B-C, and reveal that certain brain regions responded differently during stimulation with low pulse repetition rates (5 and 20 Hz), which corresponded to behaviorally ineffective optogenetic STN DBS, versus during stimulation with high pulse repetition rates (above 75 Hz), which were associated with behaviorally effective optogenetic DBS. In the Chronos group, within-group comparisons of high (75, 100, and 130Hz) and low (5 and 20Hz) pulse repetition rate ranges revealed significantly enhanced positive BOLD responses in ipsilateral SNr, SC (superior colliculus), GP, LHb (lateral habenula nucleus) and PTg (pedunculotegmental nucleus) during high-pulse repetition rate DBS (upper row in Supplementary Fig. S1C, also see Supplementary Figs. S1A-B for the contrast maps for low and high stimulation frequencies in each virus group). In the ChR2 group, the same analyses revealed significantly enhanced positive BOLD responses in ipsilateral medial forebrain bundle (MFB) and enhanced negative BOLD responses in ipsilateral CPu and contralateral SC (lower row in Supplementary Fig. S1C). Collectively, there were clear differences in the modulation of STN feedforward activity mediated by the behaviorally-effective Chronos construct and the behaviorally ineffective and kinetically limited ChR2 opsin.

Fig. 3. Brain-wide fMRI responses to optogenetic STN DBS.

(A) Schematic illustration of the study design, including stimulation paradigms. (B–C) Group-averaged response maps for each stimulation pulse repetition rate and opsin type. The red box highlights visually identifiable differences between opsin types observed at relatively high pulse repetition rates (above 75 Hz), corresponding to behaviorally effective optogenetic DBS. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

We then used a linear mixed-effects regression model coupled with ANOVA to assess the voxel-wise effects of stimulation pulse repetition rate, opsin type, and their interaction on the BOLD signal using the data including both Chronos and ChR2 groups. This analysis revealed a main effect of pulse repetition rate across five spatially distinct clusters, depicting the common effects seen in both subject groups (Supplementary Fig. S2). We also observed a significant interaction between the pulse repetition rate and opsin type in two additional clusters (Supplementary Fig. S3), suggesting that the choice of opsin was critical for fMRI outcomes, as it was for behavioral effects. It is important to highlight that the main effect of pulse repetition rate refers to a common effect shared by both groups. Considering the predominant opsin effects, we conducted further voxel-wise analyses to explore the pulse repetition rate effects within each individual subject group. In the Chronos group, we identified three brain regions, namely, ipsilateral CPu, GP and SNr (Fig. 4A), that were significantly modulated by the stimulation pulse repetition rate. Notably, the SNr cluster is located in the caudal region, which is less sensitive to susceptibility artifacts compared to the anterior region. Given that SNr is typically known to be challenging to detect with BOLD fMRI in rodent studies, the cluster’s notation was carefully evaluated using a precisely aligned isotropic BOLD template and the subject’s raw EPI, as shown in Supplementary Fig. S4. Analysis of time-courses in these regions in the Chronos group showed positive BOLD responses during 75–130Hz stimulation and negative or weak-positive BOLD responses during 5–20 Hz stimulation (Fig. 4B). Notably, BOLD responses in these regions in the ChR2 group did not exhibit a significant increase at 75Hz and above, in contrast to the Chronos group (Supplementary Fig. S5). Interestingly, the BOLD responses in GP and CPu were negative at 20Hz stimulation in the Chronos group (Fig. 4B) and at any rate in the ChR2 group (Supplementary Fig. S5), while showing positive responses during 75–130Hz in the Chronos group (Fig. 4B). In the ChR2 group, a similar analysis revealed two brain regions, the contralateral LGN and SC, that were significantly modulated by pulse repetition rate (Fig. 4C). Although these regions were not identified as significant in the Chronos group, the extracted time course data revealed similar pulse repetition rate-dependence modulation in these regions within the Chronos group as shown in Supplementary Fig. S6. In these clusters, we observed that the BOLD signals were higher during 5Hz stimulation compared to the range of 20–130Hz stimulation (Fig. 4D and Supplementary Fig. S6). This pattern suggests that the responses were light-induced visual responses, as indicated in previous research [36-41]. Therefore, these responses are likely non-therapeutic side effects of optogenetic stimulation. In addition, we assessed the voxel-level correlations between BOLD signals and pulse rate-dependent circling behavior effects (Supplementary Fig. S7) and found a general agreement with these cluster-level results.

Fig. 4. Voxel-wise analysis followed by post hoc tests depict significant pulse repetition rate-dependent BOLD responses during optogenetic STN DBS.

The 3D representations on the left display the significant clusters for Chronos (A) and ChR2 (C), with statistical parametric maps shown on the right. The BOLD response patterns from respective cluster ROIs for the Chronos and ChR2 groups are shown in (B) and (D), respectively. The stimulation epoch is indicated by the gray-shaded band. Solid lines and color shadings of the time-courses represent mean ± sem. All bar graphs extract the peak of BOLD percentage changes during the stimulation period. * denotes p < 0.05 by t-test. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

3.4. Pulse repetition rate-modulated activity in GP and CPu, but not in SNr, was related to the therapeutic effect of DBS

We further determined how STN optogenetic DBS manipulated pathological circling behavior through activity changes in the three identified brain clusters (CPu, GP, and SNr, see Fig. 4A) using a mediation model with linear mixed-effects regression [42,43]. First, the data demonstrated that the increased pulse repetition rate of optogenetic STN DBS generated the reduction of normalized angular velocity shown earlier in Fig. 2D (coefficient = −0.735, 95 % CI = −0.744 to −0.726, p < 0.01; paths c in Fig. 5). Next, in a multiple mediation model, we set the pulse repetition rate as the independent variable, normalized angular velocity as the dependent variable, and the evoked BOLD fMRI responses in ipsilateral CPu, GP, and SNr as mediators. This analysis revealed that CPu and GP had significant mediation effects (paths a1, b1, a2, b2, all p < 0.01 in Fig. 5) on the relationship between pulse repetition rate and normalized angular velocity. Critically, such an effect was absent when examining SNr (path b3, p > 0.1, in Fig. 5). Sobel test confirmed that both CPu and GP were significant mediators (Sobel z = −14.99, p < 0.01 and −21.82, p < 0.01, respectively). Finally, with CPu and GP incorporated as mediators, the relationship between pulse repetition rate and normalized angular velocity remained significant (path c’, p < 0.01 in Fig. 5), suggesting that CPu and GP are significant partial mediators for the therapeutic effect of optogenetic STN DBS.

Fig. 5. Pulse repetition rate tuning of pathological behavior amelioration is significantly mediated by optogenetic DBS-induced changes in activity in CPu and GP, but not in SNr.

Mediation analysis was conducted with direct effects (a1, a2, a3, b1, b2, b3, and c) and with mediators (c’). For each significant path, the Sobel test was performed to evaluate the significance of the mediation effect (Sobel z = −14.99 in the path through CPu and Sobel z = −21.82 in the path through GP, p < 0.01). *p < 0.05, **p < 0.01, and ***p < 0.001; n.s., not significant.

4. Discussion

We combined optogenetic DBS and fMRI to map the brain regions specifically modulated by STN feedforward activity in hemiparkinsonian rats. We first replicated the behavioral finding in a prior study [14] that high pulse repetition rate optogenetic STN DBS using Chronos can ameliorate pathological behavior in hemi-parkinsonian rats. In the same subject groups, we then identified a range of subcortical brain regions, including CPu, GP, and SNr, that were activated in a pulse repetition rate-dependent manner by optogenetic STN DBS in rats expressing the fast kinetic Chronos opsin. Notably, pulse rate-dependent activation in these areas was absent (in CPu and GP) or moderate (in SNr) in rats expressing the kinetically slow ChR2 opsin. A mediation model revealed that among the STN feedforward BOLD activity changes found in CPu, GP, and SNr, only CPu and GP exhibit significant partial mediation effects to pulse repetition rate-dependent amelioration of pathological circling, highlighting the critical roles of the CPu and GP in mediating symptom alleviation by STN DBS. We discuss how our findings may contribute to the understanding of mechanisms of DBS in PD.

4.1. Pulse repetition rate-dependent BOLD signal changes during optogenetic STN DBS

We used optogenetics to confine the effects of DBS to local STN cell body activation [14,44] and leveraged brain-wide fMRI to map the modulated territories [36,40,45-58]. Optogenetic stimulation modulated the BOLD signal in three major downstream brain areas: SNr, GP and CPu. Notably, the pulse repetition rate-dependent effects in GP, SNr, and CPu were only observed in the Chronos group, but not in the ChR2 group. The kinetics of ChR2 limit maximum firing rates to ~40 Hz [24], as the slow recovery kinetics of ChR2 can cause depolarization block. In contrast, the frequency-dependent spiking fidelity mediated by fast opsins like Chronos are much closer to those evoked by electrical stimulation [13,59]. The results in the ChR2 group thus serve as an important control for the Chronos effects in activating the STN at high frequencies (>100Hz). In both groups, we observed light-related stimulus side effects of optogenetic DBS in LGN and SC that were most robust at non-therapeutic, low pulse repetition rates. Activation in LGN and SC are broadly reported in literature studying visual-evoked fMRI responses at low stimulus frequencies as observed in our data [36-41].

The principal neurons in STN are glutamatergic [60-62], in contrast to the GABAergic neurons found in the rest of the basal ganglia nuclei [63]. STN receives inhibitory inputs from GP (GP of rodent is homologous to the primate external segment of globus pallidus, GPe) and excitatory inputs from the frontal cortex [64-66]. STN efferents target intensively the GPe, internal segment of globus pallidus (GPi), and SNr, and sparsely the CPu [61]. In the current study, we used optogenetics to stimulate the cell bodies in STN, without antidromically activating the cortex or GP (or white matter tracts near the stimulation site). Our findings extend the existing knowledge of anatomical connectivity by revealing robust, pulse repetition rate-dependent functional BOLD responses in SNr, GP, and CPu. The overall rate dependence is consistent with previous electrophysiological studies, which found high pulse repletion rate electrical DBS in STN activated efferents to GP, SNr and CPu, and modulated the firing rate and firing pattern in these downstream brain regions [22,67-77].

4.2. Pulse repetition rate modulated-activity in GP and CPu, but not in SNr, was related to the therapeutic effect of DBS

We employed a mediation model to examine how activity changes in specific brain regions contributed to the improvement of pathological circling behavior. While the BOLD signals in SNr, GP and CPu were all significantly modulated by stimulation pulse repetition rate, the mediation analysis revealed that only GP and CPu, but not SNr, contributed to the reduction of pathological circling.

GP is reciprocally connected with STN, and the GP-STN network generates oscillatory activity that regulates the output of the basal ganglia [78,79]. In PD, the GP-STN network is altered, leading to pathological beta oscillations (12–30Hz) [80,81] that accompany motor symptoms in PD patients [82], and which are suppressed by therapeutic STN DBS [83]. Critically, optogenetic or chemogenetic activation of GP alone was sufficient to attenuate pathological oscillations and/or ameliorate motor dysfunction in PD models [84-86]. Our results highlight that the change of GP activity potentially plays a causative role in the therapeutic effect of STN DBS. It calls for further studies of the changes of the GP-STN network in understanding the mechanisms of STN DBS for PD.

Our mediation analysis also showed that CPu was involved in the resolution of pathological circling by high pulse rate STN DBS. The modulation of CPu activity by STN DBS may be via direct activation, through the sparse projection from STN to CPu, or via indirect activation through pathways inside or outside the basal ganglia. Indeed, a previous study showed that high frequency STN DBS suppressed pathological oscillatory activity in CPu in a PD model [87]. In addition to directly modulating the CPu activity, STN DBS may also exert its influence by regulating residual dopamine release from substantia nigra pars compacta via the basal ganglia. This notion is supported by experiments showing increased striatal dopamine levels during STN high frequency stimulation in parkinsonian animals [87-89].

In contrast to GP and CPu, SNr was not significantly involved in the pulse repetition rate-dependent reduction of pathological circling in the present study. Clinically, SNr is not a preferred site (as compared to STN or GPi) for DBS procedures to control PD motor symptoms. One previous study showed that SNr DBS was less effective than STN DBS in improving most PD motor symptoms, including akinesia, rigidity, and tremor [90]. Another study chemically modulated activity in either SNr, GPi or STN in a parkinsonian monkey model, and compared the resulting behavioral effects. It was found that altering the activity of SNr was less effective and less consistent in reducing parkinsonian motor signs. Furthermore, the symptom relieving effects were observed only when manipulating the anterolateral SNr that projects to the ‘motor’ thalamocortical circuit, but not medial SNr, which is related to the ‘non-motor’ functions [91,92]. The current study selectively stimulated STN neurons, which led to both pulse repetition rate-dependent symptom relief and rate-dependent activation of downstream regions, including CPu, GP, and SNr, yet the induced SNr activation was not causally related to the improvement of pathological circling. Our results provide evidence that the therapeutic effects of DBS on PD symptoms vary across output nodes of the basal ganglia. The effects on motor symptoms like tremor and akinesia may be via the output of GP. Thus, future research on SNr DBS may focus on its effect on non-motor symptoms of PD [93,94] or on gait-related symptoms [95-100].

4.3. Limitation of study

The comparatively small sample sizes (5 for Chronos group and 4 for ChR2 group) is a potential limitation of the current study. It is possible that the small size limited the statistical power for some results. For example, the results of individual pulse rate activation maps (Supplementary Fig. S1) show more significant clusters than those revealed in the pulse rate and virus type modulation effects (Fig. 4). This may be because the more complex statistical analysis using the linear mixed model generally requires a larger sample size to reveal significance. The linear mixed model analysis was necessary for the repeated measures and imbalanced subject numbers, and we repeated the measures to reduce the variance (6 trials per scan, 3 scans per subject). Despite the sample size, the veracity of the major results is supported by the consistency of key clusters across different analyses (Fig. 4, Supplementary Figure S1-S3, S7), the direct anatomical connections between these clusters and the stimulation site, and the co-variation of the behavioral data and the BOLD signal as revealed by the mediation analysis.

Another limitation of the study is the lack of a third animal group with only GFP but without opsin expression. Such a control group would help to rule out confounds like visual stimulation and heating effects. Nevertheless, even without a GFP-only virus group, the results are not consistent with either visual stimulation or heating. Visual stimulation by laser light leakage did affect the BOLD signal, but these visual-evoked BOLD responses were not therapeutically relevant: the BOLD signals in visual areas were highest during low pulse rate stimulation in both the ChR2 group and the Chronos group (Fig. 4C-D and Supplemental Fig. S6), while the behavioral-related effects were observed during high pulse repetition rate stimulation and only in the Chronos group.

Similarly, the observed BOLD signals were unlikely caused by laser-induced tissue heating. First, while several studies showed that high-intensity laser light was sufficient to affect local fMRI signal around the light stimulating site [101-103], the stimulation-induced BOLD signals found in the current study were located in brain regions relatively distant from the stimulation site. Second, our analysis showed that the fMRI responses in CPu and GP (but not in SNr, which is closer to stimulation site) had significant mediation effect on the relationship between the stimulation pulse rate and the behavioral effects, suggesting that the BOLD signal was therapeutically relevant. Third, in a previous study we recorded single unit activity in STN, GPe and SNr during optogenetic STN DBS [14], using the same stimulation parameters as in the current study. We found neuron populations that were either excited or inhibited by optogenetic stimulation, and the changes of the firing rate were instantaneous upon the laser light onset. These rapid yet heterogeneous neuronal responses were distinct from those induced by laser heating [104]. Lastly, we used short pulses (1 ms duration, duty cycle = 13 % at 130Hz) of laser light to minimize heating.

Finally, while the current study focused on investigating the role of activating the feedforward pathway in the therapeutic STN DBS, we should not dismiss the possible critical involvement of antidromic cortical activation in STN DBS treatment [10,11,16,105-107]. Future studies could use optogenetics to activate the two different pathways separately or simultaneously to investigate their individual contributions, as well as possible interactions in contributing to the therapeutic effects of DBS.

5. Conclusion

Numerous studies have underscored the significance of the hyperdirect pathway in therapeutic STN DBS, while downplaying the contribution of STN feedforward circuits. In a departure from previous approaches that recruited antidromic activation of the hyperdirect pathway or failed to appropriately stimulate STN at high pulse repetition rates, we employed an unbiased, brain-wide fMRI approach coupled with optogenetics using Chronos to identify subcortical structures responsible for ameliorating pathological circling behavior in a rat model of PD. Contrary to some prevailing beliefs that subcortical regions downstream of the STN do not play a critical role in STN DBS therapy, our findings challenge this notion. Specifically, our results indicate that two nuclei within the basal ganglia, the GP and the CPu, may be causally involved in the therapeutic effects of STN stimulation. These results highlight the importance of understanding how the reshaping of network activity within the basal ganglia, and the subsequent impact on subcortical basal ganglia output, contribute to the effectiveness of STN DBS therapy. This novel perspective prompts further exploration of the STN-GP and STN-CPu pathways, potentially leading to enhanced outcomes in STN DBS interventions in PD. The interdisciplinary approach presented herein, encompassing optogenetics, fMRI, quantitative behavioral tests, and mediation analysis, serves as a model for investigating diverse neural pathways involved in DBS therapies for a spectrum of neurological and psychiatric disorders.

Data availability

The dataset generated during the current study is available on an open platform: https://openneuro.org/datasets/ds004913/versions/1.0.0

https://doi.org/10.18112/openneuro.ds004913.v1.0.0.