Orientation selective deep brain stimulation of the subthalamic nucleus in rats

Abstract

Deep brain stimulation (DBS) has become an important tool in the management of a wide spectrum of diseases in neurology and psychiatry. Target selection is a vital aspect of DBS so that only the desired areas are stimulated. Segmented leads and current steering have been shown to be a promising addition to DBS technology enabling better control of the stimulating electric field. Recently introduced orientation selective DBS (OS-DBS) is a related development permitting sensitization of the stimulus to axonal pathways with different orientations by freely controlling the primary direction of the electric field using multiple contacts. Here, we used OS-DBS to stimulate the subthalamic nucleus (STN) in healthy rats while simultaneously monitoring the induced brain activity with fMRI. Maximal activation of the sensorimotor and basal ganglia-thalamocortical networks was observed when the electric field was aligned with the STN pointing laterally in the mediolateral direction, while no cortical activation was observed with the electric field pointing medially to the opposite direction. Such findings are consistent with mediolateral main direction of the STN fibers, as seen with high resolution diffusion imaging and histology. The asymmetry of the OS-DBS dipolar field distribution using three contacts along with the potential stimulation of the internal capsule, are also discussed. We conclude that OS-DBS offers an additional degree of flexibility for optimization of DBS of the STN which may enable a better treatment response.

Graphical Abstract

1. Introduction

Deep brain stimulation (DBS) has become an increasingly important tool for the management of a multitude of neurological and psychiatric disorders. Stimulating the target region with DBS while avoiding the regions near the target is vital for maximizing the treatment effect and minimizing possible side effects (Christen et al., 2012). Segmented DBS leads and advanced stimulation paradigms (Butson and McIntyre, 2008; Chaturvedi et al., 2012a; Keane et al., 2012; Martens et al., 2011) are promising directions for expanding and optimizing the stimulation outcomes. The capability of steering electric fields to the specific side of the implanted multichannel electrode allows a wider therapeutic window by improving target selection and by at the same time avoiding structures related to adverse side effects, especially when the precision of implantation is compromised.

Recently, a novel orientation selective strategy for DBS (OS-DBS) was introduced (Lehto et al., 2017b). With this concept, the spatial gradient of electric field can be oriented along a desired orientation in space using multiple contacts with independent current sources. OS-DBS allows preferential stimulation of axonal bundles with different orientations. This strategy was first demonstrated by stimulating the corpus callosum (Lehto et al., 2017b) and then further applied in the infralimbic cortex (target for major depression) (Lehto et al., 2018) in healthy rats. The selectivity of the stimulation was monitored with functional magnetic resonance imaging (fMRI). OS-DBS is based on the known phenomenon that axons are the most sensitive to electric stimulation when the spatial gradient of the electric field is parallel to the axons (Lehto et al., 2017b; Rattay, 1989).

While the applications of DBS are currently expanding to e.g. psychiatric diseases such as depression and obsessive compulsive disorder (Kisely et al., 2014; Lakhan and Callaway, 2010) or epilepsy (Klinger and Mittal, 2016), the hallmark applications are still the treatment of movement disorders such as Parkinson’s disease (PD) (Flora et al., 2010; Pizzolato and Mandat, 2012) where the subthalamic nucleus (STN) is one of the primary targets (Benabid et al., 2009). In patients, the DBS lead is often orientated in the STN so that the target fibers, namely the STN efferents and hyperdirect pathway, are nearly perpendicular to the lead. On the other hand, the STN is closely bordered by the fibers of the internal capsule that are nearly parallel to the lead. The potential stimulation of the internal capsule during STN-DBS is associated with side effects such as involuntary muscle contractions and paresthesia (Hamani et al., 2004). Encouragingly, in our recent modelling study of human DBS leads, PD patient specific neuronal models demonstrated high potential for OS-DBS to selectively stimulate axons of different orientations in three dimensions using commercial leads with independently driven contacts (Slopsema et al., 2018).

Several reports have shown high utility of simultaneous fMRI and DBS for recording of full brain network activity in response to DBS (Lai et al., 2014; Min et al., 2012; Phillips et al., 2006). Here, we further extend our initial finding of the stimulation of the corpus callosum (Lehto et al., 2017b) and infralimbic cortex (Lehto et al., 2017a) areas to OS stimulation of STN, and demonstrate in a preclinical fMRI setting how OS-DBS can be used to modulate the network activity related to the STN. Additionally, we characterize the main directions of fibers of the STN and surrounding areas using high resolution diffusion MRI tractography combined with histological analysis, revealing the likelihood of axonal candidates for OS-DBS. Our results show that OS-DBS can differentially activate basal ganglia and cortical networks connected to the STN when changing the orientation of the stimulation.

2. Methods

2.1. Animals

For DBS-fMRI, Sprague-Dawley rats (Envigo; Madison, WI, USA; n = 10, male, approximately 300 g) were housed in pairs in a temperature and humidity-controlled vivarium with a 12-h light-dark cycle with ad libitum diet. These animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Minnesota.

2.2. Electrode implantation

After the induction of isoflurane anaesthesia (5% induction, 2–3% maintenance; carrier gas O₂/N₂O 30/70%), the animals were placed on a heating pad and into a stereotactic frame (Stoelting; Wood Dale, IL, USA). Temperature was monitored via a rectal probe and maintained at 37 °C, while respiration rate was monitored using a plastic pressure sensor and the rate remained at 60–80 per min. Burr hole with a diameter of 0.7 mm was performed over the implantation target and a three-channel electrode composed of tightly braided three polyimide-insulated (30 μm thickness) tungsten electrodes (PlasticsOne, MS333T/3C-C 3TW; Roanoke, VA, USA) with tip-only contact diameter of 127 μm was inserted unilaterally into the right STN (Figure 1A; rostrocaudal −3.7 mm, mediolateral 2.5 mm and dorsoventral 7.8 mm) (Swanson, 2018). The orientation of the electrode tip (Figure 1B) was controlled using a microscope, identifying each channel with a multi-meter. The burr hole around the electrode was filled with gelatin foam (SPONGOSTAN™, Søborg, Denmark) and covered using dental acrylic (Lang Dental, Jet Acrylic, Wheeling, IL, USA) to secure the tripolar electrode to the cranium. An Ag/AgCl grounding wire (4 cm long, diameter of 1 mm) was inserted below the skin, with the tip located at the base of the neck. Prior to the transfer into the MRI system, anesthesia was switched to urethane (4 consecutive intraperitoneal injections with the dose of 1.25–1.50 g/kg, 15 minutes apart) while gradually decreasing the isoflurane level and discontinuing it at the last urethane injection. Optical rectal temperature probe and pressure respiration sensor were employed to monitor body temperature and respiration (Small Animal Instruments Inc., New York, NY, USA), respectively, during the MRI scan. The body temperature was maintained at the level of 37 °C using heated water circulation and heated air. Respiration remained at 110–130 per min during the fMRI experiments. Heart rate was not monitored.

Figure 1. DBS setup.

(A) Illustration of the electrode in the STN on a coronal schematic of the rat brain, and (B) a corresponding T2-weighted image. The electrode represented by a grey bar in (A, B) (C) Illustration of the OS-DBS stimulation orientations on a horizontal schematic of the rat brain using three-wire electrodes. 0°/180° corresponds to the rostrocaudal direction and −90°/90° corresponds to the mediolateral direction on the horizontal plane. Blue indicates the potential field of the cathode while red indicates the field of the anode in (C). Abbreviations: zona incerta, ZI; subthalamic nucleus, STN; internal capsule, ic.

Lastly, two additional animals were used to confirm the effect of OS-DBS so that the electrode implanted on the left side instead of the right side.

2.3. MRI acquisition

All MRI scans were conducted with a 9.4-T 31-cm horizontal-bore magnet equipped with Agilent DirectDRIVE console (Palo Alto, CA, USA) using a quadrature radio frequency volume coil with full rat brain coverage. The coil was composed of ¹H MRI invisible materials, thus ensuring that no unwanted signal would fold into the field of view (FOV) from the coil itself. Prior to fMRI, anatomical images were acquired using a fast spin-echo (FSE) sequence: repetition time (TR) = 3 s, effective echo time = 48 ms, number of echoes = 8, matrix size = 192², FOV = 3.2 × 3.2 cm², slice thickness = 1 mm, number of slices = 15, no interslice gap and number of averages = 4. Next, MB-SWIFT fMRI (Lehto et al., 2017a) was conducted using the following parameters: TR = 0.97 ms, 3094 spokes per volume, resulting in temporal resolution of 3 s, bandwidth (BW) = 192 kHz, matrix size = 64³, FOV = 3.5 × 3.5 × 6.4 cm³ and flip angle = 5°. Excitation was performed with a chirp pulse gapped into four 2.6-μs sub-pulses (Idiyatullin et al., 2006; Idiyatullin et al., 2015). Two-fold oversampling was used during acquisition in the gaps of 32/BW duration. The post-correlation free induction decay (Idiyatullin et al., 2006) consisted of 32 points. MB-SWIFT is a 3D radial FID based pulse sequence with virtually zero acquisition delay. Unlike conventional EPI pulse sequences optimized for fMRI, MB-SWIFT is insensitive to T2*-decay in the timescale regime of the BOLD contrast. The functional contrast of MB-SWIFT has been related mainly to blood inflow effect (Lehto et al., 2017a). Due to the minimal acquisition delay and high bandwidth in three dimensions, MB-SWIFT is also insensitive to susceptibility artefacts due e.g. electrodes or air-tissue interfaces.

2.4. Functional paradigms of deep brain stimulation

All stimulation paradigms consisted of 3 blocks of 60 s of rest and 18 s of stimulation, ending with an additional rest period, resulting in 4 min 54 s of total paradigm. OS-DBS was achieved by controlling the orientation of an electric dipole under the tip of three-channel electrode. As the strongest electric field gradient of a dipole is aligned with its primary axis, an axon is the most excitable when the primary axis of a dipole is aligned with the axon (Lehto et al., 2017b; Rattay, 1989). The orientation of the dipole was controlled by changing the amplitudes of currents of individual channels relative to each other by choosing the amplitudes from phase offset sinusoids, as detailed in (Lehto et al., 2017b). Stimulation was applied using 100 μs symmetric biphasic square pulses without interphase delay repeated at 130 Hz. Current was set to 0.8–1.2 mA based on initial fMRI scans with monopolar stimulation with total current divided between the three channels. The stimulation angles were incremented in steps of 30° resulting in 12 separate OS-DBS experiments. The angles of stimulation were set such that 0°/180° corresponded to the rostrocaudal direction and −90°/90° corresponded to mediolateral direction (Figure 1B). The order of the individual experiments with different stimulation angles was randomized for each animal with the exception of starting the stimuli from stimulation angle 90°. Because preliminary studies indicated that 90° angle likely provides the strongest response, the intent was to test at the beginning of the study that dipolar OS-DBS is optimized. As demonstrated in Supplementary Figure 3, the beginning of stimulation at 90° was not affected the ratios between responses at different angles of the stimulation.

The stimulation was delivered using STG-4008–16mA (Multi Channel Systems, Reutlingen, Germany) 8 channel stimulus-isolator system in current controlled mode. The stimulation waveforms were computed using MATLAB 2016a (Mathworks; Natick, MA, USA) and exported to the STG-4008–16mA software. The stimulus isolator was connected to the electrodes via three twisted pair cables and routed through a radiofrequency low-pass filter plate into the MRI scanner Faraday cage to reduce radiofrequency noise of the MRI acquisition. Additional low-pass filters (cut-off 1.9 MHz) were attached to the output of the stimulus isolator to remove potential radiofrequency noise coupled to the stimulus-isolator itself. The ends of the three twisted pair cables were connected to a 50-cm coaxial cables that were stripped exposing the shielding and the center conductor. The shielding was soldered to a small copper plate connected to a subcutaneous Ag/AgCl ground electrode placed intracutaneously in the neck. The center conductors were soldered to the electrode connector (PlasticsOne, 335–000; Roanoke, VA, USA) (Lehto et al., 2018).

2.5. MRI data processing and analysis

MB-SWIFT images were reconstructed using gridding and iterative FISTA algorithm (Beck and Teboulle, 2009) with three iterations. The resulting data were analysed in SPM8 (www.fil.ion.ucl.ac.uk/spm) and custom scripts using MATLAB 2013b. The pre-processing included motion correction, co-registration and normalization to an animal without an electrode outside the fMRI group based on FSE images, and smoothing with a [2 2 1] pixel FWHM Gaussian kernel. The general linear model consisted of a block design model convolved with a 1st order gamma function (time-to-peak 3 s, width 8 s) and the baseline. For the second level analysis, a one way within subject ANOVA model was fitted to the first level β-maps using the stimulation angle as a factor resulting in 12 levels. Maps of main effects and each angle’s difference to the other angles (one vs. all) were finally computed. To check the data for anesthesia-related artifacts, individual ICA maps using Melodic ICA (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) for each animal were calculated. Out of 30 components, the cortical components were chosen and visually estimated. To estimate the overall activation pattern, group ICA maps were calculated. Components related to the stimulus were chosen and combined.

To quantify the differences of the fMRI responses to different OS-DBS angles, aggregates of MB-SWIFT fMRI data were then obtained in anatomically defined regions of interest (ROIs) drawn manually with Aedes (aedes.uef.fi) based on a rat brain atlas (Swanson, 2018). The ROIs were drawn in the major brain structures connected to the STN including the caudate putamen (CP), superior colliculus (SC), globus pallidus (GPl), substantia nigra (SNr), thalamus and motor cortex (Mx). For each ROI and stimulation angle, the time series of all voxels in an individual ROI were averaged and the corresponding fMRI response amplitude was assessed by averaging 7 time points surrounding the maximum of the response during each stimulation period (i.e. 21 points in total) from the ROI mean time series. The statistical analysis of the ROI results was conducted by comparing the highest fMRI response observed at a certain OS-DBS angle with responses to all other angles using Wilcoxon signed-rank test (p < 0.05, false discovery rate [FDR] corrected at ROI level). A tensor group ICA analysis of the individual data sets using dual regression was also performed for further evaluating maps obtained at different angles of the stimulation. For each stimulus-related component, cross-subject comparison of maps for different angles was carried out. The contrast maps were corrected for multiple comparisons.

After fMRI, rats were transcardially perfused by manually injecting 600 ml of saline followed by 600 ml of 4% paraformaldehyde (PFA) over 30 minutes. The brains were extracted, post-fixed in 4% paraformaldehyde for 24 h, and stored at 4°C in 30% sucrose. Brains were mounted on to a vibratome for cutting coronal sections at a thickness of 50 μm. Brain tissue sections were placed on to glass slides and stained with hematoxylin and eosin for visualization of the electrode tract. Finally, confirmation of the implantation of the electrodes into the STN was assessed by light microscopy.

2.6. Track density imaging

One adult male Sprague-Dawley rat outside the DBS group (300 g, Harlan Netherlands B.V., Horst, Netherlands) was used for ex vivo super-resolution tract density imaging (TDI) (Calamante et al., 2010) to investigate the directionality of axons within the STN. The rat was housed in an individual cage and kept under a normal 12 h light/12 h dark cycle with constant temperature (22 ± 1°C) and humidity (50–60%). Water and food were available ad libitum. This animal procedure was approved by the Animal Ethics Committee of the Provincial Government of Southern Finland, and carried out in accordance with the guidelines of the European Community Council Directives 2010/63/EEC.

The rat was transcardially perfused using saline (30 ml/min) for 2 min followed by 4% PFA (30 ml/min) at 4°C for 25 min. The brain was removed from the skull and post-fixed in 4% PFA at 4°C for 4 h. The brain was placed in saline for at least 12 h to remove excess PFA. During ex vivo imaging, the brains were immersed in perfluoropolyether (Solexis Galden®, Solvay, Houston, TX, USA) to avoid signal from outside the brain.

The rat brain was scanned in a vertical 9.4 T/89 mm magnet (Oxford Instruments PLC, Abingdon, UK) interfaced with a DirectDrive console (Varian Inc., Palo Alto, CA, USA) using a quadrature volume RF-coil (Ø = 20 mm; Rapid Biomedical GmbH, Rimpar, Germany) as a transceiver. The data were acquired using a 3D spin-echo-echo-planar-imaging sequence using the following parameters: TR/TE = 800/35 ms, echo spacing = 0.584 ms, number of shots = 4, bandwidth = 250 kHz, number of averages = 2, FOV = 16.2 × 12.0 × 14.1 mm³, matrix size = 108 × 80 × 94, spatial resolution = 150 × 150 × 150 μm³, number of diffusion directions = 60, b-value = 3020 s/mm², number of minimally diffusion weighted images = 1, diffusion gradient amplitude = 3.5 mT/cm and duration (δ)/separation (Δ) = 6/11.50 ms, and acquisition time = 20 h 48 min.

The data was pre-processed using ExploreDTI (Leemans et al., 2009) consisting of simultaneous motion and residual eddy current induced geometric distortion correction. In order to increase anatomical contrast, the data were then upsampled by a factor of two using B-spline interpolation. A voxel-wise fit of the diffusion tensor model was performed using MRtrix3 (www.mrtrix.org)(Tournier et al., 2012). Fiber tracks were then generated using constrained spherical deconvolution (CSD) to resolve crossing-fibers by estimating the fiber orientation distribution at each voxel. Probabilistic fiber tracking was performed using the iFOD2 algorithm (Tournier et al., 2012) with a step-size of 0.01 mm, maximum angle of 45° between steps, and lmax = 8. A CSD probabilistic-tractography dataset of 2 million short fiber tracks was calculated from the frontal part of the brain, with a minimum track length of 0.20 mm and maximum track length of 1.00 mm. The tractography dataset was then used to generate directionally encoded color TDI maps with a final resolution of 20 × 20 × 20 μm³.

The orientation of the myelinated axons in the STN have been assessed by an expert in at least 10 animals in our laboratory, and a representative animal was selected (Figure 6). In the current study we present single data set as the results are intended to be a qualitative visualization, and it is unlikely that interanimal variability in the main fiber directions in healthy animals is such that it would negate our conclusions, and be a significant factor over the angular precision of OS-DBS and the imprecision of implantation. Moreover, our findings are supported by previously reported studies (Coizet et al., 2009; Plantinga et al., 2016).

Figure 6. Main axonal direction in the STN.

Two representative examples out of 10 studied rats for the orientation of fibers in the STN. (A, D) Diffusion b0-image of the region of the STN, (B, E) close-up of the STN in super-resolution TDI maps, and (C, F) photomicrographs of myelin-stained sections in the coronal and horizontal plane, respectively. Qualitatively, the axonal direction in the STN is mainly mediolateral (white arrowheads). The size of the electrode is shown in scale in (B, E) by black dashed lines. Color coding: red, mediolateral; blue, dorsoventral; green rostrocaudal. Abbreviations: subthalamic nucleus, STN; internal capsule, ic. Scalebar in C and F is 100 μm.

3. Results

In 8 out of the group of 10 animals in the DBS study the implantation was confirmed to be in the STN based on histological analysis. Only these 8 animals were included in further analysis. Overall, OS-DBS of the STN induced modulation of the STN related networks including the basal ganglia structures of the ipsilateral substantia nigra, globus pallidus and caudate-putamen, bilateral thalamus, ipsilateral somatosensory, motor and retrosplenial cortices, and in the midbrain the reticular nucleus and superior colliculus.

The main effects of the group analysis of the fMRI contrast in response to the OS-DBS stimulation are shown in Figure 2. Stimulation angle 90° led to the strongest response in terms of spread of the activation including the whole ipsilateral midbrain and contralateral superior colliculus, the whole ipsilateral and parts of the contralateral thalamus, ipsilateral caudate-putamen, ipsilateral globus pallidus, ipsilateral motor, cingulate, somatosensory and medial prefrontal cortices. Stimulation angle 90° also showed the strongest activation in terms of t-score values. Weakest activity was seen with stimulation angles 180°, −150° and −120° with the later showing no significant voxels (Figure 2). The cingulate cortex showed group level activated voxels at stimulation angles −60°, −30°, 0°, 30°, 90°, 120°, 150°, the somatosensory cortex activated at angles −30°, 0°, 30°, 60°, 90° and 150° while the motor cortex activated only at angles −30° and 90°. With a differential contrast comparing each stimulation angle to all other angles at group level (Figure 3A), only stimulation angle 90° showed statistically significant clusters of increased activity at p < 0.001, cluster level, uncorrected. At this statistical threshold, two major clusters of activity were found with a size of 4002 and 351 voxels. The larger ipsilateral one included globus pallidus, caudate-putamen, motor and somatosensory cortices, reticular, ventral, oral and parafascicular thalamic nuclei, zona inserta, lateral hypothalamus, and substantia nigra. The smaller contralateral one included parafascicular, ventral posterior, and lateral and medial geniculate thalamic nuclei, substantia nigra, inferior colliculus and deep layers of superior colliculus. When applying these clusters as ROIs to the individual level β-maps, clear orientation selectivity was observed, and the highest cluster mean β-values were achieved using the 90° stimulation angle (Figure 3B).

Figure 2. Activation maps obtained from the group analysis at all OS-DBS stimulation angles.

Maximal activation was observed at 90° when the dipole under the electrode was aligned mediolaterally and the cathode was lateral. p < 0.05, FWE corrected. The used ROIs are shown in white. Abbreviations: superior colliculus, SCo; substantia nigra, SNr; thalamus, Tha; caudate putamen, CPu; globus pallidus, GPl; cingulate cortex, Cg; motor cortex, Mx; and somatosensory cortex, SS.

Figure 3. Comparison of response to OS-DBS at 90° to all other angles based on the group analysis.

(A) Group level activation maps comparing 90° to all other stimulation angles (p < 0.001, cluster level, uncorrected). Note that only results from OS-DBS angle 90° are shown as this was the only angle to give a statistically significant clusters. (B) Mean individual level β-values calculated using the two major group level clusters as ROIs. The peaks of these cluster are indicated by white crosses in (A). Maximal β-values were seen at 90° when the dipole under the electrode is aligned mediolaterally and the cathode is lateral. Blue lines represent the means and error bars represent standard deviations in (B). *p < 0.05 (FDR corrected), smaller than response at 90°, paired t-test.

Mean time series for the caudate putamen at different OS-DBS stimulation angles are shown in Figure 4. Mean fMRI amplitudes were further quantified using an ROI analysis (Figure 5). All ROIs showed stimulation angles that had statistically significant lower mean fMRI responses compared to the stimulation angle 90° except the cingulate cortex. In general, larger variability between the animals using the stimulation angles of the rostral half (−90° to 60°) were seen. Strongest activity was seen in the substantia nigra with an approximately 4% response at the stimulation angle 90°. Individual time series can be seen in Supplementary Figure 1. When comparing maps obtained at different angles by dual regression ICA, no clusters reached statistical significance.

Figure 4. Mean time series from an ROI in the caudate putamen (CP; middle image) in response to OS-DBS using different angles of stimulation.

The mean is shown using red lines and standard deviation is shown by the yellow shaded areas.

Figure 5. ROI analysis of the fMRI response amplitudes.

(A) Regions of interests drawn on the anatomical FSE reference images and (B) individual data points of the fMRI responses of each animal using different OS-DBS angles. *p < 0.05 (FDR corrected), smaller than response at 90°, Wilcoxon signed-rank test.

To verify that our fMRI findings accurately follow with the main axonal direction in the STN, high resolution TDI and myelin stained histology were performed in an additional ex vivo rat brain (Figure 6). The axonal directionality was shown to be mainly mediolateral in the STN, which was in a good agreement with fMRI results. Additionally, the main axonal directionality of the internal capsule was shown to be nearly perpendicular to that of the STN.

ICA analyses showed that potential abnormal brain activity and state changes due to urethane anesthesia was not a confounding factor for the results of OS-DBS. Aggregated time series based on individual cortical ICA components exhibited responses to the stimulation (Supplementary Figure 2) while a group level ICA over all the stimulation angles showed clear unilateral activation on the side of the electrode (right; Supplementary Figure 3).

When the electrode was implanted on the left hemisphere instead of the right hemisphere, the effect of OS-DBS was sustained in line with the right side stimulation with maximal activation near the stimulation angles 90° (with cathode towards left) and −90° (cathode towards right; Supplementary Figure 4).

Finally, the histological verification of the location of electrode was conducted after each MRI section. One example histological section showing precision of the implantation in the STN is presented in Supplementary Figure 5.

4. Discussion

In this work, we demonstrate for the first time how DBS can be used to differentially modulate the activity of sensorimotor and basal ganglia-thalamocortical networks by stimulating STN in different directions using OS-DBS. This technique allows to control the orientation of the stimulating electric dipole field on a plane under the tips of three independently driven contacts. (Lehto et al., 2018; Lehto et al., 2017b). Our results using simultaneous fMRI and OS-DBS in healthy animals showed maximal response at a specific angle during OS-DBS that corresponds to the main axonal direction in the STN that was verified using high resolution diffusion MRI and myelin stained histology. Our results are in agreement with the previously reported studies which utilize simultaneous DBS of the STN and fMRI in rats (Lai et al., 2014) and in the swine (Min et al., 2012), where positive responses in the basal ganglia-thalamocortical network (thalamus, caudate-putamen, midbrain and globus pallidus) and the sensorimotor network (motor, somatosensory and cingulate cortices) were detected. The activation of basal ganglia-thalamocortical network in response to STN-DBS can be attributed to the stimulation of the STN efferents, while the activation of the sensorimotor network may originate from antidromic stimulation of the hyperdirect cortical afferents to the STN (Gradinaru et al., 2009; Lai et al., 2014; Li et al., 2012) as well as by feed-forward stimulation of the thalamocortical circuit (Bolam et al., 2000). Anatomical evidence supports antidromic stimulation of the STN cortical afferents (Canteras et al., 1990; Kita and Kita, 2012; Naito and Kita, 1994; Rouzaire-Dubois and Scarnati, 1985).

DBS of STN is a well-established procedure with clinically proven outcomes, but it is also associated with significant side effects due to stimulation of unwanted pathway. The human STN in considered to have an axial gradient, such that the sensorimotor sector is located dorsolaterally, associative sector in the middle and the limbic sector inferomedially (Hamani et al., 2017). The primary target of the STN-DBS is the sensorimotor sector, providing relief of major motor disability in PD (Zaidel et al., 2010) while stimulation of the associative and limbic sectors may elicit serious adverse effects in PD patients, such as cognitive decline, mood changes, reduced impulse control, etc. (Mallet et al., 2007; Okun et al., 2009; York et al., 2008). Despite increasing accuracy of the neurosurgical instrumentation, allowing us to avoid major surrounding structures, the ability to target specifically these functionally diverse STN subsections with consistent success rates remains elusive. One limitation is the diagonal angle of approach and linear design of conventional stimulation probes, which makes it very challenging to target precisely the desired sector of STN. Another limitation is current spread to intimately located fiber tracts, ansa lenticularis and internal capsule (Tommasi et al., 2008; Xu et al., 2011). Especially the internal capsule is relatively easy to stimulate as it contains large, myelinated axons. Although direct supra-threshold stimulation of internal capsule is easily avoidable in clinical practice, current spill during STN-DBS at sub-motor threshold levels has been shown to stimulate the pyramidal tract within the internal capsule so that rigidity is improved due to STN stimulation but akinesia and bradykinesia are worsened due to simultaneous stimulation of the internal capsule (Xu et al., 2011). A recent study employing 7 T diffusion-weighted MRI revealed that STN fibers targeting nearby structures emanate from STN at different orientations both inside and outside the nucleus (Plantinga et al., 2016). This finding suggests that OS-DBS has the potential to substantially improve accuracy of targeting the stimulation effect.

Current steering using segmented DBS leads has shown great potential in increasing the therapeutic window for treating Parkinson’s disease with STN-DBS in an intra-operative setting using experimental leads (Contarino et al., 2014; Pollo et al., 2014) but also post-operatively with fully implantable commercial leads (Steigerwald et al., 2016). With three segments around the lead, the direction of the optimal stimulating segment was shown very variable between patients (Pollo et al., 2014; Steigerwald et al., 2016), and when using a single contact from a row of segmented contacts deemed less optimal in conventional ring mode, current steering was shown even more beneficial over the ring mode (Steigerwald et al., 2016). Functional and anatomical connectivity studies (Horn et al., 2017) have further shown that for a successful treatment outcome, the volume of tissue activated by DBS leads in the STN needs to be connected to specific regions, namely the motor cortex, anterior cingulate and medial prefrontal cortex, all which were activated at group level in our results. These findings underline the fact that even with the extreme care and planning used for DBS lead implantations, optimal stimulation settings will still vary between patients and methods beyond the conventional ring electrodes are needed. Using neuronal modelling of STN-DBS using multichannel human leads with independently controlled channels the feasibility of angular selectivity of OS-DBS had been demonstrated (Slopsema et al., 2018). Thus, it may be possible to use OS-DBS to select the activity of networks related to further optimize DBS outcomes.

The OS-DBS strategies of the STN can potentially provide further improvement in the current therapeutic practice, as it may allow more selective stimulation of tracts of interest (motor network), while avoiding the stimulation of areas known to cause adverse side effects. OS-DBS inherently requires three or more contacts and could be applied using the existing segmented leads given that the lead system uses multiple independent current source control. Optimal lead design for OS-DBS likely enables symmetric contact groupings (Contarino et al., 2014) that may not be possible with segmented rings. In the present study, the opportunity for using OS-DBS is substantiated by the main direction of the fibers of the internal capsule being nearly perpendicular to the main fiber direction of the STN as shown by histology and TDI. However, as the exact connectivity of the fibers of the internal capsule near the stimulation site is not known, it is difficult to discriminate when an activation in a specific area is connected to the stimulation of the internal capsule.

Based on the fMRI activation maps and the quantitative analysis of the fMRI response using voxel-wise and ROI analyses, almost all the analyzed brain regions including, caudate-putamen and globus pallidus, motor and somatosensory cortices, several thalamic nuclei, substantia nigra, and inferior and superior colliculus exhibited maximal response at 90° OS-DBS angle. This coincided with the main direction of the STN fibers within the nucleus as shown by diffusion imaging and myelin stained histology. Previously, we have shown using neuronal modelling (Lehto et al., 2017b) that in an ideal scenario axons located below the center of a three-wire electrode should be identically activated when the dipole is pointed in either −90° or 90°. Note, here the 90° shift in angle denomination was used as compared to our previous work. Notably, although not as evident as in the present results, a difference between the stimulating dipole being parallel to the axons in two opposite directions was detected using fMRI in our prior work (Lehto et al., 2017b). In that work, the difference between fMRI outcomes during the stimulation of the corpus callosum and the modelling was explained by a partial anodic block generated by the anodic lobe of the dipole (Szabó et al., 1972; Van Den Honert and Mortimer, 1979). An anodic block can be formed by anodic stimulation that hyperpolarises the cell membrane so strongly that an action potential cannot propagate through. This may partially also explain our current results in the STN-DBS. Furthermore, in the currently used setup of a dipole being divided between three contacts, at the stimulation angle of −90°, the anode is under the single lateral wire whereas the cathode is divided between two medial wires. In this case, the anode extends further ventrally compared to the cathode, meaning that potentially more fibers are blocked, and on the other hand, potentially less of the internal capsule is stimulated. In contrast, at 90° stimulation angle, the situation is reversed and fewer of the fibers are blocked and thus more of the STN and internal capsule are stimulated. Considering further the orientation of the STN on the coronal plane so that it is pointed upwards dorsolaterally (Figure 1A), more of the girth of the nucleus is within the cathodic lobe and also more of the internal capsule may be stimulated when stimulating at 90° angle. A detailed tracing study in the rat suggests that cortico-subthalamic projections leave the internal capsule and target the midrostral STN (Kita and Kita, 2012). Thus, these antidromically activated fibers would be maximally affected when stimulating at 90° angle. The selective activation of the small contralateral cluster by stimulation at 90° angle is more difficult to explain mechanistically. All STN connections and predominantly ipsilateral; however, contralateral connections with thalamic parafascicular nucleus (Castle et al., 2005), substantia nigra (pars reticulata) and superior colliculus (Cavdar et al., 2018) have been reported(Cavdar et al., 2018). The exact course of these projections is not known, but presumably the stimulation orientation that activates the largest number of STN neurons has the highest possibility to activate also the weak projections.

The present results demonstrate the capability of OS-DBS to preferentially control circuits relevant to STN-DBS. Given a more advanced lead design with a larger channel count in three dimensions, it may be possible to perform selective stimulation of various fiber tracts emanating from or impinging on the STN. It has been shown that compared to bipolar current steering, monopolar two electrode current steering may stimulate a greater fraction of simulated projection neurons of the STN simultaneously with axons of the internal globus pallidus (Chaturvedi et al., 2012b) while avoiding the axons of the internal capsule. On the other hand, the same authors showed that unbalanced bipolar current steering may be able to stimulate a greater fraction of only STN projection neurons in comparison to monopolar current steering (Chaturvedi et al., 2012b). Additionally, in simulations of cylindrical contacts of different lengths on human DBS leads, unbalanced bipolar configurations were found to be superior to monopolar configurations when selectively simulating axons parallel to the lead (Howell et al., 2015). The concept of an unbalanced bipolar stimulation could also be applied to OS-DBS and in fact it can already be seen as part of the current tripolar OS-DBS scheme in the sense that cathode/anode can be divided between contacts depending on the stimulation angle as discussed above. Further utilizing the approach of an unbalanced dipole adds an additional level of optimization in tuning the orientation selective stimulation.

An earlier study of STN-DBS, which utilized monopolar stimulation with simultaneous fMRI in rats (Lai et al., 2014), showed robust activations of the motor and the primary somatosensory cortices, which is in line with results presented here. Our results show seemingly more widespread activation throughout the brain than the study by Lai et al. We relate these differences to larger dimensions of the electrodes, higher charge delivery of the stimulation pulses and also the fMRI pulse sequences. Higher current amplitude and stimulation pulse duration were needed to achieve robust functional contrast. MB-SWIFT has been shown to provide very similar amplitudes of activation compared to conventional spin-echo-echo-planar-imaging, however, cerebral blood volume weighted fMRI (Lai et al., 2014; Lai et al., 2015; Van Den Berge et al., 2017) is likely more sensitive for DBS. On the other hand, Lai et al. (Lai et al., 2014) reported that signal, for example, in the globus pallidus and thalamus could not be observed due to susceptibility artefacts induced by the electrodes. In the present study detecting the activity in these regions near the electrode was enabled by the very use of the MB-SWIFT technique inherently minimally sensitive to susceptibility artefacts (Lehto et al., 2017a). The same authors also stimulated the internal globus pallidus and observed negative BOLD responses in the hemisphere contralateral to the electrode, and since the internal globus pallidus is surrounded by the internal capsule, this response was related to stimulating the internal capsule. During STN stimulation no negative responses were observed as also seen in our results. Functional anticorrelation of the stimulated site in the STN and the primary motor cortex has been reported, and using PET decreased activity in the primary motor cortex when the STN is active has been shown (Horn et al., 2017). It is unclear if the anticorrelation or decreased activity would reflect negative BOLD contrast, and on the other hand, MB-SWIFT fMRI is weighted by the inflow effect and thus different than the BOLD contrast. Although negative cerebral blood flow and volume in response to stimulus have been related to decreased neural activity (Boorman et al., 2010; Shmuel et al., 2006; Shmuel et al., 2002), it is currently unknown how this would be reflected by MB-SWIFT. Finally, the ROI analysis showed larger standard deviation using the rostral stimulation angles. This is likely related to the interplay between precision of implantation and the dimensions of the STN so that it extends further mediolaterally than rostrocaudally (Swanson, 2018). Hence, the same degree of implantation error can cause more severe deviation from the intended stimulation scheme as more of the tip of the electrode may be outside the STN. This could also lead to stimulation of the surrounding areas. However, it should be noted that in the present study the location of the electrode tip was verified by histology in every animal in order to control a possible implantation error.

Our findings substantiate the ability of OS-DBS to recruit neuronal pathways of distinct orientations relative to the position of the electrode. We extended the concept of OS-DBS to the STN which is a clinically critical area for DBS therapy. The activation maps presented in Fig. 2 exhibit activation in the areas connected to the STN, i.e., included the ipsilateral and contralateral superior colliculus, ipsilateral reticular nucleus and substantia nigra, and parts of the ipsilateral thalamus, caudate putamen, globus pallidus, motor and primary somatosensory cortices. Here, a significant angular orientation dependence on the electric field gradient of the activation maps in different brain regions clearly demonstrate capability of OS-DBS to specifically recruit areas to be stimulated, and thus advances the DBS to higher level of flexibility. Our cluster analysis (Fig. 3) demonstrated that stimulation angle 90° resulted in statistically significant clusters of increased activity in these areas.

Despite the proven fact that the STN is in the key nucleus of the structure and function of the basal ganglia, the main problem of complete characterization of the basal ganglia function still remains. The paucity of its understanding relies on the structural complexity of this region. To qualitatively address the modulation of brain activation as a function of orientation of the electric field gradient, we included TDI maps for visualization the orientation of the fibers in the STN. Since TDI is purely qualitative, for quantification purposes other quantitative methods able to extract orientation information, such as DTI, and histological quantitative analysis, such as Fourier analyses, can be applied, which could be the focus of future studies. Even though the STN connections are well-characterized, still a limited number of studies evaluate the anatomy of those connections exist. The few existing tracer studies (e.g., Coizet et al, (Coizet et al., 2009)) have shown that the orientation of the fibers to be mediolateral which is in agreement with our findings. Finally, in the work by Plantinga et. al (Plantinga et al., 2016) the orientation of the fiber tracks of the STN were investigated in the human brain using diffusion MRI, and it was shown that the orientation of fibers within the STN is mostly in the mediolateral orientation. Also, the majority of fiber tracks connecting the STN to the SNc and SNr are primarily oriented mediolaterally.

The results of the stimulation at different angles shown in Figure 3 and the Supplementary Figure 1 demonstrate significant differences in activation for different angles of stimulation. Therefore, we performed a tensor group ICA analysis of the individual data sets using dual regression with focusing on the comparison between maps obtained at different angles of the stimulation. However, we did not obtain statistically significant results likely due to small sample sizes and/or inter-subject variability.

The present study has several limitations. Major limitation of any fMRI study on anesthetized animals is the effect of anesthesia on brain function which is unavoidable when compared to the awake state (Paasonen et al., 2018; Schlegel et al., 2015; Schroeter et al., 2014). We chose to use urethane is widely used in electrophysiological studies, it has been demonstrated to preserve cardiovascular and respiratory activity (Field et al., 1993; Maggi and Meli, 1986a, b) and maintained resting state functional connectivity as closest to the awake state (Paasonen et al., 2018). However, it has also been shown that anesthesia including urethane can induce spontaneous changes in brain state (Clement et al., 2008; Pagliardini et al., 2013) that can be detected with resting state fMRI (Zhurakovskaya et al., 2016). Furthermore, it has been shown that electrical stimulation of the pedunculopontine tegmental nucleus (PPN) can disrupt the brain state (Clement et al., 2008). Hence, it is conceivable that due to the connectivity between the PPN and STN, STN-DBS could induce a change in the brain state and be observed as unnatural urethane anesthesia related brain activity in response to DBS rather than the desired physiological change in activity. However, the state changes have been shown to be global in the brain (Zhurakovskaya et al., 2016) whereas our results were unilateral with the exclusion of minor group level contralateral activity in the thalamus, superior colliculus and reticular nucleus. Furthermore, ICA analysis finds a strong stimulus-related cortical component (Supplementary Figure 3), and no other cortical components that could be attributed to the action of urethane. In addition, spontaneous urethane related state changes have varying duration (Zhurakovskaya et al., 2016), and brain state alteration after PPN stimulation have been shown to last long after the end of stimulus (Clement et al., 2008) which is opposite to the results presented here, where the fMRI responses started to decrease right after the end of the stimulation periods.

Although our myelin stained histology and high-resolution diffusion imaging of the STN and nearby regions gives an estimate of the main fiber directions, this may not correspond to the fibers where the stimulus should be targeted for the best treatment effect. Thus, more precise microanatomy and connectivity of the STN should be investigated for better understanding of the OS-DBS of the STN. In this study, we also used fairly large electrodes so that current spill outside of the STN is likely. Larger number of smaller contacts would help to better contain the stimulating electric field within the STN and would likely also enable better performance of OS-DBS. Finally, this study also used anesthetized healthy rats, but to truly demonstrate the benefits of OS-DBS, behavioural testing in awake Parkinsonian rat model is needed to differentiate between direct motor cortex activation via internal capsule or indirect modulation of motor cortex by altering STN activity and the signal of its afferents and efferents.

In conclusion, our results demonstrate that in a preclinical setting OS-DBS of the STN can modulate brain network activity relevant to the treatment of movement disorders and can enhance the electrical activation of appropriate neural components of the STN to provide optimal functional stimulation. Although the contacts in human DBS leads are positioned on the shaft of the lead rather than at the tip, the principles of the present results are likely transferable to segmented leads providing the possibility to fine-tune the final stimulation outcome. Future directions of research will include added numbers of contacts for better control of OS-DBS.

Abbreviations:

DBS

deep brain stimulation

OS

orientation selective

fMRI

functional magnetic resonance imaging

MB-SWIFT

multiband SWeep Imaging with Fourier Transformation

ROI

region of interest

STN

subthalamic nucleus

PFA

paraformaldehyde

FSE

fast-spin-echo

FDR

false discovery rate

FEW

family wise error