Neurophysiological mechanisms of deep brain stimulation across spatiotemporal resolutions

Abstract

Deep brain stimulation is a neuromodulatory treatment for managing the symptoms of Parkinson's disease and other neurological and psychiatric disorders. Electrodes are chronically implanted in disease-relevant brain regions and pulsatile electrical stimulation delivery is intended to restore neurocircuit function. However, the widespread interest in the application and expansion of this clinical therapy has preceded an overarching understanding of the neurocircuit alterations invoked by deep brain stimulation. Over the years, various forms of neurophysiological evidence have emerged which demonstrate changes to brain activity across spatiotemporal resolutions; from single neuron, to local field potential, to brain-wide cortical network effects. Though fruitful, such studies have often led to debate about a singular putative mechanism. In this Update we aim to produce an integrative account of complementary instead of mutually exclusive neurophysiological effects to derive a generalizable concept of the mechanisms of deep brain stimulation. In particular, we offer a critical review of the most common historical competing theories, an updated discussion on recent literature from animal and human neurophysiological studies, and a synthesis of synaptic and network effects of deep brain stimulation across scales of observation, including micro-, meso- and macroscale circuit alterations.

Keywords: deep brain stimulation; mechanisms of action; neuromodulation; oscillations; singe neurons.

Figure 1

Circuit engagements associated with subthalamic deep brain stimulation (STN-DBS). DBS refers to pulsatile electric current applied through an implanted electrode. Typically, the contact expected to yield best clinical effects is set as the cathode, with the anode either set to the case of the implantable pulse generator (monopolar setting) or to another contact of the same electrode (bipolar stimulation). (A) In movement disorders, DBS is usually only effective when applied at high frequencies, with efficacy gradually increasing from 60 Hz up to or beyond 130–180 Hz. Typical pulse widths range from 30–200 µs. The pulse shape is rectangular and pseudo-monophasic with an initial square at therapeutic amplitudes of ∼1–5 V or 0.7–4 mA, followed by a passive charge balancing pulse with a several-fold longer pulse duration and lower amplitude. The total electrical energy delivered (TEED) is a function of the stimulation pulse width, amplitude, frequency and the impedance between the electrode contact and brain tissue. These parameters have significant influences on DBS effects by governing the extent of neural tissue engagement. (B) The stimulation activates neuronal structures, producing local and distant effects, which can include ‘orthodromic’ (downstream) and ‘antidromic’ (upstream) activation of afferent and efferent neuron projections, as well as ‘invasion’ of axonal architectures (i.e. activation of axonal branches/collaterals). The schematic takes into consideration common inputs and outputs of the STN and their collateral axonal branches to demonstrate the widespread effects of DBS. (C) Each plot shows the peristimulus spike firing histogram of engaged brain structures, demonstrating that STN-DBS can produce differential responses upstream in cortical regions (adapted from Johnson et al.), as well as mixed responses in downstream structures such as globus pallidus internus (GPi) and externus (GPe) (adapted from Hashimoto et al.). Single neuron traces of cortical (red), GPe (blue) and STN (yellow) are also depicted from three anatomical non-human primate tracing studies.

Figure 2

Microscale effects of DBS. The local neuronal effects of deep brain stimulation (DBS) have been proposed to be governed by activation of afferent inputs. (A) In the thalamic ventral intermediate nucleus (Vim) the vast majority of inputs are glutamatergic from the dentatorubrothalamic tract, with some GABAergic afferents from the thalamic reticular nucleus. In the subthalamic nucleus (STN), there are similar proportions of GABAergic (from globus pallidus externus, GPe) and glutamatergic (from cortex) inputs, somewhat favouring inhibitory inputs in number. In both globus pallidus internus (GPi) and substantia nigra parts reticulata (SNr), GABAergic projections (mainly from striatum and to a lesser degree from GPe) outweigh glutamatergic projections from STN. Pie charts depict estimated distributions of excitatory (E) versus inhibitory (I) afferent inputs. (B) The balance of GABAergic versus glutamatergic afferent fibres activated defines the net response to single pulses of stimulation, producing strong excitations in Vim, weak net inhibitory responses in STN, and strong inhibitions in GPi and SNr (adapted from Milosevic et al.^,,). (C) In Vim, high-frequency stimulation (HFS) produces an initial increase of firing, followed by a suppression of firing, which is stronger when higher stimulation frequencies are applied; hypothesized to be the result of synaptic depression (adapted from Milosevic et al.; dashed time series is from n = 1 previously unpublished data demonstrating the persistence of neuronal suppression over time). HFS produces sustained inhibition of neuronal firing in STN due to persistent activation of GABAergic inputs, but incomplete neuronal suppression in SNr and GPi (n = 1; previously unpublished data depicted by dashed time series) where a return of neuronal firing can be observed over several seconds due to the depression of GABAergic inputs from striatum (STN and SNr figures adapted from Steiner et al.). SNc = substantia nigra pars compacta.

Figure 3

Mesoscale effects of DBS. Deep brain stimulation (DBS) can simultaneously suppress pathological oscillatory beta (13–35 Hz) activity in Parkinson’s disease and give rise to high-frequency evoked resonant neural activity (ERNA). (A) Time-frequency plot demonstrates changes to subthalamic beta oscillatory power with increasing DBS stimulation amplitude in an LFP recording from a parkinsonian patient OFF medication (adapted from Feldmann et al.). (B) The ERNA waveform is observable (i) after each high-frequency pulse (blue downward arrows in waveform panel during stimulation; adapted from Sinclair et al.) with decaying resonance even after stimulation is switched off (green downward arrows after stimulation). In tandem with the suppression of beta activity at higher stimulation intensities, the amplitude of ERNA increases (adapted from Sinclair et al.). The schematics below demonstrate intraoperative ERNA waveforms and peristimulus spike firing histograms in both (ii) subthalamic nucleus (STN) and (iii) globus pallidus internus (GPi) during 100 Hz stimulation, as well as hypothesized circuit activation profiles that would explain the emergence of ERNA (adapted from Steiner et al.). In STN, each stimulus would produce a net inhibitory response in STN, as well as concurrent excitation of globus pallidus externus (GPe), resulting in feedback inhibition in STN. The same is hypothesized for GPi ERNA via invasion/activation of collateral projections and axons of passage of the reciprocal STN-GPe connectivity. Thus, each of the ERNA waveform peaks is likely a substrate of inhibitory input via GPe. An important note is that the spike firing patterns in STN are only achieved when using subthreshold stimulation amplitudes, which do not cause complete suppression of neuronal firing. High-frequency stimulation (HFS) at clinically relevant intensities would result in the complete suppression of firing and the elicitation of large amplitude ERNA waveforms. In GPi, patterned firing seems to manifest when spike firing re-emerges (as shown in Fig. 2) after depression of striatal GABAergic inputs, likely unmasking inhibitory-excitatory GPe-STN competition, which could give rise to the mixed response as depicted in Fig. 1C. This phenomenon is currently under further investigation by the authors. The inset in the top right summarizes that stimulation intensity differentially modulates beta versus ERNA amplitudes, and furthermore emphasizes an additional potential use of ERNA as a closed-loop marker during NREM sleep, during which beta oscillations are attenuated.

Figure 4

Macroscale effects of DBS. In addition to neurophysiological circuit alterations at the local level (i.e. subcortical functional changes at target structures), deep brain stimulation (DBS) has an influence upon whole-brain interregional networks. The modulation of a single node of a circuit will likely lead to state changes in all circuit connections. This could, in part, be achieved by local axonal activations (bottom left), as evidenced by (A) short-latency stimulation-induced evoked potentials in the cortex^,; wherein the strength of network engagement (evoked potential amplitude) may be reflective of the achieved anti-parkinsonian efficacy with therapeutic settings (adapted from Bahners et al.). Moreover, in addition to the suppression of subcortical beta oscillations, (B) DBS can suppress beta power across the sensorimotor cortical network (adapted from Abbasi et al.), as well as (C) attenuate cortico-cortical beta-gamma coupling (adapted from De Hemptinne et al.), which is otherwise pathologically elevated in Parkinson’s disease. Despite such findings, direct antidromic activation of cortical circuitry is suggested not be able to corroborate the clinical efficacy of GPi-DBS. (D) As such, the spatial overlap in functional connectivity, which co-localizes the predictive efficacy of STN- and GPi-DBS to the M1 region may be representative of a slower timescale (functional MRI-based) functional readout of a downstream basal-ganglia-thalamo-cortical network effect, which could be generated by a convergent GPe-mediated subcortical network activation profile achieved by STN- and GPi-DBS (i.e. ERNA). GPe = globus pallidus externus; GPi = globus pallidus internus; STN = subthalamic nucleus.