Changes in Somatosensory Evoked Potentials Elicited by Lateral Cerebellar Nucleus Deep Brain Stimulation in the Naïve Rodent

Abstract

Deep brain stimulation (DBS) of the deep cerebellar nuclei has been shown to enhance perilesional cortical excitability and promote motor rehabilitation in preclinical models of cortical ischemia and is currently being evaluated in patients with chronic, post-stroke deficits. Understanding the effects of cerebellar DBS on contralateral sensorimotor cortex may be key to developing approaches to optimize stimulation delivery and treatment outcomes. Using the naïve rat model, we characterized the effects of DBS of the lateral cerebellar nucleus (LCN) on somatosensory evoked potentials (SSEPs) and evaluated their potential use as a surrogate index of cortical excitability. SSEPs were recorded concurrently with continuous 30 Hz or 100 Hz LCN DBS and compared to the DBS OFF condition. Ratios of SSEP peak to peak amplitude during 100 Hz LCN DBS to DBS OFF at longer latency peaks were significantly greater than 1, suggesting that cortical excitability was enhanced as a result of LCN DBS. Although changes in SSEP peak to peak amplitudes were observed, they were modest in relation to previously reported effects on motor cortical excitability. Overall, our findings suggest that LCN output influences thalamocortical somatosensory pathways, however further work is need to better understand the potential role of SSEPs in optimizing therapy.

Introduction

Despite advances in acute care, stroke remains a major source of long-term disability worldwide [1]. Physical rehabilitation can facilitate significant recovery of motor function but typically faces limitations in the overall extent of that recovery as well as the therapeutic window [2]. Novel neurorestorative approaches to enhance motor rehabilitation beyond physical therapy alone, which typically aim to restore diminished limb function by encouraging functional reorganization of surviving tissue, are promising avenues currently being explored in both clinical and preclinical phases [3, 4]. These approaches generally involve neurostimulation of specific anatomical targets, including perilesional cortex [3, 5–7], the vagus nerve [8, 9], and peripheral nerves of the affected limb [10, 11]. Delivery methods range from noninvasive, externally-applied techniques, such as transcranial magnetic stimulation (TMS) [12], transcranial direct current stimulation (tDCS) [5, 6], and transcranial focused ultrasound stimulation (tFUS) [13], to more invasive implanted devices, such as nerve cuffs [8] and epidural electrodes [7, 14]. A key component of any such therapy, however, is understanding how to titrate stimulus delivery to achieve patient-specific optimization of circuit modulation and therapeutic potential. The absence of acute, behavioral feedback has yielded an emphasis on surrogate markers of efficacy, including electrophysiology-based metrics [11, 15–18].

Given the high degree of connectivity present between each cerebellar hemisphere and the contralateral cerebral cortex [19], we have previously proposed chronic activation of the disynaptic, net excitatory dentatothalamocortical (DTC) pathway as a means of increasing contralateral cortical excitability and, in turn, promoting functional reorganization of surviving, perilesional cortex for patients with chronic post-middle cerebral artery (MCA) stroke motor deficits [20]. The pathway consists of dentate nucleus (DN) neurons that project to the ventroposterolateral par oralis (VPLo), mediodorsal (MD), and ventrolateral pars caudalis (VLc) nuclei of the thalamus, which, in turn, connect to widespread regions of frontal and parietal cortex [21]. Given such widespread connections, the DTC pathway provides a promising avenue for modulating cortical regions spared after stroke and promoting functional reorganization. Notably, our preclinical studies support the use of chronic stimulation of the lateral cerebellar nucleus (LCN), the rodent DN homologue, to enhance motor cortical excitability [22, 23], functional cortical reorganization [24], and motor rehabilitation [24–26] and have formed the foundation of an on-going phase I clinical trial (NCT02835443).

Previous efforts to identify a surrogate marker for DBS programming in chronic stroke models included indexing changes in cortical excitability through motor evoked potentials (MEPs). Data from both naïve and post-stroke animals have shown that, indeed, cortical excitability changes in response to LCN DBS and, more importantly, that the magnitude and duration of such changes were sensitive to pulse frequency [22, 23]. Unfortunately, the comparable, TMS-based approach, that is readily applicable in humans is a time-consuming procedure that is subject to patient fatigue; limiting its ability to serve as a primary surrogate in the face of the multitude of potential combinations of DBS stimulation location, frequency, pulse amplitude and width. The failure of motor evoked potentials as an index of cortical excitability to translate from the rodent model to human application has prompted a renewed search for other potential surrogate markers to facilitate DBS programming. Here, we report our efforts to evaluate the potential for SSEP-based quantification of cortical excitability and its response to LCN DBS in the naïve rodent model. We hypothesize that 30 Hz LCN DBS will enhance excitability of sensory cortex, as indexed by an increase in the magnitude of the SSEP response, while higher frequency LCN stimulation (100Hz) will reduce that response.

Methods

Overview

Animals were surgically instrumented with a sciatic nerve cuff, recording screws over S1 hindlimb cortex, and an electrode stereotactically implanted in the LCN. SSEPs were recorded concurrently with LCN DBS at 30 Hz or 100 Hz, with each trial consisting of data acquisition in the LCN DBS OFF followed by DBS ON state. The effect(s) of LCN DBS on the SSEP waveform was quantified using measures of RMS power over broad latency ranges as well as individual peak-to-peak amplitudes.

Animals

Experiments were performed using 10 male Long Evans rats (250–350g). The animals were housed in an approved animal facility in a climate-controlled environment that included a 12-h light/dark cycle and free access to water. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the Cleveland Clinic.

Prior to collection of SSEPs, all rodents underwent surgery to implant materials required for experimentation (Figure 1). Animals were fitted with a sciatic nerve cuff (Model NC-2-2-100umPt-1-1-SUT-300SS, Microprobes for Life Science, Inc., Gaithersburg, MD, Figure 1A), EEG recording screws bilaterally over S1 hindlimb cortex (Figure 1B), and a concentric stimulating electrode (Model MS306, P1 Technologies, Inc., Roanoke, VA) in the LCN (Figure 1C). EEG screws were prepared by filing down and scoring the tips of self-tapping bone screws (Model 19010–00, Fine Science Tools (USA), Inc., Foster City, CA). Burr holes for EEG screws were marked at coordinates −2.2 and +0.3 mm anterior-posterior (AP) and ±3.0 mm medial-lateral (ML) relative to bregma using a stereotaxic apparatus. The EEG screw coordinates correspond to the posterior and anterior edges of the S1 hindlimb cortex [27]. EEG screws were advanced 2 rotations to place the rounded screw tips in contact with the cortical surface. A reference screw was placed 3 mm anterior to bregma and a ground screw was placed over the cerebellum contralateral to the stimulated hemisphere, as described for the EEG screws but advanced 3 rotations to account for thicker skull. The LCN electrode was inserted at −2.3 mm posterior and −3.6 mm lateral to the interaural point, with the tip lowered to a depth of +3.5 mm dorsal to the interaural point [23, 28–30]. Additional screws were implanted in the skull for structural support. Tunneled connections from all implants were fed to a central head connector secured to the skull using methyl methacrylate. Prophylactic infection (cefazolin (15 mg/kg SQ) and pain management (buprenorphine (0.05 mg/kg SQ) was provided in accordance with the IACUC-approved protocol and a period of at least one week was allowed for recovery before experimentation.

Figure 1:

Illustrations of experimental targets and experimental design. (A) Overview schematic of the hardware implanted into the rat. A concentric, bipolar electrode is implanted in the lateral cerebellar nucleus (LCN) for DBS and a nerve cuff is implanted around the sciatic nerve for generating SSEPs. (B) Stereotaxic coordinates of the EEG screws used to record SSEPs. Figure adapted from [27]. (C) Rat coronal atlas section depicting the rodent LCN, the implantation target for the DBS electrode. Figure adapted from pg. 286 of [30]. (D) Experimental stimulation and recording procedure performed on all rodents. Each trial consists of a DBS OFF SSEP recording, rest, DBS ON period, and DBS ON SSEP recording.

SSEP Set Up

SSEPs were elicited using a Dual Output Square Pulse Stimulator (Model S88, Grass Instruments, Astro-Med, Inc., West Warwick, RI) with two photoelectric stimulus isolation units (Model SIU-7, Grass Instruments Astro-Med, Inc., West Warwick, RI). Stimuli were delivered to the right sciatic nerve of the rat through the implanted nerve cuff. At the beginning of each experiment, sciatic nerve stimulation motor threshold was determined in the awake, freely behaving animal, using biphasic square pulses delivered at a frequency of 2.7 Hz and with a pulse width of 500 us. Pulse amplitude was gradually increased until periodic toe flexion was observed, at which point stimulation current was logged as motor threshold. For the subsequent SSEP measurements, biphasic square pulses were delivered to the sciatic nerve at a frequency of 2.7 Hz, a pulse width of 500 us, and amplitude of 2x motor threshold.

Electrophysiological responses evoked in the S1 hindlimb cortex were recorded using a Tucker Davis Technologies (TDT) RZ5 Bioamp Processor and TDT PZ5 Neurodigitizer Amplifier. Sync pulses generated by the RZ5 were delivered to the S88 to trigger stimulation. Neural signals were sampled at a rate of 24,414 Hz and bandpass filtered between 0.4 and 11 kHz at acquisition.

DBS Setup

A second S88 stimulator and two SIU-7 stimulus isolation units were used for LCN DBS. A DBS motor threshold for each rat was independently determined for pulse rates of 30 Hz and 100 Hz frequencies using 1–5 second trains of biphasic square pulse waveforms with pulse widths of 400 us per phase [26]. A gradual increase in pulse amplitude was implemented until a forelimb response was observed. For a 30 Hz stimulation, a tremor of the paw indicated DBS motor threshold had been reached. If 100 Hz frequency was being tested, a retraction of the wrist indicated DBS motor threshold had been reached. DBS settings for subsequent experimentation had a frequency of 30 Hz or 100 Hz, pulse width of 400 us, and an amplitude that was 80% of the frequency-specific motor threshold.

Experimental Procedure

Within each experimental trial, SSEPs were recorded from the cephalic screws for a period of 5 minutes while DBS remained off, referred to as “DBS OFF”. This was followed by a 5-minute rest period, after which LCN DBS at either 30 or 100 Hz was initiated. Two minutes later, SSEP recordings were repeated for another 5 minutes while LCN DBS continued, referred to as “DBS ON”. The experimental trial design is summarized in Figure 1D.

Data Processing and Analysis

Neurological recordings from SSEP trials were converted from TDT files to MATLAB data structures and analyzed using custom MATLAB scripts. Neural recordings collected during the DBS OFF and each of the two DBS ON conditions were time locked to leg stimulation sync pulses and segmented into 175 ms windows consisting of a 25 ms baseline followed by a 150 ms post-stimulation response. Segments were averaged across each 5-minute sampling period (~810 pulses total) to generate a mean SSEP waveform. DBS ON recordings with 30 Hz or 100 Hz LCN DBS were compared to their preceding DBS OFF period using Root Mean Square (RMS) power and peak-to-peak amplitude.

Prior to RMS calculations in MATLAB, large artifacts were removed from the data by excluding time segments that exceeded a z-score of 100 and re-averaging the remaining time segments within the trial. RMS values were calculated in MATLAB (Eq. 1) over time ranges of 5–15 ms, and 15–100 ms, referred to herein as short and long-latency, respectively. Short and long-latency RMS values were quantified for DBS OFF and DBS ON recordings in both 30 Hz and 100 Hz DBS trials. Ratios of RMS for DBS ON to RMS for DBS OFF were calculated within each trial to assess the effects of LCN DBS on SSEP waveforms. RMS ratios from all subjects and trials were compared against a value of 1 using a one-sample t-test as well as a Wilcoxon signed rank test.

$$\begin{gathered} RMS=\sqrt{\frac{1}{n}\sum _i{x}_i^2} \\ RMSRatio=\frac{RMS(DBSON)}{RMS(DBSOFF)} \end{gathered}$$ Eq. 1

In order to assess features of SSEP waveforms within more specific time ranges, major peaks of the SSEP waveform were labeled across trials and subjects. Peak amplitude and latency were defined as the y coordinate (voltage) and x coordinate (time) of the identified peaks. Peak-to-peak amplitudes were calculated by subtracting the minima of a specific negative peak from the maxima of an adjacent positive peak (reference equation below). The ratio of the peak-to-peak amplitude for DBS ON to peak-to-peak amplitude of DBS OFF was calculated within each trial to assess the effects of LCN DBS on SSEP waveforms. Peak-to-peak amplitude ratios from all trials and subjects were compared against a value of 1 using a one-sample t-test, a Wilcoxon signed rank test, and mixed effects models which accounted for repeated observations.

$$PeaktoPeakAmplitudeRatio=\frac{PositivePeakAmplitud{e}_{DBSON}-NegativePeakAmplitud{e}_{DBSON}}{PositivePeakAmplitud{e}_{DBSOFF}-NegativePeakAmplitud{e}_{DBSOFF}}$$ Eq. 2

Statistical Analysis

Statistical comparisons first utilized a Wilcoxon signed rank test, with significant finding further interrogated using generalized linear mixed effect models (GLMM). GLMM are flexible extensions of linear regression which can accommodate and account for the error arising from clustering of repeated measurements within session, within animal. The models also adjust for the unbalanced data that arises when animals do not contribute the same number of data points to the analysis. Models tested the interaction between peak to peak amplitude ratio and frequency. Animals were treated as random factors and a compound symmetry covariance structure was assumed.

Results

Across the 10 animals enrolled, 38 trials of 30 Hz “DBS ON” vs. “DBS OFF” and 27 trials of 100 Hz “DBS ON” vs. “DBS OFF” were recorded. Individual animals contributed different numbers of trials for each of 30 Hz and 100 Hz stimulation (Table 1) due to hardware failures. All 10 rodents participated in 30 Hz trials while only 8 of the 10 rodents were available for the 100 Hz trials.

Table 1:

Distribution of performed trials by each rodent in the study. All rodents performed trials of 30 Hz DBS. 8 out of the 10 rodents performed 100 Hz trials. Standardization of the number of performed trials did not occur for both frequencies. 38 trials of 30 Hz and 27 trials of 100 Hz were collected.

Subject Number	Number of 30 Hz Trials	Number of 100 Hz Trials
1	1	1
2	1
3	5	1
4	1	2
5	3	3
6	6
7	6	5
8	5	5
9	5	5
10	5	5
Total:	38	27

SSEP waveforms recorded from the S1 hindlimb cortex began with a large, short-duration stimulus artifact followed by distinct positive and negative long-latency components ~25 ms and ~45 ms, post-stimulation. Although all recordings shared identifiable peaks at these times, additional peaks were present across other recording channels, with variable peak amplitudes and latencies observed across animals (Figure 2). For the current study, emphasis was placed on recordings derived from the screw placed over S1 hindlimb representation contralateral to the stimulated sciatic nerve as this signal was most robust across animals.

Figure 2:

SSEP waveforms vary by recording site. Sample waveforms recorded from all EEG recording screws over the S1 hindlimb cortex are depicted above. All waveforms have a distinct stimulus artifact starting at time 0 ms. Waveforms across all channels have an identifiable positive and negative long-latency peak displayed at ~25 ms and ~45 ms. Although all recordings shared identifiable peaks at these times, additional peaks, amplitude, and latency varied across each recoding channel. Central figure adapted from [27].

The SSEP waveform displayed both short- and long-latency peaks. Short-latency peaks were operationally-defined as those occurring between 5–15 ms while long-latency peaks were observed between 15–100 ms. Positive peaks, on average (mean ± sample standard deviation), occurred at 9.67 ± 0.48 ms (P1A), 12.82 ± 0.59 ms (P1), and 21.10 ± 1.74 ms (P2) in DBS OFF control SSEP waveforms. Negative peaks, on average, occurred at 11.77 ± 0.52 ms (N1A), 15.95 ± 1.07 ms (N1) and 36.02 ± 2.62 ms (N2) in DBS OFF control SSEP waveforms. These average times and amplitudes peaks varied across subjects and trials (Figure 3).

Figure 3:

Overview of sciatic SSEP features and labeling of peaks. The top figure illustrates an overlay of all SSEP waveforms recorded over contralateral S1 hindlimb cortex in the DBS OFF state from the pre-stimulation baseline to 150 ms post stimulation. The graph labels stimulus artifact, short-latency range, and long-latency range. The bottom image shows a magnified view of the clustered waveforms from the short-latency range. Peaks were labeled an evaluated from P1A through N2 as shown.

A box and whisker plot comparing RMS values calculated over short and long-latency ranges for 30 Hz and 100 Hz frequency trials is presented in Figure 4. During 30 Hz LCN DBS, the median RMS values were increased by 3.4% in the short-latency range and decreased by 2.85% in the long-latency range without significant changes in DBS ON/OFF RMS ratio from a value of 1 (p > 0.05, Wilcoxon). During 100 Hz LCN DBS, the median RMS values were increased by 12.2% in the short-latency range and increased by 0.04% in the long-latency range without significant changes in DBS ON/OFF RMS ratio from a value of 1 (p > 0.05, Wilcoxon). The absence of significant changes in DBS ON/OFF RMS ratios from a value of 1 signify that neither LCN DBS frequency significantly modulated the overall SSEP waveform. Further analysis of the individual peak-to-peak amplitude ratios are summarized in Figure 5. Of the 5 peak-to-peak amplitudes examined during 30 Hz LCN DBS, only the increase in P2-N1 amplitude of 4.4% relative to DBS OFF was found to be significant (p = 0.044, GLMM). For the 100 Hz condition, two peak ratios were found to change significantly between conditions, with the amplitudes of both P2-N1 (p = 0.018, GLMM) and P2-N2 (p = 0.003, GLMM) enhanced during the 100 Hz DBS ON condition. During 100 Hz LCN DBS, the median peak-to-peak amplitude of the P2-N1 component was increased by 32.5% and the median peak-to-peak amplitude of the P2-N2 component was increased by 4.2%. Peak-to-peak amplitude analysis of ipsilateral responses on channel 2 did not result in any significant changes in 30 Hz or 100 Hz DBS ON/OFF ratios from 1.

Figure 4:

Short- and long-latency RMS ratios summarized in a box and whisker plot. The ends of the box and whisker plots display the upper (75th) and lower percentiles (25th). The box spans the interquartile range (50th). The median is marked by a horizontal line within the interquartile range. The whiskers extend to the highest and lowest points observation. Circles outside of the box indicate outliers. Circles or positive signs within the boxes indicate mean. All analyzed data depicted was from channel 1 and 38 trials of 30 Hz and 27 trials of 100 Hz were included in analysis for both latencies. All RMS ratios were compared to a value of one and yielded insignificant (p value > 0.05, Wilcoxon) differences for both short and long latencies.

Figure 5:

Peak to peak amplitude ratios summarized in box and whisker plots. The ends of the box and whisker plots display the upper (75th) and lower percentiles (25th). The box spans the interquartile range (50th). The median is marked by a horizontal line within the interquartile range. The whiskers extend to the highest and lowest points observation. Circles outside of the box indicate outliers. Circles or positive signs within the boxes indicate mean. Data analyzed from channel 1. 38 trials of 30 Hz and 27 trials of 100 Hz were included for analysis of individual peak ratios. Ratios of the peak-to-peak amplitudes for DBS ON to peak-to-peak amplitude of DBS OFF were calculated and compared to 1. 30 Hz trials resulted in one peak ratio of significance. P2-N1, denoted by the “#” symbol, was significantly greater than 1 (p = 0.044, GLMM). Remaining peak-to-peak ratios had no significance (p> 0.05, GLMM). 100 Hz trials resulted in two peaks of significance. P2-N1 (p = 0.018, GLMM), denoted by the “$” symbol, and P2-N2 (p = 0.003, GLMM), denoted by the “*” symbol, had ratios statistically significant from 1. Remaining peak- to-peak ratios had no significance (p> 0.05, GLMM).

Discussion:

Cortical excitability indexes are a useful tool for programming neurorestorative therapies, such as dentate nucleus DBS. Here, we sought to characterize the effect of DTC pathway stimulation on sensorimotor cortical excitability using sciatic nerve SSEPs. In addition to enhancing our overall understanding of the effects of DTC stimulation on cortical excitability, our goal was to evaluate the potential use of this approach in guiding patient-specific therapeutic programming. Overall, our data support a modulatory effect of DTC stimulation on sensorimotor cortical excitability, as indexed by peak-to-peak amplitude changes in the SSEP response, and, in turn, interaction across the somatosensory and DTC pathways. Moreover, in contrast to our original hypothesis, both stimulation frequencies were observed to enhance one or more peak amplitudes of the long -latency SSEP response.

Although our findings in awake rodents contradict the frequency-dependent effects of LCN DBS on excitability demonstrated by MEPs recorded in rodents under propofol anesthesia [22], our findings coincide with the effects observed during MEPs recorded in rodents under ketamine anesthesia [23, 31]. The influence of anesthesia on cortical excitability and the DTC pathway, in addition to differences between MEPs and SSEPs, may account for differences in outcomes. Further, the absence of significant changes in RMS, despite changes in peak-to-peak amplitude, suggest that quantifying RMS of broad time ranges may not be sufficiently sensitive to detect amplitude changes occurring over short periods of time. Notably, however, the limited magnitude of the amplitude changes suggests that, as an index of sensorimotor cortical excitability, SSEPs may not be sufficiently sensitive for use in patient-specific DBS programming. Finally, the variability and occurrence of outliers in the peak to peak amplitude values we collected in awake rodents is a major limitation of our study, potentially attributable to the awake state of the animal, however the statistical model applied accommodates outliers without those data points having a disproportionate effect on the analysis.

Previous investigations of the effects cerebellar stimulation on SSEPs focused on the cerebellar cortex [15, 50]. Stimulation of the cerebellar cortex resulted in decreased SSEP amplitude in the cortex in both instances, as well as in VPL nucleus of the thalamus in the study by Crispino et al. Our results are consistent with the inhibitory effects of cerebellar cortex on the DN. Other studies examined the effects of cerebellum lesions on SSEPs [32, 47]. Schieppati et al. observed similar SSEP latencies and peak-to-peak amplitudes before and after complete removal of the cat cerebellum, aside from one instance in which an additional SSEP peak was present [32]. Meanwhile, Kolodziejak et al. observed increased short-latency SSEP amplitudes in cats with cerebellar cortex lesions but no effects on short-latency SSEP features with removal of both cerebellar cortex and deep cerebellar nuclei [33]. Long-latency SSEP features were reduced with either cerebellar cortical or combined cerebellar cortical and nuclear lesions. Integrating the previous studies with our data, the DN appears to have an excitatory effect on SSEPs whereas activation of cerebellar cortex generally has an inhibitory effect.

Although the overall magnitude of SSEP changes was limited, our findings suggest a potential interaction of the somatosensory and DTC pathways. There are several anatomical locations where these pathways may intersect (Figure 6). Both SSEPs and the DTC have synapses in the thalamic nuclei and subsequent synapses in the cortex. Despite synapses in the adjacent VPLo and VPLc, autoradiographic labeling and horseradish peroxide retrograde labeling by Tracey et al. demonstrated that pathways from the DN to VPLo to motor cortex and pathways from dorsal column nuclei to VPLc to somatosensory cortex do not intersect at the level of the thalamus [34]. Thus, direct innervation of dentatothalamic cells to SSEP thalamic neurons is unlikely, but less direct thalamic pathways, including via more medial nuclei, cannot be excluded. Beyond the thalamus, the various thalamic neurons of the DTC pathway proceed to innervate the primary motor cortex, supplementary motor area, premotor cortex, prefrontal, and posterior parietal cortex [20, 21, 34–36], whereas VPLc thalamic neurons of the SSEP pathway predominantly project to the S1 somatosensory cortex. Connections from cerebellar nuclei to S1 cortex via thalamus have been demonstrated via axonal tracing, but dentate-specific tracing was not reported [37]. It is possible that the DTC pathway is activating the S1 cortical cells directly involved in the SSEP pathway. Even without direct S1 activation by the DTC, additional cortical-cortical interactions may contribute to changes in excitability. Further, orthodromic depolarization during sciatic nerve stimulation produces muscle activity and activation of proprioceptors, resulting in cortical activation. A contribution of potential interactions between the DTC pathway and the traditional SSEP pathway outside of ascending sensory pathways cannot be ruled out from the current data.

Figure 6:

Possible interactions between the ascending sensory and DTC pathways. Solid arrows depict better defined interactions and dashed lines depict controversial or non-confirmed interactions. Red arrows portray the main pathway of the sciatic nerve SSEP. Following electrical stimulation of the sciatic nerve, action potentials travel up the dorsal column of the spinal cord, synapse in nucleus gracilis, proceed to the thalamus, synapse in the VPLc nucleus, proceed to cortex, and synapse in primary somatosensory cortex. Other neural generators of cortical SSEPs have been suspected in S2, M1, and frontal lobe (not depicted). The spinocerebellar tract is also thought to contribute to SSEP transmission. DBS of the LCN activates cells which project to and synapse in multiple thalamic nuclei (notably MD, VLc, VPLo, and area X). The thalamic nuclei primarily project M1, but there is evidence for additional connections to pre-frontal, pre-motor, and posterior parietal cortex. Direct connections between the DTC pathway and S1 have been suspected but not explicitly demonstrated. Additional cortical-cortical synapses may be involved in DTC/ascending sensory communications. DTC activation has also been shown to send descending signals via the reticulospinal tract.

The effects of LCN DBS were limited to peak-specific changes in SSEP waveforms, so understanding the probable sources of each would help elucidate the DTC-ascending sensory interactions. SSEP waveforms recorded at different positions along the ascending sensory pathways provide insights about nearby spinal, brainstem, and cortical structures. Despite extensive characterization of cortically recorded SSEP peaks following human median nerve [38–40] and tibial nerve [41, 42] stimulation, the neural generators of rodent SSEPs are not well understood. Several groups have measured sciatic nerve SSEPs from the cortex of rats, with varying naming systems, peak latencies, numbers of peaks observed, and relative amplitudes of peaks [7, 16, 20]. Similar waveforms to those observed in this study were presented by Hurlbert and colleagues, including peaks at 9, 12, 14, 19, and 28 ms [20], which roughly correspond to N1a, P1, N1, P2, and N2, with slightly shorter latencies and the absence of a counterpart to P1a where the signal appears to be obscured by effects of the stimulus artifact. The authors of that study hypothesized that the P9-N12-P14-N19 complex they observed in rats corresponds to the P28-N31-P37-N45 complex observed after stimulating posterior tibial nerve in humans. Assuming our N1 and P2 are derived from the same neural generators as P14 and N19 of Hurlbert et al and the equivalent human generators of P37-N45 in tibial SSEPs [22], our N1 and P2 likely originate in primary somatosensory cortex. Following similar logic, the features in our SSEP waveforms occurring before N1 would derive from subcortical structures, similar to the P30 and N33 tibial SSEP peaks measured at the scalp, which have been attributed to activity near the dorsal column nucleus and brainstem activity below thalamus, respectively [45]. Our N2 peak would be subject to the same ambiguity of origin as long-latency peaks observed after human tibial and median nerve stimulation [3, 4], but likely involve additional cortical synapses based on the longer latency. The narrow scope of LCN DBS effects to SSEP peaks associated with cortex corroborate our speculation of DTC-SSEP interactions occurring at the level of the cortex.

Despite exhibiting modest changes in response to LCN DBS, SSEPs have been utilized as a metric of cortical excitability in numerous clinical studies, including investigations of TMS [43–45], transcranial static magnetic field stimulation [46], peripheral nerve stimulation [47], repetitive TMS [48], sensory gating during voluntary activity [49, 50], and paired associative electroacupuncture with TMS [51]. Although some studies exhibited normalized peak to peak amplitude ratios as high as approximately 2.25x baseline [43] or as low as approximately 0.4x baseline [50], it is not uncommon for equivalent changes to 1.1–1.25x baseline [44, 47] or 0.8–0.9x baseline [46, 49, 50] to be regarded as significant changes. The effects of LCN DBS may be within the sensitivity limits of SSEPs on a population level but may not be distinct and consistent enough to assess favorable DBS parameters in an individual patient level.

Overall, this study offered insight into the reliability of SSEPs for detecting enhanced cortical activity in relation to LCN DBS. Although the group level findings showed significant differences, the magnitude of the effect of LCN DBS on SSEP characteristics was limited in this preclinical model and may not be the ideal modality for indexing DBS settings. Compared to our previous MEP findings, SSEPs exhibited smaller peak-to-peak amplitude ratios. MEPs had a ratio from ~1–3.5x in Baker et al. 2010, while SSEP peak-to-peak amplitude ratios in this study were ~1–1.25x [22]. MEPs, however, are an imperfect index for a clinical trial as they localize to a single muscle group, defined operationally, that provide a consistent response at the lowest amplitude. Ultimately, this localization is time consuming, particularly in patients with moderate-to-severe deficits. In the future, paired-pulse SSEPs should be considered as it is an additional system used to assess complex synaptic responses. Paired-pulse stimulation may be more sensitive to cortical excitability as it includes an additional intrinsic control by comparing the ratio of two spaced SSEP pulses [52]. Finally, future studies should consider generating SSEPs via stimulation of the median nerve as a better correlate for recovery of post-stroke upper limb deficits encouraged by dentate nucleus stimulation.