Epidural and transcutaneous spinal cord stimulation facilitates descending inputs to upper-limb motoneurons in monkeys

Abstract

Objective:
There is renewed interest in epidural and transcutaneous spinal cord stimulation (SCS) as a therapy following spinal cord injury, both to reanimate paralysed muscles as well as to potentiate weakened volitional control of movements. However, most work to date has focussed on lumbar SCS for restoration of locomotor function. Therefore, we examined upper-limb muscle responses and modulation of supraspinal-evoked movements by different frequencies of cervical SCS delivered to various epidural and transcutaneous sites in anaesthetized, neurologically intact monkeys.

Approach:
Epidural SCS was delivered via a novel multielectrode cuff placed around both dorsal and ventral surfaces of the cervical spinal cord, while transcutaneous SCS was delivered using a high carrier frequency through surface electrodes.

Main results:
Ventral epidural SCS elicited robust movements at lower current intensities than dorsal sites, with evoked motor unit potentials that reliably followed even high-frequency trains. By contrast, the muscle responses to dorsal SCS required higher current intensities and were attenuated throughout the train. However, dorsal epidural SCS and, to a lesser extent, transcutaneous SCS were effective at facilitating supraspinal-evoked responses, especially at intermediate stimulation frequencies. The time- and frequency-dependence of dorsal SCS effects could be explained by a simple model in which transynaptic excitation of motoneurons was gated by prior stimuli through presynaptic mechanisms.

Significance:
Our results suggest that multicontact electrodes allowing access to both dorsal and ventral epidural sites may be beneficial for combined therapeutic purposes, and that the interaction of direct, synaptic and presynaptic effects should be considered when optimising SCS-assisted rehabilitation.

Full text

Journal of Neural Engineering
PAPER • OPEN ACCESS
Epidural and transcutaneous spinal cord stimulation facilitates
descending inputs to upper-limb motoneurons in monkeys
To cite this article: Thomas Guiho et al 2021 J. Neural Eng. 18 046011
View the article online for updates and enhancements.
This content was downloaded from IP address 89.185.78.149 on 12/04/2021 at 18:25

J. Neural Eng. 18 (2021) 046011 https://doi.org/10.1088/1741-2552/abe358
Journal of Neural Engineering
OPEN ACCESS
RECEIVED
4 December 2020
REVISED
19 January 2021
ACC EPT ED FOR PUB LICATI ON
4 February 2021
PUBLISHED
22 March 2021
Original content from
this work may be used
under the terms of the
Creative Commons
Attribution 4.0 licence.
Any further distribution
of this work must
maintain attribution to
the author(s) and the title
of the work, journal
citation and DOI.
PAPER
Epidural and transcutaneous spinal cord stimulation facilitates
descending inputs to upper-limb motoneurons in monkeys
Thomas Guiho, Stuart N Bakerand Andrew Jackson∗
Biosciences Institute, Faculty of Medical Sciences, Newcastle University, Newcastle NE2 4HH, United Kingdom
∗
Author to whom any correspondence should be addressed.
E-mail: andrew.jackson@ncl.ac.uk
Keywords: spinal cord injury, non-human primates, spinal networks, spinal cord stimulation, upper-limb movements
Abstract
Objective. There is renewed interest in epidural and transcutaneous spinal cord stimulation (SCS)
as a therapy following spinal cord injury, both to reanimate paralyzed muscles as well as to
potentiate weakened volitional control of movements. However, most work to date has focussed on
lumbar SCS for restoration of locomotor function. Therefore, we examined upper-limb muscle
responses and modulation of supraspinal-evoked movements by different frequencies of cervical
SCS delivered to various epidural and transcutaneous sites in anaesthetized, neurologically intact
monkeys. Approach. Epidural SCS was delivered via a novel multielectrode cuff placed around both
dorsal and ventral surfaces of the cervical spinal cord, while transcutaneous SCS was delivered
using a high carrier frequency through surface electrodes. Main results. Ventral epidural SCS
elicited robust movements at lower current intensities than dorsal sites, with evoked motor unit
potentials that reliably followed even high-frequency trains. By contrast, the muscle responses to
dorsal SCS required higher current intensities and were attenuated throughout the train. However,
dorsal epidural SCS and, to a lesser extent, transcutaneous SCS were effective at facilitating
supraspinal-evoked responses, especially at intermediate stimulation frequencies. The time- and
frequency-dependence of dorsal SCS effects could be explained by a simple model in which
transynaptic excitation of motoneurons was gated by prior stimuli through presynaptic
mechanisms. Significance. Our results suggest that multicontact electrodes allowing access to both
dorsal and ventral epidural sites may be beneficial for combined therapeutic purposes, and that the
interaction of direct, synaptic and presynaptic effects should be considered when optimising
SCS-assisted rehabilitation.
1. Introduction
Spinal cord injuries (SCIs) have disastrous con-
sequences for patients who face major visceral and
motor impairments (Snoek et al 2004). Recent years
have seen considerable progress in the use of spinal
cord stimulation (SCS) to restore locomotor func-
tion. This was initially conceived as an alternat-
ive to direct functional electrical stimulation (FES)
of muscles (Taylor, 2002, Guiraud, 2006) to engage
surviving central pattern generator (CPG) circuitry
deprived of supraspinal control (Dimitrijevic et al
1998, Minassian et al 2004). However, clinical trials of
epidural lumbar SCS additionally revealed unexpec-
ted improvements in volitional control of the lower
limbs, for example the ability to voluntarily lift the
legs while lying down, even for people with injuries
classified as clinically complete (Harkema et al 2011,
Angeli et al 2018). This suggests that many clinically-
complete lesions may nevertheless be anatomically
incomplete, and that SCS can raise the excitability
of spinal circuits to unmask weakened but surviving
descending pathways (Minassian et al 2016). In con-
junction with extensive rehabilitation, these pathways
may be strengthened to further support restoration
of function (Van Den Brand et al 2012, McPherson
et al 2015, Krucoff et al 2016, Duffell and Donaldson
2020).
If one therapeutic mechanism of SCS is rais-
ing spinal excitability rather than directly driving
CPGs, this technique could also be effective for res-
toration of upper-limb function, identified as a top
© 2021 The Author(s). Published by IOP Publishing Ltd

J. Neural Eng. 18 (2021) 046011 T Guiho et al
priority by patients with quadraplegia (Anderson
2004). Control of the upper-limb in humans and non-
human primates relies heavily on corticospinal neur-
ons, many of which synapse directly onto cervical
motoneurons via the pyramidal tracts (Lemon 2008).
In principle, the appropriate pattern of cervical SCS
could raise motoneurons closer to threshold enabling
their activation by weakened descending pathways.
Although previous studies in non-human primates
have examined the ability of cervical SCS to drive
muscles directly (Moritz et al 2007, Zimmermann
et al 2011, Sharpe and Jackson 2014, Greiner et al
2021, Kato et al 2020), little is known about the
interaction between SCS and descending motor com-
mands. It is not clear whether cervical SCS proto-
cols (electrode locations, stimulus frequency, pattern
etc) optimized for transiently activating muscles dir-
ectly will necessarily prove best for providing sus-
tained elevation of spinal excitability. For example,
Sharpe and Jackson (2014) demonstrated that while
stimulation of the ventral surface of the spinal cord
was effective at driving upper-limb muscle responses,
it was less effective than dorsal surface stimulation at
facilitating the response to a subsequent intraspinal
stimulus. The primary aim of our study was there-
fore to examine the interaction between trains of
epidural SCS, delivered at different frequencies to
various sites around the cervical spinal cord, and
supraspinal inputs, elicited by stimulating the motor
cortex or pyramidal tract. To this end we used a novel
epidural cuff electrode that allowed current to be
delivered through eight contacts located around the
circumference of the spinal cord, providing a means
to stimulate both dorsal and ventral aspects.
Recently, transcutaneous SCS has emerged as
a non-invasive method for therapeutic stimulation
(Megía García et al 2020). Often this is delivered
using the ‘Russian Current’ method, in which low
stimulation frequencies modulate a higher carrier fre-
quency that is thought to reduce the pain associated
with high stimulation intensities (Ward 2009). While
transcutaneous SCS has shown promising benefits
when delivered at both lumbar (Gerasimenko et al
2015, Hofstoetter et al 2015) and cervical (Gad et al
2018, Inanici et al 2018, Benavides et al 2020) levels,
the mechanism of action is unclear. A secondary aim
of our study was therefore to examine the facilitatory
effects of transcutaneous SCS on supraspinal inputs
to the spinal cord and compare this with epidural
SCS.
2. Methods
2.1. Surgical methods
Experiments were performed on five anesthetized,
neurologically intact monkeys (monkey Si—7 years,
6.6 kg; Un—5 years, 5.9 kg, Uk—7 years, 8.8 kg, Yi—
5 years, 7.5 kg and Yu—4 years, 5.3 kg) under appro-
priate UK Home Office licenses in accordance with
the Animals (Scientific Procedures) Act 1986, and
with the approval of the Animal Welfare and Ethical
Review Board of Newcastle University. We report data
from terminal experiments involving epidural SCS,
performed under an anaesthetic regime involving
intravenous infusion of ketamine (6 mg kg−1h−1),
alfentanil (0.2–0.3 µg kg−1min−1) and midazolam
(0.14 mg kg−1hr−1). We have used this regime pre-
viously (e.g. Sharpe and Jackson 2014), since ketam-
ine does not depress spinal excitability to the same
extent as volatile agents (Kendig 2002). Temperat-
ure, heart rate, saturation, blood pressure and end-
tidal CO2were monitored throughout and at the
end of the terminal procedure animals were perfused
transcardially with buffer followed by formalin fixat-
ive. In addition we report data from recovery seda-
tion sessions involving transcutaneous SCS in Mon-
keys Uk and Yu. Sedation was maintained with an
intravenous infusion of ketamine (6.5 mg kg−1h−1),
midazolam (0.33 mg kg−1h−1) and medetomidine
(0.001 mg kg −1h−1) while temperature, heart rate,
saturation and blood pressure were monitored.
2.1.1. Epidural SCS
Epidural SCS was delivered using an eight-contact
electrode array placed around the circumference of
the cervical spinal cord (AirRay Fetz Spinal Cord 8,
Cortec Germany; 0.3 mm diameter platinum elec-
trodes, 2.5 mm pitch on a 27 ×2.5 mm silicone
substrate). Under initial sevoflurane anaesthesia, a
laminectomy was made to expose the cervical enlarge-
ment. A pliable guide-rod was used to tunnel a cath-
eter under the spinal cord at the C7 level. Sutures
passed through the catheter were used to pull the
spinal electrode underneath the cord and tie it snugly
around the dura (figure 1(a)).
2.1.2. Electromyogram recording
Pairs of Teflon-insulated stainless steel wires were
inserted percutaneously to record evoked potentials
from 12 muscles (bilaterally: 1DI, first dorsal inter-
osseous; APB, abductor pollicis brevis; ECR, extensor
carpi radialis; FCU, flexor carpi ulnaris; BB, biceps
brachii; TB, triceps brachii). Electromyogram (EMG)
signals were amplified with a gain of 1000, band pass
filtered between 100 Hz and 1 kHz (Model 1700, A-M
Systems, Carlsborg, US) and sampled at 5 kHz (Micro
1401, CED Cambridge, UK).
2.1.3. Supraspinal stimulation
Intracortical microstimulation (ICMS) was delivered
through tungsten microelectrodes (Microprobes,
US) inserted through a craniotomy over the hand
area of primary motor cortex, positioned to elicit
low threshold responses in contralateral arm/hand
muscles. In two animals (monkey Yi and Yu) we also
positioned stimulating electrodes in the pyramidal
tract (stereotaxic coordinates A2.0, L1.5 and P3.0,
L1.5). These electrodes were fixed at a height which
2

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 1. Experimental set-up. (a) Cuff electrode for epidural stimulation of dorsal and ventral cervical spinal cord (C7 spinal
level). (b) Stimulus patterns used to characterize the effect of intensity (left) and frequency (right). (c) Paired stimulation
protocols. Supraspinal intracortical microstimulation or pyramidal tract stimulation (A) was paired with spinal cord stimulation
delivered either through epidural electrodes (B1) or transcutaneous electrodes (B2). EMG responses were recorded from six
upper-limb muscles on each side.
maximized the short-latency antidromic field poten-
tials recorded from the motor cortex after stimulation
through the electrode.
2.1.4. Transcutaneous SCS
Transcutaneous SCS was delivered in two anim-
als (monkeys Uk and Yu) during repeated ses-
sions under light sedation. At the beginning of
each session, the skin was shaved and cleaned
before the cervicothoracic junction was identi-
fied by skin palpation. Two transcutaneous adhes-
ive electrodes (circular, 2.5 cm diameter, Model
J10R00, Axelgaard Manufacturing, Denmark) were
positioned above C3-C4 and C7-T1 intervertebral
spaces.
2.2. Stimulation protocols
2.2.1. Intensity series
Single biphasic, constant current pulses (0.1 ms per
phase, cathodal first) were delivered through the
spinal epidural contacts (using a return electrode
placed in nearby muscle) with an isolated, constant
current stimulator (Model DS4, Digitimer, Hertford-
shire, UK). Initially we used ten stimulus intensities
of 20, 40, …, 200 µA. For each contact on the elec-
trode, 20 repetitions of each intensity were delivered
in a pseudorandomized order with an interstimu-
lus interval of 0.5 s (figure 1(b)). Depending on the
threshold for muscle activation, additional series (5–
50 µA, 50–500 µA, 200–2000 µA) were performed as
appropriate.
3

J. Neural Eng. 18 (2021) 046011 T Guiho et al
2.2.2. Frequency series
Dorsal and ventral spinal epidural contacts were then
stimulated with trains of 15 pulses at frequencies
between 10 and 120 Hz (10 Hz increments) using a
different isolated constant-current stimulator (Model
2100, A-M Systems, Carlsborg, US). Twenty repeti-
tions of each frequency were delivered in a pseudor-
andomized order, with a one second interval between
each train (figure 1(b)). The stimulus intensity was set
slightly above threshold so as to elicit spinal-evoked
potentials in at least one muscle.
2.2.3. Paired supraspinal and epidural SCS
Interaction between supraspinal inputs and epidural
SCS was first assessed in two monkeys (monkeys
Si and Un) by pairing suprathreshold ICMS (three
biphasic pulses at 333 Hz, 0.1 ms per phase, cath-
odal first) with single pulse SCS, using interstimulus
intervals between −100, −50, −40, −30, …, 40, 50,
100 ms after compensating for the longer response
latency from ICMS so hand muscle responses coin-
cided (both delivered using Model 2100 stimulators).
Experiments were performed with an SCS intens-
ity that was suprathreshold for activating a subset of
muscles.
We next examined modulation of supraspinal
responses during one second trains of subthreshold
dorsal or ventral epidural SCS at frequencies of 10, 20,
50, 100, 143 or 200 Hz (monkeys Si and Un). ICMS
was delivered either 500 ms before the beginning of
the spinal train, between 0, 100, …, 1000 ms after the
beginning of the train, or 500 ms after the end of the
train, in pseudorandomized order.
In additional experiments (monkey Yi and Yu),
similar stimulation protocols were repeated using
single pulse pyramidal tract stimulation (PTS,
biphasic, 0.1 ms per phase) in place of ICMS
(figure 1(c)).
2.2.4. Paired supraspinal and transcutaneous SCS
Trains of transcutaneous SCS was delivered using
a Model 4100 stimulator (A-M Systems, Carlsborg,
US). We used the ‘Russian current’ approach to deliv-
ering high currents without associated cutaneous dis-
comfort, in which a high carrier frequency is mod-
ulated by a lower stimulation frequency. We tested
carrier frequencies of 1, 2, 5 and 10 kHz, delivered as
a burst of ten biphasic (0.05 ms per phase, cathodal
first) pulses, with intervals of 1, 0.5, 0.2 and 0.1 ms
respectively. We used stimulation frequencies of 10,
20, 50 and 100 Hz, i.e. with intervals of 100, 50, 20 and
10 ms between each burst of ten pulses. We delivered
1 s trains of this transcutaneous SCS pattern while
testing modulation of ICMS responses as described
above (figure 1(c)).
2.3. Analysis methods
2.3.1. Intensity series
Following the method of Sharpe and Jackson (2014),
EMG responses were rectified and averaged over
a 10 ms post-stimulus window adapted to their
latencies, and compared against an equivalent pre-
stimulus time-window using a two-tailed unpaired
t-test (p< 0.05) to construct recruitment curves,
R(I), quantifying the percentage of ipsilateral muscles
exhibiting a significant response to each stimulus
intensity, I. We then performed a least-squares fit to
these curves with a cumulative Normal distribution:
R(I) = 100%×
I
ˆ
−∞
Nx−µ, σ2dx(1)
to obtain estimates of the mean, µ, and variance, σ2
of the response threshold across muscles. Note that
this approach assumes thresholds are normally dis-
tributed, but has the advantage that mean thresholds
can be assessed across different intensity ranges and
even if not all muscles are activated by the highest
intensity.
2.3.2. Frequency series
EMG responses following each pulse of the SCS train
were rectified and averaged over an appropriate 8 ms
post-stimulus window (to allow exclusion of stimu-
lus artefacts with the highest frequency trains). For
muscles exhibiting a significant response to the first
stimulus pulse, the response to each subsequent pulse
in the train was normalized by the magnitude of the
first response. These normalized values were then
averaged across responding muscles to assess facilita-
tion and suppression of subsequent responses during
SCS trains of different frequencies.
2.3.3. Paired supraspinal and spinal stimulation
In order to assess the impact of SCS on supraspinal-
evoked responses, we compared the EMG elicited
by paired stimulation to that predicted by a linear
sum of responses to spinal and supraspinal stimuli
delivered alone. Since rectification introduces a non-
linearity, the average rectified responses cannot be
added to generate this prediction. Instead, we used
the method of (Baker and Lemon 1995) to generate
a set of surrogate paired responses based on the linear
summation of individual unrectified EMG responses
with the appropriate interstimulus intervals, prior
to rectification and averaging (see also Sharpe and
Jackson 2014). Since these surrogates contain the
summation of two sections of background activity
(in addition to the evoked responses), we added a
section of background activity (taken from a period
without stimulation) to each response to paired stim-
ulation before rectification and averaging. We used
this method even when SCS appeared subthreshold
to ensure that any occasional spinal-evoked poten-
tials could not account for an increase in the paired
response. We quantified modulation as the percent-
age ratio of the paired response to that predicted
from linear summation. Thus a modulation >100%
4

J. Neural Eng. 18 (2021) 046011 T Guiho et al
represents facilitation of the supraspinal response and
a modulation <100% represents suppression.
To test whether modulation depended on SCS
frequency, we averaged values across all interstimu-
lus intervals within the train, and across all muscles
responding to supraspinal stimulation alone. We then
constructed 10 000 surrogate datasets in which we
shuffled the responses for each interstimulus inter-
val across stimulation frequencies. We used the vari-
ance of the modulation (across frequencies) as a test
statistic and compared this with the 95th percentile
of the null distribution obtained from the shuffled
surrogates.
2.3.4. Modelling muscles responses to spinal and
supraspinal stimulation
We have previously used a simple computational
model of combined excitatory and inhibitory influ-
ences (Zimmermann et al 2011, Sharpe and Jackson
2014) to explain the frequency-dependent modula-
tion of responses to SCS trains. We adapted this
model to simulate also the modulation of supraspinal
inputs. This was not intended as a detailed biophysical
simulation, but an attempt to capture the qualitative
features of our results with minimal free parameters.
Briefly, our original model assumed that preceding
pulses in a SCS train act both to facilitate (with a short
decay time) and suppress (with a longer decay) the
responses to subsequent stimuli. In the present work,
we first assumed that this was mediated by the accu-
mulation of post-synaptic potentials within the mem-
brane potential of a population of neurons, Vmem,
with arbitrary units. In our first model, this was given
by:
Vmem (t)
=Vrest +
nAfPt−tn
τf−AsPt−tn
τs
(2)
where Vrest represents the negative baseline rest-
ing membrane potential, Afand Asrepresent the
strengths of facilitation and suppression, and τfand
τsrepresent their respective time constants. tnare the
times of 15 equally-spaced stimulus pulses, and P(x)
is an exponentially-decaying synaptic potential:
P(x) = e−xfor x⩾0
0 for x<0.(3)
In our second model, we assumed that suppression
did not act directly on the motoneurons, but instead
via an upstream mechanism that gated the facilita-
tion that reached the motoneurons, for example via
primary afferent depolarisation (PAD). We modelled
afferent depolarisation, Vaff as:
Vaff (t) = 
n
AsPt−tn
τs(4)
and assumed this gated the excitatory post-synaptic
potentials in motoneurons according to:
Vmem (t) = Vrest +
n
AfP(Vaff (tn))Pt−tn
τf.
(5)
Note that in the first model (equation (2)), Vmem
becomes progressively negative during the SCS train
as the slowly decaying inhibitory potentials accumu-
late. In the second model (equation (5)), Vmem instead
decays towards the base-line resting potential as the
facilitatory influence of SCS is gated upstream. This
leads to a key difference between model predictions
for responses to supraspinal stimulation during SCS
trains. In the first model, muscles become unrespons-
ive to both spinal and supraspinal stimuli due to dir-
ect inhibition of motoneurons. By contrast, in the
second model, motoneurons become unresponsive to
subsequent spinal stimuli (due to gating of afferent
input) but still respond to supraspinal inputs at close
to baseline levels.
To convert the membrane potential into a
response to stimulation, we assumed that the firing
threshold of the motoneuron populations was nor-
mally distributed with zero mean and unity variance
(these arbitrary scaling factors are accommodated by
the other model parameters). Thus the proportion
of motoneurons responding to stimulation was equal
to:
S(Vmem) =
∞
ˆ
−Vmem
N(0,1)dx(6)
and the modulation of that response relative to
baseline was equal to:
Mn=100%×S(Vmem (tn))
S(Vmem (t1)) .(7)
Each model comprised five free parameters: the
strengths of facilitation, Af, and suppression, As,
their respective time-constants, τfand τs, which gov-
ern the temporal modulation of responses, and the
resting motoneuron potential, Vrest which affects
only their overall magnitude. We chose Vrest such
that 50% of motoneurons responded to the first
stimulus in the train (i.e. such that S(Vmem (t1)) =
0.5). We then fit the remaining four parameters to
the experimentally-observed temporal modulation of
responses to dorsal spinal stimulation trains at dif-
ferent frequencies using least-squares regression. We
fit the first five responses in each train, and assessed
model fit over the same data using a coefficient of
determination:
R2=1−Var (Mfit −Mactual )
Var (Mactual ).(8)
Finally, we used each model to predict the modulation
of weak supraspinal inputs, based on the parameters
fit to responses to spinal stimulation alone. We first
5

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 2. Intensity series results. (a) Example mean, rectified EMG responses in right abductor pollicis brevis (APB) muscle
elicited by single-pulse ventral (contact 5) and dorsal (contact 8) epidural SCS with intensities between 20 and 200 µA in monkey
Si. Red traces indicate statistically-significant response (paired t-test, P< 0.05). Black triangles indicate time of stimulation while
blue arrows indicate response onset. (b) Proportion of ipsilateral muscles exhibiting significant responses to single-pulse ventral
(contact 4) and dorsal (contact 1) SCS. Data are fit with a cumulative normal distribution to assess average activation threshold.
(c) Average activation thresholds for all electrode contacts. The lowest thresholds are obtained for contacts on the ventral side of
the spinal cord. Error bars indicate s.e.m over the six ipsilateral muscles.
calculated the membrane potential resulting from 1 s
trains of spinal stimulation at different frequencies
(using equations (2) or (5)). We assumed that the
supraspinal inputs would be dispersed in time, so
we used the average value of Vmem across consecut-
ive 100 ms epochs through the train. We then used
equations (6) and (7) to calculate the modulation of
a weak supraspinal stimulus assumed to recruit 1%
of motoneurons at baseline (i.e. such that S(Vrest) =
0.01).
All analyses were performed using custom soft-
ware written in Matlab (Mathworks, Natick, US).
3. Results
3.1. Ventral epidural SCS effectively recruits
upper-limb muscles
We first characterized responses in upper-limb
muscle to epidural SCS delivered through different
contacts on the epidural spinal array. Spinal-evoked
potentials were quantified by average rectified EMG,
and compared to a pre-stimulus baseline period
to assess statistical significance. Figure 2(a) shows
example responses in a single muscle to stimulation
delivered at 20–200 µA through dorsal and vent-
ral contacts. Note that muscle responses to dorsal
stimulation occurred with longer latency and higher
stimulus intensities compared to ventral stimulation.
Figure 2(b) shows the proportion of all recorded
muscles exhibiting a significant evoked potential at
different stimulus intensities. We fitted these recruit-
ment curves with cumulative Gaussian distributions
to obtain average response thresholds. Response
thresholds were lowest for contacts on the ventral side
of the spinal cord (typically below 50 µA), and higher
for contacts on the dorsal side (typically greater than
200 µA) (figure 2(c)). A circular-linear correlation
analysis showed a significant relationship between
the positioning of the contact around the spinal cord
and thresholds of muscle activation (circular-linear
correlation R=0.92, P=0.035).
In addition, we examined the reliability of
muscle responses evoked by stimulus trains with
different frequencies between 10 and 120 Hz. For
6

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 3. Frequency series results. (a) Example EMG responses in left APB elicited by trains of ventral (contact 3—left panel) and
dorsal (contact 1—right panel) SCS at 120 Hz (top) and 40 Hz (bottom) in monkey Si. Triangles indicate times of stimuli.
(b) Similar data for left APB responses in monkey Un. (c) Modulation of mean, rectified EMG response to each pulse in the train,
normalized by the response to the first pulse, averaged across all muscles responding to ventral (left panel—24 muscles) and
dorsal (right panel—22 muscles) SCS in monkeys Si and Un.
ventral stimulation, consistent responses could be
observed following each stimulus even at the highest
frequencies (figures 3(a) and (b); left column). By
contrast, when delivering high-frequency stimulation
to the dorsal surface, the response to the second and
third pulses in the train was often facilitated relative
to the first pulse, while responses to later stimuli
were profoundly suppressed (figures 3(a) and (b);
right column). Figure 3(c) summarizes modulation
of responses to both ventral and dorsal stimulation
for each pulse within the trains of different
frequencies.
7

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 4. Non-linear interactions between cortical- and spinal-evoked muscle responses. (a) EMG responses to paired ICMS and
epidural SCS were compared to surrogate data generated from the linear summation of responses to each stimulus alone. Prior to
rectification and averaging, a baseline trace with no stimulation was added to the paired response to compensate for the higher
background level in the surrogate traces. This example shows supralinear facilitation of the response in 1DI muscle in monkey Si
when ICMS preceded SCS by 10 ms in monkey Si. Green and blue triangles correspond to cortical and spinal stimulation
respectively. (b) Comparison of paired responses to linear summation for all interstimulus intervals between −100 and 100 ms.
Red/black traces indicate significant/non-significant supralinear facilitation (paired t-test against linear summation surrogates,
grey traces, P< 0.05). (c) Modulation of response to paired stimulation relative to linear summation for all muscles that
responded to ICMS in monkeys Si and Un. Muscles are divided according to whether the spinal stimulus alone was
suprathreshold or subthreshold for eliciting a response. Left panel shows results for ventral epidural SCS (seven muscles). Middle
panel shows results for dorsal epidural SCS (five muscles). Right panel compares modulation for dorsal vs. ventral epidural SCS,
on the same ordinate scale. Shading indicates s.e.m over muscle-electrode pairs.
These findings are consistent with the results of
Sharpe and Jackson (2014), and suggest that stim-
ulation of the ventral surface acts directly on the
axons of the motoneurons, while stimulation of
the dorsal surface activates motoneurons transyn-
aptically, leading to higher recruitment thresholds
and temporal facilitation/suppression of subsequent
responses.
3.2. Single-pulse dorsal epidural SCS facilitates
cortical-evoked responses
Next, we examined interactions between single pulse
SCS and descending volleys elicited by ICMS. We used
interstimulus intervals between −100 and +100 ms,
adjusted according to the different conduction times
from cortex and spinal cord to the periphery. Mean
rectified EMG was calculated over a response win-
dow following the second of the two stimuli, and
compared to that predicted by a linear super-
position of the response to each stimulus alone
(figures 4(a) and (b)). We used a spinal stimu-
lation intensity that activated around half of the
recorded muscles, and assessed separately ICMS
responses in muscles for which spinal stimulation
was below or above threshold (figure 4(c)). Only
weak facilitation of the cortical response was observed
8

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 5. Facilitation of supraspinal-evoked muscle responses by trains of epidural SCS. (a) Modulation of EMG responses to
ICMS delivered at different times during a 1 s train of ventral epidural SCS. Colours indicate different frequencies of spinal
stimulation. Data from four muscle-electrode pairs in monkeys Si and Un. Shading indicates s.e.m over muscle-electrode pairs.
(b), (c) Mean modulation of cortical-evoked responses during the first and second halves of the spinal train respectively. Error
bars indicate s.e.m over intervals (d), (e), (f) Equivalent plots for dorsal epidural SCS. Data from four muscle-electrode pairs in
monkeys Si and Un. (g), (h), (i) Equivalent plots for modulation of EMG responses to PT stimulation during dorsal epidural SCS.
Data from ten muscle-electrode pairs in monkeys Yu and Yi.
for both suprathreshold (mean [standard deviation]
facilitation of 134% [11%]) and subthreshold (130%
[13%]) ventral SCS (preceding the cortical stimulus
by 10 ms). By contrast, dorsal SCS was more effect-
ive at potentiating descending inputs, with mean
(SD) peak facilitations of 298% (120%) for supra-
threshold stimuli and 248% (54%) for subthreshold
stimuli. The difference between dorsal and ventral
stimulation was statistically significant (10 ms inter-
val, paired t-test across responding muscles, t9=1.96,
P=0.041), suggesting that transynaptic activation of
motoneurons from the dorsal surface, but not direct
activation from the ventral surface, is most effective
at facilitating the response to descending inputs.
3.3. Trains of dorsal epidural SCS facilitate
supraspinal inputs in a frequency-dependent
manner
We examined the temporal profile of facilitation by
1 s trains of subthreshold SCS delivered at frequen-
cies from 10 to 200 Hz. On separate trials, ICMS was
delivered at different time-points to probe excitability
changes through the train. Figure 5(left column)
shows modulation of the cortical-evoked potential
(expressed as a percentage of the linear superposition
of cortical and spinal responses) for all time intervals,
while the histograms show the average modulation
over the first (middle) or second (right) halves of the
train.
In general, cortical-evoked potentials were facil-
itated when delivered during the SCS train, with
responses induced by paired stimulation being the
same or higher (>100%) than that predicted by linear
superposition. Facilitation was significantly greater
for SCS delivered to the dorsal surface (mean [SD]
facilitation of 316% [137%]; figures 6(d), (e) and
(f)) compared to ventral stimulation (121% [23%];
figures 6(a), (b) and (c); paired t-test across all
responsive muscles, t3=2.64, P=0.039) There was a
general trend for facilitation to decrease through the
train, especially for high frequencies of SCS. Thus,
facilitation during the first half of the dorsal epi-
dural SCS train (345% [150%]; figure 5(e)) was sig-
nificantly higher than during the second half (291%
9

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 6. Facilitation of cortical-evoked muscle responses by trains of transcutaneous SCS (5 kHz carrier frequency). Left panel
shows modulation of EMG responses to ICMS delivered at different times during a 1 s train of transcutaneous SCS. Colours
indicate different frequencies of spinal stimulation. Shading indicates s.e.m. over muscle-electrode pairs. Data from 27
muscle-electrode pairs in monkeys Uk and Yu. Middle and right panels show mean modulation of cortical-evoked response
during the first and second halves of the spinal train respectively. Error bars indicate s.e.m. over intervals.
[128%]; figure 5(f); paired t-test across responsive
muscles, t3=3.06, P=0.028).
We used a shuffling analysis to test the influ-
ence of spinal stimulation frequency (see section 2).
This revealed a statistically significant (P< 0.05)
effect of frequency for both ventral (Var(Data) =
92.2,Var (95%Shuffled) = 58.5) and dorsal epidural
SCS (Var (Data) = 2.39 ×104, Var(95%Shuffled) =
6.32 ×103). For ventral stimulation, the greatest
facilitation was seen with the highest stimulation
frequency of 200 Hz (132% [42%]). By contrast,
for dorsal stimulation, an intermediate frequency of
50 Hz was optimal (605% [391%]).
To demonstrate that facilitation occurred at the
spinal level, rather than as a result solely of affer-
ent volleys increasing cortical excitability, we also
tested facilitation of motor responses elicited by
PTS. Figures 5(g), (h) and (i) shows that facilita-
tion of PT-evoked potentials could be obtained with
dorsal epidural SCS, with muscle responses exhibit-
ing facilitation (>100%) relative to linear summation
throughout the SCS train. PTS responses were facilit-
ated less than cortical-evoked potentials when paired
with dorsal SCS (245% [115%] vs. 316% [137%]) but
these recordings were made in different animals and
this difference in any case did not reach significance
(unpaired t-test across responsive muscles, t12 =0.92,
P=0.2). The greatest facilitation was again obtained
for an intermediate stimulation frequencies, with a
maximum facilitation of 314% (252%) when stimu-
lating at 100 Hz. Again, our shuffling analysis revealed
a significant impact of SCS frequency on facilita-
tion (Var (Data) = 3.71 ×103,Var(95%Shuffled) =
1.79 ×103).
We conclude that trains of subthreshold epidural
SCS delivered to the dorsal aspect of the spinal cord
are effective at facilitating the muscle response to
supraspinal inputs. This is likely mediated by tran-
synaptic inputs to motoneurons that raise their mem-
brane potential closer to threshold, making them
more responsive to descending volleys. Moreover, this
effect is frequency-dependent, such that intermediate
frequencies (50 or 100 Hz) are more effective than
either lower or higher frequencies of stimulation.
3.4. Facilitation of descending pathways by
transcutaneous SCS
Our final set of experiments sought to establish
whether similar facilitation of supraspinal inputs
could be achieved using transcutaneous SCS. We
delivered 1 s trains of ‘Russian current’ stimulation
(low-frequency modulation of a high-frequency car-
rier) through surface electrodes placed on the skin
over the upper back (figure 6). Again we observed
facilitation of ICMS responses, although the mag-
nitude of facilitation was lower than that obtained
with epidural stimulation. As with dorsal epidural
SCS, the most effective modulation frequency was
50 Hz, (mean [SD] facilitation of 134% [39%]),
and our shuffling method again verified a signific-
ant impact of transcutaneous stimulation frequency
(Var (Data) = 242,Var(95%Shuffled) = 13). Unlike
epidural SCS, the effect of transcutaneous SCS grew
throughout the train (115% [22%] and 121% [22%]
for first and second halves respectively; paired t-test
t26 =3.38, P=0.001). Moreover, facilitation outlas-
ted the end of the train, with significantly enhanced
responses occurring 500 ms after the train com-
pared to 500 ms before (118% [18%], paired t-test
t26 =4.71, P=3.6 ×10−5).
These experiments were conducted using a car-
rier frequency of 5 kHz, based on pilot experi-
ments in our first animal (monkey Uk, three ses-
sions, figure 7). We initially compared five differ-
ent carrier frequencies embedded in a 50 Hz stim-
ulation pattern (1 kHz, 2 kHz, 3 kHz, 5 kHz and
10 kHz) and found that 5 kHz provided the best
facilitation (114% [20%]), although a single factor
analysis of variance (ANOVA) failed to show a sig-
nificant impact of carrying frequencies on facilita-
tion (single factor ANOVA, F4,57 =1.97,P=0.11).
We subsequently examined the effect of carrier
frequency in our second animal (Monkey Yu, four
sessions) and combining data from both animals
10

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 7. Effect of different carrier frequencies for transcutaneous SCS. Plots show mean modulation of cortical-evoked responses
during trains of transcutaneous SCS at 50 Hz with different carrier frequencies in monkeys Uk and Yu. Data from 27
muscle-electrode pairs in monkeys Uk and Yu. Error bars indicate s.e.m. over intervals.
confirmed the absence of significant effect of carrier
frequency (F4,107 =0.37,P=0.83).
3.5. Modelling spinal modulation as synaptic
excitation gated by upstream inhibition
Previously we have introduced a simple mathematical
description of the frequency-dependence of muscle
responses to SCS trains (Zimmerman et al 2011,
Sharpe and Jackson 2014). Briefly, we assumed that
the response to each pulse in the train was modu-
lated by an accumulation of excitatory and inhib-
itory influences from preceding stimulus pulses. By
modelling the decay of these influences as exponen-
tials with fast and slow time-courses respectively, we
could explain the transient facilitation of responses
to high-frequency SCS and later response suppres-
sion for high- and intermediate-frequency trains.
While not intended to be a realistic biophysical model
of spinal circuits, we were interested in whether
this same description could explain the frequency-
dependent modulation of supraspinal inputs in our
current dataset. Our first attempt assumed the excit-
atory and inhibitory influences of spinal stimula-
tion both acted to raise or lower the membrane
threshold of a population of spinal motoneurons
(figure 8(a)). As a result, different stimuli within
the train recruited a varying proportion of the pop-
ulation. We used least-squares regression to fit the
experimentally-observed modulation of responses to
suprathreshold dorsal epidural SCS (figure 3(c), left),
obtaining time-constants for facilitation and suppres-
sion of 7 ms and 87 ms respectively. The model
reproduced the time- and frequency-dependence of
these responses well (figure 8(b)), with a coefficient
of determination (R2) of 0.87. However, it failed to
replicate the experimental results for modulation of
ICMS responses (figures 8(c) and (d)). In particu-
lar, the model predicted strong suppression of ICMS
responses during the later period of high-frequency
spinal stimulation, as the slowly-decaying inhibition
builds up. This highlights a puzzling qualitative fea-
ture of our results. While high-frequency SCS inhib-
its the response to subsequent spinal stimuli (i.e. to
less than 100% of the response to the first stimulus),
it merely reduces the facilitation of supraspinal inputs
(i.e. declining towards 100% of the response to ICMS
alone).
Therefore we speculated that the inhibitory effect
of SCS may not act directly on the motoneur-
ons but upstream, gating subsequent spinal but not
supraspinal inputs (figure 8(e)). A possible mech-
anism for such gating would be presynaptic inhibi-
tion via PAD. We modified our original model such
that the inhibitory influence of preceding stimulus
pulses modulated the extent to which subsequent SCS
pulses excited the population of motoneurons (see
section 2). This model was again capable of repro-
ducing the frequency-dependence of motor responses
to SCS alone (figure 8(f)), with similar best-fit time-
constants for facilitation and suppression (9 ms and
74 ms respectively) and R2=0.90. However, the new
model was also able to qualitatively capture the mod-
ulation of ICMS responses by SCS trains (figures 8(g)
and (h)). For low SCS frequencies (10–50 Hz), the
influence of preceding stimuli was small, and thus
the excitatory effects produced a facilitation of the
ICMS response that became progressively larger with
increasing frequency of pulses. However, as SCS fre-
quency increased further (100–200 Hz), the inhibit-
ory influence of previous pulses diminished this excit-
atory effect, particularly during the latter part of the
train. As a result, modulation of ICMS responses was
greatest for an intermediate frequency (50 Hz) but
never reduced below 100%, as in the experimental
data.
4. Discussion
We investigated whether SCS delivered via ventral
epidural, dorsal epidural and dorsal transcutaneous
11

J. Neural Eng. 18 (2021) 046011 T Guiho et al
Figure 8. Computational model of time- and frequency-dependence of SCS effects. (a) Our first model assumed that excitatory
and inhibitory influences of SCS (decaying over different time-scales) accrue at the motoneurons. (b) The model reproduced the
experimentally-observed modulation of spinal-evoked responses to different frequencies of SCS (figure 3(c)) with a least-squares
fit, R2=0.87. (c), (d) The same model and fit parameters were then used to predict modulation of cortical-evoked responses by
SCS. This yielded suppression of cortical-evoked responses, especially for high-frequency SCS, due to accumulation of inhibition
on motoneurons, in contrast to experimental results. (e) Our second model assumed inhibition acted upstream of the
motoneurons to gate the excitatory influence, for example via presynaptic inhibition. (f) This model could also reproduce the
modulation of spinal-evoked responses with a least-squares fit, R2=0.90. (g), (h) This model could qualitatively reproduce the
experimental results of modulation of cortical-evoked responses by SCS, predicting facilitation, especially for intermediate SCS
frequencies.
electrodes could modulate muscle responses evoked
from descending pathways. The largest facilitation
was observed using dorsal epidural SCS at an interme-
diate stimulation frequency of 50 Hz, which increased
the magnitude of EMG responses to ICMS by around
six times. This was likely driven largely by an interac-
tion at the spinal level, as comparable facilitation was
also observed in motor responses to pyramidal tract
stimulation which elicits a constant descending vol-
ley unaffected by modulations in cortical excitability.
These results support the idea that subthreshold epi-
dural SCS may be beneficial in the case of anatomic-
ally incomplete spinal cord injuries, by transynaptic-
ally increasing the excitability of the spinal cord and
thus facilitating volitional control of muscles medi-
ated by surviving descending connections.
By contrast, ventral epidural SCS produced con-
siderably less modulation of muscle responses to des-
cending volleys. However, suprathreshold ventral epi-
dural SCS was more effective at driving muscles,
with lower thresholds and EMG responses that
consistently followed stimulation pulses even with
high-frequency trains. This is consistent with a direct
action on the motor unit, for example, eliciting action
potentials in the ventral roots. Thus ventral epidural
SCS may be well-suited for strong, direct activation of
paralyzed muscles for FES applications. The flexible
cuff electrode we used in this study allowed contacts
to be placed on both dorsal and ventral surfaces, and
could in future be adapted for human use, such that
both types of stimulation might be delivered in com-
bination as appropriate for patient needs. Our exper-
iments were performed using neurologically intact
animals under anaesthesia so it remains to be seen if
similar results can be obtained in awake, spinal cord-
injured subjects. However, dorsal SCS in lumbar seg-
ments has already proved effective in human indi-
viduals (Harkema et al 2011, Angeli et al 2018), while
the direct action of ventral SCS on motor units should
not be greatly influenced by upstream reorganization
of spinal circuitry.
We previously compared ventral and dorsal
epidural SCS using a ball electrode placed on the
dura (Sharpe and Jackson 2014). One discrepancy
12

J. Neural Eng. 18 (2021) 046011 T Guiho et al
worth noting is that in our previous study, thresholds
for eliciting movements with epidural SCS were puzz-
lingly higher on the ventral vs. dorsal side of the cord.
This was not the case in the present study using the
cuff electrode. We do not have a definitive explan-
ation for this, but our previous surgical approach
involved accessing the spinal cord through the ver-
tebral body, which was surgically awkward and may
have contributed to either damage to the spinal cord
or inaccurate positioning of electrodes on the ventral
side. By contrast, the epidural cuff electrode provided
efficient stimulation of the ventral spinal cord, and
could be placed with relative ease via a dorsal laminec-
tomy. In future it may be possible to reduce the need
for an extensive laminectomy using the techniques
of minimally-invasive spinal surgery (Vaishnav et al
2019).
We also found that transcutaneous SCS was
capable of facilitating cortically-evoked muscle
responses, although this was less effective than dorsal
epidural SCS. Interestingly, and in contrast to epi-
dural SCS, the degree of facilitation progressively
increased during the 1 s stimulation trains, and
was still evident 0.5 s after the end of the train. It
is possible that transcutaneous SCS accesses addi-
tional spinal circuitry compared with epidural SCS,
or even contributes to short-term plasticity mech-
anisms (Benavides et al 2020). However, it is worth
noting that unlike with epidural SCS, we observed
activation of back musculature under the electrodes
during transcutaneous stimulation. Therefore it is
likely that reafference or nociceptive effects could
have contributed to the facilitatory effects of these
stimulus trains. As with dorsal epidural SCS, facilit-
ation was greatest with an intermediate stimulation
frequency of 50 Hz.
Interestingly, these intermediate SCS frequencies
around 50 Hz were not particularly effective at driv-
ing muscles directly. As in our previous studies, we
found that muscles would follow trains of dorsal epi-
dural SCS only up to around 10 Hz, above which
responses to subsequent stimuli were progressively
attenuated. At frequencies above 50 Hz, we observed
facilitation of the first few responses followed by pro-
nounced suppression. However, a simple computa-
tional model was able to reconcile these observations,
by assuming that an excitatory synaptic influence of
SCS on motoneurons was gated by upstream inhibi-
tion. Thus, for low rates, increasing the frequency of
stimulation increased the net excitatory influence on
motoneurons. However, beyond a certain point, the
increased stimulation frequency attenuated the facil-
itation, and the excitability of motoneurons returned
towards baseline. Our fit to the experimental data
yielded excitatory decay-times of around 10 ms, close
to the motoneuron membrane time-constant (Burke
1967). However, the inhibitory influence decayed
with a much longer time-constant of 74 ms, which
seems much too slow for inhibitory post-synaptic
potentials. We suggest that the inhibitory influence
may be mediated by presynaptic mechanisms, which
have a similarly slow time-course (Eccles et al 1962,
Jankowska et al 1981) and are known to act on affer-
ents but not corticospinal inputs to motoneurons
(Jackson et al 2006). Since dorsal SCS is thought to
act primarily on the dorsal roots (Ladenbauer et al
2010), our data suggest that each spinal stimulus has
at least two effects: first it produces a fast-decaying
compound excitatory post-synaptic potential in the
motoneurons (e.g. via Ia afferents) and, second, it
produces slower presynaptic inhibition (via GABAer-
gic interneurons) of subsequent afferent input, likely
via PAD. An effect of SCS on presynaptic inhibition
would also be consistent with its efficacy at treating
spasticity (Hofstoetter et al 2014).
These results have several practical implications.
First, it is worth emphasizing that SCS patterns
optimized for driving muscles directly may not be
optimal for generating subthreshold facilitation of
spinal circuits. In future, computational models such
as the one we have introduced may be useful for
designing temporally-patterned stimulus trains to
maximize synaptic facilitation while minimizing the
gating effects of presynaptic inhibition. Alternat-
ively, using the facilitation of cortical-evoked muscle
responses (for example by transcranial magnetic
stimulation) as a biomarker of corticospinal excitabil-
ity may allow patient-specific stimulation parameters
to be optimized. Finally, the effect of GABABagon-
ists that are commonly used to treat spasticity after
SCI should be considered when developing and eval-
uating SCS therapies, since these have indirect effects
on pre-synaptic inhibition (Rudomin and Schmidt
1999).
In summary, we have shown that intermediate-
frequency trains of subthreshold dorsal epidural SCS
and transcutaneous SCS can raise the excitability of
spinal circuits and thus enhance upper-limb muscle
responses to weak supraspinal inputs. By contrast,
ventral epidural SCS is effective at activating those
muscles directly, even with low current intensities and
high stimulation frequencies. Both effects may be use-
ful in the restoration of upper-limb function follow-
ing SCI, and the epidural cuff electrode used here may
pave the way for clinical implants to deliver both types
of stimulation in patients. Consideration of the inter-
action between synaptic facilitation of motoneurons
and presynaptic gating mechanisms will help design
stimulation protocols to maximize the potential of
these therapies.
Acknowledgments
We thank Jennifer Tulip, James Kersey and Chloe
Terry for assistance with experiments, and the CER-
MOD consortium for discussions. This work was
supported by the Wellcome Trust (106149) and an
13

J. Neural Eng. 18 (2021) 046011 T Guiho et al
ERANET Neuron joint call co-funded by the Medical
Research Council (R001189).
ORCID iDs
Thomas Guiho https://orcid.org/0000-0001-8763-
9015
Stuart N Baker https://orcid.org/0000-0001-8118-
4048
Andrew Jackson https://orcid.org/0000-0001-
8701-6387
References
Anderson K D 2004 Targeting recovery: priorities of the spinal
cord-injured population J. Neurotrauma 21 1371–83
Angeli C A, Boakye M, Morton R A, Vogt J, Benton K, Chen Y and
Harkema S J 2018 Recovery of over-ground walking after
chronic motor complete spinal cord injury New Engl. J. Med.
379 1244–50
Baker S N and Lemon R N 1995 Non-linear summation of
responses in averages of rectified EMG J. Neurosci. Methods
59 175–81
Benavides F D, Jo H J, Lundell H, Edgerton V R, Gerasimenko Y
and Perez M A 2020 Cortical and subcortical effects of
transcutaneous spinal cord stimulation in humans with
tetraplegia J. Neurosci. 40 2633–43
Burke R E 1967 Composite nature of the monosynaptic excitatory
postsynaptic potential J. Neurophysiol. 30 1114–37
Dimitrijevic M R, Gerasimenko Y and Pinter M M 1998 Evidence
for a spinal central pattern generator in humans Ann. New
York Acad. Sci. 860 360–76
Duffell L D and Donaldson N D N 2020 A comparison of FES and
SCS for neuroplastic recovery after SCI: historical
perspectives and future directions Front. Neurol. 11
Eccles J C, Magni F and Willis W D 1962 Depolarization of central
terminals of group I afferent fibres from muscle J. Physiol.
160 62–93
Gad P, Lee S, Terrafranca N, Zhong H, Turner A, Gerasimenko Y
and Edgerton V R 2018 Non-invasive activation of cervical
spinal networks after severe paralysis J. Neurotrauma
35 2145–58
Gerasimenko Y, Gorodnichev R, Moshonkina T, Sayenko D, Gad
P and Reggie Edgerton V 2015 Transcutaneous electrical
spinal-cord stimulation in humans Ann. Phys. Rehabil. Med.
58 225–31
Greiner N et al 2021 Recruitment of upper-limb motoneurons
with epidural electrical stimulation of the cervical spinal
cord Nat. Commun..12 1–19
Guiraud D, Stieglitz T, Koch K P, Divoux J L and Rabischong P
2006 An implantable neuroprosthesis for standing and
walking in paraplegia: 5-year patient follow-up J. Neural.
Eng. 3268
Harkema S, Gerasimenko Y, Hodes J, Burdick J, Angeli C, Chen Y
and Edgerton V R 2011 Effect of epidural stimulation of the
lumbosacral spinal cord on voluntary movement, standing,
and assisted stepping after motor complete paraplegia: a
case study Lancet 377 1938–47
Hofstoetter U S, Krenn M, Danner S M, Hofer C, Kern H, Mckay
W B and Minassian K 2015 Augmentation of voluntary
locomotor activity by transcutaneous spinal cord
stimulation in motor-incomplete spinal cord-injured
individuals Artif. Organs 39 176–86
Hofstoetter U S, McKay W B, Tansey K E, Mayr W, Kern H and
Minassian K 2014 Modification of spasticity by
transcutaneous spinal cord stimulation in individuals with
incomplete spinal cord injury J. Spinal Cord Med. 37 202–11
Inanici F, Samejima S, Gad P, Edgerton V R, Hofstetter C P and
Moritz C T 2018 Transcutaneous electrical spinal
stimulation promotes long-term recovery of upper
extremity function in chronic tetraplegia IEEE Trans.
Neural. Syst. Rehabil. Eng. 26 1272–8
Jackson A, Baker S N and Fetz E E 2006 Tests for presynaptic
modulation of corticospinal terminals from peripheral
afferents and pyramidal tract in the macaque J. Physiol.
573 107–20
Jankowska E, McCrea D, Rudomin P and Sykova E 1981
Observations on neuronal pathways subserving
primary afferent depolarization J. Neurophysiol.
46 506–16
Kato K, Nishihara Y and Nishimura Y 2020 Stimulus outputs
induced by subdural electrodes on the cervical spinal cord in
monkeys J. Neural. Eng. 17 016044
Kendig J J 2002 In vitro networks: subcortical mechanisms of
anaesthetic action Br. J. Anaesth. 89 91–101
Krucoff M O, Rahimpour S, Slutzky M W, Edgerton V R and
Turner D A 2016 Enhancing nervous system recovery
through neurobiologics, neural interface training, and
neurorehabilitation Front. Neurosci. 10 584
Ladenbauer J, Minassian K, Hofstoetter U S, Dimitrijevic M R and
Rattay F 2010 Stimulation of the human lumbar spinal cord
with implanted and surface electrodes: a computer
simulation study IEEE Trans. Neural. Syst. Rehabil. Eng.
18 637–645
Lemon R N 2008 Descending pathways in motor control Annu.
Rev. Neurosci. 31 195–218
McPherson J G, Miller R R, Perlmutter S I and Poo M M 2015
Targeted, activity-dependent spinal stimulation
produces long-lasting motor recovery in chronic cervical
spinal cord injury Proc. Natl Acad. Sci. USA 112 12193–8
Megía García A, Serrano-Muñoz D, Taylor J, Avendaño-Coy J and
Gómez-Soriano J 2020 Transcutaneous spinal cord
stimulation and motor rehabilitation in spinal cord injury:
a systematic review Neurorehabil. Neural Repair 34 3–12
Minassian K, Jilge B, Rattay F, Pinter M M, Binder H,
Gerstenbrand F and Dimitrijevic M R 2004 Stepping-like
movements in humans with complete spinal cord injury
induced by epidural stimulation of the lumbar cord:
electromyographic study of compound muscle action
potentials Spinal Cord 42 401–16
Minassian K, McKay W B, Binder H and Hofstoetter U S 2016
Targeting lumbar spinal neural circuitry by epidural
stimulation to restore motor function after spinal cord
injury Neurotherapeutics 13 284–94
Moritz C T, Lucas T H, Perlmutter S I and Fetz E E 2007 Forelimb
movements and muscle responses evoked by
microstimulation of cervical spinal cord in sedated monkeys
J. Neurophysiol. 97 110–20
Rudomin P and Schmidt R F 1999 Presynaptic inhibition in the
vertebrate spinal cord revisited Exp. Brain Res. 129 1–37
Sharpe A N and Jackson A 2014 Upper-limb muscle responses to
epidural, subdural and intraspinal stimulation of the
cervical spinal cord J. Neural. Eng. 11 016005
Snoek G J, Ijzerman M J, Hermens H J, Maxwell D and
Biering-Sorensen F 2004 Survey of the needs of patients with
spinal cord injury: impact and priority for improvement in
hand function in tetraplegics Spinal Cord 42 526–32
Taylor P, Esnouf J and Hobby J 2002 The functional impact of the
freehand system on tetraplegic hand function. Clinical
results Spinal Cord 40 560–6
Vaishnav A S, Othman Y A, Virk S S, Gang C H and Qureshi S A
2019 Current state of minimally invasive spine surgery
J. Spine Surg. 5
Van Den Brand R, Heutschi J, Barraud Q, DiGiovanna J,
Bartholdi K, Huerlimann M and Courtine G 2012 Restoring
voluntary control of locomotion after paralyzing spinal cord
injury Science 336 1182–7
Ward A R 2009 Electrical stimulation using kilohertz-frequency
alternating current Phys. Ther. 89 181–90
Zimmermann J B, Seki K and Jackson A 2011 Reanimating the
arm and hand with intraspinal microstimulation J.
Neural. Eng. 8054001
14