Evidence For A Window of Enhanced Plasticity in the Human Motor Cortex Following Ischemic Stroke

Abstract

Background

In pre-clinical models, behavioral training early after stroke produces larger gains compared with delayed training. The effects are thought to be mediated by increased and widespread reorganization of synaptic connections in the brain. It is viewed as a period of spontaneous biological recovery during which synaptic plasticity is increased.

Objective

To look for evidence of a similar change in synaptic plasticity in the human brain in the weeks and months after ischemic stroke.

Methods

We used continuous theta burst stimulation (cTBS) to activate synapses repeatedly in the motor cortex. This initiates early stages of synaptic plasticity that temporarily reduces cortical excitability and motor evoked potential (MEP) amplitude. Thus, the greater the effect of cTBS on the MEP, the greater the inferred level of synaptic plasticity. Data were collected from separate cohorts (Australia and UK). In each cohort, serial measurements were made in the weeks to months following stroke. Data were obtained for the ipsilesional motor cortex in 31 stroke survivors (Australia, 66.6±17.8 years) over 12 months and the contralesional motor cortex in 29 stroke survivors (UK, 68.2±9.8 years) over 6 months.

Results

Depression of cortical excitability by cTBS was most prominent shortly after stroke in the contralesional hemisphere and diminished over subsequent sessions (p=0.030). cTBS response did not differ across the 12 month follow-up period in the ipsilesional hemisphere (p=0.903).

Conclusions

Our results provide the first neurophysiological evidence consistent with a period of enhanced synaptic plasticity in the human brain after stroke. Behavioral training given during this period may be especially effective in supporting post-stroke recovery.

Introduction

Stroke is a leading global cause of disability.¹ Although mortality and age-adjusted incidence rates have decreased, the annual crude incidence of stroke, number of stroke survivors and burden of disease have increased over the past two decades.² Those who survive stroke often require extensive therapy to support recovery. In humans, the majority of motor recovery happens early after stroke and appears to plateau by 3-6 months.^3–7 This recovery involves reorganization of motor output in surviving neural structures through a process known as plasticity.

Pre-clinical studies provide evidence that ischemic events give rise to a spontaneous period of increased neural plasticity, leading to heightened responsiveness to training.^8,9 A time-limited, spontaneously occurring period of upregulation in gene and protein expression to support neuronal growth, synaptogenesis, proliferation of dendritic spines, reduced peri-neuronal nets and enhanced brain excitability with changes in the excitation-inhibition balance have been observed within days of experimental stroke in the peri-infarct and contralesional cortex, suggesting widespread changes in both hemispheres.^10–21 These changes appear to support behavioral recovery. Rats exposed to enriched environments within 5 days of ischemia had greater skilled forelimb recovery and enhanced dendritic growth compared to those where therapy was delayed until 30 days post-stroke.¹⁵ Similarly, in monkeys, initiation of skill training within days of an infarct preserved the cortical hand territory which was thought to play an important role in motor recovery, while the absence of training substantially decreased size of the representations.^22,23

At present, there is no direct evidence at a neural level of a period of enhanced synaptic plasticity in human stroke patients. Behavioral data indicates that the effectiveness of rehabilitative therapy diminishes over time.^3,24 Furthermore, there is some evidence that early initiation of rehabilitation is associated with better stroke outcomes.²⁵ This would be compatible with a period of increased plasticity soon after stroke although it is difficult to disentangle other potential contributing factors such as stroke severity, medical complications, variations in service delivery and therapy dosage.

The aim of the present study was to measure synaptic plasticity at a neural level in human stroke survivors to test whether there is an early period of enhanced synaptic plasticity. We used continuous theta burst (cTBS) transcranial magnetic stimulation to repetitively activate synaptic connections in the motor cortex and engage early processes of synaptic plasticity via activation of NMDA receptors.²⁶ The effect is a temporary decrease in cortical excitability (long term depression, LTD-like) that can be quantified as a reduction in the amplitude of the MEP evoked by single pulse transcranial magnetic stimulation (TMS). Note that an LTD-like brain stimulation protocol was selected primarily for safety reasons since seizure risk is elevated in acute stroke and could be exacerbated by facilitatory stimulation. Moreover, behavioral benefits and plastic changes during learning, such as recovery following stroke, involve not only strengthening of synaptic connections but also weakening of inappropriate synapses.²⁷ Importantly, in this instance cTBS was not being used as a potential treatment, but as a way of assessing the level of synaptic plasticity within the motor cortex. We expected that if there was an increase in synaptic plasticity in the early weeks after stroke, cTBS would reduce excitability to a greater extent soon after stroke, with the magnitude of this response decreasing over subsequent sessions. Sub-analyses were conducted to investigate possible associations between cTBS response with: 1) stroke severity and upper limb recovery as we hypothesized that greater plasticity of the motor cortex might positively correlate with upper limb recovery; and 2) lesion location as we hypothesized ipsilesional neural responses may be reduced in people with cortical stroke.^28,29

Methods

Protocol

This study presents two separate cohorts that investigated plasticity of the ipsilesional and contralesional motor cortex, respectively. Plasticity was defined as the depression of corticospinal excitability produced by a short period of cTBS applied to the motor cortex. Separate cohorts were required to test both hemispheres as interhemispheric interactions would confound the cTBS response if we were to test both hemispheres in each participant. All testing took place in either the Neuromotor Plasticity and Development TMS laboratory located at the University of Adelaide, Australia or the TMS Laboratories in the Sobell Department at the National Hospital for Neurology and Neurosurgery, London, UK. The study was approved by 1) the Central Adelaide Local Health Network Human Research Ethics Committee; and 2) the Joint Ethics Committee of the Institute of Neurology, UCL and National Hospital for Neurology and Neurosurgery, UCL Hospitals NHS Foundation Trust. All participants provided written informed consent in accordance with the Declaration of Helsinki.

The Adelaide laboratory tested the ipsilesional motor cortex and the London laboratory tested the contralesional motor cortex. These datasets will be referred to as the ipsilesional data and contralesional data respectively. Ipsilesional data were obtained at eight time points (1 week, 2 weeks, 3 weeks, 4 weeks, 8 weeks, 12 weeks, 26 weeks and 52 weeks post-stroke). Contralesional data were obtained at four time points (2 weeks, 4 weeks, 6 weeks and 26 weeks post-stroke). Selection of these time points reflects a pragmatic approach at each site, with participants required to return to the neurophysiological laboratory for each experimental session. Note that a variety of reasons (transport, personal, treatment timetable, etc) prevented all patients from being tested on precisely each intended time point, but sessions were conducted as close as reasonably possible to each time point. Experimental sessions were scheduled for a similar time of day to control for physiological diurnal variation that can modify brain stimulation responses.³⁰ Stroke survivors sat in a comfortable armchair and were asked to keep their eyes open, their hand relaxed and their legs uncrossed during testing.

Participants

Participants were recruited from three stroke units across the two countries; the Royal Adelaide Hospital in Australia, the Hyper-Acute Stroke Unit at University College Hospital in the United Kingdom and the Acute Brain Injury Unit at the National Hospital for Neurology & Neurosurgery in the United Kingdom (Table 1 provides individual stroke severity and lesion characteristics data). At each site, all participants meeting inclusion criteria between September 2014 and April 2017 were invited to participate. All participants received standard care through their respective stroke unit during this study. Potential participants were included if they were >18 years of age, had experienced a first-ever ischemic stroke confirmed on imaging with upper limb motor impairment, were medically stable and had a recordable motor evoked potential (MEP; >50 μV) with single-pulse TMS. Those with a history of other neurological disease, recent craniotomy or other acute neurosurgical intervention, inability to provide informed consent, any concurrent medication known to modify seizure threshold or contraindications for TMS such as metallic implants in the skull, history of seizures or implanted permanent pacemaker were excluded.³¹

Table 1. Individual participant stroke severity and lesion characteristics.

ACA, anterior cerebral artery; MCA, middle cerebral artery; NIHSS, National Institutes of Health Stroke Scale; PCA, posterior cerebral artery; PICA, posterior inferior cerebellar artery.

Ipsilesional Data				Contralesional Data
ID	NIHSS	CVA Territory	Cortical/Subcortical	ID	NIHSS	CVA Territory	Cortical/Subcortical
1	7	Left MCA	Cortical	1	2	Right MCA	Cortical
2	2	Right MCA	Subcortical	2	8	Left MCA	Cortical & Subcortical
3	1	Right MCA	Cortical & Subcortical	3	2	Left Border Zone	Cortical & Subcortical
4	14	Left ACA/MCA	Cortical	4	2	Left MCA	Subcortical
5	16	Right ACA	Cortical	5	4	Right MCA	Cortical & Subcortical
6	3	Right MCA	Cortical	6	4	Left MCA	Cortical & Subcortical
7	13	Right MCA	Cortical	7	1	Right MCA	Subcortical
8	4	Left MCA	Cortical & Subcortical	8	5	Right MCA	Cortical & Subcortical
9	3	Right MCA	Subcortical	9	5	Right MCA	Cortical & Subcortical
10	13	Right MCA	Cortical & Subcortical	10	2	Left PCA	Subcortical
11	13	Right ACA/MCA	Subcortical	11	7	Left MCA	Cortical & Subcortical
12	6	Left MCA	Cortical	12	3	Left MCA	Subcortical
13	4	Left MCA	Subcortical	13	4	Left PCA	Subcortical
14	5	Left MCA	Cortical & Subcortical	14	3	Right MCA	Subcortical
15	13	Right MCA	Cortical	15	4	Left MCA	Subcortical
16	11	Left MCA	Cortical & Subcortical	16	5	Left MCA	Cortical & Subcortical
17	3	Right MCA	Subcortical	17	1	Left PCA	Subcortical
18	4	Left MCA	Cortical & Subcortical	18	1	Left MCA	Subcortical
19	17	Right MCA	Subcortical	19	3	Right MCA	Cortical
20	2	Right MCA	Cortical	20	2	Left MCA	Cortical
21	4	Right MCA	Cortical	21	1	Right MCA	Subcortical
22	3	Left MCA	Cortical	22	4	Left MCA	Subcortical
23	5	Right PCA	Cortical	23	9	Left MCA	Cortical
24	6	Left MCA	Cortical	24	3	Left MCA	Subcortical
25	3	Left MCA	Subcortical	25	7	Right MCA	Cortical
26	2	Right MCA	Cortical	26	6	Right MCA	Subcortical
27	3	Left MCA	Cortical	27	3	Right MCA	Subcortical
28	4	Right PICA	Subcortical	28	12	Left MCA	Subcortical
29	4	Right ACA/MCA	Subcortical	29	1	Left MCA	Subcortical
30	14	Right MCA	Cortical & Subcortical
31	7	Right MCA	Subcortical

Stroke severity and upper limb recovery

The National Institutes of Health Stroke Scale (NIHSS) was performed by a trained physician on acute hospital admission. Participants were scored from 0–42 across a series of domains based on clinical examination findings, with 0 indicating no clinical signs of stroke and 42 indicating a severe stroke. The action research arm test (ARAT) and Fugl-Meyer Upper Extremity (FM-UE) quantified upper limb recovery at each neurophysiological test session for the ipsilesional and contralesional datasets respectively. The use of different upper limb scales reflects local clinical standard care at each data collection site. Both the ARAT and FM-UE are valid and reliable assessments of motor function that are sensitive to change and recommended for measurement of upper limb motor recovery.³² The ARAT consists of 19 items grouped into four subscales of grasp, grip, pinch and gross movement, with each item scored from 0 to 3. Higher scores are indicative of greater arm activity, with total scores ranging from 0 to 57. The FM-UE consists of 33 items scored on a three-point ordinal scale (0 to 2). Total possible scores range from 0 to 66, with higher scores indicating reduced upper limb impairment.

Stroke diagnosis

Recent stroke diagnosis was confirmed by an experienced neuroradiologist using either magnetic resonance imaging (MRI) or computed tomography (CT). MRI was acquired on either a Siemens Trio 3T scanner, GE Genesis Signa 1.5T scanner or Siemens Avanto 1.5T scanner. Imaging sequences included T1-weighted, axial T2-weight, fluid-attenuated inversion recovery, and diffusion-weighted imaging as part of each participants routine stroke work-up. CT was acquired on a Siemens SOMATOM Definition AS scanner and included non-contrast CT brain and CT angiography routine stroke sequences.

Electromyography

Surface electromyography (EMG) recorded MEPs from the first dorsal interosseous (FDI) muscle of the paretic (ipsilesional data) or non-paretic (contralesional data) hand using Ag/AgCL electrodes (Ambu, Ballerup, Denmark) in a belly-tendon montage (Figure 1). Skin overlying the FDI was prepared by cleaning with alcohol and lightly abrading with NuPrep paste. A ground strap was placed on the wrist. Signals were sampled at 5kHz (CED 1401, Cambridge Electronic Design, Cambridge, UK), amplified 1000x (CED 1902, Cambridge Electronic Design, Cambridge, UK or Digitimer D360, Welwyn Garden City, Herts, UK), band-pass filtered (20-1000Hz) and stored for offline analysis (Signal software, Cambridge Electronic Design, Cambridge).

Figure 1. Experimental paradigm.

A) Continuous theta burst stimulation (cTBS) response of the ipsilesional motor cortex was assessed in 31 people with stroke over 8 experimental sessions. For the contralesional motor cortex, cTBS response was assessed in 29 people with stroke over 4 experimental sessions. B) Response to cTBS was quantified as a change in MEP amplitude from baseline (left) to post stimulation (right). Pharmacological studies indicate cTBS produces a long-term depression-like response, therefore leading to a decrease in MEP amplitude ²⁶. The decrease of MEP amplitude was used as a measure of plasticity. C) MEPs were recorded using surface EMG from the first dorsal interosseous muscle of the paretic hand (ipsilesional data) or non-paretic hand (contralesional data). D) Experimental paradigm at each session for the ipsilesional data (top) and contralesional data (bottom). B1 and B2 refer to blocks of baseline MEPs. P1, P2, P3 and P4 refer to blocks of MEPs recorded after continuous theta burst stimulation. Note that the differences in post cTBS timepoints for MEP collection reflects standard practice for neurophysiological experiments at each data collection site. This does not influence the analysis of cTBS responses as the modelling accounts for all time points in each participant

Transcranial magnetic stimulation

Single pulse TMS was delivered using a Magstim 200² stimulator (Magstim Co., Whitland, Dyfeld, UK) and figure-of-eight 70mm internal diameter Alpha coil (Magstim Co., Whitland, Dyfeld, UK). Stimulation was applied to either the ipsilesional motor cortex (ipsilesional data) or contralesional motor cortex (contralesional data; Figure 1). The coil was held tangentially to the scalp with the handle positioned 45° posterolateral to induce a posterior-anterior current across the hand motor cortex. The optimal coil position for evoking MEPs in the paretic (ipsilesional data) or non-paretic (contralesional data) FDI muscle at rest was located and marked on the scalp using a water-soluble felt tip marker (ipsilesional data) with coil position consistently monitored throughout experimental procedures, or fixed using Brainsight™ neuronavigation system (contralesional data; Rogue Resolutions Inc., Cardiff, UK). Neuronavigation was guided using surface landmarks and electrophysiological feedback in the form of EMG and without the incorporation of neuroanatomical imaging. Participants wore custom glasses for the duration of testing with a reflective Subject Tracker attached on the side opposite to that receiving stimulation. A second reflector was attached to the TMS coil using a Brainsight™ TMS Coil Tracker Fixation Adaptor. TMS Coil Tracker, Subject Tracker (reflective glasses) and anatomical landmarks were calibrated in 3D space using an NDI Polaris Vicra™ optical infrared position sensor (NDI Medical, Waterloo, Ontario, Canada) fixed to the ceiling. Brainsight™ software provided continuous real-time feedback on coil position and orientation. MEPs collected whilst the coil was more than 1 mm or 3° off target in any plane were discarded and repeated. Resting motor threshold (RMT) was defined as the minimum stimulus intensity required to evoke an MEP in the relaxed FDI with a peak-to-peak amplitude larger than 50μV in at least 5 out of 10 trials. Corticospinal excitability was quantified by recording MEPs and measuring peak-to-peak amplitudes (ipsilesional data, stimulus intensity equal to 120% RMT; contralesional data, stimulus intensity to evoke a 1mV MEP). Baseline corticospinal excitability was determined by recording two blocks of 20 MEPs, separated by a short rest interval (~2 min). Blocks of 20 MEPs were selected to provide high within- and between-session reliability of mean MEP amplitude.³³ Following cTBS, blocks of 20 MEPs were recorded at multiple time intervals (ipsilesional data, 5 min, 15 min, 30 min, 45 min post cTBS; contralesional data, 0 min, 10 min, 20 min, 30 min post cTBS; Figure 1). Difference in the post cTBS timepoints for MEP collection between the ipsilesional and contralesional datasets reflects standard practice for neurophysiological experiments at each data collection site. These experimental differences do not influence the statistical analysis of cTBS responses as the modelling accounts for all time points in each participant. Single TMS pulses were delivered at 0.2Hz ± 10%. For each trial, EMG in a 200 ms pre-stimulus window was visually inspected at high gain to ensure the FDI was at rest. Trials contaminated with pre-stimulus muscle activity were removed. Peak-to-peak amplitude of MEPs were quantified and averaged for each time point. cTBS response was quantified as a change in MEP amplitude from baseline to post stimulation (Figure 1).

Continuous theta burst stimulation

A Magstim Rapid stimulator connected to an air-cooled figure-of-eight coil (Magstim Company, Dyfed, UK) applied cTBS with a biphasic pulse waveform to the optimal scalp position for evoking responses in the FDI. The cTBS protocol consisted of 600 pulses delivered in triplets at 50Hz, repeated at 5Hz for a total of 40 s.³⁴ In healthy adults, there is evidence of good between-session reliability of cTBS.^35,36 A paired cTBS paradigm with a 10 min interval between cTBS trains was applied as it has been reported to induce a greater magnitude and more consistent plasticity response.^37,38 Between cTBS trains, participants were asked to relax and refrain from muscle contraction of the upper limb. The intensity of stimulation was set to 70% RMT, with RMT assessed prior to cTBS application using the rTMS coil.

Statistical analysis

Participant demographics and clinical characteristics were compared between the ipsilesional and contralesional datasets with independent t-tests (age and admission NIHSS) or Pearson’s chi-squared tests (sex, lesioned hemisphere, recombinant tissue plasminogen activator treatment). Changes in response to cTBS across stages of stroke recovery were assessed using linear modelling in R and R Studio using the lme4 and dplyr packages.^39–41 Taking this statistical approach allowed for more complete and accurate analysis of a longitudinal patient dataset with some missing data than would be afforded with a repeated measures ANOVA. It also allowed us to use the actual MEP amplitude at each cTBS time point (including baseline) in each individual. Separate models were run for the ipsilesional and contralesional data. Each model was constructed to have MEP amplitude as the dependent variable, predicted with time pre- and post-cTBS (TIMEPOINT) and timing of the session relative to the date of the infarct (SESSION).

Therefore, TIMEPOINT, SESSION, and the associated TIMEPOINTxSESSION interaction were included in the model as fixed effects. Random effects ensured that each individual had a unique intercept and slope across timepoints for each session. To understand whether the response to ipsilesional cTBS was influenced by lesion location, we ran a similar linear model to that just described on the ipsilesional data including lesion location as a fixed effect (TIMEPOINTxSESSIONxLESION LOCATION). Linear modelling also assessed changes in RMT over the course of stroke recovery. We first assessed whether these variables were best modelled with a fixed slope random intercepts or a random slopes and intercepts approach, before moving forward with the model that produced the lowest Akaike Information Criterion (AIC) or prevented overfitting of the data. In these models, PARTICIPANT was always included as random effect, and we assessed whether SESSION should also be included as a random effect, indicating that each individual has both a unique starting MEP amplitude and a unique slope in change in MEP amplitudes across sessions. Based on the AIC when comparing these models, a fixed effect of SESSION was used. Assumptions of normality and homoscedasticity of the residuals for each model were assessed visually using quantile-quantile normal plots and fitted-versus residual-value plots. To confirm participants with missing data points were not overtly responsible for the study findings, linear modelling for cTBS response was repeated with only participants who had complete datasets. As an exploratory analysis to provide indication of the duration of a critical period of enhanced plasticity, the change in mean MEP amplitudes from baseline to post cTBS at each session were compared with paired t-tests. If a critical period of enhanced cTBS response was identified, the change in cTBS response in that hemisphere and at that time point were correlated with upper limb recovery scores, controlling for baseline assessment of arm function. cTBS response was calculated as the mean post cTBS MEP amplitude normalized to baseline MEP amplitude. Finally, stroke severity (admission NIHSS) and baseline upper limb outcomes were correlated with plasticity responses at each session. For all statistical tests, the level of significance was set at p≤0.05.

Results

Participant demographics and clinical characteristics

A total of 333 experimental sessions were conducted across the two data collection sites with no significant adverse events (Figure 2). There were no differences in age, sex, lesioned hemisphere side or recombinant tissue plasminogen activator treatment between participants in the ipsilesional and contralesional dataset (Table 2). NIHSS scores were in the range of mild to moderate severity; the contralesional group was slightly less severe than the ipsilesional group. Baseline ARAT and FM-UE for the ipsilesional and contralesional datasets respectively are reported in table 2 with both cohorts exhibiting moderate to good levels of upper limb performance on average. The ARAT and FM-UE scores at each session are available in Supplementary Table 1.

Figure 2. Flow of participants through experimental procedures.

The mean ± SD time post-stroke for each session is reported.

MEP, motor evoked potential; PPM, permanent pacemaker

Table 2. Participant demographics and clinical characteristics.

Note, statistical values compare data between the ipsilesional dataset and contralesional dataset. Statistically significant differences are shown in bold.

ARAT, action research arm test; FM-UE, Fugl Meyer Upper Extremity; NIHSS, National Institutes of Health Stroke Scale; rtPA, recombinant tissue plasminogen activator.

	Ipsilesional	Contralesional	Group	Statistics
Age (mean ± SD)	66.6 ± 17.8	68.2 ± 9.8	67.4 ± 14.4	t (58) = 0.44, p = 0.66
Years	(range 26-93)	(range 46-82)	(range 26-93)
Sex				χ2 = 0.43, p = 0.59
Male (n)	20	21	41
Female (n)	11	8	19
Lesioned hemisphere				χ2 = 1.72, p = 0.21
Right (n)	17	11	28
Left (n)	14	18	32
rtPA treated
yes, n (%)	10 (32%)	8 (27%)	18 (30%)	χ2 = 0.16, p = 0.78
NIHSS (mean ± SD)	6.7 ± 4.9	3.9 ± 2.7	5.4 ± 4.2	t(58) = 2.75, p = 0.008
	(range 1-17)	(range 1-12)	(range 1-17)
ARAT at baseline	48.0 ± 14.0	-	-	-
(mean ± SD)	(range 8-57)
FM-UE at baseline	-	61.8 ± 2.9	-	-
(mean ± SD)		(range 56-66)

Neurophysiological data

For a variety of personal and treatment-related reasons, patients did not always return for follow-up on precisely the planned dates or may have missed a session. However, sessions were conducted as close as possible to the intended date (see Figure 2) and the statistical analysis accommodates any missing data points. However, in order to give a simple visual impression of the data we grouped the follow-up assessments into 4 periods: 0-14 days; 15-30 days; 31-60 days; and ≥61 days. This allows us to present the data from both sites in a similar format (Figure 3). In the contralesional data, depression of corticospinal excitability after cTBS was greatest shortly after stroke compared with later time points whereas there is no clear trend in the ipsilesional data. Additional figures presenting the cTBS response for each test session (Supplementary Figure 1) and the linear modelling of the TIMEPOINTxSESSION cTBS response (Supplementary Figure 2) are available in the supplementary material.

Figure 3. Continuous theta burst stimulation response for the ipsilesional hemisphere (left) and contralesional hemisphere (right).

Amplitudes of motor evoked potentials have been normalized to baseline. Error bars are standard deviation.

Note, continuous theta burst stimulation is thought to induce a suppression of cortical excitability. Therefore, a larger decrease in motor evoked potential amplitude provides indication of greater plasticity. Data points below the dashed black line indicate motor evoked potential suppression.

cTBS, continuous theta burst stimulation; MEP, motor evoked potential.

For the statistical model examining cTBS response in ipsilesional hemisphere (ipsilesional data, Table 3), there was a significant negative effect of SESSION, indicating that MEP amplitude decreased over time (β=-0.045, p=0.016). There was no significant main effect of TIMEPOINT (p=0.805), and no interaction between TIMEPOINT and SESSION (p=0.903). Thus, cTBS response appeared to remain approximately constant over the study. Follow-up analysis examining the influence of lesion location confirmed a significant effect of SESSION (β=-0.088, p=0.048), with no interaction between TIMEPOINT and SESSION (p=0.187) and no interaction between TIMEPOINT and SESSION and LESION LOCATION (p=0.173). In the contralesional hemisphere (contralesional data, Table 3) there was an interaction between TIMEPOINT and SESSION that influenced MEP amplitudes (β=0.029, p=0.030). There was a decrease in MEP amplitudes in response to cTBS at the first session (2 weeks post-stroke), but this response to stimulation was reduced at subsequent sessions. This observation was confirmed by comparing the change in mean MEP amplitude from baseline to post cTBS at each session that found a significant decrease in MEP amplitude at 2 weeks (t₍₂₈₎=2.16, p=0.039), but not 4 (p=0.152), 6 (p=0.852) or 26 (p=0.557) weeks post-stroke. Thus, in the contralesional hemisphere, cTBS response was maximum early after stroke and declined over subsequent weeks. RMT and baseline MEP amplitudes for each session and hemisphere are reported in Supplementary Table 2. There were no significant predictors in any of the models examining ipsilesional or contralesional RMT (all p≥0.336, Supplementary Table 3).

Table 3. Fixed effects of motor evoked potential response to continuous theta burst stimulation in the ipsilesional hemisphere (top) and contralesional hemisphere (bottom).

Model Parameter	Beta Estimate	Standard Error	Confidence Interval	p-value
Ipsilesional Data
Intercept	0.919	0.095	0.732 to 1.106	<2e-16*
Timepoint	-0.005	0.021	-0.042 to 0.041	0.805
Session	-0.045	0.019	-0.082 to -0.008	0.016*
Timepoint:Session	0.0005	0.004	-0.007 to 0.008	0.903
Contralesional Data
Intercept	0.877	0.097	0.685 to 1.069	5.11e-15*
Timepoint	-0.077	0.036	-0.148 to -0.006	0.034*
Session	-0.005	0.036	-0.076 to 0.066	0.879
Timepoint:Session	0.029	0.013	0.003 to 0.055	0.030*

^*indicates statistically significant at p ≤ 0.05

These results did not appear to be influenced by those participants with missing data. Re-analysis of cTBS response with only participants who had complete datasets found that there were no significant predictors of MEP amplitude for ipsilesional data. As in the model with all participants, analysis of cTBS response in the contralesional hemisphere found an interaction between TIMEPOINT and SESSION that influenced MEP amplitudes (β=0.03, SE=0.01, 95% CI 0.0004 – 0.053, p=0.050). This relationship was similar, such that there was a decrease in MEP amplitudes in response to cTBS at the first session, but this response to stimulation was reduced at subsequent sessions.

Response to cTBS as a predictor of upper limb recovery

The depression of corticospinal excitability by cTBS was strongest at 2 weeks post-stroke in the contralesional hemisphere. For the contralesional data, upper limb recovery measured with the FM-UE at each time point is reported in Supplementary Table 1. The partial correlations between cTBS response at 2 weeks post-stroke and FM-UE at 4 weeks (rho=0.265, p=0.16), 6 weeks (rho=0.108, p=0.58) or 26 weeks (rho=0.135, p=0.50) were not significant when controlling for baseline FM-UE.

Stroke severity and response to cTBS

NIHSS scores on acute hospital admission were not associated with cTBS response for any session for the ipsilesional (all p>0.16) or contralesional data (all p>0.32). Similarly, upper limb behavior at baseline was not associated with cTBS response for any session for the ipsilesional (all p>0.15) or contralesional data (all p>0.45).

Discussion

This multisite, longitudinal study quantified neural plasticity in the ipsilesional and contralesional human motor cortex over several months following stroke. Our measure of plasticity was the transient reduction in corticospinal excitability produced by cTBS. This was strongest around 2 weeks after stroke in the contralesional motor cortex, dissipating across subsequent sessions. Although we have no measures of pre-stroke cTBS response in our participants, the finding is consistent with the hypothesis that plasticity of the contralesional motor cortex is enhanced early after stroke. In the ipsilesional hemisphere there was no change in response to cTBS over time. As we note below, this could have been due to networks within the ipsilesional hemisphere not being as responsive to cTBS and/or the effect of injury on the response to TMS in that hemisphere.

Assessment of plasticity following stroke in humans

Corticospinal excitability is suppressed for about 30 min following cTBS. Since the effect is blocked by NMDA receptor antagonists, it seems likely that it is due to short-term, LTD-like changes in the efficacy of synaptic connections.^26,34 Recordings of corticospinal volleys evoked by single TMS pulses during the period of reduced excitability suggest that the likely site of action is within the cerebral cortex. These show that cTBS reduces the excitability of circuits generating I1 wave input to corticospinal neurons and imply that cTBS influences intrinsic circuits of the motor cortex.⁴² The reduced I-wave input decreases MEP amplitude following cTBS and provides some indication of the capacity to transiently change synaptic strength in the motor cortex.

Since the responses to theta burst stimulation appear to be relatively consistent in healthy adults,^35,36,43 the fact that it changes over time in the contralesional cortex after stroke suggests that an initial high level of neural plasticity at around 2-4 weeks declines over the next 6 months. This is not dissimilar to animal models, where widespread increased plasticity within days to weeks after injury have been reported, and which declines over time.^10–19 These changes within the contralesional cortex could contribute directly to recovery. For example, in humans, both hemispheres are recruited during execution or learning of unilateral motor tasks,^44–46 suggesting both motor cortices may work cooperatively during motor learning.⁴⁷ In people with stroke, increased activation of the contralesional cortex on imaging is reported during paretic upper limb movement,^48,49 while suppression of contralesional activity has been shown to impair motor performance.⁵⁰

Contrary to our hypothesis, we did not observe any change in neural plasticity over time in the ipsilesional motor cortex. While pre-clinical models have reported plasticity in both hemispheres after stroke, often the most robust effects are observed in ipsilesional hemisphere.^{11–13,17,18,20,21} In humans, reorganization within the ipsilesional hemisphere appears more prominent in mild to moderate stroke severity,^51,52 similar to patients within this study. It is therefore surprising that we did not see a change in plasticity within this hemisphere. This result may be explained by considering the role of long-term potentiation (LTP) and LTD synaptic plasticity. While both potentiation and depression of surviving neural circuits contribute to neural repair after stroke, it is possible there is a shift in plasticity to favor LTP in the ipsilesional hemisphere. In support, pre-clinical models have shown that reducing excessive GABA mediated hypoactivity in the peri-infarct zone after stroke promotes recovery through cellular excitability changes and enhanced LTP.^20,53,54 Other studies have shown upregulation of growth factors, sprouting and proliferation of dendritic spines enable recovery, with potentiation of neural circuits likely to be of paramount importance in the ipsilesional hemisphere.^10–21 In humans, cortical excitability of the ipsilesional hemisphere is reduced early after stroke and recovery appears related to the increase in excitability,⁵⁵ likely mediated through LTP processes. As a result, it is possible that synaptic plasticity within the ipsilesional hemisphere is biased toward promoting LTP, rather than LTD, perhaps explaining our inability to observe a change in the cTBS response over time. This proposed shift toward LTP in the ipsilesional hemisphere is unlikely to be similar in magnitude for the contralesional hemisphere where hyperexcitability is observed after stroke, along with the reduction in excitability correlating with behavior in mild to moderate stroke.⁵⁶ While testing both LTP and LTD plasticity in all patients would help decipher these neural mechanisms in each hemisphere, it is not possible to do so without confounding physiological responses. In addition, as noted earlier, LTP protocols could theoretically impose greater seizure risk after stroke.

There may be other explanations for our inability to observe a change in plasticity over time in the ipsilesional hemisphere. In a large majority of patients in this group, the stroke damaged cortical structures. Cortical lesions result in substantial neural loss, which influences the interaction of non-invasive brain stimulation with cortical circuits and reduces their capacity to be modified by stimulation. In support, studies have shown that although repetitive TMS applied to the ipsilesional motor cortex can modify corticospinal excitability and motor performance in people with subcortical stroke, it is less effective for cortical stroke.^28,29 Furthermore, the hemodynamic response to brain stimulation differs between cortical and subcortical lesions with an increase in blood flow velocity after stimulation in subcortical stroke, whilst this was less prominent in cortical stroke.⁵⁷ We conclude that lack of change in neural plasticity in ipsilesional cortex could be the result of either a bias toward LTP and/or cortical damage produced by the stroke.

Relation of neural plasticity to stroke recovery

There is evidence to support the relationship between synaptic plasticity and motor learning ability in human studies.^58,59 However, in the present study we did not observe an association between plasticity responses and upper limb recovery. This may be due to ceiling effects of the Fugl-Meyer as participants scored an average of ~62 points on admission to the study (maximal score 66) leaving minimal opportunity to quantify recovery. Patients with much lower scores tended to not meet inclusion criteria. Furthermore, given the relatively mild impairment of participants in this study, it could be that any plasticity change induced by the lesion were also mild, reducing opportunity to observe an association with stroke recovery. It may be that in more severe stroke, increased neural damage and greater opportunity for recovery before reaching a ceiling could promote a stronger or prolonged plasticity response. In support, reduced inhibition, a mechanism of plasticity, is more persistent in severe stroke.⁶⁰ It is possible that with a more pronounced increase in plasticity and magnitude of behavioral improvement there may be greater opportunity to observe a relationship between plasticity and recovery.

Clinical implications

A period of enhanced plasticity early following stroke suggests a therapeutic window may exist that could enable greater recovery from impairment. In support, most human studies suggest earlier rehabilitation is beneficial.^25,61–64 However, we refer readers to other work suggesting very early initiation of rehabilitation could be detrimental.⁶⁵ Perhaps of greatest concern, delays in initiating rehabilitation could result in patients receiving limited, or no therapy, within this period of increased plasticity. Although early rehabilitation is recommended in many stroke guidelines, evidence indicates less than eight minutes of daily therapy is dedicated to upper limb rehabilitation within the first 4 weeks of stroke,⁶⁶ a period that overlaps with a window of enhanced plasticity. Therapy delivered early after stroke is likely critical to maximize recovery as it would engage a period of heightened plasticity.

Future directions

A novel, but clinically important future direction would be to identify techniques to prolong, or even re-open, a period of increased plasticity. This may enhance the capacity for behavioral restoration and reduce persistent disability following stroke. There is some evidence from pre-clinical studies that re-opening a period of enhanced plasticity is beneficial for behavioral recovery. In rodents that had experienced a stroke with incomplete recovery, a second, subsequent stroke was shown to re-open a window of enhanced plasticity that enabled full recovery from the initial ischemic event.⁹ While this is not a feasible treatment in humans, there are several promising therapies such as pharmacological and cellular therapies, non-invasive brain stimulation and cardiovascular exercise that may be capable of prolonging or re-opening a period of increased plasticity.^26,67,68 These therapies could represent an exciting opportunity for greater recovery after stroke.

Limitations

There are several limitations to this study that should be acknowledged. First, there were slight differences in neurophysiological methodology at the two experimental sites. It could be that differences in plasticity between hemispheres is partially underpinned by the disparity in methodology between sites. However, we suggest this is unlikely as both sites delivered identical cTBS paradigms with change in MEP amplitude as an outcome to test for a critical period of enhanced LTD-like plasticity. In addition, statistical modelling accounted for differences in timing of MEPs recorded after cTBS and both hemispheres were analyzed separately. It is therefore highly unlikely that minor differences in methodology between sites is an explanation for different plasticity responses between hemisphere. Second, although neuronavigation equipment was not available at one site (ipsilesional data), previous literature emphasizes both navigated and non-navigated TMS to the motor cortex are similar in terms of variability and reproducibility of MEPs.⁶⁹ Nevertheless, small variations in coil placement can influence MEP measurements.⁷⁰ Therefore, it is not inconceivable that the non-navigated approach for the ipsilesional data could have confounded plasticity measurements. Third, several datapoints were missing, predominantly for the ipsilesional hemisphere reflecting practicalities and challenges of testing acutely unwell stroke survivors within a complex medical setting. That this was more common for the ipsilesional dataset may reflect the greater number of experimental sessions, starting earlier after stroke and finishing at 12 months. Different analytical approaches of including or excluding participants with missing data led to the same overall result, providing some level of confidence our results were not driven by those participants who had missing datapoints. Finally, we acknowledge that in assessing plasticity of the motor cortex, our results are specific to cTBS which is thought to resemble an LTD-like synaptic plasticity response.²⁶ It is possible that the temporal characteristics of other plasticity mechanisms may differ to that reported here.

Conclusion

In conclusion, our results provide the first physiological evidence to demonstrate a period of enhanced neural plasticity following stroke in humans. Our study provides neurophysiological support for an intense, front-end loaded approach to post-stroke rehabilitation. Therapy delivered within the first few weeks post-stroke could coincide with a critical period of enhanced plasticity and be especially effective.