Operant conditioning of the tibialis anterior motor evoked potential in people with and without chronic incomplete spinal cord injury

Abstract

The activity of corticospinal pathways is important in movement control, and its plasticity is essential for motor skill learning and re-learning after central nervous system (CNS) injuries. Therefore, enhancing the corticospinal function may improve motor function recovery after CNS injuries. Operant conditioning of stimulus-induced muscle responses (e.g., reflexes) is known to induce the targeted plasticity in a targeted pathway. Thus, an operant conditioning protocol to target the corticospinal pathways may be able to enhance the corticospinal function. To test this possibility, we investigated whether operant conditioning of the tibialis anterior (TA) motor evoked potential (MEP) to transcranial magnetic stimulation can enhance corticospinal excitability in people with and without chronic incomplete spinal cord injury (SCI). The protocol consisted of 6 baseline and 24 up-conditioning/control sessions over 10 wk. In all sessions, TA MEPs were elicited at 10% above active MEP threshold while the sitting participant provided a fixed preset level of TA background electromyographic activity. During baseline sessions, MEPs were simply measured. During conditioning trials of the conditioning sessions, the participant was encouraged to increase MEP and was given immediate feedback indicating whether MEP size was above a criterion. In 5/8 participants without SCI and 9/10 with SCI, over 24 up-conditioning sessions, MEP size increased significantly to ~150% of the baseline value, whereas the silent period (SP) duration decreased by ~20%. In a control group of participants without SCI, neither MEP nor SP changed. These results indicate that MEP up-conditioning can facilitate corticospinal excitation, which is essential for enhancing motor function recovery after SCI.

NEW & NOTEWORTHY We investigated whether operant conditioning of the motor evoked potential (MEP) to transcranial magnetic stimulation can systematically increase corticospinal excitability for the ankle dorsiflexor tibialis anterior (TA) in people with and without chronic incomplete spinal cord injury. We found that up-conditioning can increase the TA MEP while reducing the accompanying silent period (SP) duration. These findings suggest that MEP up-conditioning produces the facilitation of corticospinal excitation as targeted, whereas it suppresses inhibitory mechanisms reflected in SP.

INTRODUCTION

Restoring lost or disturbed motor function after spinal cord injury (SCI) is an important step toward regaining an active, independent life. However, recovery can be difficult, and individuals with SCI frequently have residual motor disabilities even after completing conventional therapies. Foot drop, weak dorsiflexion of the ankle during walking, is one of the most common and disabling impairments in people with chronic SCI (van Hedel et al. 2005).

Normally, corticospinal excitability for the ankle dorsiflexor tibialis anterior (TA) is high before and during the swing phase of walking (Capaday et al. 1999; Schubert et al. 1997), and the activity of the motor cortex is directly involved in the activation of the TA during walking (Petersen et al. 2001). Incomplete SCI that disturbs corticospinal pathways results in smaller motor evoked potential (MEP) amplitudes to transcranial magnetic stimulation (TMS) (Brouwer and Hopkins-Rosseel 1997; Calancie et al. 1999; Davey et al. 1999; McKay et al. 2005; van Hedel et al. 2007), indicating lower corticospinal excitability and weaker corticospinal connection, which would explain (at least partially) why foot drop occurs frequently in people after SCI (Barthélemy et al. 2010; van Hedel et al. 2005; Wirth et al. 2008a, 2008b). Several studies have shown that MEP amplitude increases in association with motor function recovery, even in the chronic stage; it increases in response to constraint-induced therapy (Liepert et al. 1998b, 2000a, 2000b), partial body weight support treadmill training (Thomas and Gorassini 2005), and functional electrical stimulation (Everaert et al. 2010; Stein et al. 2006). Thus an intervention that aims to increase the excitability of corticospinal pathways may enhance motor function recovery. Currently, the most common treatment for foot drop is an orthosis (e.g., ankle foot orthosis), which does not aim at improving corticospinal function. Operant conditioning of stimulus-induced muscle responses (e.g., reflexes) (Thompson and Wolpaw 2014a) that can target plasticity to a specific central nervous system (CNS) pathway (Thompson and Wolpaw 2015; Wolpaw 2010) may be used to increase the corticospinal excitability, and thereby improve corticospinal function.

Operant conditioning is a powerful method for modifying a behavior based on its consequences. Nearly four decades of reflex operant conditioning studies in monkeys, rats, mice, and humans have indicated that even a simple spinal reflex behavior can be gradually changed through learning and practice (Thompson and Wolpaw 2014a; Wolpaw 1997, 2007; Wolpaw and Tennissen 2001). This type of behavioral change involves a hierarchy of spinal and supraspinal plasticity (Chen et al. 2002, 2006a; Chen and Wolpaw 2002, 2005; Wolpaw and Chen 2006). Thus an operant conditioning protocol may be used as a gateway to facilitating specific patterns of CNS activity that are potentially linked to better motor function recovery after SCI (Wolpaw 1997; 2006; 2007; Wolpaw and Tennissen 2001). With the operant conditioning approach, CNS plasticity is guided (not nonspecifically induced). This is uniquely different from currently available therapies that expect the induction of beneficial CNS plasticity through practicing complex movements such as locomotion (Koski et al. 2004; Thickbroom et al. 2004; Yen et al. 2008). Both animal and human data support the value of this new strategy to induce the targeted plasticity (Chen et al. 2006b; Thompson and Wolpaw 2012, 2015).

Several years ago, we investigated operant conditioning of the soleus H-reflex in people with chronic incomplete SCI and showed that decreasing the H-reflex through down-conditioning is possible in this population and can improve walking (Thompson et al. 2013b). Encouraged by such results, in the present study, we investigated whether operant up-conditioning of the MEP can increase the corticospinal excitability for the ankle dorsiflexor TA in people with and without chronic incomplete SCI. We hypothesized that the up-conditioning protocol will increase MEP size in the majority of participants with and without SCI, whereas the control protocol, which consists of a similar number of MEP trials as the conditioning protocol without operant conditioning, will not increase MEP size. It was predicted that if up-conditioning could increase MEP size, the conditioned MEP increase would be the sum of rapid within-session increase that is thought to reflect immediate change in cortical influence and across-session MEP change that would reflect the long-term plasticity of the targeted pathway, similar to H-reflex conditioning in people with and without SCI (Makihara et al. 2014; Thompson et al. 2009a, 2013b). We also expected that the duration of silent period (SP) after MEP, which reflects cortical inhibition at least partly, would not increase with MEP increase through up-conditioning, because the conditioning protocol would target the MEP and not the SP.

MATERIALS AND METHODS

Study overview.

The MEP operant conditioning protocol used in this study is a modified version of a reflex operant conditioning protocol developed and used in people with and without SCI (Makihara et al. 2014; Thompson et al. 2009a, 2013b). The study protocol was approved by the Institutional Review Board of Helen Hayes Hospital, New York State Department of Health, and the Institutional Review Board of the Medical University of South Carolina. All participants gave written consent before participation.

At the beginning of the study, the participant was familiarized with the protocol and TMS in several preliminary sessions. Each participant then completed 6 baseline sessions and 24 up-conditioning or control sessions that occurred at a pace of three times per week (Fig. 1A). To avoid variability due to possible diurnal variation in response size (Lagerquist et al. 2006; Tamm et al. 2009), each individual’s sessions always occurred at the same time of day (i.e., within the same 3-h time window). A typical baseline, control, or conditioning session, which included 225 (for baseline and control session) or 20 + 225 (for conditioning session) MEP trials, took ~1 h. In all sessions, TA MEPs were measured while the participant sitting in a chair provided a preset level (see EMG recording and session protocol) of TA background EMG with ankle, knee, and hip joint angles fixed at approximately −10, 60, and 70°, respectively.

Fig. 1.

A: session schedule. Six baseline sessions were followed by 24 conditioning or control sessions that occurred at a rate of 3 sessions/wk. In the conditioning group participants (both those with and without spinal cord injury, SCI), 2 follow-up sessions occurred at 1 and 3 mo after the last conditioning session. Along with the session counts, the time elapsed since the beginning of motor evoked potential (MEP) conditioning (or control) is shown in weeks. B: setup view. MEPs were measured while the participant sat in a chair with ankle, knee, and hip angles fixed at approximately −10, 60, and 70° in a custom-made apparatus. C: composition of baseline, conditioning, and control sessions. Isometric maximum voluntary contraction (MVC) was measured as the tibialis anterior (TA) electromyography (EMG) amplitude. Common peroneal nerve (CPn) stimulation was used for the maximum M-wave and H-reflex measurements. D: visual feedback screens for control and conditioning trials. In all trials, the number of the current trial within its block is displayed at top right, and the background EMG panel shows the correct range (hatching) and the ongoing EMG activity (green vertical bar, updated every 200 ms). If TA EMG activity stays in the correct range for at least 2 s and at least 5.5 s has passed since the last trial, an MEP is elicited. In control trials (left), the MEP panel is not shown. In conditioning trials (right), hatching in the MEP panel indicates the rewarded MEP range for up-conditioning. The thick horizontal bar is the average MEP size of the baseline sessions, and the vertical bar is the MEP size [i.e., the average rectified EMG in the MEP interval (e.g., 35–50 ms after transcranial magnetic stimulation, TMS)] for the most recent trial (it appears 200 ms after TMS). If that MEP size reaches into the hatched area, the bar is green and the trial is a success. If it falls below the hatched area, the bar is red and the trial is a not a success. The running success rate for the current block is shown at bottom right.

In all 225 trials of baseline sessions and the first 20 trials of conditioning sessions, the participant received no feedback on MEP size (i.e., control trials). The participant simply maintained the preset level of TA electromyographic (EMG) activity while the MEP was elicited. During 225 conditioning trials of each conditioning session (i.e., in the conditioning participants), the participant was encouraged to increase TA MEP size and received immediate feedback as to whether MEP size was above a criterion (see EMG recording and session protocol and Visual feedback). For all trials, MEPs were elicited at the same absolute TMS intensity (~10–15% above active motor threshold).

Study participants.

Fifteen adults (5 men and 10 women, aged 30 ± 8 yr, mean ± SD) with no known neurological conditions and 10 adults (7 men and 3 women, aged 49 ± 11 yr) with well-defined stable impairment of weak ankle dorsiflexion (i.e., foot drop) due to a spinal cord lesion participated in this study. For participants with SCI, a physiatrist or a neurologist determined each prospective individual’s eligibility for the study. The profiles of participants with SCI are summarized in Table 1. The inclusion criteria were 1) neurologically stable (>1 yr post-SCI) 2) medical clearance to participate, 3) ability to ambulate at least 10 m with or without an assistive device (e.g., walker, cane, and crutches), 4) signs of weak ankle dorsiflexion at least unilaterally (i.e., manual muscle strength test score <5), and 5) expectation of no medication change at the time of study enrollment. Stable use of anti-spasticity medication (e.g., baclofen, diazepam, tizanidine) was permitted. Exclusion criteria were 1) motoneuron injury, 2) known cardiac condition (e.g., history of myocardial infarction, congestive heart failure, pacemaker use), 3) medically unstable condition, 4) cognitive impairment, 5) a history of epileptic seizures, 6) metal implants in the cranium, 7) implanted biomedical device in or above the chest (e.g., cardiac pacemaker, cochlear implant), 8) no measurable MEP elicited, 9) unable to produce any voluntary TA EMG activity, and 10) use of functional electrical stimulation to the leg on a daily basis.

Table 1.

Profiles of study participants with chronic incomplete SCI

Participant	Age, yr	Sex	Cause of Spinal Cord Damage	SCI Level	AIS	Time Since SCI, yr	Baclofen
1	61	F	NT	T12	D	N/A*	Yes
2	52	M	T	C5	D	3	No
3	29	M	T	C7	C	6.5	No
4	67	M	NT	T5	D	10	Yes
5	45	M	T	C4	D	10	Yes
6	47	M	T	C4	C	4	Yes
7	42	M	T	C4	D	1.5	Yes
8	54	M	T	C6	D	10	No
9	55	F	NT	C6	D	5.5	No
10	39	F	T	C4	C	6	Yes

F, female; M, male; NT, nontrauma; T, trauma; SCI, spinal cord injury; AIS, American Spinal Injury Association impairment scale.

^*Post-SCI duration cannot be determined for this participant due to her diagnosis (hereditary spastic paraparesis).

To examine the feasibility of an MEP conditioning protocol and its effects on corticospinal excitability, participants without neurological conditions were exposed to either the conditioning protocol, which consisted of 6 baseline sessions and 24 conditioning sessions (i.e., conditioning group, n = 8), or the control protocol, which consisted of 6 baseline sessions and 24 control sessions (i.e., control group, n = 7). All participants with SCI were exposed to the conditioning protocol (i.e., 6 baseline + 24 conditioning sessions). In the conditioning group participants (both with and without SCI), two follow-up sessions occurred at 1 and 3 mo after the last (i.e., 24th) conditioning session. The follow-up session procedures were exactly the same as those of the conditioning session.

EMG recording and session protocol.

The baseline, control, and conditioning session protocols are summarized in Fig. 1C. At the beginning of each session, self-adhesive surface Ag-AgCl electrodes (2.2 × 3.5 cm; Vermed, Bellows Falls, VT) were placed over the muscle bellies of TA and soleus for EMG recording (Thompson et al. 2009b, 2011). EMG signal was amplified, bandpass filtered (10–1,000 Hz), and sampled at 3,200 Hz. Maximum voluntary contraction (MVC) was measured as the absolute TA EMG amplitude during two 3-s trials of isometric maximum dorsiflexion while the participant sat in a chair with ankle, knee, and hip joints fixed at approximately −10, 60, and 70°, respectively, in a custom-made apparatus (Fig. 1B). To measure the TA maximum M-wave (M_max) and the maximum H-reflex (H_max), the common peroneal nerve (CPn) was stimulated at the neck of fibula, using surface electrodes (2.2 × 2.2 cm for cathode, 2.2 × 3.5 cm for anode; Vermed) and 0.5-ms-wide single square pulses delivered from a Grass S48 stimulator with an SIU-5 stimulation isolation unit and a CCU1 constant current unit (Natus Neurology, West Warwick, RI). This measurement was made with ankle, knee, and hip joints fixed at approximately −10, 60, and 70°, respectively (see Fig. 1B), while the sitting participant maintained ~15% MVC level of TA background EMG (participants without SCI) or was at rest (participants with SCI). Four EMG responses were averaged at each of ~10 stimulus intensities varied from below M-wave threshold to just above the M_max (Thompson et al. 2013a).

Following CPn stimulation, control MEPs were elicited while the participant simply produced the preset level of stable TA background EMG activity. This background level, which was maintained constant throughout data collection, was set at ~15% MVC in participants without SCI. In individuals with SCI, in whom the MEP size tends to be unstable and highly variable at a low level of background EMG activation and voluntary control of a small (i.e., fine) amount of muscle contraction is often difficult (Davey et al. 1999; McKay et al. 2005), the TA background EMG level was generally aimed at ~30–35% MVC. All 225 MEP trials (i.e., 3 × 75-trial blocks) of each baseline session in both the conditioning and control groups and the first 20 trials of each conditioning session in the conditioning group were done as control trials, with no feedback on MEP size. There was a short (usually 2–3 min) break between the MEP trial blocks. The minimum intertrial interval was 5.5 s.

In the control group, the control sessions were exactly the same as the baseline sessions; that is, the control session protocol consisted of TA MVC measurement, CPn stimulation, and 3 × 75 control MEP trials. In contrast, in 225 conditioning trials of each conditioning session (i.e., conditioned MEPs), the conditioning group participant was encouraged to increase TA MEP size and received immediate feedback as to whether MEP size was above a criterion (i.e., whether the trial was a success). For the details of feedback and reward criteria, see Visual feedback and Fig. 1D.

For all MEP trials, TMS at 10–15% above active threshold was used to evoke the MEP. The same absolute TMS intensity [expressed as a percentage of maximum stimulator output (%MSO)] was used throughout the study. TA and soleus background EMG levels were kept stable throughout data collection. To minimize the session-to-session variability in CPn stimulation and EMG recording, the positions of all electrodes were measured in relation to landmarks on the skin (e.g., scars or moles) during the first preliminary session, and the same measures were used to place the electrodes in all subsequent sessions.

Visual feedback.

Figure 1D shows the visual feedback provided to the participant during MEP trials. The screen presented two graphs, one for background EMG activity (left) and one for MEP size (right). The background EMG graph was the same for both control and conditioning trials. When the participant had kept the background EMG bar in the specified range for the past 2 s and at least 5.5 s had passed since the last trial, TMS was triggered.

The MEP feedback graph was shown only during conditioning trials. It displayed a vertical bar reflecting MEP size. MEP size was calculated as the average rectified EMG in the MEP interval, which was determined for each participant (e.g., for the participant in the Fig. 2A example, 35–50 ms after TMS was used). The MEP size bar appeared or refreshed 200 ms after TMS. The bar became green if MEP size was larger than the criterion (i.e., the lower border of the shaded area in Fig. 1D), and the bar became red if MEP size was smaller than the criterion. The criterion was based on the average MEP size for the previous block of trials. In each conditioning session, the criterion value for the first block of 75 conditioning trials was determined based on the immediately preceding block of 20 control trials, and the criterion values for the second and third conditioning blocks were based on the immediately preceding block of 75 conditioning trials. The criterion was selected so that if MEP values for the new block were similar to those for the previous block, 50–60% of the trials would be successful (Chen and Wolpaw 1995; Thompson et al. 2009a). For each block, the participant earned a modest extra monetary reward if the success rate exceeded 50%. [See Thompson et al. (2009a) for full details of the protocol.] No specific instructions on how to increase MEP size were given. During the conditioned trials, the participant was encouraged to come up with his or her own strategies to increase MEP size and was reminded to try for success on every conditioning trial. If requested by the participant, subjective strategies for changing the H-reflex size reported by previous H-reflex conditioning participants (Thompson et al. 2009a) were provided (with a reminder that those strategies reported by the others may or may not apply to the participant.) The common subjective strategies included “meditation” and “anticipation of stimulus occurrence.”

Fig. 2.

Typical examples of tibialis anterior (TA) motor evoked potential (MEP) in a conditioning group participant (A and C) and a control group participant (B and D) without known neurological conditions. A: peristimulus raw electromyography (EMG) sweeps from the second 75-trial block of the 6th baseline session and the last (i.e., 24th) conditioning session. For each, 75 sweeps are superimposed. B: peristimulus raw EMG sweeps from the second 75-trial block of the 6th baseline session and the 24th control session. For each, 75 sweeps are superimposed. C: rectified EMG signals from the 6th baseline session (dashed) and the 24th conditioning session (solid). D: rectified EMG signals from the 6th baseline session (dashed) and the 24th control session (solid). For each sweep of C and D, 225 responses were averaged together. A shaded band indicates the time window for each participant’s MEP size calculation. The horizontal dashed line indicates the background EMG level. Arrows (B6 and C24) indicate the ends of silent period.

Transcranial magnetic stimulation.

TMS was performed using Magstim 200⁻² and a 110-mm double-cone coil (in participants with SCI) or a custom-made batwing coil (in participants without SCI) with radii of 9 cm (Jali Medical, Woburn, MA) held over the scalp such that the induced current flowed in the posterior–anterior direction in the brain. When the absolute TA EMG level was maintained within the preset range (~15% MVC level for participants without SCI and ~30–35% MVC level for participants with SCI), an MEP was elicited by TMS, with a minimum interval of >5.5 s between stimuli. For each participant, this absolute background EMG level was determined during preliminary sessions and was not changed throughout the entire study with the expectation that the M_max would not change over the course of study (Thompson et al. 2009a, 2013b). In preliminary sessions, the optimal TMS location at which the lowest stimulus intensity elicited the TA MEP was determined by moving the coil over the scalp. At the optimal location, the input-output curve of the corticospinal pathway was measured by increasing the TMS intensity, represented as %MSO, in steps of 5% until the MEP reached its plateau. Four MEPs were collected at each intensity, and the peak-to-peak and mean rectified amplitudes were plotted against the stimulus intensity (Kido Thompson and Stein 2004; Knash et al. 2003; Thompson et al. 2006, 2011). From the input-output curve measurement in several preliminary sessions, a stimulus intensity that was ~10–15% above threshold was determined for each participant, and the same absolute TMS intensity was used throughout all the baseline and control/conditioning sessions.

Along with control and conditioning MEP size measurements, we also measured the SP as the period from the end of the MEP (i.e., absolute EMG amplitude fell below the prestimulus level) to the recovery of EMG activity to the prestimulus level in each MEP trial. For each block of 75 trials or 20 trials, the median value was then calculated and determined as the SP duration for that block. This method measured the SP similarly to several previous studies (Garvey et al. 2001; Kido Thompson and Stein 2004; Knash et al. 2003; Thompson et al. 2011).

Data analysis.

In all participants, MEP sizes were calculated for each session. Regardless of whether a participant belonged to the control group or conditioning group, for each participant, MEPs from all three 75-trial blocks were averaged together and called “conditioned MEPs,” and MEPs from the first 20 control trials (i.e., the first 20 trials of the first block of 75 trials for baseline and control sessions and 20 within-session control trials for conditioning sessions) were averaged together and called “control MEPs.” For these calculations, MEP size was defined as average absolute EMG in the MEP interval (determined for each participant; e.g., for the individual in the example of Fig. 2A, 35–50 ms after TMS) minus average background EMG. In addition, the extent of within-session task-dependent MEP change was calculated as the difference between the conditioned MEP and the control MEP [i.e., task-dependent adaptation; see Thompson et al. (2009a) for further explanation]. Changes in these MEP sizes across sessions were quantified as percentages of their average values for the six baseline sessions. We also determined for each participant the final effects of conditioning on the conditioned MEP by averaging the conditioned MEPs for conditioning sessions 22–24 (i.e., the last 3 conditioning sessions) and expressed the result as a percentage of the average MEP for the six baseline sessions. (Thus a value of 100% indicates no change in the MEP.) The final effect on the control MEP was calculated by averaging the control MEPs for conditioning sessions 22–24, and the result was expressed as a percentage of the average control MEP of the six baseline sessions (see Thompson et al. 2009a).

Similar to the MEPs, SPs from all three 75-trial blocks were averaged together and called “conditioned SPs,” and SPs from the first 20 control trials (i.e., the first 20 trials of the first block of 75 trials for baseline and control sessions and 20 within-session control trials for conditioning sessions) were called “control SPs,” regardless of whether a participant belonged to the control group or conditioning group.

Statistical analyses.

To determine for each participant whether MEP up-conditioning was successful, the average conditioned MEPs of the final six conditioning sessions were compared with the average MEPs of the six baseline sessions by unpaired t-test (two-tailed) with the α level set at 0.05 (Thompson et al. 2009a, 2013b). To evaluate the stability of the M_max and background EMG levels over all sessions and assess systematic changes in MEP size, MVC, and SP, a one-way repeated-measures ANOVA (sessions × participants) was used across successive six-session blocks (i.e., baseline sessions 1–6 and conditioning sessions 1–6, 7–12, 13–18, and 19–24; Thompson et al. 2009a). (Note that an ANOVA was applied to the data from the baseline and conditioning/control sessions only, not including the data for the follow-up sessions.) The α level was set at 0.05. When the ANOVA revealed a significant effect of sessions, Dunnett’s test was applied post hoc, to compare a value of each six-session block with the baseline value. In results, we report the statistical results for the conditioning group of individuals without SCI in whom MEP up-conditioning was successful, the control group of individuals without SCI, and the conditioning group of individuals with SCI in whom MEP up-conditioning was successful.

RESULTS

TA M_max, background EMG, and MVC.

In both the control and conditioning groups of participants without known neurological conditions and in participants with SCI, the TA M_max value and TA and soleus background EMG levels in control and conditioning trials remained stable throughout the study. Typically, TA M_max (averaged 4.5 ± 0.6 mV in participants without SCI and 2.9 ± 0.5 mV in participants with SCI during the baseline sessions; means ± SE) remained within ±10% of the baseline average across all the sessions, TA background EMG remained within ±10% of the baseline average, and soleus background EMG remained within the resting EMG level (i.e., <8 µV) throughout the study. In all three groups, F values for these parameters evaluated by one-way repeated-measures ANOVA were 0.4–2.0, with P values of 0.12–0.97. These results confirmed the stability of EMG recording and nerve stimulation conditions in this study, in support of the validity of the methodology.

TA MVC did not change across the sessions in participants without SCI; the average MVC over conditioning sessions 19–24 was 100 ± 5% of the baseline for the control group [F_(4,24) = 0.5, P = 0.73 by one-way repeated-measures ANOVA] and 103 ± 6% of the baseline for the conditioning group [F_(4,16) = 0.9, P = 0.46]. In participants with SCI, the MVC increased insignificantly; the average MVC over conditioning sessions 19–24 was 112 ± 6% of the baseline [F_(4,32) = 1.9, P = 0.13 by one-way repeated-measures ANOVA].

Changes in the TA MEP in participants without neurological conditions.

In five of eight conditioning participants, the average conditioned MEP for conditioning sessions 19–24 was significantly larger than that for the six baseline sessions. In the other three conditioning participants, the MEP did not increase significantly. The success rate for these MEP conditioning participants (i.e., 5/8, or 63%) is slightly less than that for the H-reflex up-conditioning participants (i.e., 6/8, or 75%; Thompson et al. 2009a) or for normal H-reflex conditioning monkeys, rats, and mice (i.e., 75–80%; Carp et al. 2006; Chen and Wolpaw 1995; Wolpaw 1987; Wolpaw et al. 1983). In the control group, none of the individual participants showed a significant increase in MEP (P > 0.05 in each of 7 participants), and the control group as a whole showed no increase in the conditioned MEP (final MEP size was 96 ± 10% of baseline). The conditioning group (n = 8) and the control group (n = 7) differed significantly in final MEP size (P = 0.03 by unpaired t-test). Thus, MEP increase was specific to the conditioning group.

Figure 2 shows typical examples of TA MEPs in a conditioning group participant and a control group participant. Figure 2, A and B, displays peristimulus raw EMG sweeps from the second 75-trial block of baseline session 6 and conditioning (or control) session 24. For each panel, 75 sweeps are superimposed. Figure 2, C and D, shows average rectified EMG signals from baseline session 6 and conditioning session 24, with each sweep being an average of 225 responses. These clearly illustrate an increase in the size of conditioned MEP over 24 up-conditioning sessions, while little change is seen after 24 control sessions. In the individual of Fig. 2, A and C, and other conditioning group participants, a similar change was also observed in the control MEP (not shown).

Successful conditioning participants reported that they tried several different strategies for increasing the MEP and identified the ones that would work within the first two to three conditioning sessions, as noted in our previous H-reflex conditioning studies (Thompson et al. 2009a, 2013b). In later conditioning sessions, the participants frequently commented that they needed to add modifications or adjustments to their strategies to further increase the MEP. The most common strategy mentioned by these participants was intense focus (on something, which varied among participants). In contrast, three conditioning participants in whom the MEP did not increase significantly rarely mentioned their strategies during casual conversations that usually occurred during the conditioning sessions. Topics of those conversations are often something unrelated to conditioning, unless redirected by the investigators.

Figure 3 shows the average courses of MEP changes for successful conditioning group participants (A, C, E) and for control group participants (B, D, F). Figure 3, A and B, shows the conditioned MEP change. In the conditioning group, the conditioned MEP gradually increased over the course of 24 conditioning sessions; the final average value of the conditioned MEP (i.e., the average of conditioning sessions 22–24) reached 150 ± 13% of baseline value. In contrast, in the control group, the MEP did not increase systematically (the final MEP size was 96 ± 10% of baseline value). Figure 3, C and D, shows the control MEP change. The control MEP increased in parallel to the conditioned MEP in the conditioning group; the final control MEP size was 133 ± 7% of baseline value. Whereas in the control group it did not change; the final control MEP size was 101 ± 12% of baseline. Figure 3, E and F, shows the change in the within-session difference between the conditioned and control MEPs. For the conditioning group, this difference represents task-dependent adaptation; that is, the MEP increase that the participants were able to produce immediately when they were asked to increase it. For the control group, this difference represents acute effects of repeated single-pulse TMS, if they exist. In the conditioning group, there seemed to be some within-session increase toward the end (+12 ± 10%), although it was statistically not significant. There was no within-session increase in the control group. Time courses of these changes with results of repeated-measures ANOVA and post hoc tests are summarized in Table 2.

Fig. 3.

Time course of motor evoked potential (MEP) changes over the course of study in participants without known neurological conditions. A, C, and E: average (±SE) MEP values for baseline and conditioning sessions for conditioning participants in whom the MEP increased significantly (n = 5). B, D, and F: average (±SE) MEP values for control participants (n = 7). A and B: average conditioned MEP size. C and D: average control MEP size. E and F: average of conditioned MEP size minus control MEP size (i.e., task-dependent adaptation; for details see Thompson et al. 2009a).

Table 2.

TA MEP values in participants with no known neurological conditions

	F Value	C1–6	C7–12	C13–18	C19–24
Conditioned MEP, %baseline
Conditioning group	F(4,16) = 5.2	118.3 ± 9.4	120.6 ± 14.3	127.8 ± 11.4*	143.7 ± 12.0*
Control group	F(4,24) = 0.4	99.0 ± 4.8	105.5 ± 5.3	97.1 ± 6.3	102.0 ± 8.6
Control MEP, %baseline
Conditioning group	F(4,16) = 3.6	110.4 ± 2.6	119.1 ± 7.5	126.3 ± 14.7*	132.1 ± 6.5*
Control group	F(4,24) = 0.9	103.9 ± 4.9	111.5 ± 7.3	97.8 ± 8.4	108.8 ± 11.5
Within-session change, %baseline
Conditioning group	F(4,16) = 0.8	7.9 ± 10.1	1.5 ± 10.9	1.5 ± 11.1	11.7 ± 9.7
Control group	F(4,24) = 1.2	−5.0 ± 3.3	−6.1 ± 4.3	−0.7 ± 3.9	−6.9 ± 4.0

Values are group means ± SE expressed as percentages of baseline tibialis anterior (TA) motor evoked potential (MEP) for the conditioning group of individuals in whom up-conditioning was successful (n = 5) and for the control group (n = 7). C1–6, C7–12, C13–18, and C19–24 are conditioning (or control) sessions 1–6, 7–12, 13–18, and 19–24, respectively.

^*P < 0.05, significant difference from the 6 baseline sessions (Dunnett’s post hoc after one-way repeated-measures ANOVA).

Overall, these results clearly show that the TA MEP increase in the conditioning participants is due to their exposure to the up-conditioning paradigm; the MEP did not change in the control participants, who received a similar number of MEP trials over the same study duration (i.e., 10 wk) without operant conditioning. In the follow-up sessions in the conditioning group, MEP changes that were present at the end of the conditioning sessions remained present. The conditioned MEP was 154 ± 11% at the 1-mo follow-up and 150 ± 20% at the 3-mo follow up, and the control MEP was 136 ± 18% at the 1-mo follow-up and 131 ± 17% at the 3-mo follow up.

The time course of MEP changes in the conditioning group was somewhat unexpected. Although the conditioned MEP gradually increased over the course of 24 conditioning sessions as expected, its composition differed from that of the previous soleus H-reflex conditioning, where the conditioned reflex change consisted of a significant amount of early onset within-session reflex change (~13% increase) and delayed onset control reflex change (~27% increase) (Thompson et al. 2009a). With MEP conditioning, the conditioned MEP increase was predominantly from the control MEP change from early on until conditioning session 17–18, and the control MEP increase was not necessarily preceded by the development of within-session increase (see Table 2).

MEP changes in participants with incomplete SCI.

In 9 of 10 participants with chronic incomplete SCI, the average conditioned MEP for conditioning sessions 19–24 was significantly larger than that for the six baseline sessions (P < 0.05, unpaired t-test). In the remaining participant, the final conditioned MEP size was 99% of the baseline. This MEP conditioning success rate (i.e., 9/10) appears better than that for MEP conditioning in people without SCI (i.e., 5/8, or 63%).

Similarly to the conditioning group without SCI, these participants with SCI also reported that they tried several different strategies for increasing the MEP and adopted the ones that worked. This process was noted within the first one or two conditioning sessions. As conditioning progressed (typically, after conditioning session 12), these participants frequently mentioned that they needed to refine their strategies or change the allocation of attention (e.g., “focus on the toe, instead of thinking of the whole leg,” “focus on dorsiflexion while relaxing the calf”). The most common comments in the later stage of conditioning (i.e., conditioning session 19 and higher) were that the participant had become more aware of co-contracting antagonists and/or upper leg muscles (when occurring) during dorsiflexion and that isolating the TA contraction from activation of other muscles was a part of the strategies to increase the TA MEP.

Figure 4A shows typical examples of peristimulus TA EMG sweeps from a participant with chronic incomplete SCI. In Fig. 4A (top), 75 EMG sweeps from the second 75-trial block of baseline session 6 are superimposed. In Fig. 4A (bottom), 75 sweeps from the second 75-trial block of conditioning session 24 are superimposed. Averaged rectified EMG sweeps for baseline session 6 and conditioning session 24 are presented in Fig. 4B. Two hundred twenty-five responses were averaged together for each sweep. These exemplify changes in MEP size as well as SP (see Changes in silent period).

Fig. 4.

Typical examples of tibialis anterior (TA) motor evoked potential (MEP) in a participant with chronic incomplete spinal cord injury. A: peristimulus raw electromyography (EMG) sweeps from the second 75-trial block of the 6th baseline session and the last (i.e., 24th) conditioning session. For each, 75 sweeps are superimposed. B: rectified EMG signals from the 6th baseline session (black) and the 24th conditioning session (red). For each sweep, 225 responses were averaged together. A shaded band indicates the time window for this participant’s MEP size calculation.

Figure 5, A–C, shows the average courses of MEP changes for successfully conditioned participants with SCI (n = 9). The conditioned MEP (Fig. 5A) increased over the course of 24 conditioning sessions; the final value (i.e., the average of conditioning sessions 22–24) reached 151 ± 4% of baseline value. This conditioned MEP increase was the sum of the control MEP change (124 ± 6% of baseline value; Fig. 5B) and within-session MEP change (27 ± 5%; Fig. 5C). Time courses of these changes with results of repeated-measures ANOVA and post hoc tests are summarized in Table 3.

Fig. 5.

Motor evoked potential (MEP) changes over the course of study in participants with chronic incomplete spinal cord injury who were exposed to MEP up-conditioning. A–C: average (±SE) MEP values for baseline and conditioning sessions for participants in whom the MEP increased significantly (n = 9). A: average conditioned MEP size. B: average control MEP size. C: average of conditioned MEP size minus control MEP size. D: TA maximum voluntary contraction (MVC).

Table 3.

TA MEP values in participants with chronic incomplete SCI in whom conditioning was successful

	F Value	C1–6	C7–12	C13–18	C19–24
Conditioned MEP, %baseline	F(4,32) = 12.1	127.6 ± 7.8*	133.0 ± 3.9*	130.3 ± 4.7*	144.7 ± 5.0*
Control MEP, %baseline	F(4,32) = 2.7	113.3 ± 2.6	110.7 ± 4.3	112.1 ± 4.3	123.5 ± 6.7*
Within-session change, %baseline	F(4,32) = 4.5	14.4 ± 6.6	22.2 ± 5.2*	18.3 ± 4.3*	21.2 ± 4.3*

Values are group means ± SE expressed as percentages of baseline tibialis anterior (TA) motor evoked potential (MEP). C1–6, C7–12, C13–18, and C19–24 are conditioning (or control) sessions 1–6, 7–12, 13–18, and 19–24, respectively. SCI, spinal cord injury.

^*P < 0.05, significant difference from the 6 baseline sessions (Dunnett’s post hoc after one-way repeated-measures ANOVA).

When we compared the time course of MEP changes between the present groups of participants with SCI and without SCI, there was no statistically significant difference between the groups. A two-way repeated-measures ANOVA (groups × sessions × participants) revealed no significant effect of groups for conditioned MEP changes, control MEP changes, or within-session MEP changes [F_(1,48) = 0.50, 0.99, and 2.49, respectively, P > 0.15 for all].

Six participants with SCI attended at least one follow-up session. At the 1-mo follow-up session (n = 6), the conditioned MEP size was 134 ± 10% of baseline, which was the sum of control MEP (108 ± 10% of baseline) and 26 ± 2% within-session increase. At the 3-mo follow up (n = 3), the conditioned MEP was 134 ± 2% of baseline, which consisted of 117 ± 2% (of baseline) control MEP and 17 ± 0.4% within-session increase. Overall, these limited follow-up data suggest that MEP changes that were present at the end of the conditioning sessions were partially preserved for at least 3 mo after the conditioning ended.

TA H-reflex.

In each session, before MEP trials, the CPn was stimulated to measure the TA H-reflex/M-wave recruitment curve while the participant maintained an ~15% MVC level of TA background EMG (in participants without SCI) or at rest (in participants with SCI). TA H_max during the baseline sessions averaged 0.4 ± 0.1 mV in participants without SCI and 0.7 ± 0.4 mV in participants with SCI. TA H_max did not change across the sessions in the control and conditioning groups of participants without SCI and in participants with SCI. The results of one-way repeated-measures ANOVA and H_max values across sessions are summarized in Table 4. These data suggest that the excitability of spinal excitatory reflex was not affected systematically by the operant up-conditioning or control protocol.

Table 4.

TA H_max in participants with and without chronic incomplete SCI

	F Value	C1–6	C7–12	C13–18	C19–24
Conditioning group without SCI	F(4,16) = 0.9	97.7 ± 4.8	102.0 ± 5.4	110.7 ± 10.3	98.4 ± 6.5
Control group without SCI	F(4,24) = 0.8	100.5 ± 3.4	102.1 ± 8.7	93.2 ± 2.6	95.4 ± 2.3
Conditioning group with SCI	F(4,32) = 1.6	100.5 ± 5.3	111.4 ± 5.6	105.5 ± 2.8	108.6 ± 5.0

Values are group means ± SE expressed as percentages of baseline tibialis anterior (TA) maximum H-reflex (H_max) for the conditioning group of individuals without neurological conditions in whom up-conditioning was successful (n = 5), for the control group of individuals without neurological conditions (n = 7), and for the conditioning group of individuals with SCI (n = 9). C1–6, C7–12, C13–18, and C19–24 are conditioning (or control) sessions 1–6, 7–12, 13–18, and 19–24, respectively.

Changes in silent period.

In the conditioning group of participants without SCI, the SP after conditioned MEP (i.e., conditioned SP) gradually decreased over the course of 24 conditioning sessions; SP duration was 94 ± 10 ms for the 6 baseline sessions (116 ± 11 ms when calculated from the MEP onset) and 77 ± 12 ms for conditioning sessions 22–24 (Fig. 6A). SP values (normalized to the baseline values) across sessions are summarized in Table 5, together with the results of one-way repeated-measures ANOVA and Dunnett’s test. The SP was reduced by 20% in conditioning sessions 19–24, significantly less than that in the baseline sessions. This gradual decrease of SP duration was in stark contrast to gradual increase of MEP. Figure 6D shows changes in the ratio of SP duration to MEP (SP-MEP ratio). The average baseline ratio is 1 (i.e., SP duration of 100% baseline value: MEP size of 100% baseline value), and the ratio would remain close to 1 if the MEP and SP increase (or decrease) in parallel. In these conditioning group participants, the SP decreased while the MEP increased, resulting in a significant decrease in the SP-MEP ratio; the ratios for conditioning sessions 7–12, 13–18, and 19–24 were significantly less than that for the baseline sessions (P < 0.05 for all by Dunnett’s test after ANOVA). The SP after the control MEP (i.e., control SP) also decreased over the course of study (Table 5).

Fig. 6.

Time course of silent period (SP) changes over the course of study in neurologically normal participants and in participants with chronic incomplete spinal cord injury (SCI). Average (±SE) conditioned SP durations (A–C) and the SP-to-motor evoked potential (MEP) ratios (D–F) for baseline and conditioning sessions A and D are for conditioning participants without known neurological conditions in whom MEP size increased significantly (n = 5), B and E are for control participants without known neurological conditions (n = 7), and C and F are for conditioning participants with SCI in whom MEP size increased significantly and SP could be measured (C and F, n = 7).

Table 5.

SP duration in participants with no known neurological conditions and participants with chronic incomplete SCI

	F Value	C1–6	C7–12	C13–18	C19–24
Conditioning group without SCI
Conditioned SP	F(4,16) = 3.7	96.4 ± 3.8	91.1 ± 8.7	85.7 ± 8.4	80.1 ± 7.6*
Control SP	F(4,16) = 2.4	96.0 ± 3.4	97.5 ± 5.6	92.2 ± 6.4	87.8 ± 4.1
Within-session change	F(4,16) = 1.0	0.4 ± 5.3	−6.3 ± 7.0	−6.5 ± 7.2	−7.7 ± 8.5
Control group without SCI
Conditioned SP	F(4,24) = 0.9	97.6 ± 3.0	98.8 ± 4.0	99.8 ± 3.9	95.1 ± 2.5
Control SP	F(4,24) = 0.2	99.0 ± 2.9	98.7 ± 2.0	100.1 ± 6.4	102.1 ± 4.5
Within-session change	F(4,24) = 1.2	−1.4 ± 2.9	0.1 ± 2.5	−0.3 ± 4.2	−7.0 ± 3.8
Conditioning group with SCI
Conditioned SP	F(4,24) = 3.7	86.0 ± 8.2	84.5 ± 7.5*	88.6 ± 8.2	80.1 ± 7.1*
Control SP	F(4,24) = 2.1	96.3 ± 3.4	94.3 ± 5.3	91.2 ± 7.1	88.2 ± 6.8
Within-session change	F(4,24) = 3.2	−10.3 ± 5.4*	−9.8 ± 2.7*	−2.6 ± 2.7	−8.0 ± 2.7

Values are group means ± SE expressed as percentages of baseline silent period (SP) duration for the conditioning group of individuals without neurological conditions in whom up-conditioning was successful (n = 5), for the control group of individuals without neurological conditions (n = 7), and for the conditioning group of individuals with SCI in whom the SP could be evaluated (n = 7). C1–6, C7–12, C13–18, and C19–24 are conditioning (or control) sessions 1–6, 7–12, 13–18, and 19–24, respectively.

^*P < 0.05, significant differences from the 6 baseline sessions (Dunnett’s post hoc after one-way repeated-measures ANOVA).

In contrast, in the control group of participants without SCI, the conditioned SP did not change systematically (Fig. 6B and Table 5), and the control SP did not change, either; the SP duration for control sessions 22–24 remained 105 ± 12 ms, similar to that for the 6 baseline sessions [110 ± 15 ms (133 ± 16 ms when calculated from the MEP onset)]. With no consistent changes in MEP or SP, the SP-MEP ratio showed no significant changes over the course of study in these control participants [F_(4,24) = 1.8, P = 0.17 by ANOVA; Fig. 6E). Altogether, these results indicate that the SP decreased in response to MEP operant up-conditioning, not to the control MEP protocol.

For participants with SCI, the SP could be measured over the course of study in seven of the nine successfully conditioned participants. In the other two participants, the end of SP could not be determined (i.e., background EMG did not recover within 500 ms from the time of stimulus). In the seven participants with SCI, the SP decreased from 126 ± 12 to 99 ± 8 ms (159 ± 11 to 132 ± 9 ms when calculated from the MEP onset). The ANOVA results and the SP values in participants with SCI are summarized in Table 5. The SP duration decrease is quite obvious in Fig. 6C, also. With clear changes in MEP and SP, the SP-MEP ratio was significantly smaller across all 24 conditioning sessions (Fig. 6F); the SP-MEP ratios for conditioning sessions 1–6, 7–12, 13–18, and 19–24 were significantly less than that for the baseline sessions [F_(4,24) = 13.7, P < 0.0001 by ANOVA and P < 0.05 by Dunnett’s post hoc test). As summarized in Table 5, the control SP also decreased, although the change did not reach statistical significance. Instead, the SP decreased significantly within session, especially in the earlier conditioning sessions (i.e., see conditioning sessions 1–6 and 7–12 in Table 5).

When the time course of SP changes was compared between the conditioning groups of participants with SCI and without SCI, there was no significant difference between the groups. A two-way repeated-measures ANOVA (groups × sessions × participants) revealed no significant effect of groups for conditioned SP changes, control SP changes, within-session SP changes, or SP-MEP ratios [F_(1,40) = 0.11, 0.01, 0.18, and 0.41, respectively, P > 0.54 for all].

Although all conditioning participants (both conditioning groups of participants with and without SCI) showed a decrease in the SP-MEP ratio over the course of study, there was no correlation between the amount of MEP increase and amount of SP decrease across subjects (Fig. 7). Correlation between the final MEP increase and the final SP decrease was not statistically significant for either group (P > 0.17). Many showed clear changes in both MEP and SP (as in examples of Figs. 2 and 4), yet a person who had a larger MEP increase did not necessarily show a larger SP reduction.

Fig. 7.

Relationship of motor evoked potential (MEP) to silent period (SP) duration in conditioning groups of participants with spinal cord injury (SCI) and of participants without known neurological conditions. Final conditioned SP duration (i.e., average conditioned SP duration for conditioning sessions 22–24) is plotted against the final conditioned MEP size (i.e., average conditioned MEP size for conditioning sessions 22–24). Triangles indicate values for participants with SCI; circles indicate values for participants without known neurological conditions.

DISCUSSION

The present study yielded several key findings. First, this study demonstrates that people can increase MEP size through operant up-conditioning. Second, the study of control group proves that the observed MEP increase is not caused by repeated applications of single-pulse TMS but is indeed due to operant conditioning. Third, people with chronic incomplete SCI can increase MEP in the weakened muscle through operant up-conditioning. Fourth, the conditioning-induced MEP increase is accompanied by a reduction of SP. These help to define the neuromodulatory effects of MEP conditioning and its ability to induce targeted plasticity at the corticospinal pathway. Below we discuss corticospinal plasticity and MEP operant conditioning as a means to guide targeted neuroplasticity.

Operant conditioning of the MEP.

Through up-conditioning, the TA MEP increased in the conditioning groups of participants with and without SCI. Although not identical, the observed time course of MEP changes was similar to that of the previous H-reflex up-conditioning (Thompson et al. 2009a). That is, the gradual increase of conditioned MEP was the sum of task-dependent adaptation (i.e., within-session change from the control MEP to the conditioned MEP) and long-term change (i.e., across-session change in the control MEP). The former is thought to reflect immediate change in cortical influence, whereas the latter reflects the long-term plasticity of the targeted pathway (see discussion in Thompson et al. 2009a). Overall similarities in the time course of conditioned response change between MEP conditioning and H-reflex conditioning and the existence of these two components in both conditioning paradigms likely stem from the fact that operant conditioning of a stimulus-induced EMG response, regardless of whether it targets a reflex or an MEP, is fundamentally a simple motor skill learning (Thompson and Wolpaw 2014b; Wolpaw 2006). Rapid (and reversible) components and long-term components have also been reported in acquisition of other motor skills associated with plasticity in the cortex and subcortical areas (Floyer-Lea and Matthews 2005; Floyer-Lea et al. 2006; Jenkins et al. 1994; Lehéricy et al. 2005; Penhune and Doyon 2002), supporting the generalizability of the two-phase (or component) hypothesis in motor learning across the CNS (Adkins et al. 2006; Dancause and Nudo 2011; Kleim et al. 2004; Thompson et al. 2009a). An important implication of the present study is that people can learn to increase the corticospinal excitability over many weeks of practice and repetition, and that the induced changes can be consolidated and become present outside of the conditioning paradigm. It is probable that this long-term plasticity includes changes in electrophysiological and anatomical properties of corticospinal and spinal pathways including motoneurons and interneurons (Carp et al. 2001a, 2001b; Carp and Wolpaw 1994, 1995; Halter et al. 1995; Pillai et al. 2008; Wang et al. 2009, 2012) and their synaptic connections with descending corticospinal tract (CST) neurons (Long et al. 2017; Urbin et al. 2017). Further studies are needed to determine exact locations and timings of these long-term changes. Regardless of the precise mechanisms, the observed sustained change in corticospinal excitability is of clinical significance. Because enhancement of corticospinal excitability and connectivity is essential in motor function recovery after SCI (Bunday and Perez 2012; Everaert et al. 2010; Long et al. 2017; Stein et al. 2007; Thomas and Gorassini 2005) and other CNS disorders (Everaert et al. 2010; Kale et al. 2009; Mainero et al. 2004; Stinear et al. 2007), MEP operant conditioning that induces targeted plasticity in the corticospinal pathways may be used as a therapeutic tool to augment rehabilitation outcomes in the future (Thompson et al. 2018). Assessment of the effects of MEP conditioning on locomotion that accompanied the present study supports this possibility; TA MEP up-conditioning in sitting appears to increase the swing-phase TA activity, increase ankle dorsiflexion, and thus alleviate foot drop during locomotion in people with chronic incomplete SCI (Thompson et al. 2016).

MEP conditioning in people with and without incomplete SCI.

As shown in Figs. 3 and 5, the TA MEP increased through up-conditioning in participants with and without SCI, by almost exactly the same amount (i.e., final MEP increase of +50%). With a closer look, however, the time course and the composition of these changes appear to differ between the groups. In participants without SCI, the total MEP increase (i.e., conditioned MEP increase) was largely from the long-term change (i.e., control MEP increase). For conditioning sessions 19–24, the total 44% increase consisted of 32% long-term change and 12% task-dependent adaptation (see Table 2) with significantly larger contribution from the long-term component (P = 0.05 by paired t-test). This may suggest a highly plastic nature of corticospinal pathways targeted by this conditioning protocol. The fact that the significant amount of sustained changes could be induced in the corticospinal pathways from earlier conditioning sessions than in the H-reflex pathway (i.e., control MEP increase was significant from conditioning sessions 13–18 on (Table 2), whereas control H-reflex increase was significant only in conditioning sessions 19–24) (Thompson et al. 2009a), may further support this interpretation.

In participants with SCI, a significant amount of task-dependent adaptation (~20% MEP increase) contributed to the conditioned MEP increase from early on (i.e., from conditioning session 4 on; Fig. 5C). Such robust task-dependent adaptation in this population was unexpected for the probable partial damage to CST (caused by SCI), which is known to affect reflex operant conditioning in rats (Chen and Wolpaw 2002; Chen et al. 1996, 1999b, 2002). Contrary to our prediction, overall, these participants produced task-dependent adaptation more than participants without SCI (compare Fig. 3E with Fig. 5C); for conditioning sessions 7–24, participants with SCI produced +21 ± 3% task-dependent adaptation, whereas participants without SCI produced +5 ± 10% task-dependent adaptation (P < 0.05 by unpaired t-test). Furthermore, they reached the same final MEP size over the same conditioning duration (i.e., 24 conditioning sessions over 8 wk) as the participants without SCI. The conditioning success rate was also high in participants with SCI (9 of 10) compared with those without SCI (5 of 8). These findings suggest that neural pathways that are critical for generation of MEP conditioning-induced changes were likely preserved and accessible in the present participants with SCI. The intracortical transmission, reflected in the late I-waves (measured by a paired-pulse TMS method), is presumed to have a critical role in the recruitment of motoneurons (Cirillo et al. 2016; Thickbroom 2011) and thus would be a key for motor function recovery (Silbert et al. 2011). In fact, a recent study by Long et al. (2017) showed that strengthening the late I-wave volleys can improve voluntary hand motor output in people with chronic cervical incomplete SCI. Thus, it is possible that corticospinal pathways that convey the late I-wave volleys were available also for the TA motoneuron excitation in the present participants with incomplete SCI and involved in the MEP conditioning-induced plasticity. Altogether, these results strongly support the applicability of TA MEP conditioning in people with chronic incomplete SCI.

Effects of MEP conditioning on corticospinal excitation and inhibition.

In the present study, the SP after MEP, known to reflect cortical inhibition at least partly (Bertasi et al. 2000; Chen et al. 1999a; Fuhr et al. 1991; Schnitzler and Benecke 1994; Shimizu et al. 2000; Wu et al. 2000; Ziemann et al. 1993), became shorter as MEP grew larger in both participants without known neurological conditions and participants with SCI. These results indicate the potential parallel occurrence of increased corticospinal excitability and decreased cortical inhibition. Because TMS activates separate populations of inhibitory and excitatory interneurons within the cortex (Liepert et al. 1998a; Ziemann et al. 1996b) and different neural circuits underlie MEP and SP (Ashby et al. 1999; Inghilleri et al. 1993; Trompetto et al. 2001), distinct effects of MEP conditioning on excitation and inhibition are not surprising. Rather, what is remarkable here is how consistently the SP duration decreased over the course of study. If MEP up-conditioning simply increased the general excitability of the cortex, MEP and SP would have increased together (e.g., Knash et al. 2003). It was not the case with the present study. Furthermore, a lack of systematic change in the TA H-reflex over the course of study supports the view that long-term changes were likely focused along the corticospinal pathways including the cortex and corticospinal-motoneuron synapses, and would not cause a diffused general increase of spinal excitability.

Presuming that the cortical excitatory and inhibitory neurons interact to produce CST output (Di Lazzaro and Ziemann 2013; Ziemann et al. 1996b), changing the CST output excitability would almost inevitably involve changing the inhibition that converges onto CST neurons. In fact, it is well established that the reduction of intracortical inhibition is critical for inducing cortical plasticity (Bütefisch et al. 2000; Jacobs and Donoghue 1991; Kaelin-Lang et al. 2002; Levy et al. 2002; Ridding and Rothwell 1999; Ziemann et al. 1998). Cortical inhibition changes through modulation of GABAergic inhibitory interneurons (Nielsen et al. 2002). For examples, GABA_A receptor positive allosteric modulator lorazepam decreases MEP (Boroojerdi et al. 2001), prolongs cortical SP (Ziemann et al. 1996a), and increases the excitability of inhibitory circuits in the cortex (Di Lazzaro et al. 2000). Local blockade of GABA_A receptors results in expansion of sensory receptive field (Capaday and Rasmusson 2003) or motor representation (Jacobs and Donoghue 1991). Intrathecal administration of baclofen (GABA_B receptor agonist) prolongs cortical SP (Siebner et al. 1998). Are these GABAergic inhibitory mechanisms involved in the plasticity induced through MEP conditioning? For its duration, the persistent and significant decrease of SP observed in this study is most probably due to the decrease in GABA_B-mediated intracortical inhibition (Chen et al. 1999a; McDonnell et al. 2006; Werhahn et al. 1999). The fact that the above-mentioned late I-waves reflect pyramidal neurons’ connections with local GABAergic interneurons (Di Lazzaro et al. 2012; Di Lazzaro and Ziemann 2013) further encourages us to consider a high likelihood of GABAergic inhibition involved in the conditioning-induced plasticity.

Among the present participants with SCI, some had been orally taking stable doses of baclofen as anti-spasticity medicine (see Table 1) and their medication schedule did not change over the course of study (however, see Consideration for methodological limitations and operant conditioning for one exception). Two participants (subjects 5 and 6 in Table 1) showed clear SP decreases (by 30–45%); one participant (subject 4) decreased SP by <10%; and SPs could not be determined in three other participants (subjects 1, 7, and 10). Thus it seems that the chronic stable amount of baclofen that had been a part of these participants’ physiological norm would not necessarily interfere with conditioning-induced changes in GABA_B-mediated inhibition (Barry et al. 2013). It would also be possible that GABA_A-mediated mechanisms were involved, either independently or through interaction with GABA_B-mediated mechanisms. Such possibilities need to be confirmed through careful examinations in the future.

Short-interval cortical inhibition (SICI) can change rapidly and task dependently (Federico and Perez 2017; Liepert et al. 1998a; Reynolds and Ashby 1999). Thus it could be one of the essential mechanisms of acute, within-session MEP increase (i.e., task-dependent adaptation) observed in the first conditioning session in the present participants with SCI (Fig. 5, A and C). Whether and how SICI may be involved in the within-session SP change and the across-session control SP decrease are yet to be determined through future investigations.

It has become increasingly recognized that inducing targeted plasticity might be one of the essential strategies for enhancing function recovery after CNS damage (Dancause and Nudo 2011; McPherson et al. 2015; Nudo 2015; Thompson and Wolpaw 2015). Thus it would be critically important to understand what mechanisms are involved in task-dependent adaptation (i.e., immediate target plasticity) and how they relate to long-term plasticity of the cortical, corticospinal, and spinal pathways through future investigations, to guide the plasticity toward improving rehabilitation outcomes.

Consideration for methodological limitations and operant conditioning.

The primary goal of this study was to investigate whether operant up-conditioning of the MEP can increase the corticospinal excitability for the TA in people with and without chronic incomplete SCI. To achieve this, the session protocol was designed to maximize the relative weight of conditioning trials while minimizing the number of TMS trials that were not conditioning trials. Because earlier reflex conditioning studies with the session protocols that included many non-conditioning trials and longer session hours resulted in predominantly unsuccessful conditioning (unpublished data), in this study, we kept the combined number of CPn stimulation and non-conditioning TMS trials to be less than a single block of 75 MEP trials. This resulted in not including several important mechanistic measures in the session protocol. MEP recruitment curve and active motor threshold measurements, F-wave measurement, short- and long-interval cortical inhibition (SICI and LICI), and cervicomedullary motor evoked potentials would have provided valuable information on the mechanisms of conditioning-induced plasticity. Because we now know that MEP conditioning can increase the corticospinal excitability, most certainly, these measures should be implemented in future MEP conditioning studies, either in the day-to-day session protocol or as the pre- and postintervention measures, to better understand the physiological mechanisms of conditioning-induced corticospinal plasticity.

In this study, we investigated MEP up-conditioning in participants without known neurological conditions and in participants with chronic incomplete SCI. One surprising finding is that the conditioning success rate was higher in participants with SCI (90%) than in participants without (63%). How could the participants with intact CNS (thus intact CST) have a higher failure rate? One common fact across the three unsuccessful participants without SCI is that they all were constantly using a smartphone throughout study sessions (even for a few minutes of break between blocks of MEP trials), except for the periods of actual MEP trials. None of the participants with SCI or successful conditioning participants used their phones to the same extent during conditioning sessions. It is possible that these individuals’ extensive smartphone use could be a reflection of their distractedness or could have negatively impacted their focus and attention on the task (Matar Boumosleh and Jaalouk 2017; Wilmer et al. 2017), which is essential for operant conditioning approaches (Thompson et al. 2009a; Thompson and Wolpaw 2014b, 2015). Another difference between unsuccessful participants without SCI and successful participants with or without SCI is the compliance with the three-sessions-per-week schedule (see Study overview). We have found that normal participants with busy lifestyles filled with irregular events and plans deviated from a consistent three-sessions-per-week schedule (e.g., M-W-F for 8 wk) more often, but each deviation was not severe enough (e.g., not attending the session for 10 days) to withdraw the participant from the study. Although yet to be confirmed through systematic investigations, the consistency in the session frequency and/or the session frequency itself might affect the conditioning success rate in general. Altogether, these observations suggest the following recommendations for the future operant conditioning studies or similar skill-learning or training intervention studies that are intended for inducing and guiding the neural plasticity: 1) smartphone use should be limited during the study sessions, and 2) the conditioning (i.e., training) session schedule should be strictly maintained, to test the study hypothesis appropriately.

Regarding the potential impact of medication on operant conditioning, we encountered a unique episode during this investigation, which is worthy of reporting. One day, one of the participants with SCI came in for his conditioning session, about 1 h after taking his regular dose (20 mg) of baclofen, out of schedule (all other times he took baclofen 4–5 h before his session). It was his 19th conditioning session. In that session, his within-session MEP increase was minimum, the conditioned MEP was 30% smaller than consecutive 9 prior and 5 subsequent sessions, whereas the SP duration was 30% longer. Because of such anomalies in MEP and SP, and because of a lack of within-session MEP increase (which was always present in this participant), we regarded this session as an example of a session potentially influenced by baclofen. Because this is a single episode of this kind that we have experienced, it would not be appropriate to further speculate potential mechanisms of the observed response. This should, however, serve as a reminder of the importance of maintaining the medication and session schedule throughout many weeks of study.

Conclusion.

This study showed that MEP operant up-conditioning can increase MEP size in people with and without chronic incomplete SCI. The conditioning-induced MEP increase was accompanied by gradual decrease in the SP duration, indicating that MEP up-conditioning increases the corticospinal excitability while suppressing the cortical inhibition. These specific effects of MEP conditioning are in contrast to nonspecific facilitation of excitation and inhibition (Knash et al. 2003) and are not easily explainable by mechanisms known to take place in relatively short time frames, such as homeostatic plasticity (Jung and Ziemann 2009; Müller et al. 2007; Pötter-Nerger et al. 2009; Turrigiano 1999; Ziemann et al. 2004), that work toward maintaining and/or stabilizing the function of the pathways and networks. Further investigations are clearly needed to understand what mechanisms are involved in acute task-dependent plasticity and how they relate to long-term plasticity of the targeted corticospinal pathways. Regardless of the exact physiological mechanisms, observed changes in corticospinal excitation are desirable for enhancing corticospinal function. Thus, when applied appropriately, an MEP conditioning protocol may be able to effectively enhance motor function recovery after SCI and other CNS disorders.

GRANTS

This work was supported in part by the Morton Cure Paralysis Foundation (to A. K. Thompson), National Institute of Neurological Disorders and Stroke Grant NS069551 (to A. K. Thompson), New York State Spinal Cord Injury Trust Fund Awards C023685 and C029131 (to A. K. Thompson), and National Institute of General Medical Sciences Institutional Development Award (IDeA) Grant GM104941 (to S. Binder-MacLeod).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.K.T. conceived and designed research; A.K.T., R.H.C., J.M.S., and J.A.B. performed experiments; A.K.T., R.H.C., and J.A.B. analyzed data; A.K.T., R.H.C., J.M.S., and J.A.B. interpreted results of experiments; A.K.T. prepared figures; A.K.T. drafted manuscript; A.K.T., R.H.C., J.M.S., and J.A.B. edited and revised manuscript; A.K.T., R.H.C., J.M.S., and J.A.B. approved final version of manuscript.