Effect of coil orientation on motor‐evoked potentials in humans with tetraplegia

Abstract

Key points

• Although corticospinal function changes following spinal cord injury (SCI), the extent to which we can activate the corticospinal tract after injury remains poorly understood.

• To address this question, we used transcranial magnetic stimulation over the hand representation of the primary motor cortex to elicit motor‐evoked potentials (MEPs) using posterior–anterior and anterior–posterior induced currents in the brain and compared them with responses evoked using lateral–medial currents in participants with and without cervical incomplete SCI during small levels of index finger abduction.

• We found prolonged MEP latencies in all coil orientations in SCI compared to control subjects. However, the latencies of MEPs elicited by posterior–anterior and anterior–posterior compared to lateral–medial stimulation were shorter in SCI compared to controls, particularly for MEPs elicited by anterior–posterior currents.

• Our findings demonstrate for the first time that corticospinal responses elicited by different directions of the induced current in the brain are differentially affected after SCI.

Abstract

The corticospinal tract undergoes reorganization following spinal cord injury (SCI). However, the extent to which we can activate corticospinal neurons using non‐invasive stimulation after injury remains poorly understood. To address this question, we used transcranial magnetic stimulation over the hand representation of the primary motor cortex to elicit motor‐evoked potentials (MEPs) using posterior–anterior (PA) and anterior–posterior (AP) induced currents in the brain and compared them with the responses evoked by direct activation of corticospinal axons using lateral–medial (LM) currents. Testing was completed during small levels of index finger abduction in humans with and without (controls) cervical incomplete SCI. We found prolonged MEP latencies in individuals with SCI in all coil orientations compared to controls. However, latencies of MEPs elicited by PA and AP stimulation relative to those elicited by LM stimulation were shorter in SCI compared to control subjects. Notably, the largest difference between SCI and control subjects was present in MEPs elicited by AP currents. Using a novel controllable pulse parameter transcranial magnetic stimulation, we also found that MEPs elicited by AP currents with 30 μs compared to 60 and 120 μs pulse width had increased latency in controls but not in SCI subjects. Our findings demonstrate that differences between corticospinal responses elicited by AP and PA induced currents were not preserved in humans with tetraplegia and suggest that neural structures activated by AP currents change largely after the injury.

Key points

• Although corticospinal function changes following spinal cord injury (SCI), the extent to which we can activate the corticospinal tract after injury remains poorly understood.

• To address this question, we used transcranial magnetic stimulation over the hand representation of the primary motor cortex to elicit motor‐evoked potentials (MEPs) using posterior–anterior and anterior–posterior induced currents in the brain and compared them with responses evoked using lateral–medial currents in participants with and without cervical incomplete SCI during small levels of index finger abduction.

• We found prolonged MEP latencies in all coil orientations in SCI compared to control subjects. However, the latencies of MEPs elicited by posterior–anterior and anterior–posterior compared to lateral–medial stimulation were shorter in SCI compared to controls, particularly for MEPs elicited by anterior–posterior currents.

• Our findings demonstrate for the first time that corticospinal responses elicited by different directions of the induced current in the brain are differentially affected after SCI.

Introduction

Following spinal cord injury (SCI), extensive reorganization occurs in the corticospinal tract (Oudega & Perez, 2012; Moxon et al. 2014). Animal models of SCI showed that corticospinal neurons sprout rostral and caudal to the lesion (Hill et al. 2001; Bareyre et al. 2004) and their pattern of synaptic connectivity changes (Kim et al. 2006; Ghosh et al. 2010). Studies using non‐invasive transcranial magnetic stimulation (TMS) of the primary motor cortex showed that corticospinal responses in humans with SCI have different characteristics compared to uninjured controls, including decreased amplitude, longer latency and higher threshold (Ellaway et al. 2007; Perez, 2012) and expanded representations (Levy et al. 1990; Topka et al. 1991; Streletz et al. 1995). Although the corticospinal tract is a major descending pathway contributing to the control of voluntary movement (Lemon, 2008) and recovery of motor function after injury (Oudega & Perez, 2012), the extent to which we can activate the corticospinal pathway in humans with incomplete SCI remains poorly understood.

A number of studies in uninjured controls have shown that changes in the direction of the TMS induced current in the brain result in corticospinal responses with different characteristics (Di Lazzaro et al. 2012, 2018; Di Lazzaro & Rothwell, 2014). For example, motor‐evoked potentials (MEPs) elicited when the coil is in the anterior–posterior (AP) orientation have a longer latency, larger latency dispersion and higher threshold than MEPs elicited with the coil in the posterior–anterior (PA) orientation. MEPs elicited by PA and AP stimulation are also modulated to a different extent during different voluntary motor behaviours, motor learning and plasticity protocols (Di Lazzaro et al. 2011; Hamada et al. 2013, 2014; Federico & Perez, 2017; Hannah & Rothwell, 2017; Long et al. 2017). Recordings from the spinal cord of humans implanted with epidural electrodes (Di Lazzaro et al. 2001) and single motor unit studies (Day et al. 1989; Sakai et al. 1997; Hanajima et al. 1998) showed that PA‐induced currents evoke highly synchronized corticospinal activity, whereas AP‐induced currents evoke less synchronized activity with peak latencies that are later than those evoked after PA stimulation. To gain insight into changes in corticospinal transmission after SCI, we examined MEPs elicited by PA and AP stimulation in humans with SCI.

Although little is known about how changes in coil orientation affect corticospinal responses after SCI, earlier studies allow some predictions to be made. Paired‐pulse facilitation results showed that responses reflecting activation of AP compared to PA inputs were more different from controls in humans with SCI (Cirillo et al. 2016). MEPs elicited by AP stimulation probably involve contributions from multiple brain areas (Groppa et al. 2012; Volz et al. 2014), which might be difficult to recruit after SCI because of changes in functional connectivity between cortical regions reported after injury (Min et al. 2015). MEPs elicited by AP stimulation are modulated by afferent input to a greater extent compared to the MEPs elicited by PA stimulation (Hannah & Rothwell, 2017) and afferent input is altered after SCI (Ozdemir & Perez, 2018). Thus, we hypothesized that differences between MEPs elicited by PA and AP stimulation would not be preserved to the same extent in humans with incomplete SCI compared to uninjured control subjects. To test this hypothesis, we examined the latencies of MEPs elicited by PA and AP currents in a finger muscle during small levels of tonic voluntary activity and compared them with MEP latencies evoked by direct wave (D‐wave) activation using lateral–medial (LM) currents.

Methods

Subjects

Seventeen subjects with SCI (mean ± SD age = 49.0 ± 15.7 years, two females) and 17 age‐matched, right‐handed control subjects (mean ± SD age = 41.2 ± 15.9 years, six females, P = 0.16) participated in the present study. All subjects provided their written informed consent prior to participation, which was approved by the local ethics committee at the University of Miami (20140997) in accordance with the guidelines established in the Declaration of Helsinki. The study was not registered in a database. Individuals with SCI had a chronic (≥1 year) cervical injury (C2–C8), with an intact (score = 2) or impaired (score = 1), but not absent, innervation in dermatome C6, using the International Standards for Neurological Classification of Spinal Cord Injury sensory scores, and residual hand motor function (Table 1). Three out of 17 individuals with SCI were categorized by the American Spinal Cord Injury Impairment Scale (AIS) as AIS A (complete injury) as a result of the lack of sacral sparing (Marino et al. 2003), despite being able to elicit voluntary force with hand muscles. Fourteen individuals were classified as incomplete AIS C and D (Table 1). We measured isometric maximal voluntary contraction (MVC) into index finger abduction (controls = 0.66 ± 0.19 mV, SCI = 0.30 ± 0.17 mV; P < 0.001). To test our hypothesis, we examined PA and AP MEP latencies in the first dorsal interosseous muscle (FDI) during small levels of tonic voluntary activity and compared them with MEP latencies evoked by direct wave (D‐wave) activation using LM currents. Because recent work showed that the latency of MEPs elicited by AP stimulation changes according to the duration of TMS pulses (D'Ostilio et al. 2016), we also used a recently developed controllable pulse parameter TMS to assess MEPs.

Table 1.

Spinal cord injury participants

SCI subjects	Age (years)	Sex	Injury level	AIS	Aetiology	Time since injury (years)	Maximum FDI MVC (mV)
1	32	M	C2‐C4	C	T	9	0.15
2	68	M	C2‐C3	C	T	6	0.14
3	52	F	C5‐C8	D	T	14	0.33
4	80	M	C3‐C4	C	T	53	0.18
5	36	M	C7‐C8	C	T	3	0.39
6	50	M	C5‐C8	C	T	13	0.31
7	24	M	C6‐C7	A	T	5	0.32
8	33	M	C5‐C8	C	T	2	0.84
9	65	M	C2‐C6	D	T	5	0.20
10	64	M	C5‐C7	A	NT	5	0.40
11	49	M	C4‐C6	D	NT	3	0.21
12	66	M	C5‐C7	C	T	11	0.44
13	50	M	C4‐C5	D	T	4	0.20
14	36	M	C2‐C5	C	T	4	0.16
15	54	M	C5‐C8	A	T	18	0.42
16	39	F	C2‐C7	D	NT	25	0.22
17	35	M	C2‐C5	C	T	1	0.15

M, male; F, female; C, cervical; T, traumatic; NT, non‐traumatic.

Experimental procedures

Subjects were seated with both arms flexed at the elbow by 90°. At the beginning of experiments, subjects performed two or three brief MVCs for 3–5 s into index finger abduction (Fig. 1 A) separated by 30 s. MVC was measured by calculating the highest mean rectified electromyographic (EMG) activity found in the FDI muscle in 1 s during the MVC burst. EMG activity from the FDI muscle was displayed continuously on a computer screen and verbal feedback was provided to the subjects to ensure that physiological measurements were acquired at 5% of MVC across conditions. A familiarization trial was completed to ensure that all subjects were able to complete the task using the required level of EMG activity. In total, 5.8 ± 3.2% (5.4 ± 2.2% in controls and 6.1 ± 3.9% in SCI participants) of trials in which the mean rectified EMG activity was ± 2 SD of the mean EMG, as measured 100 ms before the stimulus artefact, were excluded from the analysis (Bunday et al. 2014).

Figure 1. TMS.

A, schematic of the hand posture used during testing of index finger abduction. Raw EMG traces show activity in the FDI muscle in a control and SCI subject. B, traces show the average of 20 MEPs in the FDI tested during index finger abduction tested with the TMS coil in the LM, PA and AP orientation in a control subject. Note that latencies of MEPs elicited by AP and PA directed currents were longer compared to LM. Also note, that latencies of MEPs were longer with AP compared to PA directed currents.

EMG recordings

EMG activity was recorded from the FDI muscle via electrodes secured to the skin over the muscle belly (Ag–AgCl, 10 mm in diameter). We tested the right hand in control subjects, as well as the more affected hand in individuals with SCI. However, in SCI individuals, if the TMS responses were not elicited in the more affected side, the other side was tested. The signals were amplified (× 200), filtered (30–2000 Hz) and sampled at 10 kHz (CED 1401 in conjunction with Signal software; Cambridge Electronic Design, Cambridge, UK) and stored in a computer for offline analysis.

TMS

TMS was applied using a figure‐of‐eight coil (loop diameter of 70 mm) via a Magstim 200² magnetic stimulator (Magstim, Whitland, Dyfed, UK) with a monophasic current waveform. As in previous studies, single‐pulse TMS was used to elicit MEPs by changing the current flow across the hand area of the primary motor cortex (Sakai et al. 1997; Di Lazzaro et al. 1998, 2001; Hamada et al. 2013; Volz et al. 2014). The three coil orientations tested were (Fig. 1 B): (i) a figure‐of‐eight coil held tangentially on the scalp at an angle of 45° to the midline with the handle pointing laterally and posteriorly [PA induced current in the brain (PA inputs)]; (ii) a figure‐of‐eight coil handle in reverse position around the intersection of coil windings by placing the coil 180° to the PA currents [AP induced current in the brain (AP inputs)]; and (iii) with the handle of the coil held leftwards 90° from the mid‐sagittal line (LM induced current in the brain) in all subjects during 5% of MVC (LM = 5.13 ± 0.84% of MVC, PA = 5.09 ± 0.67% and AP = 4.97 ± 0.64% of MVC; F _2,66 = 0.9, P = 0.4) and in a subgroup of participants at rest (controls, n = 9; SCI, n = 7). Measurements were performed at the hotspot determined by PA currents because the direction of the current does not significantly influence the position of the hotspot (Sakai et al. 1997; Arai et al. 2005). This optimal scalp position determined by PA currents was then marked on a cap placed on the head with a pen for reference and used as a reference to rotate the coil to elicit MEPs in other coil orientations (Hamada et al. 2013, 2014; Federico & Perez, 2017). The coil was held manually by the experimenter with the head secured to a headrest by straps to limit head movement. The different coil orientations were tested in a randomized order. TMS measurements included resting (RMT) and active (AMT) motor threshold and MEP onset latency as tested for each coil orientation.

MEPs

RMT was determined at the minimum stimulus intensity required to elicit MEP ≥ 50 μV peak‐to‐peak amplitude above the background EMG activity in at least five out of 10 consecutive trials in the relaxed FDI muscle. AMT was determined as the minimum stimulus intensity required to elicit MEP ≥ 200 μV peak‐to‐peak amplitude above the background EMG activity in at least five out of 10 consecutive trials in the contracting FDI muscle (Rothwell et al. 1999). When this is tested during low levels of voluntary contraction where spinal motoneuron excitability is increased, the effect of temporal summation of descending volleys at the spinal level is minimized (Wilson et al. 1996; Sakai et al. 1997; Hamada et al. 2013). Therefore, MEPs were tested during 5% of MVC using TMS intensities of 150% of the AMT for LM and 110% of the AMT for the PA and AP coil orientations (Hamada et al. 2013). A higher stimulus intensity was used for LM to ensure that corticospinal neurons were directly stimulated (D‐wave) at this coil orientation. For two SCI subjects in whom the intensity for LM exceeded the maximum stimulator output (MSO), the intensity was set at 100% MSO. MEP onset latency was measured for individual trials in each subject and condition. The MEP latency was defined as the time point where rectified EMG signals exceeded 2 SD of the mean background EMG, measured 100 ms before the stimulus artefact. The latency of MEPs elicited by PA and AP directed currents was compared with the LM to calculate the difference in latencies between PA‐LM and AP‐LM as a measure of activation of PA and AP‐sensitive inputs, respectively. AP and PA MEP latencies were also compared. We quantified the variability of MEP latencies during each coil orientation in all participants by analysing MEP latency SD, referred as to latency dispersion. Single TMS pulses were delivered at 0.2 Hz and 20 MEPs were recorded in each coil orientation. Central (CCT) and peripheral (PCT) conduction times were calculated from latencies of MEPs, F‐wave and M‐max [CCT: MEP latency – (PCT + M‐max latency); PCT: (F‐wave latency – M‐max latency) × 0.5]. F‐waves were measured using the supramaximum stimulus intensity to the ulnar nerve at the wrist (200 μs pulse duration, DS7A; Digitimer, Welwyn Garden City, UK) via a monopolar bar electrode with the cathode positioned proximally. The stimuli were delivered at 1 Hz at an intensity of 150% of the maximal motor response (M‐max). For each trial, we quantified the latency of each F‐wave and the shortest response was used for estimations.

Transcranial electrical stimulation (TES)

In a subgroup of subjects (control, n = 7; SCI, n = 5), MEPs evoked by TES were tested to ensure direct activation of corticospinal neurons by comparing the latencies of MEPs evoked by TMS in the LM direction and TES. Electrical stimulation was delivered with an interval of 5 s via gold cup electrodes with the cathode placed on Cz and the anode 7 cm lateral to Cz. With this position, it is possible to evoke direct descending volley (D‐waves) in the corticospinal neurons at threshold intensity (Rothwell et al. 1994). The stimulation intensities used for testing were 100%, 110%, 130% and 150% of AMT during 5% of MVC into index finger abduction. Five MEPs were recorded at each intensity. In those two SCI subjects with higher LM AMT (80% and 86% of the MSO), the latencies of MEPs evoked by TMS with the coil in the LM direction were comparable to those evoked by TES and were included in the study.

Controllable pulse parameter TMS

Single TMS pulse stimulation was performed using a custom built TMS device that generates triangular monophasic magnetic pulses inducing near‐rectangular electric field pulses with independent control of the pulse width (Peterchev et al. 2014). MEPs were elicited with pulse widths of 30, 60 and 120 μs with the coil in the PA and AP orientation. Note that the pulse used in the previous experiments with the Magstim 200² has a pulse width of 82 μs and exhibits a sinusoidal pulse shape (Rothkegel et al. 2010). The stimulation intensity was limited by the TMS device to 100%, 71% and 49% of maximum amplitude at pulse widths of 30, 60 and 120 μs, respectively (Table 2). AMT was higher in SCI compared with control subjects at 30 and 60 μs. We also found significant effects of coil orientation (PA < AP) and pulse widths (30 > 60 > 120 μs) in both groups (Table 2). Twenty MEPs were recorded from the FDI muscle during 5% MVC at 110% AMT for each pulse width and orientation to measure the latency of MEPs.

Table 2.

AMT with controllable pulse parameter TMS

Controls	30 μs	60 μs	120 μs	P
PA	63.3 ± 12.5	36.9 ± 7.5	27.9 ± 5.6	<0.001
AP	75.5 ± 9.0*	43.6 ± 5.5*	33.1 ± 4.6*	<0.001
SCI	30 μs	60 μs	120 μs
PA	75.5 ± 9.0	43.6 ± 5.5	33.1 ± 4.6	<0.001
AP	90.1 ± 9.7*	55.0 ± 8.8*	39.4 ± 5.9*	<0.001

^* P < 0.05 for PA and AP comparison within each group.

Statistical analysis

Normal distribution was tested by Shapiro–Wilk's test and homogeneity of variances by Levene's test of equality and Mauchly's test of sphericity. When normal distribution could not be assumed, data were log transformed. When sphericity could not be assumed, the Greenhouse‐Geisser correction statistic was used. Repeated measures ANOVAs were performed to determine the effect of GROUP (controls and SCI) and COIL ORIENTATION (LM, PA and AP) on background EMG activity, AMT, MEP latency and dispersion, CCT and PCT. The same test was performed to determine the effect of GROUP and STIMULATION (LM TMS and TES) on MEP latency; the effect of GROUP and LATENCY DIFFERENCE (PA‐LM and AP‐LM) on MEP latencies; and the effect of GROUP and PULSE DURATION (30 μs, 60 μs and 120 μs) on MEP latency and MEP latency differences. A Pearson correlation analysis was used as required. P < 0.05 was considered statistically significant. Group data are reported as the mean ± SD.

Results

MEP latency in all TMS coil orientations

AMT was different across groups (controls: PA = 44.1 ± 6.7%, AP = 59.9 ± 13.5%, LM = 51.8 ± 9.8%; SCI: PA = 57.4 ± 10.3%, AP = 76.6 ± 11.8%, LM = 64.1 ± 9.0%; F _1,32 = 19.9, P < 0.001) and coil orientations (F _1.5,48 = 76.9, P < 0.001). Here, AMT was lower for PA than for LM (controls, P < 0.001; SCI, P < 0.001) and AP (controls, P < 0.001; SCI, P < 0.001) and for LM than for AP (controls, P < 0.01; SCI, P < 0.001) directed currents in control and SCI subjects.

Figure 2 A shows MEPs in the FDI muscle in a representative control (black traces) and SCI (red traces) subject during index finger abduction when testing was conducted with the TMS coil in the LM orientation and with TES (grey and pink traces). Repeated measures ANOVA showed an effect of GROUP (F _1,10 = 10.2, P = 0.01) but not STIMULATION (F _1,10 = 0.01, P = 0.9), nor their interaction (F _1,10 = 1.5, P = 0.2), on MEP latency. We found that LM MEP latencies were longer in individuals with SCI compared to the mean ± SD value of controls in the majority of subjects (12/17) (Fig. 2 B). Similarly, TES MEPs latencies were longer in SCI compared to controls (P = 0.01) in most subjects (5/5). Note that MEP latencies were similar for LM and TES MEPs in both groups (controls: LM MEPs = 21.6 ± 2.0 ms, TES MEPs = 21.7 ± 2.0 ms, P = 0.9; SCI: LM MEPs = 25.8 ± 2.4 ms, TES MEPs = 25.7 ± 2.6 ms, P = 0.9) (Fig. 2 C).

Figure 2. LM MEPs latency.

A, MEPs elicited in the FDI muscle during index finger abduction when the TMS coil was oriented in the LM direction and using TES in a representative control (black and grey traces, respectively) and SCI (red and pink traces, respectively) subject. Waveforms represent the average of 20 trials for LM MEPs and five for TES MEPs. For TES MEPs, the four different lines indicate the different stimulation intensities used for testing (100%, 110%, 130% and 150% of the active motor threshold). B, group data (control, n = 17; SCI, n = 17) shows the latencies of MEP evoked by TMS in LM direction in controls (black bar and dots) and in individuals with SCI (red bar and dots). Dots indicate an individual subject within each group. The abscissa shows the MEP latency (ms) and the ordinate shows the group tested (control = black bar, SCI = red bar). C, group data (control, n = 7; SCI, n = 5) shows the latencies of LM MEPs and TES MEPs during index finger abduction. The abscissa shows the group tested (control and SCI subjects) and the ordinate shows the MEP latency (ms). Error bars indicate the SD. *P < 0.05, comparison between groups.

Figures 3 A and B show MEPs in representative subjects tested with the TMS coil in the PA and AP orientation. Repeated measures ANOVA showed an effect of GROUP (F _1,32 = 16.1, P < 0.001) and COIL ORIENTATION (F _2,64 = 232.5, P < 0.001), as well as their interaction (F _2,64 = 13.7, P < 0.001), on MEP latency. Post hoc analysis revealed that MEP latencies were longer in the PA (controls = 23.1 ± 1.3 ms; SCI = 26.4 ± 3.2 ms, P < 0.001) (Fig. 3 A) and AP (controls = 24.7 ± 1.5; SCI = 27.4 ± 3.3 ms, P < 0.01) (Fig. 3 B) orientations in SCI (PA: 12/17; AP: 11/17) compared to the mean ± SD value of control subjects. In both groups, LM MEP latencies were shorter compared to the PA (controls, P < 0.001; SCI, P < 0.001) and AP (controls, P < 0.001; SCI, P < 0.001) and AP MEP latencies were longer than PA (controls, P < 0.001; SCI, P < 0.001), indicating that, after SCI, it is possible to engage these different sets of excitatory inputs to corticospinal neurons. Note that CCT was prolonged in SCI compared to control subjects in all coil orientations (controls: LM = 4.6 ± 0.5 ms, SCI: LM = 7.6 ± 2.0 ms, P < 0.001; controls: PA = 6.2 ± 0.5 ms, SCI: PA = 8.5 ± 2.1 ms, P < 0.001; controls: AP = 7.7 ± 1.1, SCI: AP = 9.6 ± 2.0, P < 0.001). However, no differences were found in PCT between SCI and control subjects (P = 0.11). CCT correlated with MEP latency (LM: r = 0.86, P < 0.001; PA: r = 0.83, P < 0.001; AP: r = 0.82, P < 0.001) but not with MEP amplitude (LM: r = 0.22, P = 0.41; PA: r = 0.12, P = 0.65; r = 0.02, P = 0.93) in all coil orientations.

Figure 3. PA and AP MEPs latency.

Group data (control, n = 17; SCI, n = 17) shows the latencies of MEP elicited in the FDI muscle during index finger abduction when the TMS coil was oriented in the PA (A) and AP (B) direction in controls (black traces) and in individuals with SCI (red traces). Dots indicate individual subjects within each group. Waveforms represent the average of 20 MEPs in a representative control and SCI subject. The abscissa shows the MEP latency (ms) and the ordinate shows the group tested (control = black bar; SCI = red bar). Error bars indicate the SD. *P < 0.05, comparison between groups.

MEP latency differences across coil orientations

Figure 4 shows rectified MEPs in a representative control (Fig. 4 A) and SCI (Fig. 4 B) subject with the TMS coil in the LM, PA and AP orientation. Repeated measures ANOVA showed an effect of GROUP (F _1,32 = 16.2, P < 0.001) and LATENCY DIFFERENCE (F _1,32 = 102.1, P < 0.001), as well as their interaction (F _1,32 = 4.3, P < 0.05), on MEP latencies. Here, we found the PA‐LM (controls = 1.6 ± 0.5 ms; SCI = 0.9 ± 0.4 ms, P < 0.001) (Fig. 4 C) and AP‐LM (controls = 3.2 ± 1.0 ms; SCI = 2.0 ± 0.7 ms, P < 0.001) (Fig. 4 D) MEP latency differences were decreased in SCI compared to control subjects. Note the larger decrease in AP‐LM compared to PA‐LM in SCI compared to control subjects (P < 0.01). This difference is also observed in individual data showed for each coil orientation. Note that the gap between control (black circles) and SCI (red circles) participants is larger for AP‐LM compared to PA‐LM. Therefore, it led to smaller latency difference between AP‐PA in SCI compared to control subjects (controls = 1.6 ± 0.9 ms; SCI = 1.0 ± 0.6 ms, P < 0.05).

Figure 4. MEP latency differences.

Comparison of MEP latencies elicited with the coil in the LM, PA and AP orientation during index finger abduction in the FDI muscle in a representative control (A) and SCI (B) subject. Waveforms represent the average of 20 trials. Group data (control, n = 17; SCI, n = 17) show PA‐LM (C) and AP‐LM (D) MEP latency differences during index finger abduction in controls (black bars) and SCI (red bars). The abscissa shows the group tested (control = black; SCI = red) and the ordinate shows the MEP latency differences (ms). Dots indicate individual subject within each group. Error bars indicate the SD. *P < 0.05, comparison between groups.

In a subgroup of subjects (control, n = 10; SCI, n = 9), we matched AMT across groups (controls: PA = 46.8 ± 5.8%, AP = 61.6 ± 13.5%, LM = 55.7 ± 7.7%; SCI: PA = 50.7 ± 5.2%, AP = 68.9 ± 9.6%, LM = 59.1 ± 6.3%; F _1,17 = 2.1, P = 0.2) and found similar results [GROUP (F _1,17 = 17.9, P < 0.01) and LATENCY DIFFERENCE (F _1,17 = 66.5, P < 0.001), as well as their interaction (F _1,17 = 4.8, P < 0.05) on MEP latencies]. We found that PA‐LM (controls = 1.6 ± 0.4 ms; SCI = 0.9 ± 0.3 ms, P < 0.01) and AP‐LM (controls = 3.4 ± 1.1 ms; SCI = 1.9 ± 0.7 ms, P < 0.01) MEP latency differences decreased in SCI compared to control subjects. The decrease in AP‐LM was larger compared to PA‐LM (P < 0.01) and a decreased latency difference was also present in this subgroup of subjects (controls = 1.9 ± 0.8 ms; SCI = 1.1 ± 0.7 ms, P < 0.05). These results suggest that changes in stimulus intensity across groups probably did not contribute to our results. Note that AP MEP latencies were positively correlated with AP‐LM latency difference in controls (r = 0.62, P < 0.01) but not in SCI (r = 0.12, P = 0.6) participants.

Additionally, in a subgroup of subjects, we found that PA‐LM (controls: rest = 1.6 ± 0.7 ms, 5% of MVC = 1.5 ± 0.6 ms, P = 0.2; SCI: rest = 1.2 ± 0.8 ms, 5% of MVC = 1.2 ± 0.7 ms, P = 0.9) and AP‐LM (controls: rest = 3.5 ± 0.7 ms, 5% of MVC = 3.2 ± 1.0 ms, P = 0.1; SCI: rest = 2.2 ± 1.1 ms, 5% of MVC = 2.3 ± 0.9 ms, P = 0.7) MEP latency differences were similar when measured at rest and during 5% of MVC in both controls and SCI participants.

MEPs latency dispersion

Figure 5 A shows the latency dispersion for MEPs tested with all coil orientation in both groups. Repeated measures ANOVA showed an effect of GROUP (F _1,29 = 18.5, P < 0.001) and COIL ORIENTATION (F _2,58 = 24.4, P < 0.001) but not their interaction (F _2,58 = 0.3, P = 0.7), on MEP latency dispersion. Post hoc analysis revealed that latency dispersion was higher for LM (controls = 0.42 ± 0.11 ms; SCI = 0.95 ± 0.38 ms, P < 0.001), PA (controls = 0.57 ± 0.16 ms; SCI = 1.09 ± 0.42 ms, P < 0.001) and AP (controls = 0.94 ± 0.43 ms; SCI = 1.37 ± 0.60 ms, P < 0.05) MEPs in SCI compared with control subjects (Fig. 5 B). The latency dispersion was higher for MEPs tested with AP compared to PA (controls, P < 0.01; SCI, P = 0.047) and LM (controls, P < 0.01; SCI, P < 0.01) stimulation in both groups (Fig. 5 B). MEP latency dispersion was also higher in the PA compared to the LM orientation in controls (P = 0.03) but not in SCI (P = 0.16) subjects.

Figure 5. MEP latency dispersion.

A, latency dispersion for MEPs tested with the coil oriented in the LM (left), PA (centre) and AP (right) direction during index finger abduction in controls (black histograms) and individuals with SCI (red histograms). B, group data shows relative MEP latency dispersion during index finger abduction when the coil was oriented in the LM (left), PA (centre) and AP (right) direction. The abscissa shows the coil orientation tested (LM, PA and AP) and the ordinate shows the MEP latency dispersion (ms). Error bars indicate the SD. *P < 0.05, comparison within group. ¥P<0.001, comparison between groups.

Controllable pulse parameter TMS

Figure 6 A shows raw MEPs elicited by the controllable pulse parameter TMS with the TMS coil in the PA and AP orientation in a representative control and SCI subject. We found that PA MEPs (30 μs: controls = 23.0 ± 1.2 ms, SCI = 25.7 ± 2.1 ms, P < 0.01; 60 μs: controls = 23.1 ± 1.2 ms, SCI = 25.8 ± 2.3 ms, P < 0.01; 120 μs: controls = 23.2 ± 1.3 ms, SCI = 25.8 ± 2.2 ms, P < 0.01) had longer latencies in SCI compared with control subjects at all pulse widths tested. PA MEPs had a similar latency across pulse widths in controls (P = 0.2) and SCI (P = 0.5) subjects (Fig. 6 B). Our results also revealed that AP MEPs (30 μs: controls = 25.2 ± 1.1 ms, SCI = 26.9 ± 2.3 ms, P < 0.05; 60 μs: controls = 24.4 ± 1.4 ms, SCI = 26.6 ± 2.2 ms, P < 0.05; 120 μs: controls = 24.1 ± 1.3 ms, SCI = 26.6 ± 2.2 ms, P < 0.05) had longer latencies in SCI compared with control subjects at all pulse widths tested. Importantly, AP MEPs had a longer latency at 30 μs compared to 60 μs and 120 μs in controls (P = 0.001) but not in SCI (P = 0.4) subjects (Fig. 6 C). Finally, the AP‐PA MEP latency difference was different at 30 μs (controls = 2.2 ± 0.6 ms, SCI = 1.2 ± 0.6 ms, P < 0.05) and 60 μs (controls = 1.3 ± 0.8 ms, SCI = 0.8 ± 0.5 ms, P < 0.05) but not at 120 μs (controls = 0.9 ± 0.4 ms, SCI = 0.7 ± 0.6 ms, P = 0.5) between groups (Fig. 6 D).

Figure 6. Controllable pulse parameter TMS.

A, MEPs elicited with a controllable pulse parameter TMS in the FDI muscle during index finger abduction with the TMS coil in the PA and AP orientation in a control (black traces) and SCI (red traces) subject. Waveforms represent the average of 20 trials with pulse widths of 30 μs in PA and pulse widths of 30, 60 and 120 μs in AP orientation. Group data (controls, n = 11; SCI, n = 7) shows the latencies of MEPs when the TMS coil was oriented in the PA (B) and AP (C) orientation in controls (black bars) and in individuals with SCI (red bars). D, AP‐PA MEP latency difference for 30, 60 and 120 μs pulse widths. Error bars indicate the SD. ^* P < 0.05, comparison within group. ^¥ P < 0.05, comparison between groups.

Discussion

To gain insight into changes in corticospinal transmission after SCI, we tested MEPs in an intrinsic finger muscle with the TMS coil in the LM, PA and AP orientation in humans with and without chronic incomplete cervical SCI. We found prolonged MEP latencies in all coil orientations in SCI compared with control subjects. However, MEPs elicited by PA and AP stimulation relative to those elicited by LM stimulation had a shorter latency in SCI compared with control subjects and the largest difference was present in MEPs elicited by AP currents. Using a controllable pulse parameter TMS, we also found that MEPs elicited with a pulse duration of 30 μs pulse width had increased latencies in controls but not in SCI participants. Overall, our findings demonstrate that differences between corticospinal responses evoked by PA and AP induced currents in the brain are not preserved in humans with tetraplegia and suggest that neural structures activated by AP currents change largely after SCI.

PA‐ and AP‐induced currents in the brain in humans with SCI

Here, for the first time, we show that MEPs elicited by AP stimulation had a longer latency, larger latency dispersion and higher threshold in SCI compared with control subjects. This agrees with previous results showing similar changes in MEPs elicited by PA stimulation after SCI (Ellaway et al. 2007; Perez, 2012). To gain further insight into activation of the corticospinal tract, we also examined latencies of MEPs induced by both AP and PA stimulation in relation to each other. A number of studies in uninjured controls have shown that MEPs elicited with the coil in the AP orientation have a longer latency, larger latency dispersion and higher threshold than MEPs elicited in the PA orientation (Di Lazzaro et al. 2012, 2018; Di Lazzaro & Rothwell, 2014). In agreement, we found in controls that MEPs elicited by PA currents and those elicited by AP currents were 1.6  and 3.2 ms longer, respectively, than the MEPs elicited by LM stimulation (Day et al. 1989; Volz et al. 2014; Wiethoff et al. 2014; Federico & Perez, 2017). Notably, these latency differences decreased to 0.9 ms for PA and to 1.9 ms for AP currents in humans with SCI and the largest difference was observed for responses elicited by AP currents.

A critical point is how we interpret these results in humans with tetraplegia. The evidence shows that a single TMS pulse over the primary motor cortex evokes temporally synchronized descending waves in the corticospinal tract, which can be recorded from the epidural space (Patton & Amassian, 1954; Di Lazzaro et al. 2012). The shortest wave is a result of direct stimulation of the corticospinal neuron (D‐wave) at some distance from the cell body, whereas the later indirect (I) waves (termed I1, I2 and I3) possibly arise from transsynaptic activation of corticospinal neurons by intracortical circuits (Di Lazzaro et al. 2012). The interval between I‐waves in primates (Kernell & Chien‐Ping, 1967; Amassian et al. 1987; Edgley et al. 1997; Maier et al. 1997) and humans (Di Lazzaro et al. 1998, 2001) is similar and the characteristics of the volleys evoked by PA and AP current in humans resemble the I‐waves elicited in primates by TMS (Edgley et al. 1997) or by intracortical electrical stimulation (Maier et al. 1997). The evidence shows that TMS‐induced PA currents across the central sulcus preferentially evoke synchronized corticospinal activity, whereas AP currents preferentially evoke less synchronized activity with their peaks, partially matching the latency of PA‐evoked later activity (Day et al. 1989; Sakai et al. 1997). Thus, it has been proposed that the neural structures activated by PA and AP currents differ, even though it is still unclear whether independent circuits are activated or a single circuit is activated at different sites (Di Lazzaro et al. 2012, 2018; Di Lazzaro & Ziemann, 2013; Di Lazzaro & Rothwell, 2014). One possible interpretation of our results is that neural structures contributing to the generation of AP compared with PA MEPs are affected more after SCI. Some studies suggest that AP MEPs might involve contributions from multiple brain areas and that AP stimulation might be particularly effective with respect to recruiting inputs from premotor areas to the primary motor cortex (Groppa et al. 2012; Volz et al. 2014). Excitatory inputs from non‐primary motor cortical areas affect the responses evoked from the primary motor cortex at times coinciding with the peak of I‐waves in primates (Shimazu et al. 2004) and humans (Groppa et al. 2012). In addition, functional connectivity between cortical regions decreases after SCI (Min et al. 2015; Rao et al. 2016), including premotor areas (Nishimura et al. 2007), which might have contributed to our results. Also, sensory areas might be involved in AP induced responses in that AP MEPs have been shown to be modulated by afferent inhibition to a larger extent than PA MEPs (Hannah & Rothwell, 2017). This might explain why AP MEPs are effected more after SCI because the afferent input is altered after the injury (Ozdemir & Perez, 2018). However, these hypotheses should be considered with caution because direct recordings in animal models are still lacking, which would help us to more precisely understand the origin of PA and AP MEP latency differences reported in human subjects (Di Lazzaro et al. 2001, 2012, 2018).

Regardless of the precise mechanism contributing to our results, our findings support the view that MEPs elicited by AP currents are not as long as expected after SCI compared to MEPs elicited by PA stimulation. This is supported by the smaller difference between AP and PA MEP latency differences in SCI compared to controls at matched and unmatched TMS intensities. This is also supported by our results obtained using a recently developed controllable pulse parameter TMS (Peterchev et al. 2014). This device allows the duration of the TMS pulses to be changed to provide information about the properties of the stimulated axons (Boinagrov et al. 2010; Bostock, 1983; Geddes, 2004; Rattay et al. 2012). As in other studies, we found in control subjects that MEPs elicited by AP stimulation with a pulse duration of 30 μs had increased latencies compared with AP MEPs elicited with pulse durations of 60 μs and 120 μs, suggesting that this short current activated late inputs to corticospinal neurons (Hannah & Rothwell, 2017). By contrast, the latency of MEPs elicited by AP stimulation with all pulse durations was similar in individuals with SCI. We also found that the dispersion of MEP onset latencies was larger for AP compared with PA stimulation in controls but not in SCI participants. Another important difference to note is that the latencies of MEPs elicited by LM stimulation were prolonged in SCI compared to control participants. Electrical stimulation of the primary motor cortex probably activates axons of pyramidal tract neurons in the subcortical white matter at the axon initial segment and generates D‐wave and I‐waves (Day et al. 1987). Because we found that the latency of MEPs evoked by LM and electrical stimulation was similar, our results probably reflect activation of the corticospinal axons directly. As in previous results (Bunday & Perez, 2012), we also found that CCT was prolonged in SCI compared with control subjects. The latencies of MEPs elicited by LM currents, as well as CCT, were ∼4 ms longer in SCI compared with control subjects, which makes the CCT a possible factor contributing to our results. In humans with SCI, however, CCT correlated with the latencies of MEPs elicited in all coil orientations. Therefore, it is less probable that specific differences observed in latencies of MEPs elicited by AP stimulation were related to changes in CCT.

The clinical relevance of the changes in PA and AP MEP latency that we observed in SCI is still unclear. However, several investigations have revealed that MEPs elicited by PA and AP currents can be modulated to a different extent during voluntary motor tasks, motor learning and plasticity protocols (Di Lazzaro et al. 2011; Hamada et al. 2013, 2014; Federico & Perez, 2017; Hannah & Rothwell, 2017; Long et al. 2017), suggesting that an evaluation of the responses evoked by both forms of stimulation might provide some useful insights to understand behaviors and stimulation‐induced modulation of central motor circuits. In uninjured humans, larger differences in latencies between MEPs elicited by AP and PA stimulation have been associated with a more pronounced response to protocols of repetitive non‐invasive brain stimulation capable of inducing motor cortical plasticity (Hamada et al. 2013). In humans with chronic incomplete cervical SCI, evidence shows that repeated non‐invasive brain stimulation using an interval involved in the generation of AP MEPs resulted in improvements in voluntary motor output and hand dexterity (Long et al. 2017). Note that, in the present study, corticospinal responses elicited by AP but not by PA currents were sufficiently sensitive to detect the changes in corticospinal excitability after stimulation, highlighting the relevance of responses elicited by AP currents with respect to the after‐effects of plasticity protocols. The results of the present study provide evidence showing that differences between corticospinal responses elicited by AP‐ and PA‐induced currents are not preserved after SCI and suggest that neural structures activated by AP currents show more pronounced changes after the injury.

Additional information

Competing interests

The authors declare that they have no competing interests.

Author contributions

HJJ and MAP are responsible for the conception and design of the study, as well as the acquisition, analysis or interpretation of data. HJJ, MAP and VD are responsible for drafting and revising the manuscript. All authors have approved the final version of the manuscript submitted for publication and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding

The study was supported by the National Institutes of Health (R01NS090622), US Department of Veterans Affairs (I01RX002474, I01RX000815 and I01RX001807) and Paralyzed Veterans of America (PVA17_RF_0024).

Biography

Hang Jin Jo is a postdoctoral associate in the Department of Neurological Surgery at the University of Miami, FL, USA. She received a PhD in Kinesiology from Pennsylvania State University, where she studied motor control and coordination in individuals with neurological disorders. To study the contributions of cortical and spinal cord circuits in the control of movement in humans, she joined The Miami Project to Cure Paralysis as a research fellow working with Dr Monica A. Perez. Her research is focused on understanding changes in residual motor circuits and enhancing voluntary motor output in humans after spinal cord injury.