Corticomotor representation to a human forearm muscle changes following cervical spinal cord injury

Participants

We examined nine male participants with a cervical SCI, all of whom fulfilled the following inclusion criteria: (i) bilateral upper and lower limb impairment; (ii) no head or brain lesion associated with the trauma leading to the injury; (iii) no seizure, and no medical or mental illness; and (iv) no contraindications for MRI.

All SCI participants were right-handed prior to injury and remained so thereafter, except for one. For this participant, we mapped the right hemisphere during contraction of the left EDC. On the day of the experiment, no medication (i.e. baclofen) was taken.

Fourteen age-matched and gender-matched right-handed healthy participants [mean age, 39.3 years; standard deviation (SD), 15.8 years] without any history of neurological or psychiatric illness were recruited.

All participants were assessed on the nine-hole peg test (9HPT) bilaterally (Goodkin et al., 1988). Furthermore, the maximum grip strength and the action research arm test (ARAT) (Lyle, 1981) were assessed with the dominant hand in SCI subjects. To obtain the grip strength, we asked all participants to perform a dynamic isometric handgrip with a handgrip device that consisted of two force transducers (Honeywell FSG15N1A; Honeywell, Morristown, NJ, USA) situated between two moulded plastic bars spanning a width of 6 cm. Compression of the two bars then resulted in the generation of a differential voltage signal (mV) that was linearly proportional to the exerted handgrip force. The signal was then fed into a signal conditioner (CED 1902; Cambridge Electronic Design, Cambridge, UK), digitized (CED 1401; Cambridge Electronic Design), and stored on a PC. The averages of two trials for each hand of the 9HPT and of the grip strength of the dominant hand were calculated. Three SCI participants were scored with the maximum time allowed for the 9HPT (300 s), as they were unable to perform the 9HPT with their non-dominant hand within this time window (Hoogervorst et al., 2002). The Mann-Whitney test was used to assess differences in motor performance between the groups, and a P-value of < 0.05 was considered to be significant.

The study was approved by the Joint Ethics Committee of the UCL Institute of Neurology and the National Hospital for Neurology and Neurosurgery (ref. 08/0243). All participants gave informed, written consent before the study.

MRI protocol

Navigated TMS mapping procedure

We mapped, at 110% AMT, the cortical muscle representation of the EDC in each participant with the Brainsight Frameless Navigation system (Rogue Research, Montreal, QC, Canada), in order to monitor the position of the TMS coil throughout the experiment and to assess topographical changes in cortical excitability and inhibition. Therefore, we adapted a protocol that was initially developed for the assessment of training-induced changes in CS output (Kleim et al., 2007). In brief, we imported the anatomical scans into the neuronavigation system, and superimposed a target grid (1 × 1 cm) on the hemisphere to be stimulated (Fig. 1A) (Kleim et al., 2007). The grid was centred at the point where the central sulcus meets the longitudinal fissure at the dorsal aspect of the brain. The grid extended 7 cm anterior, 2 cm posterior and 8 cm lateral to this point, with a dimension of 8 × 9 cm (72 sites). We then used the stereotaxic neuronavigation system to position the TMS coil to each grid point, and applied 10 TMS pulses at each position, in pseudorandom order. Following each block of 10 trials, participants were given a short break of approximately 30 s, during which the TMS coil was positioned over the next grid point. During the experiment, participants were asked to maintain 10% of their maximum muscle contraction for a duration of 10 trials. A stable background activation level was achieved by online visual feedback of their individual current EMG activity. Throughout the experiment, the TMS coil was orientated at a 45° angle to the midline, with the handle of the coil pointing posteriorly at a scalp site (Orth et al., 2003). From these data, the peak-to-peak amplitude and CSP were measured.

Statistical analysis of TMS physiology

We used the Stata 9.2 software package (http://www.stata.com) for all analyses. Results associated with P < 0.05 are reported. For group comparison, we applied the non-parametric Mann–Whitney test. Two multiple linear regression models were designed to explore associations between physiological measures (response variables) and (i) spinal changes (e.g. SCA = predictors) and (ii) clinical measures of upper limb function. For the first regression model, we included age, height, weight and scalp–cortex distance as covariates of no interest. Specifically, we included scalp–cortex distance because it has been shown to directly influence the magnitude of cortical stimulation in TMS (Stokes et al., 2007). For the second regression model, we included age and scalp–cortex distance as confounds. In order to determine whether these associations were different in SCI participants and controls, subject type indicator and type × predictor interaction terms were added; however, there was no evidence that associations were different in SCI participants and controls, and therefore the combined group was used in these analyses.

Clinical results

Nine SCI participants (mean age, 46.7 ± 11.3 years) had a chronic traumatic lesion between cervical levels C5 and C8. The mean period following SCI was 145.9 months (±59.8 months; range, 57–288 months). Seven of the nine SCI participants had incomplete lesions of the cervical cord and two had complete lesions of the cervical cord, based on the American Spinal Cord Injury Association (ASIA) Impairment classification. As a result of trauma, all SCI participants had severe upper and lower limb impairment, as confirmed by reduced ASIA motor scores of upper and lower limbs and impairment on the ARAT (mean score of 38.72 of a maximum of 57). SCI participants were impaired bilaterally on the 9HPT [dominant hand, SCI, mean = 98.77 s (SD, 84.45) vs. dominant hand, controls, mean = 16.37 s (SD, 1.69), Mann–Whitney U = 117, number of controls (n_c) = 14, number of participants (n_p) = 9, P < 0.001, two-tailed; non-dominant hand, SCI, mean = 141.71 s (SD, 121.44) vs. non-dominant hand, controls, mean = 17.72 s (SD, 2.12), Mann–Whitney U = 117, n_c = 14, n_p = 9, P < 0.001, two-tailed]. The grip strength of the dominant hand was also lower than normal [SCI, mean = 0.16 mV (SD, 0.15) vs. controls, mean = 0.61 mV (SD, 0.19), Mann–Whitney U = 112, n_c = 14, n_p = 9, P < 0.001, two-tailed]. More details on SCI participants’ characteristics are summarized in Table 1 and also in a previous report (Freund et al., 2011).

TMS results

The background level of EMG activity of the EDC (approximately 10% of the maximum EDC muscle contraction) did not differ significantly between SCI participants and controls (SCI, median = 0.11 mV [interquartile range (IQR), 0.04] vs. controls, median = 0.13 mV (IQR, 0.04); Mann–Whitney U = 49, n_c = 14, n_p = 9, P > 0.05, two-tailed) (Fig. 2A–C). The AMT in the EDC was higher in SCI participants [SCI, median = 52% (IQR, 25.5) vs. controls, median = 46% (IQR, 12.5); Mann–Whitney U = 22, n_c = 14, n_p = 9, P = 0.01, two-tailed; Table 2]. The duration of the CSP, calculated as the mean over all stimulation sites, was longer in SCI participants [SCI, median = 76 ms (IQR, 20) vs. controls, median = 47 ms (IQR, 20); Mann–Whitney U = 23, n_c = 14, n_p = 9, P = 0.012, two-tailed], and the CSP duration measured at the motor cortical hot spot was also longer [SCI, median = 130 ms (IQR, 60) vs. controls, median = 96 ms (IQR, 30); Mann–Whitney U = 26, n_c = 14, n_p = 9, P = 0.02, two-tailed; Table 2].

To account for intersubject anatomical variability, we expressed the location of the individual CoG as the ED between the CoG and the anatomical hand knob (Yousry et al., 1997; Dechent & Frahm, 2003). The ED of the CoG of the EDC–MEP map was smaller in SCI participants than in controls [SCI, median = 0.8 cm (IQR, 0.4) vs. controls, median = 1.45 cm (IQR, 0.68); Mann–Whitney U = 29, n_c = 14, n_p = 9, P = 0.031, two-tailed; Table 2].

Relationship between cervical cord atrophy and TMS physiology

We applied multiple regression analysis including all subjects, as there was no a priori reason why controls would have a different relationship between the dependent parameters (TMS parameters) and independent parameters (SCA and clinical parameters) than SCI participants (see Materials and methods) (Fig. 3A and B). The SCA in this patient cohort was reduced by more than 30%, as reported previously (Freund et al., 2011). Here, we relate these same data to the electrophysiological data derived from the cortical maps acquired in the present study; this, in contrast to our previously published work, allows for direct comparison of an MRI-based measure of cord atrophy and cortical excitability and inhibition. Crucially, we found a negative association between SCA and AMT (r² = 0.39, P = 0.015), and negative associations between SCA and mean CSP duration (r² = 0.26, P = 0.022) and hot spot CSP duration (r² = 0.4, P = 0.005). In other words, AMT was higher and CSP duration longer in SCI participants with greater cord atrophy.

Corticomotor representation is changed following SCI

In a previous study on the same group of patients, we observed that voluntary movement of the hand was associated with an increased BOLD signal, not only in the somatotopic hand area of M1, but also in the adjacent leg area. We had therefore expected to see a similar shift in the CoG of the CS projection to the EDC when tested with TMS. In fact, this did not occur. Stereotaxic TMS mapping showed that the location of the EDC CoG was shifted posteriorly towards the region of the anatomically defined hand knob in the central sulcus. The CoG is thought to approximate to the location of the highest density of CS projections, and in non-SCI individuals it has been shown to correspond with the CoG identified with fMRI imaging (Ruohonen & Karhu, 2010). Why, then, did it not show a shift in direction similar to that of the BOLD signal after SCI?

The MEP responses evoked by a single TMS pulse are attributable to conduction in large-diameter CS axons that have relatively direct monosynaptic or polysynaptic inputs to spinal motoneurons (Rothwell, 1997). Thus, the small shift in the EDC CoG that we observed might be caused by limited reorganization of the immediately surrounding neuronal networks, which normally innervate the intrinsic hand muscles (Lemon, 2008). The low cervical lesion in our patients had damaged the CS outputs to intrinsic hand muscles to a much greater extent than those to the EDC in the forearm. However, CS axons to intrinsic hand muscles often send branches to forearm muscles, and could therefore contribute to MEPs evoked in the EDC by TMS. If the strength of these connections had increased after SCI, then this would shift the CoG into the M1 hand area, which is predominantly represented in the anterior wall of the central sulcus (Rathelot & Strick, 2009). This type of reorganization of CS neurons might have increased motor output and possibly could improve voluntary movement of the upper limb.

However, this is not the only explanation. For example, the higher TMS thresholds observed in the SCI group might recruit more of the EDC representation located within the sulcus, and shift the balance of representation more posteriorly. Second, the posterior location of the CoG might have been attributable to an increased contribution of the primary sensory cortex to the recovery of hand function (Green et al., 1998, 1999).

However, this does not explain why we did not evoke any MEPs in the EDC when we stimulated over the leg area of M1, when the fMRI images showed such clear activation of that area during volitional hand movement (Freund et al., 2011). One possibility is that the outputs activated during voluntary movement include polysynaptic inputs from the leg area that are not activated strongly by TMS. These could, for example, be unmasked cortico-cortical connections linking neurons in the leg area to CS neurons in the arm area; alternatively, they might involve CS outputs from the leg area that had contacted spinal interneuronal networks with indirect access to the EDC motoneurons. If the excitability of these indirect connections was low, then they would not have been detected by single TMS pulses. Future experiments using multiple TMS pulses to improve conduction through the network might be able to detect such connections. Alternatively, it may be possible to use spinal reflex testing to activate the networks and test whether these are modulated by inputs from TMS over the leg area. Nevertheless, the complete lack of any evidence for fast-conducting CS connections from the leg area to the EDC in SCI humans suggests that any reorganization of motor output may be relatively minor in functional terms.

If this is so, then what is represented by the BOLD signal activation of the leg area in fMRI? One possibility is that this represents sensory input from the hand, which has reorganized to innervate neurons in the leg representation. This would be consistent with the BOLD signal activity in the leg area after electrical stimulation of the median nerve. Another possibility is that, during hand movement, there is a potential for uncontrolled ‘overflow’ of activation to the leg area, which is normally prevented in intact individuals, but in SCI individuals is no longer cancelled because leg movements no longer occur. One way to test for the importance of leg area activation to voluntary hand movement in future experiments would be to use TMS as a ‘virtual lesion’, and examine whether leg area stimulation interferes with hand movement in SCI but not in intact individuals.

Relationship between cord atrophy and physiological measures

Reduced SCA is associated with impaired upper limb function following SCI (Freund et al., 2011). Here, we provide further evidence for this, as we demonstrate a negative correlation between SCA and increased AMT and CSP duration. In other words, greater damage to the spinal cord correlates with decreased CS excitability and prolonged duration of CSP. As the corticomotor system is directly connected to the spinal cord via CS projections (Lemon, 2008), these findings may be caused by retrograde degeneration (Cohen-Adad et al., 2011; Freund et al., 2011) of these projections that might, in turn, lead to physiological remodelling and changes to the output properties of the injured CS neurons, as observed in the rodent model of SCI (Ghosh et al., 2010).

CS motor output is reduced following SCI

The cortical AMT of the EDC was significantly increased in SCI participants. This was not attributable to a difference in scalp–brain distance, which affects the efficiency of TMS (Stokes et al., 2007), as this was equal in both groups of subjects. One possibility is that increased AMT is attributable to a reduced density of motoneurons or interneurons in low segments of the cervical cord that innervate the EDC. However, this seems unlikely to be the only explanation, given the negative association between increased AMT and lower SCA at C2. At this level, the major reason for changes in SCA is degeneration of axons and not motoneurons. Moreover, we accounted for potential influences of changes in excitability of spinal motoneurons or cortical neurons, as this was equalized, as far as possible, across subjects by their maintaining a constant background voluntary contraction. It therefore seems likely to reflect the reduced density of surviving CS neurons, which would require a higher intensity of stimulation to recruit a given CS output (Di Lazzaro et al., 2004). In addition, because the response evoked by TMS results from activation of facilitatory inputs to CS neurons (Amassian et al., 1987; Shimazu et al., 2004), changes in cortical AMT may reflect structural and functional alterations in cortico-cortical connectivity that result from degenerative processes affecting CS neurons (Aguilar et al., 2010).

In line with a previous study on complete SCI (Lotze et al., 2006), we observed increased CSP duration. The CSP is thought to be mediated by GABA_B receptors on pyramidal cells (McCormick, 1989). Degeneration of CS axons and soma shrinkage of pyramidal cells (Hains et al., 2003; Beaud et al., 2008) may lead to reduced afferent excitatory cortico-cortical and thalamo-cortical drive to the motor cortex. Consequently, a shift of excitability towards inhibition could occur. However, we cannot exclude the possibility that the increased AMT may also contribute to the prolonged CSP. If SCI reduces the number of CS neurons without changing the number of inhibitory interneurons, then the higher intensity required to recruit sufficient CS output may activate proportionally more inhibitory interneurons within the cortex. Finally, it should be noted that, because of high thresholds in some participants, we were only able to measure CSP at a relatively low intensity of stimulation (110% AMT). At such intensities, the duration of the CSP in controls is short, and this may be partially attributable to changes in excitability of spinal motoneurons that are refractory from having discharged in the preceding MEP (Fuhr et al., 1991). Under normal circumstances, spinal motoneurons become excitable after 100 ms, so that increased duration of the hot spot CSP in patients to 130 ms seems likely to have been caused by changes in cortical excitability. However, as we did not test for changes in spinal motoneuron behaviour in these patients, this must be regarded as a tentative conclusion that requires further testing in future experiments.

Limitations of the study

The present study has limitations that could have contributed to differences in the CoG EDC between the two groups. First, the study was conducted on a relatively small number of SCI participants in the chronic stage of SCI. Thus, longitudinal studies are required to validate the present findings and to investigate the temporal development of parameter changes associated with SCI. Second, the muscle studied may also lead to different results. Initially, we intended to also examine the first dorsal interosseous muscle, but, owing to muscle atrophy, this turned out not to be feasible. Therefore, in future studies, additional proximal arm and trunk muscles should be added to the analysis in order to obtain a more detailed image of changes to the cortical muscle representations that are (relatively) unaffected by the injury.