Time-Dependent Central Compensatory Mechanisms of Finger Dexterity After Spinal Cord Injury

See allHide authors and affiliations

Science  16 Nov 2007:
Vol. 318, Issue 5853, pp. 1150-1155
DOI: 10.1126/science.1147243


Transection of the direct cortico-motoneuronal pathway at the mid-cervical segment of the spinal cord in the macaque monkey results in a transient impairment of finger movements. Finger dexterity recovers within a few months. Combined brain imaging and reversible pharmacological inactivation of motor cortical regions suggest that the recovery involves the bilateral primary motor cortex during the early recovery stage and more extensive regions of the contralesional primary motor cortex and bilateral premotor cortex during the late recovery stage. These changes in the activation pattern of frontal motor-related areas represent an adaptive strategy for functional compensation after spinal cord injury.

Neurorehabilitation has its basis in the concept that training recruits the remaining neuronal systems to compensate for partial injury of the central nervous system (CNS). However, the neuronal basis of these compensation mechanisms is poorly understood. Brain imaging studies in human stroke patients show increased activity in various cortical regions, including the side ipsilateral to the affected extremity (1, 2). However, in these case studies, the extent of the lesion varies between patients, and thus identifying the damaged pathways is often difficult. Moreover, it is unclear whether the brain regions showing increased activity causally contribute to the recovery. In addition, because the tests are usually performed at a specific time after the lesion, longitudinal information is lacking. To assess the neuronal mechanism of functional compensation, we need a longitudinal study that applies quantitative behavioral evaluation to an animal model, preferably macaque monkeys (3), with a defined lesion of the particular neuronal system. Dexterous finger movements can be restored within a few weeks to 1 to 3 months after a lesion of the direct cortico-motoneuronal (CM) connection via the lateral corticospinal tract (l-CST) at the border between the C4 and C5 segments of the spinal cord (4). This lesion site (“C4/C5 l-CST lesion”) is rostral to the segments where motoneurons of hand muscles are located. These results suggest that indirect cortico-motoneuronal pathways, mediated by subcortical or spinal interneuronal systems, can mediate commands for the control of dexterous finger movements in primates. In the present study, we examined the neuronal mechanism of this functional recovery from spinal cord injury. We hypothesized that, in addition to the plasticity of neural circuits in the spinal cord (4, 5), adaptive learning by higher order structures may contribute to the recovery. We applied positron emission tomography (PET) scanning using H 152O to measure changes in brain activity during precision grip tasks at different stages of recovery.

Five monkeys were trained to reach for a small piece of food through a narrow vertical slit and to grasp it between the pads of the index finger and thumb (Fig. 1A, preop). In the monkey shown in Fig. 1A, the precision grip was completely impaired immediately after the C4/C5 l-CST lesion (Fig. 1A, day 7). On day 14, the monkey was able to grasp the food with the index finger and thumb, but the independence of the fingers remained impaired. All abilities gradually recovered (Fig. 1A, day 99), as previously reported (4). Figure 1B shows the time course of recovery of the success rate for precision grip. The success rate recovered to more than 80% of that before the lesion within 3 weeks in all five monkeys. On the basis of these observations, we defined postoperative days 1 to 45 (about 1 month postoperative) as the early recovery stage and postoperative days 90 to 143 (more than 3 months) as the late recovery stage. We performed PET scanning and inactivation experiments during the following three stages: (i) preoperative stage, (ii) early recovery stage, and (iii) late recovery stage.

Fig. 1.

Recovery time course of precision grip. (A) Three representative frames showing the retrieval of a small piece of food by monkey S. These images were taken at 0.1-s intervals. (B) The success ratio of food retrieval from a vertical slit in the five monkeys. A success trial was defined as any trial that resulted in the successful precision grip and removal of the food from the pin without dropping it.

Three monkeys (monkeys H, T, and K) were examined in the PET study. In multiple comparisons, their performance in the precision grip task during the preoperative stage was associated with an increased activity in visuomotor-related regions, including the sensorimotor cortex, premotor cortex, and intraparietal sulcus in the contralateral hemisphere and the early visual cortices, putamen, and the cerebellum on the ipsilateral side, as previously shown (6) (Fig. 2). To identify the cortical regions that showed an increase in activity during postoperative stages, we compared the regional cerebral blood flow (r-CBF) during the postoperative stages with that during the preoperative stage. Activity increased in the bilateral primary motor cortex (M1) during early recovery (Fig. 3, A to C and E, and table S1). Furthermore, activity increased in the bilateral early visual cortices (V1/V2), contralateral S2, contralateral accumbens, and the vermis of the cerebellar cortex (Fig. 3, A to G, and table S1). During the late recovery stage, increased activation was observed in the contralateral M1 (co-M1) and ipsilateral ventral premotor cortex (ip-PMv) (Fig. 3, I to M, and table S1). Activity in the bilateral insula, contralesional accumbens, and the cerebellar vermis also increased (Fig. 3, L to O, and table S1). The area of co-M1 with increased activity expanded during the late recovery stage compared with that during the early recovery stage (compare Fig. 3, A and I, B and J, and E and M), and the increased activity extended into co-PMv (Fig. 3J and table S1).

Fig. 2.

Brain areas activated during the precision grip task at the preoperative stage were quantified as increased rCBF using PET. rCBF on the control task at the preoperative stage was subtracted from rCBF on the precision grip task at the preoperative stage. Brain areas with significantly increased rCBF are indicated (P < 0.01, uncorrected for multiple comparisons in three monkeys). The activations are superimposed on a three-dimensional reconstruction of a template brain magnetic resonance image (MRI) of macaque monkeys that was produced by our group. The significance level is given in terms of a Z score represented on a colored scale. (A) Top view, (B) view from the contralesional hemisphere, (C) view from the ipsilesional hemisphere, (D to H) coronal sections, and (I) lateral view of the brain. The lines D to H in (I) correspond to the respective coronal sections. Contra indicates contralesional hemisphere; Ipsi, ipsilesional hemisphere.

Fig. 3.

Increased brain activation related to functional recovery. Results were obtained from three monkeys (monkey H, T, and K). Twenty scans were conducted on monkeys H and K in both tasks for every stage, and 24 scans were conducted on monkey T. Activation during early and late recovery stages was compared with the preoperative stage (precision grip task at postoperative stage – control task at postoperative stage) – (precision grip task at preoperative stage – control task at preoperative stage). Brain areas with significantly increased rCBF (P < 0.01, uncorrected for multiple comparisons) are superimposed on a three-dimensional reconstruction of a template brain MRI of macaque monkeys that was made by our group. The significance level is given in terms of a Z score represented on a colored scale. (A to G) and (I to O) are results during early and late stages of recovery, respectively. (A) and (I), top view; (B) and (J), view from contralesional hemisphere; (C) and (K), ipsilesional hemisphere; (D) to (G) and (L) to (O), coronal sections. (H) and (P) show lateral views of the brain. Lines D to G and L to O in (H) and (P) indicate the levels of coronal sections of (D) to (G) and (L) to (O), respectively.

To clarify whether the increased activity of those cortical regions observed in the PET study were causally involved in the functional recovery, we performed focal inactivation of individual cortical regions by using microinjections of muscimol, a γ-aminobutyric acid type A (GABAA) receptor agonist (5 μg/μl), at various recovery stages and observed the precision grip in two monkeys (monkeys S and C). Because we observed increases in the activity in M1 and PMv of both hemispheres during the recovery stages in the PET study, we chose the digit areas of these cortical regions as targets of muscimol injection. We performed intracortical microstimulation (ICMS) mapping before the inactivation study to obtain a topographical map of these regions and to determine the injection site (Fig. 4A, a and b). In the preoperative trials, inactivation of the digit area of co-M1 with 0.8 or 1.5 μl of muscimol resulted in clumsy finger movements, accompanied by a loss of the independent control of each digit. Both monkeys reached for the food piece but were not able to achieve a precision grip. The success rate for retrieval was reduced to zero (Fig. 4B, d and j). On the other hand, inactivation of either ip-M1 (Fig. 4B, a and g, 5.0 μl of muscimol) or co-PMv (Fig. 4B, p and v, 1.5 μl of muscimol) or ip-PMv (Fig. 4B, m and s, 5.0 μl of muscimol) resulted in no impairment of digit movements. Precision grip during the early recovery stage was severely impaired by inactivation of co-M1. The monkeys reached for the target but showed muscular hypotonia in the hand. Monkey C showed total paresis of the hand. Monkey S was able to move its digits, but the thumb and index finger could not be inserted into the slit. The success rate for retrieval remained zero in both monkeys (Fig. 4B, e and k). Interestingly, after inactivation of ip-M1, the capacity to retrieve the food piece was impaired. The success rate for retrieval with precision grip in monkeys S and C decreased by 33 and 15%, respectively (Fig. 4B, b and h), from that before inactivation. Even in successful trials, both monkeys achieved the grip not with the pads of index finger and thumb but with the pad of index finger and the nail of thumb. After inactivation of co-PMv (Fig. 4Bq) or ip-PMv (Fig. 4Bn), the capacity to retrieve the food piece was impaired, and digit movements became clumsy in monkey C but not in monkey S. During the late recovery stage, the effect of inactivation of co-M1 was greatly reduced, even in comparison with the preoperative trials (Fig. 4B, f and l). In contrast to the early stage, inactivation of ip-M1 during the late recovery stage resulted in no impairment of digit movement in either monkey (Fig. 4B, c and i). Inactivation of co-PMv did not cause impairment of digit movements in either monkey (Fig. 3B, r and x). We confirmed these results by administering additional injections of muscimol into ip-M1 or co-PMv, and no impairment of digit movements was observed. Inactivation of ip-PMv led to a marked slowing of movements in both monkeys (Fig. 4Bu). Monkey C repeated several grips before retrieving the food piece.

Fig. 4.

The effect of inactivation of M1 or PMv on the precision grip. (A) Somatotopic map revealed by ICMS performed before the CST lesion. The location of muscimol injections are indicated on a surface map of the precentral regions shown in a and b. Each electrode penetration is represented with a character indicating the body territory activated at threshold: D, digit; W, wrist; E, elbow; S, shoulder; T, trunk; and F, face. The size of characters indicates the threshold for induction of movements (inset). The sites and volumes of muscimol injection are shown by colored circles (concentration, 5 μg/μl; volume, light purple dot, 0.8 μl; green dots, 1.5 μl; orange dots, 5.0 μl). Anterior-posterior and medial-lateral orientation are indicated in the inset (A, anterior; M, medial). (B) The effect of inactivation of M1 or PMv on food retrieval at preoperative, early, and late recovery stages in two monkeys. The success rates for target retrievals and retrieval time obtained before (Cont) and after muscimol injection (Mus, 2 hours after muscimol injection). The prehension time was defined as the time interval between the first insert of any digit to the tube and the timing of release of all the digits from the tube. Only trials with successful retrieval were used to measure the prehension time. a to c and g to i, ip-M1; d to f and j to l, co-M1; m to o and s to u, ipsilesional PMv; p to r and v to x, co-PMv; a to f and m to r, success rate for retrieval; g to l and s to x, prehension time; a, d, g, j, m, p, s, and v are preoperative trials; b, e, h, k, n, q, t, and w are trials at the early stage of recovery; c, f, i, l, o, r, u, and x are trials at the late stage of recovery. Error bars indicate standard deviation. *P < 0.05, **P < 0.01 (corrected t test). CS, central sulcus; Ipsi, ipsilesional hemisphere; Contra, contralesional hemisphere; Preop, preoperative trials; Early, trials during the early stage of recovery; Late, trials during the late stage of recovery; R, rostral; C, caudal; and N.A., not available.

During surgery, we tried to achieve a complete lesion of the direct CM connection, leaving the ventrally located interneuronally mediated pathways intact (fig. S1A). The extent of the lesion was similar among all five monkeys (fig. S2). The extent of the lesion was assessed by counting the number of anterogradely labeled l-CST axons caudal to the lesion in monkey T (fig. S1C) versus rostral to the lesion (fig. S1B) after biotinylated dextran amine (BDA) injection into the co-M1. A small number of corticospinal fibers remained in the ventral region (box in fig. S1C). Thus, the extent of interruption of the l-CST fibers was estimated to be 98.7%. The extent of the lesion was also assessed by electrophysiological recordings in three monkeys (monkeys T, H, and S) under anesthesia. Extracellular field potentials in the lateral motor nuclei in C6 (a train of three stimuli was applied to the contralateral medullary pyramid) and the cord dorsum potential (CDP) at the same segment on the intact and lesion sides were recorded (fig. S3). The extent of the lesion estimated by the amplitude of the negative volley in CDP was 96.4% in monkey T, 100% in monkey H, and 99.6% in monkey S (relative to the intact side). Thus, we conclude that a near-complete or complete lesion was made to l-CST in all five monkeys.

Our results indicate that the inactivation of co-M1 resulted in a severe deficit in finger movements during the early recovery stage (Fig. 4B, e and k). This suggests that early recovery depends strongly on increased activity of the co-M1. Interestingly, during the late recovery stage, the effect of inactivation of co-M1 was greatly reduced in comparison with the early recovery period (Fig. 4B, f and l). As shown in Fig. 3, I, J, and M, the activated area in the co-M1 greatly expanded and appeared to extend into co-PMv. Additionally, increased activation was found in ip-PMv. The present result suggests that the recovery process was also assumed by regions outside the preoperative digit area in co-M1 and that the effect of inactivation of the digit area by injection of the same amount of muscimol could be compensated by the strong activity in these surrounding regions and ip-PMv. Our finding of an increased area of activation in co-M1 agrees with previous observations that the representation of trained movement in M1 expanded with learning (7, 8) and with recovery after injury (911). Neuronal pathways from these activated areas to the hand motoneurons are not clear. Propriospinal neurons with cell bodies in the C3-C4 segments and with axons passing through the ventral part of the lateral funiculus can mediate the excitation from the co-M1 to digit motoneurons (1214). However, the contribution of reticulospinal neurons cannot be excluded (15, 16).

The inactivation of ip-M1 resulted in no deficit in the preoperative trials but caused a deficit during the early recovery stage (Fig. 4Bb) and no deficit during the late recovery stage (Fig. 4Bc). Ip-M1 thus transiently contributes to the recovery during the early recovery stage. Increased activity in M1, premotor area, and supplementary motor area on the ipsilateral side to the affected hand has been reported in stroke patients (1, 2). The activation of ip-M1 was reported to have a negative correlation with the outcome after stroke (1). However, in the above studies, it was not clear whether the increased activity was controlling the digit movements or was simply a side effect of the increased activity of the co-M1. Inactivation of ip-M1 after hemisection at the cervical spinal cord of the monkey produced no effects (17). However, these examinations were performed over 3 months after the lesion, which may have been too late to show effects, judging from our observation presented herein. The ip-M1 may exert effects either through the co-M1 via callosal fibers (1820) or by relay through subcortical pathways (15, 16, 21, 22). Furthermore, axons descend through the l-CST that originate from the ip-M1 and recross the midline at the spinal level (23). This commissural projection may also drive hand motoneurons. Similar indirect excitatory pathways from the ip-M1 to hand motoneurons may exist. Transmission through these pathways should be inhibited normally but may be disinhibited after injury (24).

A contribution from the PMv has been indicated for the recovery from an M1 lesion through observations of changes in the topographic map (25). Concerning the co-PMv, it has been shown to influence motoneuronal activity mainly via the co-M1 (26, 27). The co-PMv may also control the spinal interneuronal system directly via its direct projection to the mid-cervical segments (28). The pathway from the ip-PMv is difficult to explain; however, it is likely that the ip-PMv can exert its effect via the co-PMv, co-M1 (1820), or a direct projection to the spinal cord and commissural projection at the spinal level.

The present study demonstrates that functional recovery after lesion of the corticospinal tract involves specific networks, including not only interneuronally mediated subcortical pathways downstream of the co-M1, but also other parallel pathways including the ip-M1 and higher-order cortices like the PMv. The contribution of each cortical region changes depending on the postoperative recovery stage. We thus suggest that the brain uses existing systems by reducing inhibition during the early recovery stage and gradually enhancing the original neural systems or recruiting other systems by synaptic plasticity during the late recovery stage for more stable control. Clarification of the neural pathways from these regions to the relevant motoneurons will lead to a deeper understanding of the strategy for functional compensation and, in addition, will provide a good indication of the prospects for recovery after spinal cord injury.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Table S1


References and Notes

Stay Connected to Science

Navigate This Article