Large-Scale Sprouting of Cortical Connections After Peripheral Injury in Adult Macaque Monkeys

See allHide authors and affiliations

Science  06 Nov 1998:
Vol. 282, Issue 5391, pp. 1117-1121
DOI: 10.1126/science.282.5391.1117


Distributions of thalamic and cortical connections were investigated in four macaque monkeys with long-standing, accidental trauma to a forelimb, to determine whether the growth of new connections plays a role in the reorganization of somatosensory cortex that occurs after major alterations in peripheral somatosensory inputs. In each monkey, microelectrode recordings of cortical areas 3b and 1 demonstrated massive reorganizations of the cortex related to the affected limb. Injections of tracers in area 1 of these monkeys revealed normal patterns of thalamocortical connections, but markedly expanded lateral connections in areas 3b and 1. Thus, the growth of intracortical but not thalamocortical connections could account for much of the reorganization of the sensory maps in cortex.

The reorganization of somatosensory cortex that has been observed in monkeys with forelimb amputation (1) or sensory deafferentation (2), and in human amputees (3, 4), is presumed to be the basis for the sensation of phantom limbs (5) and perhaps phantom pain (4). A critical issue is how such large-scale changes are mediated in the adult brain. We previously showed that sprouting of peripheral nerve axons in the brainstem could account for some of the changes in cortical organization after forearm amputation (1), but additional mechanisms might be necessary for complete reactivation of deprived cortex. To investigate the possibility that new growth at other levels of the pathway contributes to the cortical reorganization, we studied thalamocortical and corticocortical connections in monkeys that had long-standing injury to the forearm, including arm amputation and wrist fracture. Electrophysiological maps of the cortical forelimb representation in the same monkeys allowed us to relate the patterns of connections to the functional changes produced by the injury.

Small injections of a bidirectional tracer, either wheat germ agglutinin conjugated to horseradish peroxidase (WGA-HRP) or fluoro ruby, were made into somatosensory cortical area 1 of three normal macaque monkeys and four monkeys that had suffered accidental forelimb injuries (6) at other primate facilities 1 to 10 years before the terminal experiments were performed (7). The injuries resulted in amputation of all the digits on one hand in one monkey, arm amputations in two monkeys and, in a fourth monkey, a wrist fracture that healed with the hand in a ventrally flexed position that rendered it useless (6). Although the injuries differed in each of the monkeys, they greatly altered the nature of the effective sensory input to cortex. Thus, these monkeys presented a valuable opportunity to learn about mechanisms of large-scale cortical reorganization that follow major changes in afferent drive.

Injections of similar size and location were placed in the hand representation of the control monkeys and in the reorganized representation of the experimental monkeys (7). The locations for the injections were determined from surface landmarks in the normal and injured monkeys and were confirmed by electrophysiological recordings in one of the controls and in all experimental monkeys. Subsequently, extensive electrophysiological recordings were made throughout the deprived zone in cortical areas 3b and 1 of the experimental monkeys. All procedures were performed in accordance with both National Institutes of Health and university guidelines for the care and use of animals in research. This report describes the patterns of label in cortical area 3b (primary somatosensory cortex) and area 1 because we have evidence for reorganization after peripheral injury in these fields. (Labeled neurons and processes also were apparent in other cortical somatosensory areas, including areas 2, 3a, 5, SII, and PV, but it remains uncertain whether or how these connections differ from those in normal animals.)

The distributions of label in areas 3b and 1 in the normal monkeys were similar across animals (Fig. 1). Both in area 1 and in the adjoining somatosensory field, area 3b, clusters of labeled cells and processes were separated from other clusters by zones of little or no labeling (Fig. 1). The label extended across much of the anteroposterior widths of areas 1 and 3b, but this distribution involved limited shifts in representation across the hand, from the palm or the proximal portion of a digit onto the distal digit tips. However, in the mediolateral dimension, where large topographic shifts can occur across the forelimb representation, the distribution of the label was limited (Fig. 1). This results in a pattern of lateral connections that limits the representational spread of activity to near neighbors in normal cortex.

Figure 1

(A) Lateral view of a macaque brain showing the approximate location of the cortical injections into area 1, which occupies the posterior lip of the central sulcus (CS); anterior is to the left. Area 3b is situated ventral to area 1. The hand representations in areas 1 and 3b are located medial to the tip of the intraparietal sulcus (IP). (B) Three-dimensional reconstruction of the spatial distribution of retrogradely labeled neurons in cortical areas 3b and 1 of a normal macaque monkey in an en face view of the posterior bank of the central sulcus, following methods described by Pons et al. (10). Each vertical line indicates the extent of label collapsed across all cortical layers. The distribution of label in every other section has been illustrated. The solid black region indicates the injection site, and the gray fill indicates the halo of dense reaction product around the injection. Lateral is to the right and dorsal is to the top. Scale bar, 1 mm. (C) Summary reconstruction of the hand representation in area 3b and a partial representation of the hand in area 1, based on electrophysiological recordings in the same monkey. The reconstructions were generated using methods described in (10). The injection site is shown for reference. Numerals indicate digits on the hand. Dashed lines indicate the borders between sensory representations; at every recording location within a sensory representation, neurons had receptive fields that involved similar skin locations. Fine solid lines indicate the borders of cortical area 3b. Stippled regions indicate hairy skin representations.

The findings from the normal monkey in which the hand representation was mapped electrophysiologically (8) show the normal relationship of the labeled neurons to the representation of the hand. The injection of the representation of digit 2 in area 1 produced label in area 1 that extended only into the representations of the adjacent digit 3 and the palm (Fig. 1). In area 3b, the main focus of label was in the digit 2 representation and extended predominantly into the adjacent representation of digit 3, with sparse labeling extending into the representation of digit 4 (Fig. 1). Thus, neurons in somatosensory cortex are connected predominantly with other neurons that represent that same general region of the body surface; this is consistent with the results of studies that used fluorescent tracers to show cortical connections in macaque monkeys (9). Measurements of the extent of retrograde labeling (labeled neurons projecting into the injection sites) in area 1 of the normal monkeys ranged from 2.9 to 4.3 mm with a mean spread of 3.8 mm, and in area 3b the retrograde label ranged from 2.8 to 4.6 mm in mediolateral extent with a mean spread of 3.8 mm. The extents of anterograde labeling (axons projecting from the injection site) were similar to the distribution of retrograde label, but measurements were not made.

In the monkeys with long-standing forelimb injury, the extents of labeled connections in areas 3b and 1 contralateral to the injured forelimb were markedly more widespread than in the normals, even though the sizes of the injections were similar (7). The size of the retrogradely labeled zone in area 1 in the experimental monkeys ranged from 5.7 to 8.3 mm (mean, 7.1 mm), and the size of the retrograde projection in area 3b was even more expanded, ranging from 6.9 to 9.2 mm (mean, 7.8 mm). Relative to the labeling in areas 3b and 1 in the normal monkeys, the expansions of the extents of labeling in the monkeys with forelimb injury were highly significant (P < 0.01) (7). The most numerous and most densely labeled neurons typically were located in the central core of the labeled region. Fewer well-labeled neurons were found beyond the core zone; however, even at the outer extremes of the labeled zones, labeled neurons were readily apparent, and assessment of the labeled extents could be made without ambiguity (Fig. 2). Anterograde labeling usually was located in the same general area as the labeled neurons.

Figure 2

(A) Illustration of the distribution of labeled neurons in a series of parasagittal sections through the postcentral gyrus of a 17-year-old monkey that had a long-standing amputation of one arm at the midhumeral level. (B) Summary reconstruction of all labeled sections from the case illustrated in (A) (conventions as in Fig. 1). The total mediolateral extent of the label in areas 3b and 1 is twice as wide as the largest projections in the normal monkeys. (C) Summary reconstruction of the mapping data from areas 3b and 1 of the same monkey as in (A) and (B), showing extensive topographic reorganization of the region where the hand normally is represented. (D) Reconstructions of the extent of labeled neurons in areas 3b and 1 (above) and of the receptive field data (below) in area 3b of a monkey that had amputation of all the digits at the metacarpal/phalangeal joints. (E) Reconstructions of the distribution of labeled neurons in areas 3b and 1 (above) and of the receptive field data (below) in a monkey that had an uncorrected fracture of the wrist. Numerals indicate digits on the hand. Asterisks indicate recording sites where multiple receptive fields were detected. Stippled regions indicate hairy skin representations (other conventions as in Fig. 1). (F) Light-field photomicrograph illustrating well-labeled neurons at the extreme edge of the labeled zone for the case illustrated in (A). (G) Dark-field photomicrograph of a labeled axon from the case in which fluoro ruby was injected. The axon bifurcates near layer III in area 3b, well away from the injection core, and is similar to others seen throughout the expanded area of labeling. FA, forearm; P, palm; W, wrist.

Substantial modifications of the electrophysiological maps in the deprived regions of cortical areas 3b and 1 were also found in the monkeys with forelimb injury (Fig. 2). In the monkeys with the arm amputations, the regions that normally respond to sensory stimulation of the hand had become responsive to the face (laterally) and the remaining upper arm (medially). The monkey that survived 10 years after an arm amputation displayed the largest shift of the face representation into the region where the hand representation had been (Fig. 2). In this case, neurons at many sites responded to stimulation of both the face and the upper arm (Fig. 2C). Neurons with such receptive fields are never found in normal monkeys (10). Also, the broad expanse of labeled neurons in areas 3b and 1 of this monkey extended laterally well into the face representation. A comparable expansion of the retrogradely labeled zone was also present in the other monkey with arm amputation (not illustrated). As a result of the expanded distribution of lateral cortical connections, neurons in the normal face representation could activate neurons in the deprived hand representation, and they could cause the neurons to respond to stimulation of the face. The label also extended medially toward the normal location of the arm representation (Fig. 2), so that neurons activated by the arm could also drive deprived neurons in the hand cortex.

In the monkey with wrist fracture, there was a complete representation of the hand, with the digits represented in a lateromedial sequence (Fig. 2E) at least roughly similar to that in normal macaque monkeys (10). However, receptive field sizes were strikingly large and often extended across two or more digits, or involved noncontiguous portions of the palm and one or more digits. The labeled neurons in this animal extended throughout most of the hand representation in both areas 3b and 1. In area 3b, and perhaps to a lesser extent in area 1, the label extended beyond the representation of the hand into the forearm representation (Fig. 2E). This broad distribution of cortical connections could cause neurons to respond to stimulation over large regions of the hand, rather than the very precise receptive field distribution that is normally seen in the hand representation of areas 3b and 1 in macaques.

Finally, in the monkey with amputations of digits 1 to 5, the zones in areas 3b and 1 where the digits are normally represented contained expanded representations of the palm (medially) and perhaps an expanded representation of the face (laterally) (Fig. 2D). Many neurons with abnormal receptive fields also were present in this monkey. Neurons typically had very large receptive fields and, at a number of sites in both areas 3b and 1, responded not only to stimulation of the palm but also to stimulation of the face or the back of the hand. The injection produced large foci of labeled neurons in areas 3b and 1 that spanned the medial edge of the face representation and the lateral edge of the hand representation (Fig. 2). These may represent normal distributions of label, but there were also additional patches of label at the medial edge of the hand representation (Fig. 2). These connections may reflect new inputs from the palm and wrist to neurons in the deprived hand representation, because widely separated clusters of label such as this have not been seen in areas 3b or 1 of normal monkeys. Connections between topographically distant regions of the hand such as these may lead to the large receptive fields that were observed in this monkey.

The injections also labeled thalamocortical relay neurons in the ventroposterior nucleus (VP) of the thalamus. In normal monkeys, the labeled thalamocortical connections occupy a dorsoventral column in the subnucleus (Fig. 3) where the representation of the hand is located (11). The labeled VP neurons in the experimental monkeys had a columnar distribution similar to that seen in the normals. If the forelimb injuries had caused the cortical terminations of VP neurons to grow and activate deprived neurons in hand cortex, we would have expected labeled neurons to be broadly distributed in the hand portion of VP, and even in the more medial face portion of VP (VPM). The measurements suggest that the extents of the labeled zones might be larger in the experimental monkeys than in the normals, but the difference was not significant (P < 0.05). The volume of label in the normal monkeys averaged 2.2% of the total volume of the nucleus, with a range of 0.7 to 3.4%. In the experimental monkeys, the average volume of the labeled zone was 3.9%, with a range of 2.0 to 5.1% (12). The important feature of these findings is that the thalamocortical projections were at least relatively normal, whereas the changes at the cortical level were immense. Similarly in the visual system, reorganizations in primary visual cortex of cats and monkeys after focal bilateral retinal lesions were not accompanied by comparably extensive reorganizations in the lateral geniculate nucleus (13).

Figure 3

Frontal sections through the ventroposterior nucleus (VP) of the thalamus, showing the distribution of retrogradely labeled neurons after an injection of area 1 in a normal monkey (A) and in monkeys with forelimb injury (Bto D). The sensory representation in VP normally progresses from the face, medially, across the forelimb representation to the hindlimb and tail representation laterally. The locations of the face, hand, and foot subnuclei are indicated in (A). In each panel, black dots represent labeled neurons. In all cases, a column of label is apparent in the lateral subdivision of VP that contains the hand subdivision. Variations in the distribution of label from dorsal to ventral reflect shifts from proximal to distal on the hand representation, which could not account for the physiological changes detected in cortex. In contrast, differences in the mediolateral location of label could involve very large shifts across the body representation, including from hand to face or from hand to foot. Thus, significant changes in the thalamocortical projection would likely involve modifications either in the mediolateral location of the labeled neurons or in their total mediolateral extent. However, few or no differences were detected. Dorsal is to the top and medial is to the left. Heavy solid lines indicate the borders of VP, fine solid lines distinguish the hand subdivision within VP, and dashed lines indicate the borders between VP and ventroposterior superior (VPS) dorsally, and ventroposterior inferior (VPI) ventrally.

Our study provides evidence of widespread expansions of lateral connections in the region of somatosensory cortex where substantial changes in functional organization have occurred. The expansions of the connections probably reflect intracortical sprouting. The only other demonstration of sprouting in cortex, without direct insult, is in visual cortex of adult cats that had retinal lesions (14). The deactivations produced by the retinal lesions were followed by an increase in the density of connections but no growth beyond the preexisting territory. Our results suggest that sprouting can be both within and beyond the framework of the preexisting connections. An alternative explanation for these results is that sparse widespread cortical connections that normally exist in the hand representation of somatosensory cortex were unmasked by the injury. However, if widespread lateral connections are normally present, they should be able to reactivate cortex relatively rapidly after sensory deprivation. But an earlier study has shown that regions of deactivated cortex can remain silent for months after denervation (15).

Major cortical reorganizations likely take time to emerge. The present monkeys all had long-standing injuries with the common effect of depriving a broad expanse of cortex of the normal patterns of activation. The implication is that the chronic deprivation of cortex, whether by extensive finger loss, forearm amputation, or wrist fracture, is the key variable in the cortical reorganizations. Apparently, long recovery periods provide an opportunity for new connections in cortex to grow and become effective. Undoubtedly there are additional consequences of the injuries on cortical organization that are specific to the injury. These could include major changes at lower levels of the somatosensory pathway. For example, new inputs expand into the deprived representation of the forelimb in the brainstem after limb amputation (1); in this case, the role of new cortical connections may be to synergistically potentiate the other inputs. In contrast, in the case of the wrist fracture where the major impact is disuse, the new cortical connections may serve as an important source of information about the sensory environment, and they may largely account for cortical reactivation. In either case, new growth in the adult brain has important implications for recovery of function after injury, including direct damage to the central nervous system, such as spinal cord injury or stroke. Thus, an important direction for future research will be to identify the mechanisms that lead to such profound changes in intracortical connections.


View Abstract

Navigate This Article