Smaller spinal cord injuries often allow some degree of spontaneous behavioral improvements because of structural rearrangements within different descending fiber tracts or intraspinal circuits. In this study, we investigate whether rehabilitative training of the forelimb (forced limb use) influences behavioral recovery and plastic events after injury to a defined spinal tract, the corticospinal tract (CST). Female adult Lewis rats received a unilateral CST injury at the brainstem level. Use of the contralateral impaired forelimb was either restricted, by a cast, or forced, by casting the unimpaired forelimb immediately after injury for either 1 or 3 weeks. Forced use of the impaired forelimb was followed by full behavioral recovery on the irregular horizontal ladder, whereas animals that could not use their affected side remained impaired. BDA (biotinylated dextran amine) labeling of the intact CST showed lesion-induced growth across the midline where CST collaterals increased their innervation density and extended fibers toward the ventral and the dorsal horn in response to forced limb use. Gene chip analysis of the denervated ventral horn revealed changes in particular for growth factors, adhesion and guidance molecules, as well as components of synapse formation suggesting an important role for these factors in activity-dependent intraspinal reorganization after unilateral CST injury.
Injury to the spinal cord leads to a disruption of ascending and descending fiber tracts followed by loss of sensation and voluntary movements below the level of the lesion. Whereas large injuries often lead to permanent disabilities, smaller lesions are followed by some degree of spontaneous functional recovery (Blight, 1993; Burns et al., 1997; Little et al., 1999), which can be increased by rehabilitative therapies. The benefits of training on the recovery of sensory and locomotor function after spinal cord injury (SCI) have been demonstrated in animal models (Rossignol et al., 1999; Edgerton et al., 2004) and SCI patients (Barbeau and Rossignol, 1994; Dietz et al., 1998), but the underlying mechanisms are still poorly understood.
Regeneration and adaptive changes in response to injury are limited in the adult mammalian CNS because of intrinsic neuronal mechanisms and environmental factors (Schwab, 2002; Yiu and He, 2006). Nevertheless, spontaneous reorganization has been reported on different levels and within different descending fiber tracts (Raineteau and Schwab, 2001; Bareyre et al., 2004) and might contribute to spontaneous functional improvements (Dobkin, 2000).
The corticospinal tract (CST), as one of the most important descending motor pathways for skilled movements in all mammalian species (Nudo and Masterton, 1988, 1990), has been a frequent target to investigate injury-induced plasticity within the adult CNS (Weidner et al., 2001; Bareyre et al., 2004). Within recent years, there have been many attempts to increase the plastic potential of the injured CST. Neutralization of growth inhibitors such as Nogo-A or the Nogo-A receptor NgR (Thallmair et al., 1998; Schwab, 2004; Cafferty and Strittmatter, 2006) as well as the expression of neurotrophic factors (Zhou and Shine, 2003; Zhou et al., 2003) after unilateral CST injury were followed by increased growth of collaterals from the intact tract across the midline. Additionally, stimulation of the intact CST promoted outgrowth of ventral fibers into the denervated gray matter (Brus-Ramer et al., 2007).
Research on plasticity and neurorehabilitation after SCI has mostly focused on the recovery of hindlimb/leg function. Assessing the effect of training on forelimb/hand function and the underlying structural and molecular mechanisms has proven to be much more difficult and complex. In one study, Girgis et al. (2007) showed that training of a reaching task was paralleled by increased sprouting of lesioned CST fibers above the injury followed by behavioral recovery of grasping.
In this study, we investigate the effect of forced limb use after unilateral CST injury on behavioral recovery, compensatory growth, arborization, and synapse formation of the uninjured CST. We also use a gene chip approach to define some of the key factors within the denervated spinal cord that could account for structural and cellular rearrangements in response to injury and forced limb use.
Materials and Methods
Adult female Lewis rats (180–200 g) were obtained from a specific pathogen-free breeding colony (R. Janvier) and kept as groups of four animals in standardized cages (type 4 Macrolon) at a 12 h light/dark cycle on a standard regimen with food and water ad libitum. After biotinylated dextran amine (BDA) tracer injections into the contralesional forelimb motor cortex (described below), rats were divided into the following experimental groups: unlesioned animals (intact), animals receiving a sham operation (sham), and animals with unilateral CST injury (Pyx) at the brainstem level. Sham-operated as well as injured animals were further subdivided: One group of animals returned to their home cage in which they were able to freely move without any restrictions (free use). In all other animals, a plaster of paris cast immobilized either the limb ipsilateral or contralateral to the lesion, forcing the animals to completely rely on their impaired (forced use) or their unimpaired (forced nonuse) forelimb for either 1 or 3 weeks (Fig. 1). Additional groups of animals were operated, casted, and used for gene chip analysis. All experiments were performed according to the guidelines of the Veterinary Office of the Canton of Zurich, Switzerland.
All surgical procedures were performed under aseptic conditions. One week before injury, animals were deeply anesthetized with a subcutaneous injection of Hypnorm (0.3 mg/kg body weight; VetaPharma) and Dormicum (0.6 mg/kg body weight; Roche Pharmaceuticals). For anterograde tracing of the unlesioned CST, the caudal forelimb area of the contralesional sensorimotor cortex was exposed (Neafsey et al., 1986). A 10% solution of BDA (10,000 molecular weight; Invitrogen) in 0.01 m PBS (0.1 PB) was pressure injected stereotactically into the right motor cortex (coordinates: 1.5 mm anterior to bregma, 2 mm lateral to bregma, 1.5 mm depth) using a 5 μl Hamilton syringe driven by an electric pump with a flow rate of 5 nl/s. A total volume of 3 μl was injected at five injection sites, separated by 500 μm. The syringe remained in place for 3 min after completion of each injection.
One week after BDA injection, all animals were anesthetized as described above. Unilateral pyramidal tract lesions were made according to the method described previously (Thallmair et al., 1998; Z'Graggen et al., 1998) (Fig. 1A). Using an anterior approach, the ventral midline was transected, trachea and esophagus were carefully displaced to the left, and the surface of the occipital bone was exposed by a deep blunt dissection. A small hole was drilled into the bone, which revealed the medullary pyramids. The dura was removed and a fine tungsten needle inserted into a depth of ∼1 mm below the ventral brainstem surface just lateral to the basilar artery. The needle was laterally displaced and gently lifted, transecting the left CST just rostral to its decussation. In sham-operated animals, the dura was removed but the CST was left undamaged. Finally, esophagus and trachea were replaced, and the muscles and skin were closed by suture. Completeness of the injury was confirmed by histology approach in all animals and in a sample of animals by anterograde tracing of both sides of the CST with BDA. Animals with incomplete lesions were excluded from additional studies.
Immediately after pyramidotomy surgery while animals were still anesthetized, one forelimb was restricted as previously described (Jones and Schallert, 1994). Animals were randomly assigned to two different groups in which either their impaired or their unimpaired limb was immobilized by a plaster of paris cast that resembled a one-holed vest around their upper torso and one limb (Fig. 1C). Animals were forced to completely rely on either their impaired (forced use) or their unimpaired forelimb (forced nonuse) for either 1 or 3 weeks starting immediately after injury. To facilitate grooming and body hygiene, all rats were randomly assigned to groups of a minimum of four animals per cage. Correct position of the cast was checked on each animal twice per day. Occasionally, animals were able to remove their cast but were recasted immediately under light isoflurane anesthesia. All casts were removed either 1 or 3 weeks after injury, and animals were able to recover for 48 h before their performance was tested on the horizontal ladder test.
Horizontal ladder test
Two weeks before baseline “recordings,” animals were accustomed to the testing apparatus every other day until each animal voluntarily crossed the ladder in a slow consistent speed. The ladder was 1 m long and elevated 1 m over ground. To prevent habituation to a specific bar distance, bars were irregularly spaced (1–6 cm) and each animal had to cross the ladder in both directions. For behavioral testing, three runs over a defined 60 cm stretch were filmed and evaluated using frame-by-frame video analysis (Virtualdub; www.virtualdub.org). In unconstrained animals (free use), locomotor performance and spontaneous recovery of the impaired forelimb was evaluated immediately after injury and then once per week. These testing sessions were few and short to prevent additional forelimb training effects. Casted animals were tested only once, 48 h after cast removal.
We judged forepaw placement as normal when all four digits were placed in front of the rung (see Fig. 2A). After injury, the following mistakes were observed and counted as errors: (1) animals misplaced one or more digits on the backside of the rung, (2) rats failed to place the palm of their paw directly onto the rung but used their wrist or forearm for support, (3) animals misplaced their paw on the rung and slipped off the rung or completely missed single rungs (see Fig. 2B–D). Success rate was expressed as percentage of correct steps of all steps taken by the impaired limb.
Immunohistochemistry and histological analysis
After the completion of behavioral testing, all animals were deeply anesthetized with pentobarbital (450 mg/kg body weight, i.p.; Abbott Laboratories), perfused transcardially with 100 ml of Ringer's solution [containing 100,000 IU/L heparin, Liquemin (Roche), and 0.25% NaNO2] followed by 200 ml of 4% phosphate-buffered paraformaldehyde, pH 7.4 containing 5% sucrose. Spinal cords and brains were dissected and postfixed in the same fixative overnight at 4°C before they were cryoprotected in phosphate-buffered 30% sucrose for an additional 5 d.
Diaminobenzidine staining (BDA).
The caudal parts of the cervical enlargement (C6–C8) and the brainstems were embedded in a gelatin–chicken albumin solution polymerized with 25% glutaraldehyde and cut in 50-μm-thick sections on a cutting vibratome as previously described (Z'Graggen et al., 1998). Sections were processed using the nickel-enhanced diaminobenzidine (DAB) protocol according to the semi-free-floating technique of Herzog and Brösamle (1997). Briefly, sections were rinsed three times for 30 min each time in TBS-TX (50 mm Tris, 0.9% NaCl, 0.5% Triton X-100, pH 8.0) and incubated overnight with an avidin–biotin–peroxidase complex (Vectastain ABC Elite kit; Vector Laboratories; 1:100 in TBS-X) at 4°C. After three 30 min washings in TBS-TX, sections were rinsed with 50 mm Tris-HCl, pH 8.0, and preincubated with 0.4% ammonium nickel sulfate (Sigma-Aldrich) followed by a second preincubation in ammonium nickel sulfate and 0.015% DAB (Sigma-Aldrich). The tissue was reacted in 0.4% ammonium nickel sulfate, 0.015% DAB, and 0.004% H2O2 in 50 mm Tris buffer, pH 8.0. After 5–10 min, the reaction was stopped with 50 mm Tris-HCl, sections were rinsed three times for 10 min each time in 50 mm Tris-HCl, air-dried overnight, and coverslipped with Eukitt (Kindler).
Immunofluorescence staining (2G13, vGlut1, BDA).
The rostral part of the cervical enlargements (C2–C5) as well as one postnatal rat brain [postnatal day 4 (P4)] were embedded in Tissue Tek OCT (Kindler) and frozen in isopentane (Sigma-Aldrich) at exactly −40°C. Forty-micrometer-thick sections were cut on a cryostat and transferred into ice-cold PBS. Sections were transferred into postfixation (4% paraformaldehyde, 0.1% glutaraldehyde, 0.1% saturated picric acid in PB buffer) for 10 min, endogenous peroxidase activity was quenched with ethanol peroxide (50% ethanol plus 0.3% hydrogen peroxide in ddH2O) followed by 50 mm glycine (in PBS; 10 min) to reduce autofluorescence. After microwave irradiation (600 W; 30 s) in 0.1 m Tris buffer, slides were blocked in TNB (0.1% casein, 0.25% bovine serum albumin, 25% Top block, 0.15 m NaCl, 0.05% Tween in 0.1 m Tris buffer) for 1 h.
Sections were incubated overnight at 4°C with primary antibodies against vesicular glutamate transporter 1 (vGlut1) (1:1000; rabbit; Synaptic Systems) or 2G13 (1:40; mouse; Acris). After washing, they were incubated with high-affinity-purified goat anti-rabbit (1:200) and anti-mouse (1:100) secondary antibodies (Jackson ImmunoResearch) coupled to CY2 or FITC, respectively. After washing in PBS, sections were blocked in TNB for 30 min and incubated with an avidin–biotin–peroxidase complex (ABC kit).
Conventional protocols for immunofluorescence failed to detect fine CST collaterals in the cervical gray matter. We therefore developed an improved protocol for the detection of CST collaterals using tyramide signal amplification first described by Adams (1992), and further established by Müllner et al. (2008). Slides were transferred into a self-made tyramide amplification buffer (0.02% H2O2, 1% biotinylated tyramide, 0.05% Tween in PBS) for 5 min. Another blocking step in TNB was followed by 30 min incubation with Cy3-conjugated streptavidin (1:4000 for double labeling with vGlut1; 1:500 for double labeling with 2G13; Jackson ImmunoResearch). Finally, sections were washed, mounted on glass slides, dried overnight, and coverslipped with Mowiol (Calbiochem).
Pictures of vGlut1 or 2G13 colocalization with BDA-positive fibers within the denervated ventral gray matter (Rexed's lamina VII) were taken with a confocal microscope (Leica TCS SP2; Leica confocal software, version 2.61).
Quantification of midline-crossing CST fibers.
Growth and sprouting in response to injury was evaluated in 40 adjacent cross sections at the caudal cervical enlargement (C6–C8), where motoneuron-columns innervating muscles specifically required for skilled forelimb movements are located (McKenna et al., 2000). Midline-crossing fibers were counted at a final magnification of 400× in the dorsal and ventral comissure at the central canal (see Fig. 3F, level M), and branching of these fibers was evaluated at two defined regions within the gray matter (see Fig. 3F, levels D1 and D2).
Four vertical (M, D1, D2, L) and one horizontal line (H) were superimposed on each spinal cord section (Neurolucida 7.0; MicroBrightField) as reference points for crossing axons. The first vertical line M was drawn through the central canal; H was also drawn through the central canal, perpendicular to M. L was drawn parallel to M and crossed H at the lateral rim of the gray matter. D1 and D2 were drawn parallel to M at one-third and two-thirds of the distance between M and L (see Fig. 3F). Collateral CST fibers in the gray matter have an irregular course, passing in and out of the plane of the section. To prevent multiple counting of single collaterals, only fibers that crossed M, D1, or D2 were counted on each section.
To correct for variations in BDA uptake by CST neurons in the sensorimotor cortex, we normalized the quantitative data by counting BDA-labeled axons in the main pyramidal tract in three rectangular areas (150 μm2) per slide on four adjacent sections at the level of the brainstem (facial nerve). Results are expressed as mean number of fibers at the cervical level crossing M, D1, and D2 divided by the mean of counted fibers at the brainstem level for each animal.
Camera lucida reconstructions of three representative consecutive cross sections of the cervical spinal cord (C6) were performed for three groups (intact, lesioned_3 weeks_forced nonuse/forced use) to visualize the extent and specificity of CST fiber growth in response to injury and forced limb use.
Quantification of BDA-positive synaptic boutons.
In each animal, BDA-positive boutons along the length and at the tip of midline-crossing fibers were counted at a defined area (400 μm2) on six sections within the denervated gray matter (intermediate zone, Rexed's lamina VII, C6) at a 400× magnification. The number of boutons was normalized for tracer efficacy as described above.
Quantification of vGlut1 immunoreactivity.
Pictures of the contralesional (denervated) as well as the ipsilesional (intact) ventral gray matter (intermediate zone, Rexed's lamina VII, C5) were taken at a 400× magnification with a confocal microscope on five sections per animal, and vGlut1-positive varicosities were counted on each photograph within a randomly chosen area (400 μm2). vGlut1 immunoreactivity was expressed as percentage of vGlut1-positive varicosities within the denervated side compared with the contralateral intact ventral gray matter.
Gene chip analysis
For the gene chip analysis, one group of rats received a unilateral pyramidotomy and the other group a sham surgery as described above. Both groups were further subdivided into animals that were forced to rely either on their ipsilesional or their contralesional forelimb for 1 week as described above (total: four groups; n = 4). One week after injury, rats were decapitated, and the spinal cords were dissected on a cold plate and frozen in liquid nitrogen. Cervical enlargements were mounted on a cutting microtome, covered with dry ice, and cut into sections of 250 μm. In each section, the ventral horn of the denervated side was extracted by punch dissections with a blunt syringe (diameter, 1.2 mm) (see Fig. 9A,B). The tissue was immediately transferred into RNA-later (Roche Diagnostics) and subsequently used for total RNA extraction (RNeasy Lipid Tissue kit; QIAGEN). Total RNA was quantified by NanoDrop (ND-100; NanoDrop Technologies), and quality was assessed using a bioanalyzer (2100 Bioanalyzer; Agilent Technologies). For probe preparation, procedures described in the Affymetrix GeneChip Expression Analysis Manual (Affymetrix) were followed. Biotinylated cRNA was hybridized onto Affymetrix Genome arrays (Rat Expression Array 230 2.0; Affymetrix), which represent >45,000 probe sets in the Affymetrix Fluidics Station 450, and chips were scanned with the Affymetrix Scanner 3000. Each chip was used for hybridization with cRNA isolated from one spinal cord sample from a single animal. Thus, there was a total number of 16 samples. Results were subsequently analyzed using the Affymetrix Microarray Suite 5, followed by the Genespring 7.2 (Silicon Genetics). We applied a present call filter (at least in three of four animals), fold change thresholds (1.2/0.8), and ANOVA (with p ≤ 0.05). Additional information about the regulated genes was obtained from PubMed (http://www.pubmed.com).
All data were analyzed using parametric ANOVA of the appropriate design, followed by restricted analyses or Bonferroni's post hoc pairwise comparisons whenever a main effect or interaction attained statistical significance. All statistical analyses were conducted using the statistical software SPSS (release 14.0). Data are presented as means ± SEM, single data points represent single animals, and asterisks indicate significances as follows: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.
Spontaneous functional recovery after unilateral CST injury
Adult female Lewis rats were trained to cross a horizontal ladder with irregular spacing 2 weeks before injury until they were able to perform this task without difficulties. Uninjured animals crossed the ladder at a slow but consistent speed and grasped each rung precisely with all four digits placed in front of the metal bar (Fig. 2A) and had very few misplacements. Errors were always attributable to incorrect placing of single digits but hardly ever to misplacements of the paw on rungs or steps between rungs. The success rate of uninjured animals was high at baseline levels (90.1 ± 1.9%; n = 7). After unilateral CST injury, the contralesional forepaw was severely impaired (Fig. 2E). The following mistakes could be observed: animals misplaced one or more digits behind the rung (Fig. 2B); rats failed to place the palm of their paw directly onto the rung but used their wrist or forearm for support (Fig. 2C); animals misplaced their paw and slipped off the rung or placed their paw between single rungs (Fig. 2D).
To prevent additional training on the horizontal ladder, animals were tested rarely, immediately after injury, and then once per week. Pyramidotomy led to significantly more errors of the impaired forelimb on the horizontal ladder test (ANOVA; F = 113.67; p ≤ 0.001). Spontaneous behavioral improvement was minimal within the next 3 weeks and the success rate remained low (mean success rate, 57.4 ± 7.6%; Bonferroni's post hoc, p ≤ 0.01) (Fig. 2E).
Spontaneous growth and arborization of the intact CST across the midline into the contralateral denervated gray matter
CST fibers originate in layer V of the primary motor cortex and project to all spinal cord levels through a crossed dorsal component that contains 95% of all descending axons (Brown, 1971; Schreyer and Jones, 1982) and an ipsilateral ventral component containing <5% of all CST axons (Vahlsing and Feringa, 1980; Joosten et al., 1992; Brösamle and Schwab, 2000). Unilateral lesion of the pyramidal tract at the level of the medulla oblongata just rostral to its decussation interrupts both the contralateral as well as the ipsilateral projections to the spinal cord. Lesion completeness was carefully assessed by histological analysis in all animals and in a sample of animals by tracing both sides of the CST (Fig. 3A–C). Animals with incomplete lesions were excluded.
To answer the question whether spontaneous recovery on the horizontal ladder test was paralleled by compensatory growth of intact CST fibers, the anterograde tracer BDA was injected into the caudal forelimb area of the contralesional sensorimotor cortex. The number of midline-crossing fibers from the labeled intact CST to the denervated side as well as branching of collaterals was evaluated in three experimental groups: unlesioned animals (intact; n = 7), lesioned animals 1 week after injury (Pyx_1 week; n = 5), and lesioned animals 3 weeks after injury (Pyx_3 weeks; n = 7).
Growth and sprouting in response to injury was evaluated at the caudal cervical enlargement (C6–C8) where motoneurons innervating forelimb and paw muscles are located (McKenna et al., 2000). Midline-crossing fibers were counted in the dorsal and ventral commissure at the central canal (Fig. 3F, level M); branching of these fibers was evaluated at two defined regions within the gray matter (Fig. 3F, levels D1 and D2). In intact animals, only few CST fibers crossed the midline and projected into the contralateral gray matter (Fig. 3D,G). One week after injury, there was no significant increase in labeled CST fibers at the midline or within the contralateral denervated gray matter, whereas after 3 weeks the number of CST collaterals projecting and branching within the denervated dorsal, intermediate, and ventral horn was significantly increased in response to injury (ANOVA, Bonferroni's post hoc: M, p ≤ 0.01; D1, p ≤ 0.001; D2, p ≤ 0.01) (Fig. 3E,G). Representative pictures of BDA-labeled CST fibers in the contralateral gray matter are shown for intact (Fig. 3D) or injured animals 3 weeks after CST lesion (Fig. 3E).
Colocalizing the CST tracer BDA and growth cone marker 2G13
To investigate whether the increase of labeled CST fibers in the denervated gray matter was attributable to newly grown fibers, we examined whether BDA-positive collaterals expressed the growth cone marker 2G13. 2G13 antibodies have been shown to label axonal growth cones in the developing rat brain in vitro as well as in vivo (Stettler et al., 1999). In P4 rat brain, we observed a high density of 2G13-positive, large, growth cone-like structures (Fig. 4A).
In contrast to this, we did not find a constitutive 2G13 expression in the intact adult rat spinal cord or on the intact, nonlesioned side (Fig. 4B). Within the denervated contralateral gray matter of lesioned animals (Fig. 4C,D), 2G13 exclusively labeled short regions of the terminal part of BDA-positive CST collaterals (Fig. 4E,F). Confocal analysis in double-labeled sections revealed a consistent colocalization of BDA and 2G13 1 week (Fig. 4G,H) as well as 3 weeks after injury (Fig. 4I).
Behavioral recovery after pyramidotomy and forced limb use
To investigate whether spontaneous behavioral recovery after unilateral CST injury could be enhanced by forced use of the impaired forelimb, we used constraint-induced movement therapy (CIMT), a training paradigm previously described in animal models of stroke (Schallert et al., 1997) as well as stroke patients (Taub et al., 1999). Initially, all rats were trained on the horizontal ladder until their success rates were >90%. They then received a unilateral CST injury by section of one pyramid, as described above. Immediately after injury, one-holed vests of plaster cast were formed around their upper torso, which forced the animals to completely rely either on their impaired or their nonimpaired forelimb. To investigate the influence of time on behavioral recovery, animals were subdivided into the following four groups: rats that could not use their impaired limb for either one (Pyx_1 week_nonuse; n = 7) or 3 weeks (Pyx_3 weeks_nonuse; n = 8) and animals that were forced to rely on their impaired side for either one (Pyx_1 week_use; n = 7) or 3 weeks, respectively (Pyx_3 weeks_use; n = 8). Finally, the cast was carefully removed and all animals were allowed to recover for 48 h before locomotor performance of their impaired limb was tested once on the horizontal ladder. Unilateral CST injury led to more errors on the horizontal ladder test (ANOVA; F = 134.21; p ≤ 0.001) with a significant difference between groups (F = 59.27; p ≤ 0.001) (Fig. 5A). One week after injury, animals that were not able to use their impaired forelimb had a lower success rate (60.4 ± 5.5%) compared with animals that were forced to rely on their impaired forelimb (80.2 ± 2.2%; Bonferroni's post hoc, p ≤ 0.01). This was even more prominent when rats had to wear their cast for a longer time period: Animals that were forced to completely rely on their impaired forelimb for 3 weeks regained their prelesion baseline performance level (success rate, 91.9 ± 3.3%), whereas animals that could not use their impaired forelimb remained significantly impaired (success rate, 63.0 ± 3.7%; Bonferroni's post hoc, p ≤ 0.001).
To control for the effect of possible muscle atrophy and stiffness of the casted limb on locomotor performance, two additional groups of animals received a sham operation followed by a cast for 3 weeks. The cast was removed and either the restricted limb (sham_nonuse; n = 6) or the unrestricted limb (sham_use; n = 9) was tested for errors on the horizontal ladder (Fig. 5B). Our results show no differences in the success rate between the casted or the noncasted intact forelimb and no decrease of locomotor performance in response to sham surgery (ANOVA, p > 0.05) (Fig. 5B).
Growth and arborization of CST fibers in response to forced limb use
To investigate whether forced limb use can influence fiber growth in the denervated spinal cord, the intact CST was traced in all animals and midline-crossing fibers as well as arborization of fibers in the ventral gray matter was evaluated as described above (Fig. 3F). We did not observe differences between spinal cords 1 week after injury (Fig. 5C); also, fiber counts were very similar to those of lesioned but uncasted animals (compare Figs. 5C, 3G). After 3 weeks, significantly more CST collaterals crossed the midline and arborized within the denervated gray matter in response to injury (ANOVA, p ≤ 0.001). Forced limb use led to an additional, significant increase of labeled fibers in the denervated gray matter compared with spinal cords of animals that could not use their impaired limb (Bonferroni's post hoc: D1, p ≤ 0.01; D2, p ≤ 0.05) similar to the amount of growth and arborization previously observed in lesioned, unconstrained animals (compare Figs. 5C, 3G). Interestingly, 3 weeks of forced use alone did not increase the number of midline-crossing fibers or axonal arbors in intact sham-operated animals (Fig. 5D) (p > 0.05).
To give an overall impression of the extent and specificity of CST fiber growth in intact and lesioned animals, representative pictures and camera lucida reconstructions of three consecutive cross sections are shown in Figure 6. BDA injections into the sensorimotor cortex label projections to the dorsal as well as ventral horn of the spinal cord. In intact animals, very few CST collaterals crossed the midline to innervate the contralateral gray matter (Fig. 6A–A″). Injury induced spontaneous growth of new fibers across the midline as well as arborization of these collaterals and/or preexisting fibers within the denervated gray matter. In addition, camera lucida reconstructions clearly demonstrate a contribution of ipsilateral ventral projections to the increased innervation in the denervated gray matter. CST axons not only increased their innervation area in response to injury by local arborization but they also extended arbors deeper into dorsal or ventral laminas as shown in animals that could not use their impaired side for 3 weeks (Fig. 6B–B″). Casted animals that were forced to completely rely on their impaired limb for 3 weeks showed similar growth across the midline but significantly more arborization within the intermediate zone as well as longer branches into the denervated dorsal and ventral horn (Fig. 6C–C″).
Midline-crossing CST fibers form glutamatergic synapses
CST collaterals within the denervated spinal cord (Fig. 7A) frequently showed bouton-like structures along their length (en passant) and at their tips (bouton terminal) (Fig. 7C,G,K). We colocalized BDA with vGlut1, a marker enriched in corticospinal synapses (Varoqui et al., 2002; Persson et al., 2006). BDA efficiently filled collaterals of corticospinal axons up to their presumed terminal boutons (Fig. 7D,H,L).
vGlut1 immunoreactivity was found throughout the gray matter being strongest in superficial laminas and weaker in intermediate and ventral laminas (Fig. 7B). The distribution pattern was consistent with termination zones of primary afferents, CST, probably brainstem–spinal and other fiber systems (Persson et al., 2006). Medium- to large-sized vGlut1-positive varicosities were present at a moderate density throughout most of lamina VII (Fig. 7E,I,M). Confocal microscopy revealed that BDA-positive en passant and terminal varicosities were positive for vGlut1 immunoreactivity, suggesting they are presynaptic glutamatergic boutons (Fig. 7F,J,N).
Synapse formation after forced limb use
To answer the question whether forced limb use has an influence on synapse formation, bouton-like structures were counted in lamina VII of the denervated gray matter in five different groups: intact animals (intact; n = 7), injured animals without any constraints (Pyx_3 weeks_free use; n = 7), as well as animals that could not use their impaired limb for either 1 week (Pyx_1 week_nonuse; n = 7) or 3 weeks (Pyx_3 weeks_nonuse; n = 8) and animals forced to rely on their impaired side for 1 week (Pyx_1 week_use; n = 7) or 3 weeks (Pyx_3 weeks_use; n = 8). In accordance with the low number of midline-crossing CST fibers in intact rats, we found only few BDA-positive synapses in the contralateral gray matter. One week after injury, the number of varicosities per labeled fiber did not increase in animals that used their impaired limb or animals that could not use their impaired limb. After 3 weeks, animals that were forced to rely on their impaired limb showed a significant increase in the number of boutons (ANOVA, followed by Bonferroni's post hoc, F = 8.826; p ≤ 0.001) (Fig. 8A–C). Interestingly, injured rats without constraints as well as animals with their impaired forelimb casted completely lacked this increase.
The density of vGlut1-positive varicosities was determined in lamina VII of the intact and the denervated gray matter. Because vGlut1 does not exclusively label CST terminations but also primary afferents and other descending fiber tracts, our analysis reflects a general change of excitatory input at the intermediate zone. Uninjured animals showed an equal distribution of vGlut1 on both sides of the cervical spinal cord. One week after CST transection, the density of vGlut1-positive varicosities on the denervated intermediate gray matter was reduced to ∼40%. This reduction persisted up to 3 weeks after injury (ANOVA; F = 8.38; p ≤ 0.01). The density of vGlut1-positive structures appeared higher in animals that were forced to train their impaired side, but this increase did not reach significance (p > 0.05), probably because of the relatively small contribution of midline-crossing CST fibers to excitatory input in the intermediate zone (Fig. 8D–H).
Gene expression differences in the denervated ventral horn of animals forced to use their impaired limb and animals that could not use their impaired limb
To identify genes within the denervated cervical gray matter that might play a role in activity-dependent growth, arborization, and synapse formation, we used gene chip arrays on sham-operated as well as lesioned animals either forced to rely on their impaired or unimpaired limb for 1 week. The ventral gray matter of the denervated side was dissected and used for analysis (Fig. 9A,B).
Comparison of lesioned with sham-operated animals showed injury-induced expression of specific sets of genes in the denervated gray matter. After injury, 441 genes were differently regulated in animals that could not use their impaired limb and 872 genes were regulated in injured animals that relied on their impaired forelimb (fold change, ≥1.2 or ≤0.8). Injury led to the upregulation of inflammation-related molecules, ion channels, and transporters, as well as the regulation of growth factors, guidance molecules, extracellular matrix (ECM) molecules, and molecules involved in synapse formation as shown in previous studies (Bareyre and Schwab, 2003). A table of all regulated genes in response to injury is provided in the supplemental material (available at www.jneurosci.org).
Microarrays of sham-operated animals showed 311 differentially expressed transcripts (fold change, ≥1.2 or ≤0.8) within the used compared with the nonused denervated side of which 223 were coding for known proteins. A large proportion (37.9%) of these differentially regulated genes were associated with cytoskeletal functions (4.9%), neurite growth (0.4%) and guidance (0.9%), cell adhesion and ECM (7.1%), synaptic function (0.9%), signaling (13.4%), and transcription (10.3%) (Fig. 9C, Table 1). Analysis of microarrays in lesioned animals revealed similar numbers; 337 genes coding for 249 known proteins were differently regulated (fold change, ≥1.2 or ≤0.8) within the used compared with the nonused denervated gray matter (Fig. 9C, Table 2). Interestingly, the proportion of genes related to growth (2.8%), synapse formation (5.2%), and adhesion (8.0%) was much higher in lesioned than in sham-operated animals after forced limb use (Fig. 9C).
The present study shows that growth and synapse formation of CST fibers from the intact side into the denervated spinal cord after unilateral CST injury is enhanced by forced use of the impaired limb. Forced limb use led to the upregulation of mRNAs involved in neuronal outgrowth, cytoskeletal rearrangements, adhesion and guidance, as well as synapse formation in the denervated cervical gray matter. Growth and arborization of CST fibers was accompanied by marked behavioral improvements, which led to full restoration of skilled forepaw movements in the horizontal ladder test, whereas animals that could not use their impaired limb remained permanently and severely impaired.
Forced limb use leads to functional recovery
Interruption of CST input to the cervical spinal cord resulted in a permanent impairment of skilled forelimb function as shown in previous studies (Whishaw et al., 1993; Piecharka et al., 2005). The same result was obtained in injured rats with immobilized impaired forelimbs. The cast allowed minor movements of the restricted forelimb, which may have helped to prevent muscle atrophy and allowed some degree of spontaneous behavioral improvements as shown in sham-operated animals. Nevertheless, an impaired forelimb might be more susceptible to cast restriction than the intact forelimb of a sham-operated animal, and the poor recovery of forelimb function might be attributable to some degree of muscle atrophy. Interestingly, we did not observe a difference between control animals without a cast that were therefore able to voluntarily use their impaired forelimb and animals that were actively prevented from using their limb. This might be explained by an increased reliance on the intact forelimb and paw, a phenomenon known as “learned disuse” in freely moving animals (Kartje-Tillotson and Castro, 1980; Jones and Schallert, 1992) and stroke patients (Taub et al., 1999). In sharp contrast, full recovery of forelimb function after pyramidotomy was observed by 3 weeks of forced limb use. This is in line with previous studies in which forced use of the impaired forelimb led to behavioral recovery in animal models of stroke (Nudo et al., 1996; Bland et al., 2001) as well as in stroke patients (CIMT) (Taub et al., 1999). The cast forced animals to completely rely on their impaired side for everyday behavior, thereby training a broad range of movements. Forced limb use might, therefore, lead to improvements in a variety of locomotor movements in contrast to other training paradigms in which training of one specific task came at the cost of a nontrained task (De Leon et al., 1998; Girgis et al., 2007). Forced use immediately after injury did not lead to locomotor deficits or increased tissue damage as seen after cortical lesion (Humm et al., 1998). In our model, we injured descending fibers at the medullary level, which left cell bodies undamaged and possibly less vulnerable to elevated glutamate levels in response to training (Kozlowski et al., 1996; Humm et al., 1999). To show behavioral recovery immediately after injury and training, we tested our animals on the horizontal ladder. The horizontal ladder test requires precise movements of forelimb and digits and is a sensitive test to study forelimb recovery after CST injury (Metz and Whishaw, 2002; Starkey et al., 2005; Bolton et al., 2006). Although the pellet reaching test allows a very detailed analysis of various movement components after pyramidotomy (Whishaw et al., 1993; Thallmair et al., 1998), it would have required some time for retraining (Z'Graggen et al., 1998).
Forced limb use enhances CST plasticity
Unilateral pyramidotomy induced growth of axons from the intact CST into the contralateral, denervated gray matter where fibers arborized and established glutamatergic synapses. We were able to demonstrate CST outgrowth by colocalization of midline-crossing fibers with the growth cone marker 2G13 as early as 1 week after injury. 2G13 exclusively labels growing axons (Stettler et al., 1999) and provides a valuable alternative to GAP-43 which also stains nongrowing fibers (Curtis et al., 1993; Kapfhammer and Schwab, 1994).
Three weeks after injury and forced limb use, the number of labeled axons within the denervated gray matter was significantly increased compared with animals that could not use their impaired limb. Although fiber density was also high in nonconstrained animals, functional reinnervation was increased only in animals that had to rely on their impaired limb as shown by the significantly higher number of BDA-labeled boutons as well as a slight but not significant increase in the total density of vGlut1-positive structures indicating that physical activity might lead to significant changes especially at the synaptic and connectivity level.
The increased fiber density is probably reflecting local sprouting of new, as well as preexisting fibers. Camera lucida reconstructions also clearly demonstrated a contribution of ipsilateral ventral projections in accordance with previous studies (Weidner et al., 2001; Brus-Ramer et al., 2007). Sprouting of midline-crossing as well as ipsilateral fibers occurs spontaneously in response to CST lesion; however, it is enhanced by interventions that increase CNS fiber growth and plasticity (Thallmair et al., 1998; Zhou et al., 2003). In injured animals, CST fibers also grew toward deeper laminas of the ventral and the dorsal horn, thereby resembling the pattern of normal CST innervation. A remarkable finding was the preference of midline-crossing fibers for the ventral part of the gray matter, which has been observed in previous studies (Raineteau et al., 2002; Brus-Ramer et al., 2007). It suggests a higher growth potential of motor CST fibers and/or the presence of local signals that support target-directed growth of axon collaterals toward ventral, motor circuits. Candidates for such signals were indeed found in our Affymetrix Chip analysis (see below). Although we cannot completely exclude the possibility that the higher density of labeling in the denervated spinal cord is attributable to an increased BDA transport into preexisting small diameter collaterals, the total number of BDA-labeled axons in the CST was equal in all treatment groups, and there is no evidence so far for changes in tracer transport in an intact tract in response to physical activity.
CST and functional recovery
Improvements of forepaw placement on the horizontal ladder in response to forced limb use were seen as early as 1 week after injury, whereas changes in axonal growth and branching were only detected at a later stage. This suggests that forced limb use has a beneficial influence on motor performance by involving a variety of intraspinal and supraspinal systems (Hess and Donoghue, 1994; Wolpaw, 1997; Tillakaratne et al., 2000). Intraspinal changes might contribute to behavioral improvements observed within the first week after injury, whereas structural reorganization probably requires a certain amount of time after the onset of training to be detected as suggested by Kleim et al. (2004).
Because of its accessibility and defined morphology, we demonstrate morphological changes in response to injury and forced limb use in the intact CST, but it is clear that structural plasticity of other descending tracts may also contribute to the behavioral recovery. Rubrospinal and corticospinal tracts can compensate for each other to a certain extent after injury (Martin and Ghez, 1988; Raineteau et al., 2002). The reticulospinal system is left intact in our lesion paradigm and might also contribute to the training effect. Nevertheless, control of precision movements, in particular of hand/forepaw, is a major function of the motor CST (Castro, 1972; Kalil and Schneider, 1975), and the contribution of uninjured CST fibers on behavioral improvements has been shown in previous studies (Kartje-Tillotson et al., 1987; Thallmair et al., 1998).
Intraspinal changes induced by CST denervation and forelimb training
Complete transection of one CST and forced limb use led to increased growth/stabilization of midline-crossing CST fibers, suggesting a mechanism of competition for synaptic space and trophic support as previously described after pyramidotomy and/or electrical stimulation of the intact CST (Martin et al., 2004; Brus-Ramer et al., 2007). Interestingly, electrical stimulation leads to increased outgrowth of ventral CST fibers in intact animals, whereas forced limb use alone was not sufficient to induce structural rearrangements.
Pyramidotomy was shown to induce the expression of specific sets of genes in the denervated gray matter of the spinal cord (Bareyre et al., 2002; Bareyre and Schwab, 2003). These results are in line with our present observations. The effect of training on gene expression in the cervical spinal cord is still poorly understood. Forced limb reduces growth-inhibitory factors and extracellular matrix molecules and regulated growth-promoting factors and cytoskeletal dynamics. There was regulation of adhesion molecules, axonal guidance molecules, and components of synapse formation [e.g., glial cell line-derived neurotrophic factor receptor α 1 (Gfra1), brain-derived neurotrophic factor (Bdnf), as well as Neuregulin1 (Nlgn1) and Piccolo (Pcl)] (Garcès et al., 2000; Phelps et al., 2002; Liu et al., 2005; Tao-Cheng, 2006; Fischbach, 2007). Because of the variability in lesion models, training paradigms, investigated time points, or spinal levels, there is no agreement about the importance or specificity of different factors. BDNF for example plays an important role in synapse formation and stabilization (Poo, 2001) and physical exercise in intact and spinal cord-injured animals increases mRNA levels of BDNF (Cotman and Berchtold, 2002; Gómez-Pinilla et al., 2002; Ying et al., 2005). Training after cervical lesion, however, did not increase BDNF levels (Girgis et al., 2007), and 1 week after lesion and forced limb use we even observed a decrease within the denervated gray matter.
Our study clearly shows that rehabilitative forelimb training regulates compensatory plastic changes at the spinal level after injury followed by behavioral recovery. The analysis of the specificity and importance of individual molecules involved in activity-dependent reorganization will be an exciting task for future studies.
This work was supported by Swiss National Science Foundation Grant 31-63633.00, National Center of Competence in Research “Neural Plasticity and Repair” of the Swiss National Science Foundation, and Spinal Cord Consortium of The Christopher and Dana Reeve Foundation (Springfield, NJ). We thank Laura Montani and the Functional Genomics Center Zurich for assistance with the gene chips, Michelle Starkey for critical review of this manuscript, and Roman Willi for assistance with the statistics.
- Correspondence should be addressed to Irin C. Maier, Brain Research Institute, Winterthurerstrasse 190, 8057 Zurich, Switzerland.