Anatomical plasticity such as fibre growth and the formation of new connections in the cortex and spinal cord is one known mechanism mediating functional recovery after damage to the central nervous system. Little is known about anatomical plasticity in the brainstem, which contains key locomotor regions. We compared changes of the spinal projection pattern of the major descending systems following a cervical unilateral spinal cord hemisection in adult rats. As in humans (Brown-Séquard syndrome), this type of injury resulted in a permanent loss of fine motor control of the ipsilesional fore- and hindlimb, but for basic locomotor functions substantial recovery was observed. Antero- and retrograde tracings revealed spontaneous changes in spinal projections originating from the reticular formation, in particular from the contralesional gigantocellular reticular nucleus: more reticulospinal fibres from the intact hemicord crossed the spinal midline at cervical and lumbar levels. The intact-side rubrospinal tract showed a statistically not significant tendency towards an increased number of midline crossings after injury. In contrast, the corticospinal and the vestibulospinal tract, as well as serotonergic projections, showed little or no side-switching in this lesion paradigm. Spinal adaptations were accompanied by modifications at higher levels of control including side-switching of the input to the gigantocellular reticular nuclei from the mesencephalic locomotor region. Electrolytic microlesioning of one or both gigantocellular reticular nuclei in behaviourally recovered rats led to the reappearance of the impairments observed acutely after the initial injury showing that anatomical plasticity in defined brainstem motor networks contributes significantly to functional recovery after injury of the central nervous system.
After injury to the CNS, varying degrees of functional recovery are observed depending typically on the location and size of the lesion. Identification of the underlying mechanisms of spontaneous recovery and the factors that impede further functional improvements is an important prerequisite for the development of effective new therapies for brain and spinal cord injury (Raineteau and Schwab, 2001; Courtine et al., 2011; Rossignol and Frigon, 2011). Anatomical plasticity in the CNS ranging from sprouting of lesioned and unlesioned fibres to the de novo formation of detour pathways has been shown to mediate functional recovery after brain and spinal cord injury in rodents and primates (Bareyre et al., 2004; Dancause et al., 2005; Courtine et al., 2008; Zorner and Schwab, 2010). After one-sided spinal cord injury (e.g. Brown-Séquard syndrome), spared contralateral tracts are believed to mediate functional recovery through pre-existing or newly formed fibres that cross the spinal cord midline below the level of the injury (Ghosh et al., 2009). Thus far, beneficial, spontaneous neuroanatomical adaptations are almost exclusively described in the sensorimotor cortex and its projections as well as in intraspinal networks (Nudo, 2006; Courtine et al., 2008). Much less is known about adjustments in other brain regions after CNS damage, in particular in the brainstem, which contains the phylogenetically oldest and functionally most important centres for basic movement control.
Key structures for the initiation and execution of locomotion are located in the midbrain, pons and medulla oblongata (Shik et al., 1969; Grillner, 1996). From fish to mammals, locomotor command regions located in the rostral brainstem such as the mesencephalic locomotor region (MLR) are directly connected to bulbar output systems, for example the medial reticular formation (Garcia-Rill et al., 1986; Webster and Steeves, 1991; Matsuyama et al., 2004; Ryczko and Dubuc, 2013). The reticular formation, in turn, projects to the spinal cord to initiate and coordinate limb movements and postural support, and to modulate the locomotor rhythm produced by the spinal central pattern generators (Grillner and Wallen, 1985; Matsuyama et al., 2004; Grillner et al., 2008; McCrea and Rybak, 2008). The rubrospinal system is involved in the execution of precise limb movements whereas vestibulospinal tracts primarily control balance and posture (Markham, 1987; Armstrong, 1988; Whishaw et al., 1998). Despite the fact that these phylogenetically highly conserved brainstem centres are crucial components of the CNS motor network, only limited information is currently available on their responses to CNS damage (Raineteau et al., 2001; Ballermann and Fouad, 2006; Courtine et al., 2008; Weishaupt et al., 2013). In particular, a comparative analysis of neuroanatomical changes in the different central nervous systems after injury is missing.
In this study, we assessed the anatomical plasticity in the major descending central nervous systems after cervical unilateral spinal cord hemisections in adult rats. We investigated whether changes in the spinal projection pattern of a plastic region are accompanied by anatomical remodelling of its input from other CNS centres and we evaluated the functional significance of these adaptations.
Materials and methods
Adult female Lewis rats, 10 weeks of age (200–250 g, R. Janvier, France) were housed in groups of three to five rats in standardized cages (type 4 Macrolon) under a 12-h light/dark cycle with food and water provided ad libitum. All experiments were approved by the Veterinary Office of the Canton of Zurich, Switzerland.
Spinal cord injury and animal care
Rats were deeply anaesthetized with subcutaneous injections of Hypnorm/Dormicum (Hypnorm: 120 μl/200 g body weight, Janssen Pharmaceutics; Dormicum 0.75 mg/200 g body weight, Roche Pharmaceuticals) for all surgical procedures described below. For cervical hemisection injury at C4 spinal level, the rats’ skin was cut between the occipital bone and the prominent process of the T2 vertebra with a scalpel. Laminectomy of the third cervical vertebra was performed and the dura was opened along the rostrocaudal axis with a fine needle to expose the entire dorsal surface of the spinal cord. A 27-gauge needle was lowered at the spinal cord midline and served as boundary for the hemisection. As we found no paw preference in the cylinder test in this rat strain (see ‘Results’ section), all rats received a right-sided unilateral hemisection injury at cervical level C4. The right spinal hemicord was cut with a sharp sapphire knife (WPI). To ensure completeness of the lesion, cuts were repeated until the ventral surface of the spinal canal was visible. Following the lesion, the muscle layers were sutured and the skin was closed with surgical staples. One day before and for 2 days after surgery, rats received subcutaneous injections of the analgesic Rimadyl (5 mg/kg, Pfizer). To prevent bladder and wound infections, rats were treated with daily subcutaneous injections of the antibiotic Baytril (5 mg/kg body weight, Bayer) for 1 week. All rats were checked twice daily for the entire period of the experiment.
To assess a possible role of the myelin-associated neurite growth inhibitor Nogo-A, all chronic spinal cord injured animals received a 2-week intrathecal, thoracic spinal infusion of antibodies [control-IgG or anti-Nogo-A (11C7) antibodies; total amount infused: 6 mg] by way of an osmotic mini-pump (5 µl/h, 3.1 µg IgG/µl, Alzet 2ML2; Charles River Laboratories) starting at the time of injury. As shown by densitometry of immunostained sections on different levels of the spinal cord rostral to the lesion, antibodies did not efficiently diffuse into the spinal and brain parenchyma, most probably due to a continuous, lesion-induced dura leak at the cervical hemisection site during the time of infusion (data not shown). The subdural spaces were collapsed all along the spinal cord, supporting the hypothesis of a long-lasting, major CSF leak in this lesion paradigm. Clinically, large dura mater lesions are known to heal slowly (Sayad and Harvey, 1923). Because of the unavailability of the antibody in the tissue, antibody-treatment was disregarded. When control-IgG and anti-Nogo-A antibody treated animals were compared, no differences could be observed for any of the read-outs shown in the present study.
Perfusion and Nissl staining
After completion of the experiments, rats were anaesthetized with an overdose of pentobarbital (intraperitoneal) and transcardially perfused with 100 ml heparinized Ringer’s solution followed by 400 ml Ringer’s solution containing 4% paraformaldehyde and 5% sucrose. Brains and spinal cords were removed, post-fixed in the same fixative overnight at 4°C, cryoprotected in 30% sucrose for 5 days, embedded in Tissue-Tek® and frozen in isopentane at −40°C. The tissue was cut on a cryostat and 40-μm thick cross-sections were collected on-slide. Completeness of hemisections was confirmed histologically on Nissl stained spinal cord cross-sections of the lesion centre. For staining, dried sections were bathed in Cresyl violet solution for 3 min, dehydrated in ethanol, washed in xylol and coverslipped with Eukitt® (Kapitza et al., 2012). Rats with incomplete hemisections were removed from analysis.
Behavioural testing and quantification
Rats were familiarized with the behaviour set-ups and trained on the tasks for a total of 5–7 sessions within 14 days. Baseline measurements were performed 1 week before spinal cord injury. To characterize the course of functional recovery after hemisection injury (n = 20; five rats were excluded retrospectively due to an incomplete hemisection), rats were tested 1, 2, 4, 8 and 12 weeks after lesion. Grooming behaviour was first assessed 2 weeks after injury and paw preference was tested only once, 1 week before hemisection.
The ‘paw preference test’ (‘cylinder test’) was performed to detect asymmetries in forelimb use for each individual rat before injury and evaluated according to earlier studies (Gensel et al., 2006). Also, air-righting was assessed and scored as published previously (Laouris et al., 1990; Pellis et al., 1991). In brief, rats were held in supine position ∼30 cm above a surface covered with foam material. After being released intact rats land on their paws due to a successive rotation of head, shoulder girdle, thorax and pelvis. Air-righting reaction was tested 10 times per time point and video recorded. Scoring was performed based on the posture of the rat at landing (Laouris et al., 1990): 0 = no rotation, 1 = head rotation only, 2 = head and thorax rotation only, 3 = landing on the side, 4 = full rotation except hindlimbs, 5 = full rotation. Values represent the mean of 10 trials per rat. Grooming was assessed immediately after the swimming test (with wet fur) in a transparent Plexiglas cage filled with bedding. Grooming activity was recorded with a video camera for 3 min. Each forelimb was evaluated separately. Scoring was performed according to a previously published rating scheme (Bertelli and Mira, 1993; Gensel et al., 2006) based on the most remote part of the head that could be reached by the forepaws: 0 = no contact with the head, 1 = contact with snout (underneath), 2 = contact with the dorsal surface of the snout, 3 = contact with the area between eyes and the front of the ears (forehead), 4 = contact with the ears (front), 5 = contact with the area behind the ears. Over-ground walking, wading through 3-cm deep water, swimming and walking over a horizontal ladder was tested in a behavioural testing set-up (‘MotoRater’) described previously (Zorner et al., 2010). Before testing, the rats’ skin overlying bony anatomical landmarks on fore- and hindlimbs was tattooed with a commercially available tattooing kit (Hugo Sachs Elektronik, Harvard Apparatus) to ensure reliable measurements over time. Locomotor behaviour was filmed with a high-speed video camera (Basler A504kc Color Camera) at 200 Hz and analysed with the ClickJoint software (version 5.0; ALEA Solutions). Parameters assessed during over-ground walking and wading were defined as follows: shoulder and hip height as the vertical distance between the joint and the runway; pro- and retraction of the fore- and hindlimbs as the maximal horizontal distance (forwards and backwards) between the wrist and the shoulder joint or between the toe (MTP) and the hip, respectively; paw dragging was present if toes touched the ground during the swing phase; the paw rotation angle (‘external rotation’) was defined as the angle between the body axis and the paw axis (line between the third MTP and the heel) and assessed at mid-stance; step width (‘base of support’) was calculated by adding the distances measured between the body axis and the left and the right heel for two consecutive (left-right) steps. Precise paw placement of fore- and hindlimbs of rats crossing a horizontal ladder with irregularly spaced rungs was assessed (Metz and Whishaw, 2002; Zorner et al., 2010). The percentage of functional placements, i.e. weight-bearing steps without slipping, was used as a parameter and reported for each limb separately. For swimming, the mean swimming velocity was calculated from at least three swims (swimming distance of 60 cm per run) per time point. The MTP peak velocity was defined as the maximal velocity of toe movement during hindlimb strokes.
After Week 12 post-lesion, rats with spinal cord injury (n = 15) used for the characterization of recovery after hemisection injury (Fig. 1) were further trained on the locomotor tasks (over-ground walking, wading and swimming) once a week. These rats received an electrolytic microlesion to either the ipsilesional (n = 5) or the contralesional (n = 10; one rat had to be excluded retrospectively due to incorrect positioning of the lesion) gigantocellular reticular nucleus (with reference to the side of the hemisection) 16 weeks after spinal cord injury. Rats without hemisection (n = 7) but electrolytic microlesion of the left gigantocellular reticular nucleus served as controls. Functional performance was assessed 1 day before and 2 days after the focal brainstem lesion.
Retrograde tracing from the spinal cord
Intact and spinal cord injured rats were deeply anaesthetized (Hypnorm/Dormicum, see above) and fixed in a stereotactic frame after partial laminectomy of the C5, C6 and C7 (spinal segments C6–T1) and T12 and T13 vertebrae (spinal segments L1–L4) (Gelderd and Chopin, 1977). Rats with hemisection injury received unilateral injections of two different fluorescent retrograde tracers into the ipsilesional cervical (tetramethylrhodamine, 3000 MW dextran; 10% in injectable water; Invitrogen) and ipsilesional lumbar spinal cord (diamidino yellow dihydrochloride, 2% suspension in 0.1 M phosphate buffer and 2% dimethylsulphoxide; Sigma-Aldrich) at 1 week (n = 5), 4 weeks (n = 10) or 12 weeks post-lesion (n = 10). Intact rats (n = 5) were traced accordingly. Injections were made stereotactically with Nanofil syringes attached to a MicroPump (WPI). A total of 2 µl of tetramethylrhodamine administered through 10 stereotactic injections of 200 nl was injected with a 33-gauge needle along the cervical spinal cord (700 µm lateral from the spinal cord midline, in a depth of 1 mm from the dorsal surface and a spacing of 1 mm between injections). Similarly, a total of 2 µl of diamidino yellow dihydrochloride was injected with a 28-gauge needle into the lumbar spinal hemicord (10 injections with 200 nl/injection, 500 µm lateral from the spinal cord midline, in a depth of 700 µm from the dorsal surface and a spacing of 1 mm between injections). Rats were perfused 14 days after tracing. Exclusive unilateral spread of the tracers was confirmed post-mortem for all injection sites on 40-µm thick spinal cord cross-sections under a fluorescence microscope. Coronal 40-µm thick cryostat sections of the whole brain and cross-sections of the spinal cord rostral to the lesion site were obtained for each rat. Prominent landmarks along the brains’ rostrocaudal axis were used as a reference: the anterior commissure, the rostral end of the red nucleus, the central branch of the facial nerve, the rostral end of the inferior olive and the pyramidal decussation. Each brain region was identified using these landmarks and a rat brain atlas (Paxinos and Watson, 1998). Retrogradely labelled cell bodies were counted bilaterally in 19 different brain regions by a blinded experimenter. The same number of sections was analysed for each region and each rat. Single and double labelled cell bodies were quantified on every fourth brain section (i.e. one section per 160 µm) under a fluorescence microscope and the result was multiplied by four to approximate the actual number of traced neurons. In the present study, we report absolute cell numbers for each region (Supplementary Tables 1–3). However, since absolute cell counts demonstrated high within-group variability due to the tracing procedure, absolute cell numbers for a given region were normalized to the total number of retrogradely labelled cells of the respective brain for each tracer.
Rostrocaudal coordinates (‘+’ rostral to landmark, ‘−’ caudal to landmark) used for the individual regions were: the rostral motor cortex (start at 4800 µm + anterior commissure, end at 2560 µm + anterior commissure, total 2240 µm), the caudal motor cortex (start at 2560 µm + anterior commissure, end at 4640 µm − anterior commissure, total 7200 µm), the secondary somatosensory cortex (start at 1120 µm + anterior commissure, end at 4640 µm − anterior commissure, total 5760 µm), the red nucleus (start at red nucleus, end at 1920 µm − red nucleus, total 1920 µm), the deep mesencephalic reticular formation (start at 1280 µm + red nucleus, end at 1920 µm − red nucleus, total 3200 µm), the pontine reticular nucleus, oral part (start at 1920 µm − red nucleus, end at 1120 µm + facial nerve, total 2080 µm), the pontine reticular nucleus, caudal and ventral part (start at 1120 µm + facial nerve, end at facial nerve, total 1120 µm), the gigantocellular reticular nucleus (start at 160 µm − facial nerve, end at 1440 µm − inferior olive, total 3200 µm), the lateral paragigantocellular nucleus, the gigantocellular reticular nucleus (ventral part) and the gigantocellular reticular nucleus part alpha (start at 160 µm − facial nerve, end at 1440 − inferior olive, total 3200 µm), the dorsal paragigantocellular nucleus (start at 320 µm − facial nerve, end at 320 µm + inferior olive, total 1440 µm), the medullary reticular nucleus, ventral part (start at 1440 µm − inferior olive, end at pyramidal decussation, total 1440 µm), the medullary reticular nucleus, dorsal part (start at 800 µm − inferior olive, end at pyramidal decussation, total 1920 µm), the medial vestibular nucleus (start at 320 µm + facial nerve, end at 1440 µm − inferior olive, total 3680 µm), the lateral vestibular nucleus (start at 160 µm − facial nerve, end at 320 µm + inferior olive, total 1600 µm), the locus coeruleus (start at 1760 µm + facial nerve, end at facial nerve, total 1920 µm), the raphe magnus nucleus (start at 480 µm + facial nerve, end at inferior olive (IN), total 2560 µm), the raphe obscurus nucleus (start at 640 µm + inferior olive, end at pyramidal decussation, total 3520 µm) and the raphe pallidus nucleus (start at 160 µm + inferior olive, end at 640 µm + pyramidal decussation, total 2080 µm).
For propriospinal neurons located in the rostral cervical spinal cord at C3 and C4 level, labelled neurons were counted on every second to third spinal cord cross-section (thickness of 40 µm). As the number of analysed sections was different for each spinal segment and rat (10–40 sections per segment), the mean number of labelled cells per section was calculated. One intact rat was traced bilaterally from the cervical and lumbar spinal cord and each brain region analysed in the present study was reconstructed in 3D using Neurolucida 8.0 (MicroBrightField).
Anterograde tracing of reticulospinal fibres
Rats with cervical unilateral hemisection injury (n = 15) received a single stereotactic injection of 50 nl of the tracer mini-emerald (Fluorescin and biotin tagged 10 000 MW dextran, 10% in injectable water; Invitrogen) into the contralesional gigantocellular reticular nucleus 4 weeks after injury. Using a dorsal approach via the cerebellum, a 35-gauge needle was placed slightly dorsolaterally to the centre of the gigantocellular reticular nucleus (4.6 mm caudal to lambda, 1.5 mm lateral to lambda, 2 mm above the base of the skull) to avoid diffusion of the tracer over the midline and damage to the target neurons. In intact rats (n = 10, two rats were excluded due to incorrect tracing), the left gigantocellular reticular nucleus was traced accordingly. All rats were sacrificed 3 weeks after tracing and accurate placement of the injections was confirmed post-mortem. Series of 40-µm thick cryostat cross-sections of the cervical (C6–T1) and lumbar (L1–L6) spinal cord were analysed under a fluorescence microscope (every fourth section, 10 sections per segment). We counted fibres entering the contralesional (left for intact rats) spinal cord grey matter, midline crossing fibres and fibres in the ipsilesional (right for intact rats) grey matter. The latter were defined as fibres intersecting a virtual vertical line placed in the centre of the spinal cord grey matter at a distance of 600 µm from the central canal. To account for the variability induced by the tracing procedure, normalization was performed (Starkey et al., 2012; Lindau et al., 2014). Therefore, the number of midline-crossing fibres was normalized to the number of fibres entering the contralesional/left spinal cord grey matter for each segment. To assess the degree of arborization of reticulospinal fibres crossing the midline, the number of fibres counted in the ipsilesional/right spinal cord grey matter (see above) was divided by the number of midline-crossing fibres.
Retrograde tracing from the gigantocellular reticular nucleus
As described for anterograde tracing, intact rats (n = 10) and rats with chronic hemisection injury (n = 10, 4 weeks after injury) received stereotactic tracer injections into the ipsilesional (right in intact rats) and contralesional (left in intact rats) gigantocellular reticular nucleus. Tetramethylrhodamine (3000 MW dextran; Invitrogen) was injected into the ipsilesional/right and mini-emerald (10 000 MW dextran; Invitrogen) into the contralesional/left nucleus. Correct positioning of the needle was checked for each injection site and injections that were off target were excluded from further analysis. For the rostral cortex, retrogradely-labelled cells were counted on every second 40-µm thick coronal section (in total 17 sections per rat, starting at the most rostral section with labelled cells). For the MLR, labelled cells were quantified in the area (1.4 mm2) of the cuneiform and pedunculopontine nucleus located laterally to the mesencephalic trigeminal neurons. Again, cells were counted in this area on every second 40-µm thick coronal section (in total 12 sections per rat). To normalize for differences in tracing efficiency cell counts were expressed as a lateralization index (number of cells contralateral to injection/number of cells ipsilateral to injection).
Focal lesions of the gigantocellular reticular nucleus
Rats received a local, anodal, electrolytic lesion of the gigantocellular reticular nucleus in the rostral medulla oblongata using platinum-iridium electrodes (118 kΩ; Science Products,) attached to a stereotactic frame. The tip of the electrode was placed in the ventromedial part of the gigantocellular reticular nucleus, 4.6 mm caudal to lambda, 1 mm lateral to lambda and 1.5 mm above the base of the skull. An electric current of 200 µA was applied continuously for 35 s. Only data obtained from rats with histologically confirmed accurate positioning of the lesion were analysed.
Statistical analysis was performed using SPSS (V21; SPSS, Inc) and GraphPad Prism 5 (V5.01; GraphPad Software, Inc). The level of statistical significance for all tests was set a priori at P < 0.05. The exact P-values obtained from statistical analyses are given. For analysis of the behavioural data characterizing the time course after spinal cord injury (Fig. 1), we used one-way ANOVA for repeated measurements. If significant, ANOVA was followed by post hoc Bonferroni’s Multiple Comparison Test comparing the performance at baseline and 1 week (or 2 weeks for grooming) after injury (‘lesion effect’) and 1 week (or 2 weeks for grooming) and 12 weeks after injury (‘recovery’). For retrograde tracing from the cervical and lumbar spinal cord (Figs 3 and 4), the absolute numbers of retrogradely labelled neurons found in a given CNS region were normalized to the total cell number counted per brain (for the same tracer) to reduce within-group variability due to tracing. Statistical analysis on the number of normalized cell counts was performed with one-way ANOVA and followed, if significant, by Bonferroni’s Multiple Comparison Test. For anterograde tracing from the contralesional gigantocellular reticular nucleus (Fig. 5), repeated measures two-way ANOVA with one repeated factor (‘Segment’) and one non-repeated factor (‘Injury’) was applied to detect significant differences between each spinal cord segment and between intact rats and rats with hemisection injury. Total mean values were analysed with Student’s unpaired t-test (two-sided). For changes of the lateralization index (Fig. 6), one-way ANOVA was used to test for significant differences between animal groups. Here, repeated measures ANOVA could not be applied because of the differences in group size due to the exclusion of inaccurate injection sites. If significant, ANOVA was, again, followed by Bonferroni’s Multiple Comparison Test. For statistical analysis of the behavioural effects of microlesions in the brainstem (Fig. 7), repeated measures two-way ANOVA with one repeated factor (‘Time’, i.e. before and after brainstem lesion) and one non-repeated factor (‘animal group’) was used. Only if ANOVA was significant for the repeated measures factor ‘Time’, post hoc analysis was performed with Bonferroni’s Multiple Comparison Test (Supplementary Table 4).
Behavioural outcome after cervical unilateral spinal cord injury
Adult Lewis rats were tested for their paw preference in the cylinder test and results indicated equal use of both paws (data not shown). After several days of training and baseline testing in a number of behavioural tasks (see below), all rats received a hemisection injury at cervical level C4 (Fig. 1A). Only data from rats with a histologically confirmed complete hemisection (n = 15) were included in the analysis.
The air-righting reaction (Fig. 1B), i.e. the ability to turn in the air from a supine position in order to land on the paws, was strongly impaired 1 week after injury (P < 0.0001, repeated measures one-way ANOVA). At this time point, rats were only just able to rotate their head and shoulder girdle and occasionally their hindlimbs before landing. This was followed by significant recovery of this response starting 1–2 weeks after lesion (P < 0.0001, repeated measures one-way ANOVA).
The ipsilesional, right forepaw was rigid and with flexed, contracted digits (Fig. 1C). Starting 2–4 weeks after lesion, rats used their ipsilesional forelimb like a ‘walking stick’ with only minimal movements in the elbow and wrist, as described previously (Zorner et al., 2010; Filli et al., 2011). All assessed parameters including grooming scores (Fig. 1D), shoulder height (Fig. 1E) and limb excursions (Fig. 1F and G) as well as precise paw placement on the horizontal ladder (Fig. 1H) indicated very severe deficits in the ipsilesional forelimb acutely after lesion. Starting 2 weeks post-lesion, a small but significant degree of improvement of crude forelimb movements, such as increased shoulder height (Fig. 1E) and limb protraction (Fig. 1F), occurred (P < 0.0001, repeated measures one-way ANOVA). However, performance of the ipsilesional forelimb in general, and on the horizontal ladder in particular, remained far below baseline levels up to 12 weeks post-lesion (<20% of baseline performance) indicating persistent deficits of distal forelimb function. The contralesional forelimb was not significantly affected by the lesion (Fig. 1H; P = 0.45, repeated measures one-way ANOVA).
Hindlimbs demonstrated moderate to severe deficits acutely after cervical hemisection injury with the ipsilesional hindlimb being more affected than the contralesional one in most of the behavioural tasks. Locomotor parameters assessed during over-ground walking [toe clearance (Fig. 1I), paw rotation (Fig. 1J), base of support (Fig. 1K)], wading through shallow water [hindlimb excursion (Fig. 1L and M), hip height (Fig. 1N)] and swimming [swimming velocity (Fig. 1O), maximal toe velocity during hindlimb strokes (Fig. 1P)] showed significant recovery, often close to pre-lesion levels, 4–12 weeks after spinal cord injury [for all parameters P < 0.01, repeated measures one-way ANOVA, except for paw rotation of the contralesional hindlimb (Fig. 1J)]. Precise paw placement on the horizontal ladder was initially very poor for both hindlimbs illustrated by 20–40% (contralesional limb) and <10% (ipsilesional limb) functional paw placements on the ladder rungs 1 week after lesion (Fig. 1Q). In contrast with the good recovery of basic locomotor functions, only limited recovery was observed for both hindlimbs on the horizontal ladder.
Anatomical plasticity of spinal descending tracts after spinal cord hemisection assessed by retrograde tracing
Cervical (C4) hemisection disrupts all descending tracts on the lesioned side. To re-establish functional control over the denervated spinal hemicord below the lesion, fibres from the intact side may sprout across the spinal cord midline or activate pre-existing midline crossing axonal branches. This was studied by injecting retrograde neuroanatomical tracers with different colours into the denervated ipsilesional cervical and lumbar spinal cord grey matter below the lesion (Fig. 2A). These tracers were taken up very efficiently by axon terminals, transported retrogradely and accumulated in the cell bodies. Four groups of rats received such tracer injections: (i) intact rats (n = 5); (ii) rats traced 1 week after injury (n = 5); (iii) rats traced 4 weeks after injury (n = 10) and; (iv) rats traced 12 weeks after injury (n = 10). Retrogradely labelled cell bodies were counted bilaterally in 19 different brain regions (Fig. 2B–E) and in the cervical spinal cord rostral to the injury. Although stereotactic tracer injections were performed identically for all rats, absolute cell counts showed high within-group variability (Supplementary Tables 1–3). Therefore, cell counts for a given region were normalized to the total number of retrogradely labelled cells of the respective brain (Figs 3 and 4).
After retrograde tracing from the denervated ipsilesional cervical spinal cord below the lesion (Fig. 3), we found a significant increase of labelled normalized cells in the contralesional gigantocellular reticular nucleus at 4 weeks after injury compared to the acute state (P = 0.023, one-way ANOVA). In terms of absolute numbers, we found 345 ± 146 labelled cells in the contralesional gigantocellular reticular nucleus in rats traced 1 week after injury compared to 402 ± 158 and 685 ± 280 labelled cells in those rats traced 4 and 12 weeks post-lesion, respectively (Supplementary Table 1). There was a tendency for higher cell numbers in the ipsilesional red nucleus (17 ± 11 cells at 1 week versus 59 ± 20 cells at 12 weeks after lesion, Supplementary Table 1) and the contralesional locus coeruleus (50 ± 50 cells at 1 week versus 104 ± 61 cells at 12 weeks after lesion, Supplementary Table 1), however, with large interindividual variations. Spinal projections to the denervated cord were unchanged for the cortex, other reticular nuclei and the vestibular nuclei. The number of raphe neurons projecting to the denervated cervical hemicord from the contralesional raphe was reduced 4 weeks after injury (P = 0.026, one-way ANOVA; Fig. 3). After retrograde tracing from the lumbar denervated hemicord (Fig. 4), we again found significantly higher normalized cell numbers in the contralesional gigantocellular reticular nucleus in the chronic rats (P = 0.041, one-way ANOVA). Accordingly, absolute cell numbers for the contralesional gigantocellular reticular nucleus increased from 434 ± 119 labelled cells at 1 week after injury to almost twice the amount at 4 and 12 weeks post-lesion (Supplementary Table 2). Also, normalized cell numbers were increased in the ipsilesional dorsal medullary reticular nucleus 12 weeks after injury compared to rats traced 1 week after hemisection (P = 0.0016, one-way ANOVA; Fig. 4). Again, the number of normalized cells in the raphe nuclei was reduced in chronic rats after lumbar tracing, however, this time, statistical significance was found for the ipsilesional side (P < 0.0226, one-way ANOVA; Fig. 4). There was a tendency for increased projections to the denervated lumbar hemicord from the contralesional cortex and pontine reticular nuclei. No significant changes were detected for the other brain regions. Most double-labelled cells, i.e. cells with collaterals in the cervical and lumbar spinal cord, were found in the pontine and medullar reticular formation in intact rats (Supplementary Table 3). After hemisection, the number of double-labelled cells counted in the contralesional reticular formation increased proportionally with the number of single-labelled neurons. Absolute cell counts for all analysed brain and spinal cord regions are given in Supplementary Tables 1–3.
In summary, these retrograde tracing results show plastic reactions in some but not all brainstem regions. In particular, this is seen in the contralesional gigantocellular reticular nucleus with many neurons whose axons cross the spinal cord midline in response to the hemisection injury. Anterograde tracing of this nucleus was performed to directly visualize these axons.
Anterograde tracing of reticulospinal projections of the gigantocellular reticular nucleus
Intact rats (n = 8) and rats with chronic cervical unilateral hemisection injury (n = 15; 4 weeks after lesion) received a stereotactic tracer injection into the contralesional (left nucleus in intact rats) gigantocellular reticular nucleus [Fig. 5A(I–III)]. Traced fibres entering the contralesional spinal cord grey matter, midline-crossing fibres and fibres in the ipsilesional, denervated grey matter at lower cervical (C6–T1) and lumbar (L1–L6) spinal levels were assessed [Fig. 5B(I–IV)]. In general, there were significantly more midline-crossing fibres in the lumbar spinal cord than in the cervical segments (P = 0.0003 for the factor ‘Segment’, repeated measures two-way ANOVA; Fig. 5C). Compared with intact rats, we found a significant increase of the number of midline-crossing fibres for all segments in rats with spinal cord injury (P = 0.0444 for the factor ‘Injury’, repeated measures two-way ANOVA; Fig. 5C). In intact rats, there were 1–2 fibres in the ipsilesional spinal cord grey matter [Fig. 5B(IV)] per midline crossing fibre. These fibres were primarily found in the ventral spinal cord grey matter in intact and lesioned animals. After spinal cord injury, there was a tendency towards a reduction of fibres in the ipsilesional cervical spinal cord per midline-crossing fibre but an unchanged ratio in the lumbar spinal cord (Fig. 5D).
These findings from anterograde tracing confirmed the previous results of the retrograde tracing experiments demonstrating sprouting of spinal projections of the contralesional gigantocellular reticular nucleus across the spinal cord midline to the denervated side after hemisection. We therefore investigated whether plastic changes including a side-switch were also present in the input systems to the contralesional gigantocellular reticular nucleus in response to spinal cord injury.
Retrograde tracing of the inputs of the gigantocellular reticular nucleus
Intact rats (n = 10) and rats 4 weeks after cervical unilateral hemisection (n = 10) received stereotactic injections of the fluorescent tracers mini-emerald (green) into the contralesional (left in intact rats) and tetramethylrhodamine (red) into the ipsilesional (right in intact rats) gigantocellular reticular nucleus. In intact rats, retrogradely labelled cell bodies were found primarily in the reticular formation in the midbrain, pons and medulla oblongata on both sides, as well as in the forebrain cortex. Two brain regions that are functionally and anatomically closely linked to the gigantocellular reticular nucleus were analysed in detail: the rostral motor cortex (Antal, 1984; Matsuyama et al., 2004), containing the rostral forelimb area and also considered as the premotor region in rats (Passingham et al., 1988; Rouiller et al., 1993; Smith et al., 2010) and the MLR, a phylogenetically ancient and highly conserved command centre for locomotion (Steeves and Jordan, 1984; Garcia-Rill et al., 1986) [Fig. 6A(I–IV)]. Based on retrogradely labelled cell counts in intact rats, we found that the gigantocellular reticular nucleus received bilateral projections from the rostral motor cortex with about two-thirds of the traced neurons located in the contralateral and one-third in the ipsilateral cortex [Fig. 6A(I and III)]. Projections from the MLR were also bilateral, but with stronger ipsilateral than contralateral projections [Fig. 6A(II and IV)]. As analysis of absolute cell numbers revealed high within-group variability for the same retrograde tracer, and tracers showed different tracing efficiencies, we calculated a lateralization index for each traced region as a measure for contra- versus ipsilateral innervation [Fig. 6B(I and II)]. Compared with intact rats, no statistically significant changes of the projection pattern between the rostral motor cortex and the ipsi- and contralesional gigantocellular reticular nucleus were found as a consequence of spinal cord injury [Fig. 6B(I)]. However, for the MLR, the lateralization index for projections to the contralesional gigantocellular reticular nucleus increased significantly after hemisection injury [P = 0.0302, one-way ANOVA; Fig. 6B(II)]. These results suggest that spinal sprouting of reticulospinal fibres originating from the contralesional gigantocellular reticular nucleus is accompanied by modifications of its supraspinal input after spinal hemisection.
Behavioural changes after micro-lesioning of the gigantocellular reticular nucleus in rats with chronic spinal cord injury
To assess the functional relevance of the observed anatomical changes in the reticular mesencephalic-medullar-spinal system, rats were allowed to recover from unilateral cervical hemisection injury for a period of 16 weeks. Then, rats with spinal cord injury received electrolytic microlesions of either the ipsilesional (n = 5) or contralesional (n = 9) gigantocellular reticular nucleus (Fig. 7A and B). For comparison, intact rats of the same age (n = 7) were also trained on the behavioural tasks and received a microlesion of the left gigantocellular reticular nucleus. Functional testing was performed 1 day before and 2 days after the focal brainstem lesion. Following brainstem lesion, intact rats did not demonstrate any functional impairment (Fig. 7C–Q and Supplementary Table 4).
Rats with chronic hemisection injury and acute damage to the ipsilesional gigantocellular reticular nucleus showed a significantly reduced shoulder height during wading (see results of the repeated measures two-way ANOVA for all assessed parameters in Supplementary Table 4; Fig. 7C) and a significantly reduced toe clearance of both hindpaws during walking (Fig. 7F and G). There was a tendency towards an increased external rotation of the ipsilesional (referring to the side of the hemisection), but not the contralesional hindpaw (Fig. 7H and I). In these rats, a broader step width (Fig. 7J) was mainly due to a laterally displaced ipsilesional hindlimb (Fig. 7L).
Rats with chronic hemisection injury and acute damage to the contralesional gigantocellular reticular nucleus demonstrated similar functional deficits to those rats with acute damage to the ipsilesional nucleus including reduced shoulder height during wading as well as dragging of both hindpaws, increased external rotation of the ipsilesional hindpaw and a broader step width during over-ground walking (Supplementary Table 4 and Fig. 7C–L). However, only rats with chronic spinal cord injury and damage to the contralesional gigantocellular reticular nucleus showed a significantly decreased swimming velocity and a reduced peak stroke velocity of the ipsilesional (again, referring to the side of the hemisection) hindlimb during swimming (Supplementary Table 4 and Fig. 7P and Q). Hip height and basic limb excursions (pro- and retraction) of the ipsilesional fore- and hindlimb during wading were not affected by focal brainstem lesions on either side (Fig. 7M–O). Thus, acute lesion of either the ipsi- or the contralesional gigantocellular reticular nucleus in rats with chronic hemisection injury affected bilateral hindlimb function. Damage to the contralesional gigantocellular reticular nucleus produced specific deficits of the ipsilesional hindlimb during swimming.
In humans, unilateral spinal cord hemisection causes Brown-Séquard syndrome, which is known to be associated with good recovery of locomotor functions, but poor recovery of hand movements (Little and Halar, 1985). The rat spinal cord injury model used in the present study reproduced these behavioural findings and showed anatomical plasticity in specific parts of the CNS following the cervical hemisection. The different descending motor tracts were found to vary greatly in their plastic responses; major changes were observed mainly in the medullary reticular system. Retrograde and anterograde tracing showed that reticulospinal fibres originating from the contralesional gigantocellular reticular nucleus sprouted across the spinal cord midline in response to the injury to innervate the denervated hemicord, in particular in the lumbar spinal cord (Fig. 8). These anatomical changes were accompanied by a high degree of recovery of basic locomotor functions which are believed to be at least partially mediated by the reticulospinal system innervating the spinal central pattern generators (Grillner, 1996; Matsuyama et al., 2004; Hagglund et al., 2010). Consistent with this, sparing of reticulospinal fibres after contusion injury was found to be associated with a better functional outcome (Basso et al., 2002). However, others failed to demonstrate an increase of reticulospinal midline-crossing fibres in rodents after spinal cord injury (Ballermann and Fouad, 2006; Courtine et al., 2008; Weishaupt et al., 2013). We believe that discrepancies between these and our experiments can be mainly attributed to methodological differences (tracing efficacy, type of lesion and analysis).
The rubrospinal system showed a tendency towards an increased number of fibres re-crossing the CNS midline which was, however, not statistically significant. Also, the corticospinal and vestibulospinal system demonstrated no statistically significant increase of midline-crossing fibres after hemisection. However, due to high within-group variability, the data do not allow us to draw the conclusion that these tracts did not respond to the lesion at all. In addition, we did not investigate other forms of anatomical plasticity, such as sprouting onto long propriospinal neurons or commissural interneurons that could also provide cortical reinnervation via detour circuits (Bareyre et al., 2004; Jankowska et al., 2006; Courtine et al., 2008). We observed a fast recovery of the air-righting reaction within a few days after injury. This reaction is triggered by signals from the labyrinth and mediated through the (lateral) vestibulospinal tract (Pellis et al., 1991). As recovery was not associated with significant sprouting of vestibulospinal axons over the midline below the lesion site, optimal use of the spared vestibulospinal-propriospinal connections may allow for rapid recovery of vestibulospinal function in a similar way as the crossed phrenic phenomenon (Goshgarian, 2003). The decrease of serotonergic projections to the denervated hemicord after hemisection is in-line with previous findings in the same spinal cord injury animal model (Filli et al., 2011) and may explain the beneficial effects of treatment with serotonergic agonists after spinal cord injury (Courtine et al., 2009).
After hemisection injury, the number of reticulospinal midline-crossing fibres increased in cervical as well as lumbar spinal segments suggesting growth promoting factors originating from the denervated hemicord and acting on the intact reticulospinal fibres. In development, crossing the midline is regulated by an interplay of attractive and repellent factors (Evans and Bashaw, 2010). Not much is known about such factors in the adult and injured CNS (Omoto et al., 2011), and more detailed molecular studies are needed to reveal the molecular mechanisms underlying axonal midline crossing in the adult CNS. In the lumbar spinal cord, crossed reticulospinal fibres established a similar structural organization to pre-existing reticulospinal midline crossings in intact rats with an analogous degree of arborization and a preference for the ventral grey matter (Zemlan et al., 1984; Martin et al., 1985). Interestingly, in contrast with the lumbar cord there was a tendency towards a reduced arborization of the midline-crossing reticulospinal fibres in the ipsilesional cervical spinal grey matter. This anatomical finding correlates with function: the ipsilesional forelimb remained more impaired and recovered less than the ipsilesional hindlimb after the hemisection injury. The rat model is comparable with the situation in humans, but the mechanisms underlying these segmental differences in plasticity require further investigation.
Crossed reticulospinal neurons innervating new targets in the ipsilesional hemicord may face the problem of an inappropriate or even side-inverted input. After retrograde tracing from the gigantocellular reticular nucleus, we found labelled neurons widely-spread throughout the pontine and medullar reticular formation and, especially, in the contralateral counterpart of the traced gigantocellular reticular nucleus. In addition, the gigantocellular reticular nucleus received bilateral input from higher control centres in particular the MLR, as previously described (Steeves and Jordan, 1984; Garcia-Rill et al., 1986) and the rostral motor cortex (Antal, 1984; Matsuyama et al., 2004). Electrical stimulation experiments provided evidence that the MLR is capable of initiating basic locomotor behaviours in mammals (Shik et al., 1969; Ross and Sinnamon, 1984) via reticulospinal-interneuronal pathways controlling motor neuron pools of proximal and axial muscles (Noga et al., 2003; Matsuyama et al., 2004). Recent data in primates support the theory that the reticulospinal system is also important for the execution of precise finger and hand movements (Davidson and Buford, 2006; Alstermark and Isa, 2012), a task typically requiring the involvement of cortical premotor regions (Kaas et al., 2012). In rats, the rostral motor cortex accommodates primary motor neurons involved in controlling skilled forelimb movements (rostral forelimb area) but is also considered as a premotor region (Passingham et al., 1988; Rouiller et al., 1993). In the present study, hemisection injury was followed by poor recovery of fine motor control, and retrograde tracing revealed that the input from the rostral motor cortex to the gigantocellular reticular nuclei remained unchanged. In contrast, innervation of the contralesional gigantocellular reticular nucleus from the MLR (with a typically stronger ipsilateral input) changed to a more equal bilateral projection pattern, thus extending the influence of the ipsilesional MLR on the contralesional reticular nucleus (Fig. 8). We were not able to characterize these changes in more detail due to the limitations of retrograde tracing. As the measure of lateralization was a ratio and analysis of absolute cell counts was not useful because of high within-group variability, a higher lateralization index may be due to an increase in axons emanating from the ipsilesional MLR or to a reduction of projections from the contralesional MLR. Nevertheless, the balance of the MLRs’ innervation pattern was modified in response to hemisection injury which may reflect the input adjustments needed by newly crossed reticulospinal neurons to convey purposeful information to their novel targets in the ipsilesional hemicord. Whether input modification occurred as a consequence of the midline crossing at spinal levels to provide crossed neurons with appropriate information (‘bottom-up’) or whether, in fact, the changed input from supraspinal control centres enforced the spinal adaptations (‘top-down’) remains to be elucidated.
Electrolytic microlesioning of the ventromedial portion of the gigantocellular reticular nucleus was performed to determine the functional relevance of the anatomical changes in rats recovered from hemisection injury. Intact rats did not show any behavioural deficits after unilateral destruction of the gigantocellular reticular nucleus and this was most likely due to compensation through other, spared descending CNS systems. In contrast, damage to either the ipsi- or contralesional gigantocellular reticular nucleus in behaviourally recovered, walking rats with chronic unilateral hemisection injury abolished the recovered behaviour. Interestingly, a lesion of the ipsilesional gigantocellular reticular nucleus that had lost all of its ipsilesionally descending projections below C4 led to similar devastating consequences during walking and wading as damage to the contralesional nucleus. Thus, the ipsilesional gigantocellular reticular nucleus may exert some influence on basic locomotor functions after hemisection either by way of reciprocal connections with its contralesional counterpart or through detour pathways including relay interneurons located in the cervical spinal cord above the lesion (Reed et al., 2006; Cowley et al., 2008). However, functional assessment during swimming demonstrated behavioural deficits of the ipsilesional hindlimb that were specific to the inactivation of the contralesional gigantocellular reticular nucleus. This confirms that the observed anatomical changes in the contralesional reticulospinal system made a substantial contribution to functional recovery after hemisection injury.
In conclusion, functionally-relevant anatomical plasticity, which can be as drastic as inducing a left-right side switch of axonal projection in the spinal cord and brainstem, can also occur in the oldest part of the CNS with its functionally basic neuronal circuits. A deeper knowledge of such elementary adaptations mediating spontaneous recovery after CNS damage is fundamental for the development of new therapeutic approaches aiming for improved recovery in humans (Lu et al., 2012).
This study was supported by grants from the Swiss National Science Foundation (31-63633.00 and 3100AO-122527/1), the National Center for Competence in Research ‘Neural Plasticity and Repair’ of the Swiss National Science Foundation, the Spinal Consortium of the Christopher and Dana Reeve Foundation and the Framework Program 7 EU Collaborative Project Spinal Cord Repair.
Supplementary material is available at Brain online.
mesencephalic locomotor region