Anatomical Plasticity of Rostrally Terminating Axons as a Possible Bridging Substrate across a Spinal Injury

Abstract

Transfer of information across a spinal lesion is required for many aspects of recovery across diverse motor systems. Our understanding of axonal plasticity and which subpopulations of neurons may contribute to bridging substrates following injury, however, remains relatively incomplete. Most recently, attention has been directed to propriospinal neurons (PSNs), with research suggesting that they are capable of bridging a spinal lesion in rodents. In the current study, subpopulations of both long (C5) and short (T6, T8) PSNs—as well as a supraspinal system, the rubrospinal tract (RST)—were assessed following low thoracic (T9) hemisection in the cat using the retrograde tracer Fluoro-Gold. Acutely, within 2 weeks post-hemisection, the numbers of short and long PSNs, as well as contralateral RST neurons, with axons crossing the lesion were significantly decreased relative to uninjured controls. This decrease persisted bilaterally and was permanent in the long PSNs and the contralateral red nucleus (RN). However, by 16 weeks post-hemisection, the numbers of ipsilesional and contralesional short PSNs bridging the lesion were significantly increased. Further, the number of contralesional contributing short PSNs was significantly greater in injured animals than in uninjured animals. A significant increase over uninjured numbers also was seen in the ipsilateral (non-axotomized) RN. These findings suggest that a novel substrate of undamaged axons, which normally terminates rostral to the lesion, grows past a thoracic lesion after injury. This rostral population represents a major component of the bridging substrate seen and may represent an important anatomical target for evolving rehabilitation approaches as a substrate capable of contributing to functional recovery.

Introduction

Early studies proposed three potential mechanisms for achieving connectivity across a spinal cord injury (SCI). They are regeneration, spared fiber tracts, and collateral sprouting, also known as plasticity.^1–5 Regeneration remains relatively elusive⁶ but a continually expanding body of literature shows that reorganization of circuitry above and below a SCI occurs and may be quite substantial.^7-9 Experimental studies using SCI models provide anatomical evidence that reorganization and plasticity may occur across a large number of diverse systems.

The corticospinal tract (CST) is one of the most well-studied pathways due to its importance for voluntary movements. It is accepted that following a partial SCI, spared corticospinal axons have the capacity to increase the number of collaterals crossing the midline at multiple levels of the spinal cord and extending into denervated areas.^10,11 Sprouting, however, is not unique to the CST and also occurs in other long descending supraspinal pathways associated with motor control, including the reticulospinal pathway^12–15 and the rubrospinal tract (RST).¹⁶ Of particular interest is evidence that the spared rostral components of both the corticospinal¹⁷ and reticulospinal tracts¹² can form novel connections with populations of propriospinal neurons (PSNs) in the cervical spinal cord, creating by-pass circuitry around more caudal lesion sites.

Other studies further illuminate the contributions that PSNs may make by looking at more local circuitry responses. For example, in the adult cat, Fenrich and Rose showed that when cut, the commissural axons of PSNs in the cervical spinal cord have the capacity to grow back across the midline and make functional connections.^18,19 There also is behavioral and anatomical evidence to support propriospinal plasticity within the rodent thoracic spinal cord^20-22 and behavioral evidence in the cat lumbar spinal cord.²³

Propriospinal neurons appear well positioned to serve as a mechanism for intersegmental coupling across short and long distances following SCI.²⁴ They are present in large numbers at all levels of the spinal cord, and have ipsilateral, contralateral, or bilateral projections.^25,26 Although they can have connections within their spinal segment of origin, by definition they must project to and influence other spinal segments.²⁷ Depending upon whether the distance of their axonal projection is more or less than approximately six spinal segments, they are divided into long and short propriospinal populations.²⁸ They are a diverse population functionally as well as morphologically. For example, the long descending propriospinal system connects and coordinates the activity of the cervical and lumbosacral spinal cord levels during reflex and locomotor activities—and in particular is known for its role in forelimb-to-hindlimb coordination.^28–36 In sharp contrast are the short, propriospinal populations housed completely within the thoracic spinal cord, which contribute to the control of trunk musculature important for postural control and respiration^37-39 or those within the lumbar spinal cord associated with the central pattern generator.⁴⁰

Given that most human lesions are anatomically incomplete,^41–45 plasticity of spared substrates is a relevant topic of great translational interest.⁴⁶ In particular, evolving approaches using activity-dependent approaches, epidural stimulation, and transcutaneous stimulation^46–52 are likely to capitalize on spared and plastic pathways. Experimental assessment of changes in spinal circuitry using a more straightforward cut injury model, particularly the lateral hemisection, presents the opportunity to assess spared and axotomized axons within multiple neuronal subpopulations at the same time. Due to the diameter and length of the feline spinal cord, the distance requirements to achieve meaningful connections are likely to be more similar to those required in the human than would be obtained in a rat model. Given these considerations, in the studies presented here, we chose to use the adult cat, low thoracic hemisection model to assess the plasticity of three distinct and diverse pathways using the retrograde tract tracer Fluoro-Gold. Specifically, we evaluated the rubrospinal tract, the long descending propriospinal pathways and short thoracic PSNs to compare plasticity of a descending supraspinal tract, a long intraspinal system, and a short intraspinal pathway. We observed significant changes consistent with plasticity in both the rubrospinal and short propriospinal pathways but not the long propriospinal system, indicating that not all pathways are equally plastic. Comparisons across post-injury time-points and with naïve controls indicated that axons that normally terminate rostral to the lesion site extend collaterals into more caudal spinal segments following lateral hemisection. These axons typically have not been considered in traditional assessment of plasticity.

Methods

Experimental design

All animal procedures were conducted in accordance with the National Institutes of Health guidelines for the care and use of experimental animals and were approved by the University of Florida, the Malcom Randall Veterans Affairs Medical Center (VAMC), and the Robley Rex VAMC Institutional Animal Care and Use Committees. Three groups of purpose-bred, female, adult, specific-pathogen free cats were used in this study: normal controls (n = 9), acute thoracic hemisections (n = 7), and chronic thoracic hemisections (n = 6). Bilateral tracer injections were performed at T11 and animals survived for 13 days following injections. In the acute injury group, tracer injections were made at the time of injury. For the chronic injury group, injections were made 16 weeks post-injury. Cats received daily exercise on weekdays by walking on a treadmill and/or a variety of runways pre-injury and, for the chronic injury group, resumed exercise beginning 1-2 weeks post-injury.

Surgical procedures

All spinal hemisection procedures were performed as previously described.^53,54 In summary, animals received 0.1 cc of both atropine sulfate (0.04-0.06 mg/kg) and acetylpromazine (0.4-0.5 mg/kg) subcutaneously (SQ). Next, they were anesthetized with a combination of isoflurane (2-5%) and oxygen (1-2 LO₂) and then intubated and maintained at a surgical plane of anesthesia (no peripheral reflexes) throughout the procedure. Bilateral laminectomies were performed to expose the spinal cord followed by a left lateral hemisection made at spinal T9 with iridectomy scissors. Any fibers adhering to the ventral or lateral dura were gently lifted with suction and cut. The dura was sutured, then durafilm (Codman-Shurtleff, Inc., Randolf, MA) and gelfoam (Pharmacia and Upjohn, Inc., Peapack, NJ) were placed over the sutures. Muscle and skin were closed in layers. Throughout the duration of the surgical procedure temperature, electrocardiography, respiratory rate, and expired CO₂ were monitored and maintained within normal physiological limits. Cats recovered in temperature-controlled incubators and buprenorphine (0.01 mg/kg, SQ) was delivered after injury (on the surgical table during incision closure), as well as every 6-12 h for the next 48 h. Cats were housed singly or in pairs on thick beds of shredded newspaper or egg-crate foam. Bladders were emptied using the Credé maneuver for several days after surgery and the cats’ health closely monitored throughout the study.

Fluoro-Gold (FG, 0.5% in sterile water, Fluorochrome, LLC, Denver, CO) was used to examine PSN and rubrospinal tract (RST) neurons with terminations caudal to the injury. Fluoro-Gold robustly labels the cell bodies of descending axons and using immunohistochemistry, provides a permanent marker, which facilitates cell counts.^54-56 As above, animals received 0.1 cc of both atropine sulfate (0.04-0.06 mg/kg) and acetylpromazine (0.4-0.5 mg/kg, SQ), were intubated, and maintained at a surgical plane of anesthesia with a mixture of isoflurane and oxygen. All cats received bilateral laminectomies at the injection site. Injections were performed using a 33-gauge Hamilton syringe (Hamilton Company, Reno, NV), held in a stereotaxic device, at spinal T11 (normals) or 13-14 mm (∼T11) below the lesion at the time of injury (acute) or 16 weeks post-injury (chronic). Four needle tracks were made (Fig. 1A, 1B). Within each track 0.25 μl was injected ventrally, the needle partially retracted, and ∼3 min later 0.25 μl injected dorsally. A total of 2 μl (eight injections) were distributed across the spinal cord. In order to determine normal ipsilateral versus contralateral projections of these neuronal populations, four of the nine normal cats received injections of FG on the left side of the spinal cord (four injections, totaling 1 μl). Following these procedures, Buprenorphine was given, bladders emptied and cats' health closely monitored as described above following hemisection.

FIG. 1.

Spinal injection site and retrograde labeling. A three-dimensional rendering of a typical injection site shows the areal extent of tracer spread and four injection locations from a dorsal view (*; A). Most cats showed tracer coverage throughout the cross-sectional extent of the spinal cord with the exception of varying degrees of labeling in the dorsal columns as shown in this example (B; dorsal orientation is up). This composite image (B1) is simplified to a single coronal slice (B2) showing that tracer is present across the entire spinal cord cross-section (black) with the exception of an area in the right dorsal column (white). Propriospinal neurons (C) and red nuclei neurons (D) were counted only if Fluoro-Gold was seen in the somas (scale bars = 100 μm). The inset in C is a higher magnification (scale bar = 20 μm) of a labeled neuron identified by the arrow (white) at lower magnification. CC, central canal.

Immunohistochemistry and neuron counts

All cats were transcardially perfused with saline (0.9%), followed by 4% buffered paraformaldehyde (pH 7.4) at 13 days post-injections. Spinal cords and brains were removed and blocked into segments. Three spinal segments, C5, T6, and T8, along with the red nuclei and lesion site (T9), were cryoprotected in 30% sucrose in 4% paraformaldehyde (pH 7.4) and sectioned on a cryostat (25 μm). Every 40th section (1000 μm) through each spinal segment and every 8th section (200 μm) through the red nuclei was processed for Fluoro-Gold immunoreactivity. The distance between sections assured that neurons would not be double counted (present in two sections) and allowed representative counts of the entire T6 and T8 spinal segments and red nuclei. The greater distance between spinal sections was chosen based upon the relatively long lengths of the spinal segments. Immunohistochemistry was performed using a rabbit anti-FG (1:10,000 for midbrain sections and 1:40,000 for spinal cord sections; Fluorochrome, LLC, Denver, CO) followed by an anti-rabbit goat secondary antibody (1:200, Vector labs) and the ABC Vectastain kit (Vector Labs, Burlingame, CA).

Sections were then incubated in 3,3′-diaminobenzidine (Sigma Aldrich, St Louis, MO) to visualize neurons containing FG. FG positive neurons, defined by punctate labeling, were counted using a Nikon Eclipse E600 microscope keeping left and right side counts separate (Fig. 1C, 1D). Neurons were counted by visual inspection over the entirety of each cross-section assessed. All injection sites and lesions were sectioned (25 μm). Fluoro-Gold spread was assessed by auto-fluorescence in every 10th section (250 μm) to verify that tracer had covered the cross-sectional extent of T11, but had not spread into the lesion site. Cats were excluded from this study if FG spread was identified in the lesion site. Cats also were excluded from specific neuronal subpopulation counts if tracer spread was absent or weak in areas associated with axons of interest—for example, the dorsal lateral funiculus for the rubrospinal tract^57-59 and populations of dorsal PSNs functionally contiguous with the mesencephalic locomotor region and referred to as the stepping strip.^60-62

The first section out of every 10 throughout the lesion segment was mounted onto a subbed slide and processed with cresyl violet (cresyl violet with acetate, Sigma Aldrich, St. Louis, MO) and myelin (Eriochrome Cyanine R, Fluka, NY) stains to view basic lesion morphology and verify each hemisection. The extent of each hemisection was determined by first screening through the stained sections to identify the section best representing the lesion epicenter. This section, together with rostral and caudal sections, spanning 2000 μm across the epicenter were used to generate a pictorial representation of tissue damage on a template of a coronal section of spinal cord for each animal (Fig. 2).

FIG. 2.

Cross-sectional representations of lesions. Tissue sparing and loss can be readily identified using cresyl violet and myelin stained sections. Comparison of histological sections from T9 in a normal control (A) and an animal with a lateral spinal hemisection (B) illustrate the typical asymmetrical loss seen. Using a drawing of the normal cord as a template, the extent of tissue loss and sparing for each acute (C-I) and chronic (J-O) injury was determined for each animal in the study. The representative drawing composite for the histological cross-section shown (B) lies directly below it (E) and shows the modest sparing of the medial part of the ipsilesional dorsal columns. Although there is some variation at the ventral and dorsal midlines across animals, only one lesion (J) showed notable contralesional damage. This animal was included as the damage did not appear to extend into the area of the rubrospinal tract.

Statistical analysis

All statistical analyses were completed using SPSS (IBM, Armonk, NY). Non-parametric Mann-Whitney U tests were performed to compare the number of neurons between time-points. In all cases a two-tailed (bi-directional) hypothesis was used. Non-parametric Spearman correlational analyses were performed to determine any potential relationships between tracer spread within the dorsal columns and number of labeled neurons. Significance level was set at p ≤ 0.05 for all statistical tests.

Results

Qualitative assessment of tracer spread

Fluoro-Gold spread was assessed in all of the injection sites (Fig. 1A, 1B). Unilateral and bilateral injections were made at T11. In most cats, spread was seen throughout the cross-sectional extent of the spinal cord with the exception of some dorsal column areas (Fig. 1A, 1B), which had varying amounts of tracer. We did not perceive this to be a major issue because the cell bodies of the current neurons of interest lie in gray matter areas of the midbrain and spinal cord and their descending axons run primarily through the lateral and ventral columns in the cat.^35,63–69 Further, the axons of descending dorsal PSNs located in the dorsolateral funiculus also should not be affected.^61,62 The subset of dorsal propriospinal axons located in the dorsal columns at the sacral levels of the rat spinal cord, if present in the cat, would not be affected by our lesions (Fig. 2) or affect the intended cell counts.^26,70 To determine, however, if there was a correlation between tracer spread into the dorsal columns and the number of labeled neurons, animals were ranked according to the extent of dorsal column spread and a Spearman correlation analysis performed. There were no significant correlations with the exception of one inverse correlation suggesting that the number of contralateral RN (axotomized) neurons increases with less spread into the dorsal columns [rs(5) = -0.845, p = 0.035]. This finding along with the remaining non-significant correlations are in Table 1. It also was confirmed that cats with unilateral injections (Fig. 3) did not have tracer spread across the midline.

Table 1.

Neuron Labeling and Tracer Spread

Group	Neural level	Side	Spearman correlation
Normal controls	RN	right	rs(4) = 0.334, p = 0.497
		left	rs(4) = 0.213, p = 0.658
	C5	right	rs(3) = 0.112, p = 0.683
		left	rs(3) = 0.112, p = 0.683
	T6	right	rs(3) = 0.112, p = 0.783
		left	rs(3) = 0.0, p = 1.0
	T8	right	rs(3) = -0.112, p = 0.783
		left	rs(3) = -0.112, p = 0.783
Acute hemisection	RN	Contralesional	rs(5) = -0.845, p = 0.035
		Ipsilesional	rs(5) = -0.304, p = 0.564
	C5	Contralesional	rs(5) = 0.12, p = 0.781
		Ipsilesional	rs(5) = -0.256, p = 0.545
	T6	Contralesional	rs(5) = -0.547, p = 0.181
		Ipsilesional	rs(5) = -0.571, p = 0.150
	T8	Contralesional	rs(5) = -0.447, p = 0.255
		Ipsilesional	rs(5) = -0.473, p = 0.255
Chronic hemisection	RN	Contralesional	rs(3) = -0.676, p = 0.233
		Ipsilesional	rs(3) = 0.051, p = 0.95
	C5	Contralesional	rs(4) = 0.112, p = 0.683
		Ipsilesional	rs(4) = 0.112, p = 0.683
	T6	Contralesional	rs(4) = 0.112, p = 0.783
		Ipsilesional	rs(4) = 0.0, p = 1.0
	T8	Contralesional	rs(4) = -0.112, p = 0.783
		Ipsilesional	rs(4) = -0.112, p = 0.783

Spearman correlational analyses indicate that labeling of the neuronal populations of interest is not related to the spread of tracer into the dorsal columns with one exception. There is a negative correlation at the acute time point post-hemisection for the contralateral RN, which is axotomized by the injury.

RN, red nucleus.

FIG. 3.

Ipsilateral and contralateral projections in the normal spinal cord. Unilateral injections (n = 4) labeled more long propriospinal neurons on the contralateral side than ipsilateral, whereas bilateral injections (n = 5) yielded similar numbers on both sides (A). Short propriospinal neurons at both T6 (B) and T8 (C) have similar numbers of neurons that project contralaterally and ipsilaterally. In both cases, bilateral injections labeled more neurons. Following a unilateral injection, labeled neurons were seen almost exclusively in the contralateral red nucleus with <5 neurons ipsilaterally. Bilateral injections labeled neurons in both red nuclei and numbers in each nuclei were similar to that seen in the contralateral nucleus after a unilateral injection (D). Similarity of groups determined by Wilcoxon Signed Test (all p values >0.68). Note: To allow comparison, the use of ipsilateral and contralateral labels for the bilateral injection group are matched to the corresponding unilateral injection side.

All cats included in the study showed similar tracer spread rostral and caudal to the injection sites (Fig. 1A). In three cats, (two chronic animals and one normal), the tracer did not cover the ventral funiculi, which may have affected labeling of the long PSNs. Thus, the long PSN, C5 neuron counts were not conducted in these three animals.

Propriospinal and rubrospinal tract projections in the normal spinal cord

Unilateral injections (n = 4) were performed to determine the populations of ipsilaterally and contralaterally projecting neurons from the red nuclei, C5, T6, and T8. Bilateral injections (n = 5) were performed to identify if a neuron could have both ipsilateral and contralateral projections, as well as establish a baseline of neuronal counts for comparison with injured animals. Similar to early reports,^33,71 long PSNs at C5 have a greater number of contralaterally projecting neurons (Fig. 3A). Short PSNs appear to have more equal numbers of ipsilaterally and contralaterally projecting neurons at both T6 (Fig. 3B) and T8 (Fig. 3C). However, it is likely that all three populations (PSNs at C5, T6, and T8) have bilaterally projecting neurons as bilateral injections yielded counts less than double the unilateral injection numbers. Specifically, this reduction in number is because any neuron with projections to both the right and left sides of the spinal cord will be seen as a single labeled cell body following both unilateral and bilateral tracer injections. As expected, unilateral injections yielded neurons in the contralateral RN (Fig. 3D) because the RST is primarily a crossing tract.^72-74 Following bilateral injections, there were similar numbers of neurons in both red nuclei. These counts also were similar to the number in the unilaterally injected contralateral RN. Due to the similarity of right and left neuronal counts in the bilaterally injected group, as determined by the Wilcoxon signed rank test (p values >0.68), these counts were combined into a single normal group for each neuronal subpopulation (C5, T6, T8, and RN) for comparison with post-injury counts.

Quantification of PSNs

The contributions of PSNs with projections descending caudal to the hemisection were assessed. The ipsilesional and contralesional neuronal counts in injured animals were kept separate to determine the specific population contributing to any observed plasticity. When comparing normal and injured animals, only the bilaterally injected animals were used because all injured animals received bilateral injections. In comparison to uninjured animals (n = 10 sides from five animals), the number of long PSNs at C5 (n = 7; Fig. 4D) with axons below the level of the lesion was significantly decreased on both the ipsilesional (U = 9.0, p = 0.011) and contralesional (U = 2.0, p = 0.011) sides acutely following injury. This significant decrease persisted bilaterally at the chronic time-point when compared with normal numbers (U = 4.5, p = 0.028 on ipsilesional; U = 5.5, p = 0.04 on contralesional). No significant changes were observed between the acute and chronic time-points (ipsilesional side, U = 12.0, p = 0.703; contralesional side, U = 12.0, p = 0.704) suggesting that the population of long PSNs at C5, with axons below the lesion, remained significantly decreased and unaltered for at least 16 weeks post-injury.

FIG. 4.

Changes in the long and short propriospinal neurons (PSN) numbers following injury. Neurons were counted at C5 (long PSNs), T6 and T8 (short PSNs). Tracing of a T6 section is shown in (A) to indicate that counts were made in the grey matter. Neurons typically were present in the intermediate and ventral lamina (scale bar = 500 μM). High magnification (20 × ) photomicrographs in the normal (control, B) and injured (C) T6 spinal cords show examples of Fluoro-Gold–labeled PSNs that were counted (scale bar = 100 μM). The median number of retrogradely labeled ipsilateral and contralateral long propriospinal neurons at C5 (D) shows that following injury, the number of the long propriospinal neurons significantly decreases (*) bilaterally at both acute (n = 7) and chronic (n = 4) time-points compared with normal controls. As illustrated in (E) the number of T6 PSNs decreased significantly on the ipsilateral side of the spinal cord (*) acutely compared with normal controls, but significant increases (‡) were seen bilaterally at the chronic, compared with the acute, time-point. T8 PSNs (F) also showed a significant ipsilateral decrease (*) acutely with significant bilateral increases chronically (‡) from the acute time-point. The number of T8 neurons on the contralateral side at the chronic time-point were significantly greater than in normal controls (†). See the Results section for specific p values.

Two populations of short PSNs were assessed; three segments above the lesion at T6 (Fig. 4A-C, 4E) and one segment above the lesion at T8 (Fig. 4F). In contrast to the long PSNs, both populations of short PSNs exhibited marked post-injury changes in neuronal contribution over time below the lesion. Both populations of short PSNs showed significant decreases in labeled neurons at the acute time-point on the ipsilesional side (T6: U = 4.0, p = 0.002; T8: U = 0.0, p = 0.001) in comparison to normal controls (Fig. 4E, 4F). This was followed by a significant increase in the number of ipsilesional short PSNs with axons below the lesion site in the chronic time-point group relative to acute time-point animals (T6: U = 4.0, p = 0.015; T8: U = 2.5, p = 0.008) such that the chronic numbers were not different from those in normal controls (T6: U = 17.5, p = 0.174; T8: U = 17.5, p = 0.175; Fig. 4E, 4F). A significant decrease was not noted in the contralesional short T6 PSNs acutely, although there was a strong trend towards significance at T8 (T6: U = 23.0, p = 0.241; T8: U = 15.0, p = 0.051; Fig. 4E, 4F). Similar to the ipsilesional side, contralesional neuron counts in the chronically injured group were significantly greater than in the acute group (T6: U = 6.0, p = 0.032; T8: U = 0.0, p = 0.003), suggesting significant plasticity over time post-SCI (Fig. 4E, 4F).

As with the ipsilesional side, chronic counts at T6 were no longer significantly different from the normal controls (U = 22.0; p = 0.385; Fig. 4E). Notably, the number of T8 PSNs with axons below the lesion showed a strong trend towards being greater than in normal controls (U = 12.0, p = 0.051; Fig. 4F). If two control outliers identified by quartile and Grubb's test were removed, this trend became significant (T8: U = 0.0, p = 0.002; outliers shown in Fig. 3C) suggesting contralesional, chronic T8 neuron numbers were significantly greater than the same uninjured population. A comparison of chronic ipsilesional and contralesional neuronal counts indicated that these populations were significantly different from one another (U = 2.3, p = 0.013) which corroborated an increase in the number of T8 contralesional neurons, relative to those found in normal controls, with axons below T10 (Fig. 4F). Comparison of ipsilesional and contralesional chronic neuron counts showed a trend but did not find a significant difference at T6 (U = 6.0, p = 0.053; Fig. 4E).

Quantification of rubrospinal tract neurons

As expected, there was a significant decrease in the number of labeled neurons in the contralateral (axotomized) RN after injury compared with normal animals (Fig. 5B-D; acutely: U = 0.0, p = 0.004; chronically: U = 0.0, p = 0.006). Due to the crossing nature of this tract, the presence alone of neurons in the axotomized RN (Fig. 5C) suggests an increase in their post-injury contributions at the level of the injection site. Comparisons of labeled neurons in the contralateral (axotomized) nuclei shows a significant increase in the number of neurons with axons below the lesion level at the chronic, relative to acute, time-point (U = 4.0, p = 0.044; Fig. 5D). Interestingly, neuron numbers in the ipsilateral (non-axotomized) RN were significantly greater than normal control numbers at both the acute (U = 0.5, p = 0.005) and chronic time-points (U = 0.0, p = 0.006), nearly doubling in number by the chronic time-point (Fig. 5D; note axis scale). Further, these numbers in the acute and chronic injured groups were not significantly different from one another (U = 10.0, p = 0.361; Fig. 5D). Thus, both axotomized and non-axotomized RST neurons undergo plastic changes in neuronal contributions below the lesion following a spinal injury.

FIG. 5.

Changes in rubrospinal tract (RST) neurons following injury. A line drawing of the midbrain (A) depicts the areas (red nuclei) where Fluoro-Gold–labeled RST neurons were counted (hatched area, scale bar = 500μM). Photomicrographs show examples of RST neurons counted in the normal control (B) and injured (C) animals. As illustrated in (D), the ipsilateral (non-axotomized) red nucleus had significantly more (†) neurons at both the acute and chronic time-points compared with normal controls. The contralateral (axotomized) red nucleus had significantly fewer (*) neurons at both time-points compared with normal controls, but there was a significant increase from the acute to chronic time-point (‡). See the Results section for specific p values.

Discussion

By sparing and axotomizing every pathway, the lateral hemisection allows for evaluation of potential growth from intact and damaged axons within rostral populations to areas caudal to the lesion. Consequently, it provides a unique platform for assessing regeneration and plasticity. In this study, we assessed contributions of three neuronal populations (long and short PSNs and RST neurons) to axonal projections caudal to a thoracic hemisection. Our results demonstrate that, although the numbers of long PSNs with axons below the lesion are significantly and permanently decreased, there are significant increases in the numbers of short PSNs and RST neurons. This suggests that these axons possess the ability to develop numerous new connections below the lesion level. Particularly striking are the contributions of non-axotomized T8 short PSNs and ipsilateral RST neurons.

The implications of these findings are noteworthy for two reasons. First, these results demonstrate that following SCI, uninjured neurons that normally terminate above the lesion can sprout and extend to more caudal segments below the lesion (Fig. 6E). Although the concept of axonal sprouting from uninjured neurons is not new, it has been attributed largely to axons of passage that already project below the lesion (Fig. 6C).^15,75 The increased number of rostral PSNs labeled after injury indicates that axons that are normally not present to take up tracer at T11 (Fig. 6A, 6B), either sprouted or regenerated caudally to T11 following hemisection (Fig. 6D, 6E). Second, this finding also is important in terms of scientific interpretation beyond this work. The RN is considered a model motor system for assessing axonal growth following unilateral injuries. As the RST is primarily contralateral, many tract tracing studies compare neuronal counts from the contralateral RN (primarily axotomized neurons) with the ipsilateral RN (primarily non-axotomized neurons) under the assumption that the number of labeled neurons in the ipsilateral RN is equivalent to the number in the normal animal. However, our results clearly show that this may not be the case and, in fact, these numbers may be higher following lateral hemisection, which could skew interpretation.

FIG. 6.

Retrograde labeling and mechanisms of growth. In the intact spinal cord, neuronal cell bodies with axons extending into the area injected with Fluoro-Gold will be labeled (A). The immediate effect of a lateral hemisection (Hx) is to cut all axons on one side at the lesion level and spare those on the opposite side (B). The neurons with spared axons extending into the area of Fluoro-Gold, caudal to the Hx level, will be labeled. There are several potential mechanisms of growth that may occur after Hx to re-innervate areas below the lesion level. The first is that spared axons of passage, left intact contralateral to the Hx, may develop axonal collaterals (C). This growth is not captured in the neuronal counts as the tracer also is transported by the originally spared axon. Increases in the number of contributing neurons compared with normal controls seen in this study can occur by a combination of at least two potential mechanisms. The first is regeneration of neurons that were axotomized (D). The second mechanism is growth of a non-axotomized subset of neurons with axons that normally terminate rostral to the lesion site (E). The cell bodies of these neurons were not labeled in the normal (intact spinal cord) control group because their axonal termination sites are normally rostral to the injection site. Following injury however, they extended axons into more caudal spinal segments, thus bridging the lesion site and placing them at the level of the injection site.

Changes in propriospinal pathways following hemisection

Following disruption of descending motor tracts after injury, several groups have demonstrated that local spinal circuitry, in addition to spared descending pathways, can facilitate functional recovery.^76–79 Short PSNs connect between one and six segments^35,80 and many descending motor tracts terminate on PSNs making them primed to assist with motor recovery following injury.⁸¹ Courtine and colleagues²⁰ found that following a thoracic hemisection, the number of short PSNs in the thoracic rat cord with axons below the lesion acutely decreased, but then increased over time.²⁰ This is consistent with the current study, which demonstrates an even greater potential for circuit amplification at chronic time-points. Studies that have focused on the neural substrates underlying functional recovery recognize two mechanisms through which plasticity occurs—collateral sprouting from local neurons^3,82,83 and sprouting from spared axons of passage (Fig. 6C).^20,84 However, the current study shows that another mechanism for plasticity exists—the ability of uninjured neurons, which normally terminate above the level of the lesion, to develop collaterals capable of bridging an injury site and potentially innervate segments caudal to the lesion (Fig. 6E). Their contributions are clear in this study and it is likely that they also are represented in neuronal counts from previous studies.

In contrast to the short PSNs, the number of FG-labeled long PSNs at C5 significantly decreased after injury and remained so chronically. The C5 segment was targeted as it is a main contributor to the long, descending PSN pathway in the cat.^35,85 Our results do not discount that spared long PSNs could contribute to plasticity following SCI. Axons of the long PSNs travel bilaterally through the length of the spinal cord in the ventral and lateral funiculi. These axons terminate throughout the lumbar enlargement, although there are more terminations in the rostral lumbar cord (L1-3), making them well-positioned to promote functional recovery of locomotion^35,64,85,86 Spared long PSN pathways are reported to serve as a relay between descending supraspinal pathways and lumbosacral circuitry and/or expand their terminal arbors in rats after cervical hemisection.^12,22,84 The retrograde tracer approach used in the current study cannot address these aspects of plasticity.

Supraspinal reorganization

Similar to T8 short PSNs, the increase in ipsilateral RST neurons compared with uninjured animals suggests recruitment of neurons that normally terminate more rostrally. This contrasts with an earlier study in rats, which found that the increase in the number of ipsilateral RST neurons after cervical hemisection was not statistically significant.¹⁴ There are several differences in study design between the current study and the previous study. Zorner and colleagues used cervical injuries and unilateral tracer injections, and it is unclear how far the tracer spread relative to midline. The bilateral tracer approach applied in this study allows us to capture RST fibers that may have crossed and terminated in gray matter close to midline. Zorner and colleagues allowed animals to spontaneously recover after injury whereas in this study, some of the animals were exercised daily on treadmills and runways.

Although a training effect was not the focus of this study, the daily exercise may have influenced the degree of plasticity demonstrated in the chronic group of animals.^52,87-89 Training has been linked with axonal growth,^90–94 as well as alterations in neurotrophin levels.^95–100 Changes associated with training should not have been an issue, however, for the acute survival group, as they were not exercised after injury. Despite this, significant increases were seen in contributions from non-axotomized red nucleus neurons at this acute (2-week) survival time-point. As has been pointed out by other groups, animals may self-exercise in their cages.^101,102 This is a viable possibility as animals were socially housed in large cages (30″ × 60″) with sufficient room to move around and step.¹⁰³ Lastly, there is a species difference between the two studies. Specifically, rats have a higher percentage of uncrossed RST fibers (10-28%) compared with the cat.^{65,66,104,105} Thus, the lack of a significant difference in the Zorner and colleagues' study may be due to a greater loss of these uncrossed RST neurons as a result of their higher percentage in the rat.

An interesting similarity between the current study and the study by Zorner and colleagues¹⁴ is the apparent proximity of the sprouting from spared rubrospinal axons. Specifically, there was an increased number of tracer-labeled RST neurons in the ipsilateral RN when tracer injections were made closer to the lesion site in the cervical cord, as opposed to the lumbar cord.¹⁴ This is in agreement with the current study in which tracer injections made within two segments of the lesion resulted in a significantly greater number of ipsilateral RST neurons. This suggests that rostral neurons, which bridged the lesion, may terminate within close proximity to the lesion.

Finally, it should be noted that there was a small but significant increase in contralateral FG-labeled neurons in the RN at the chronic time-point. This suggests that sprouting also occurs in the axotomized RST and may facilitate recovery. This is consistent with other studies that demonstrated the role of injured descending axons in plasticity following SCI.^12,22,84

Implications of remodeling on functional recovery

This study sheds light on potential underlying plasticity and substrates that may contribute to the substantial functional recovery reported by our group^53-55,106 and others following thoracic hemisection in the adult cat.^89,107–111 Recovery of locomotion progresses with increasing voluntary control over time, suggesting that recovery is not solely controlled by neurons spared by the hemisection, but by reorganization of supraspinal and intraspinal systems. The integration of both intraspinal and supraspinal systems is required for the recovery of postural control, which is severely impacted following SCI.^112-114 Recovery of postural control during basic locomotor tasks occurs within the first 2 weeks after thoracic hemisection,^{53,106,107,115} which coincides with the current findings showing a significant decrease in ipsilesional, but not contralesional, PSN populations, and an increase in non-axotomized RST neurons. More challenging locomotor tasks, which require greater balance and more precise limb movements, take longer to recover.^106,107This longer recovery correlates with the recruitment and probable activation of additional rostral short PSNs and RST neurons terminating in the caudal cord, as well as recovery of axotomized neurons at the chronic time-point. Although aberrant sprouting¹¹⁶ cannot be ruled out as contributor to the neuronal counts, the increased counts are temporally correlated with functional recovery reported by a number of laboratories, pointing to positive roles for the reported plasticity.

Conclusions

Results of this study support the idea that the substantial recovery intrinsic to the caudal cord below an injury^{23, 89, 117} is likely assisted by the plasticity of thoracic PSNs and supraspinal systems. Not only may this plasticity be facilitated by axotomized neurons, but significant plastic changes are occurring in non-axotomized neurons that normally terminate above the lesion. Future studies are needed to identify the degree to which this category of non-axotomized neurons are available across diverse pathways and how injury level and magnitude impacts availability, recruitment and their contributions to features of functional recovery following SCI. Understanding neuronal populations available to contribute to repair or generation of novel circuitry is critical in designing recovery-based approaches following spinal injury.

Funding Information

Rebecca F Hammond Endowment, Kentucky Spinal Cord and Head Injury Trust. DVA RR&D B7165R, B9249S, NS050699.

Author Disclosure Statement

No competing financial interests exist.