The pattern of remyelination after traumatic spinal cord injury remains elusive, with animal and human studies reporting partial to complete demyelination followed by incomplete remyelination. In the present study, we found that spared rubrospinal tract (RST) axons of passage traced with actively transported dextrans and examined caudally to the lesion 12 weeks after mouse spinal cord contusion injury were fully remyelinated. Spared axons exhibited a marginally reduced myelin thickness and significantly shorter internodes. CASPR (contactin-associated protein) and Kv1.2 channels were used to identify internodes and paranodal protein distribution properties were used as an index of myelin integrity. This is the first time the CNS myelin internode length was measured in a mouse. To better understand the significance of shortened internodes and thinner myelin in spared axons, we modeled conduction properties using McIntyre's et al. model of myelinated axons. Mathematical modeling predicted a 21% decrease in the conduction velocity of remyelinated RST axons attributable to shortened internodes. To determine whether demyelination could be present on axons exhibiting a pathological transport system, we used the retroviral reporter system. Virally delivered green fluorescent protein unveiled a small population of dystrophic RST axons that persist chronically with evident demyelination or abnormal remyelination. Collectively, these data show that lasting demyelination in spared axons is rare and that remyelination of axons of passage occurs in the chronically injured mouse spinal cord.
Experimental strategies such as cell transplantation designed to preserve myelin or to induce remyelination have lead to successful improvements in function after spinal cord injury (SCI) if administered in acute or subacute phases after SCI (Li et al., 1997; Akiyama et al., 2002; Cao et al., 2005; Cummings et al., 2005; Keirstead et al., 2005; Lepore and Fischer, 2005). However, remyelination strategies fails if transplants are delivered chronically after SCI (Keirstead et al., 2005; Karimi-Abdolrezaee et al., 2006). The primary data regarding myelin status in chronic injury is incomplete and a poor correlation between rodent and human studies exists. The majority of human cases do not show progressive chronic demyelination, whereas only a few cases report persistent demyelination (Kakulas, 1999; Norenberg et al., 2004; Guest et al., 2005). Because previous studies did not correlate fiber integrity with myelin condition, the key question remains unanswered: what is the chronic myelin status of spared or intact white matter? Because myelin sheaths not only ensure successful propagation of action potentials, but also participate in axonal maturation, transport and survival (Brady et al., 1999; Edgar et al., 2004) it is important to understand the pattern of remyelination within individual axons to develop the most appropriate myelin repair strategies.
Rodent models of SCI reveal electrophysiological and morphological signs of demyelinated axons (Gledhill et al., 1973; Blakemore, 1974; Gledhill and McDonald, 1977; Blight, 1985; Gensert and Goldman, 1997; Cao et al., 2005). Initial demyelination is thought to be caused by necrosis of oligodendrocytes; however, it is not exactly clear what causes prolonged oligodendrocyte death weeks and months after injury (Blight, 1985; Crowe et al., 1997). In accordance with morphological studies of myelin disintegration (Balentine, 1978; Griffiths and McCulloch, 1983), proteolipid protein (PLP) and myelin basic protein (MBP) expression is significantly reduced after injury (Wrathall et al., 1998). In addition, immunohistochemistry studies demonstrate that sodium and potassium channels as well as contactin-associated protein (CASPR) distribution lose their compact and restricted staining pattern in demyelinated fibers (Karimi-Abdolrezaee et al., 2004). Myelin damage results in decreased action potential conduction velocity or a conduction block and transcranial magnetic motor-evoked potentials show longer latency responses (Fehlings and Nashmi, 1995). Weeks after SCI, demyelination is evident mostly in the medial area rostral and caudal to the lesion site; however, thin profiles of new myelin are observed and their number increases with time (Griffiths and McCulloch, 1983). Electron microscopy of remyelinated nodes shows normal morphology, with myelin lamellae terminating in cytoplasm-filled loops on each side of the node (Gledhill et al., 1973). Collectively, these studies demonstrate dynamic changes in myelin status including significant demyelination but also myelin regeneration.
In the present study, we used multiple methods to discriminate spared axons (that pass through the lesion) versus damaged axons simultaneously evaluating myelin health within individual nerve segments in the chronically injured mouse spinal cord. We selectively studied the RST because it is a heavily myelinated tract and important for rodent locomotion (Waldron and Gwyn, 1969; Holstege, 1987). We quantified the amount of demyelination, myelin thickness, and internodal lengths in spared versus severed axons. To our knowledge, this is the first study to combine functional axonal labeling and the simultaneous examination of myelin indices. The results establish a tight correlation between myelin status and axon integrity.
Materials and Methods
Female C57BL/6 mice (8 weeks of age) were used in this study. All surgical procedures and experimental manipulations were in accordance with a protocol approved by the Animal Care and Use Committee of the University of Washington.
Mice, anesthetized with intraperitoneal injections (0.5 ml/20 g) of avertin (tribromoethnol in tert-amyl alcohol), underwent midthoracic (T9) laminectomy after which an impact probe was lowered onto the dura and the cord was rapidly displaced 0.5 mm (moderate SCI) using the Ohio State University contusion device. After contusion injury, muscle and skin were closed in layers, and then animals were kept on a heating plate (38°C) until fully awake. Subcutaneous saline, gentomicin (APP, Schaumburg, IL) and intramuscular buprenorphine (Reckitt Benckiser Healthcare, Hull, UK) were administered for 7 and 2 d, respectively. Bladder expression was performed two times daily until voluntary control returned.
Tracing of the rubrospinal tracts.
To examine myelin characteristics around rubrospinal fibers, the Rubrospinal tracts were anterogradely traced immediately after SCI or 8 weeks after contusion injury or in uninjured controls. Animals received bilateral pressure injections totaling 2 μl per animal of either a 10% solution of biotinylated dextran amine [BDA 10,000 molecular weight (MW); Invitrogen, Carlsbad, CA] or rhodamine dextran amine (Fluoro-Ruby 10,000; Invitrogen) into the red nuclei (coordinates: 3.4–3.6 mm posterior to bregma, 0.5–1 mm lateral to midline, 3.6 mm depth) delivered by Hamilton 5 μl syringe fitted with a 33 gauge Hamilton needle. The needle was left in place for 2 min to ensure proper dye distribution. Sixteen to 18 d after tracer injections, animals were perfused with appropriate fixatives.
We produced vector stocks by transient transfection of 293T cells grown in DMEM (Invitrogen), 10% FBS (HyClone, Logan, UT), 2 mm glutamine (Invitrogen), and 1% of penicillin-streptomycin (Invitrogen) with the pCMV-eGFP expression vector, envelope plasmid pMDG, packaging plasmid pMDLg/p RRE, and transfer vector plasmid pRSV.Rev (a gift from F. Galimi, University of Sassari, Sassari, Italy) by calcium phosphate DNA precipitation as described previously (Dull et al., 1998). Supernatants were collected 48 and 72 h after transfection, filtered, and concentrated by two successive ultracentrifugations. p24 antigen was assayed by ELISA (PerkinElmer, Wellesley, MA).
Viral transfection of the red nuclei.
To label RST axons we used a retroviral reporter system that does not rely on an active axonal transport, but rather on passive diffusion. pCMV-GFP lentivirus (titer of 109-1010, total volume 5 μl per animal) was pressure injected bilaterally into the red nuclei as described above. Animals were perfused 30 d after viral injections with 4% paraformaldehyde.
After an overdose with beupanasia-D (Schering-Plough, Union, NJ), animals were killed by intracardiac perfusion with 0.5% glutaraldehyde (Ted Pella, Redding, CA) and 3% paraformaldehyde (Ted Pella) in 0.1 m phosphate buffer, pH 7.4. Spinal cords were removed and sectioned at 60 μm on a Vibratome. The BDA was detected by the following protocol: washed three times for 10 min in 0.1 m PB, then washed three times for 10 min a 0.1 m Tris buffer, 3 h incubation in an avidin-peroxidase complex (ABC elite; Vector Laboratories, Burlingame, CA) according to the manufacturers suggestion, washed three times for 10 min in Tris-based buffer (TB), first preincubation for 10 min in 0.4% ammonium nickel sulfate, second preincubation for 10 min in 0.4% ammonium nickel sulfate and 0.7 mg/ml diaminobenzidine, and reaction in 0.4% ammonium nickel sulfate and 0.7 mg/ml diaminobenzidine with 0.004% hydrogen peroxide in TB. The reaction was stopped by washing extensively in TB. The sections were then washed three times for 10 min in 0.1 m PB buffer. After osmication in 1% OsO4 for 1h, sections were washed again in 0.1 m PB buffer, dehydrated through an ascending alcohol series ending in 2× 10 min in propylene oxide, and subsequently embedded using the Eponate 12 kit (Ted Pella). Then, semithin sections were cut at 1 μm, mounted on glass slides and stained with toluidine blue or Richardon's stain. Ultrathin sections were cut at 60 nm on a Reichert Ultracut S microtome by Leica (Nussloch, Germany), mounted on nickel grids, uranyl acetate and lead citrate stained, and viewed under a Philips (Aachen, Germany) TEM/CM 10 electron microscope.
Animals were killed with an overdose with beupanasia-D (Schering-Plough) followed by intracardiac perfusion with 4% paraformaldehyde in 0.1 m phosphate buffer, pH 7.4. Spinal cords were removed, rubrospinal tracts were dissected rostral and caudal to the lesion and incubated overnight with primary monoclonal or polyclonal CASPR (1:100, a generous gift from J. S. Trimmer, University of California, Davis, CA), and monoclonal Kv1.2 (1:200; NeuroMab, Davis, CA) antibodies. Then, tissue was incubated with secondary antibodies against mouse and rabbit Alexa 488 (1:250, Invitrogen) and Cy5 (1:250; Jackson ImmunoResearch, West Grove, PA). Some caudal parts of the rubrospinal tracts were transferred to slides with a small pool of 0.1 m PB where single positively labeled rubrospinal fibers were mechanically teased for single fiber analysis. Single fibers were stained by the same method as whole rubrospinal tracts. Some of the lesion sites were cryoprotected in 30% sucrose in 1× PBS overnight and flash frozen in OCT, and then 16-μm-thick horizontal sections were cut on a cryostat. The tissue was incubated overnight with primary guinea pig GFAP antibody (1:500). Stained rubrospinal tracts or single fibers were viewed under Nikon (Tokyo, Japan) Eclipse TE200 confocal microscope fitted with Bio-Rad (Hercules, CA) Laser Sharp 2000. Later, fibers were reconstructed with Volocity software (Improvision Lexington, MA).
Spinal cords were removed from killed animals that had undergone moderate contusion spinal cord injury (as described above) 24 h (n = 2), 5 d (n = 3), or 2 weeks (n = 3) earlier, and RNA was extracted with the Invitrogen PureLink Micro-to-Midi RNA kit. RNA was transcribed to DNA with Qiagen (Hilden, Germany) reverse transcription kit. DNA samples were analyzed in triplicate by using the ABI PRISM 7900 sequence detection system (Applied Biosystems, Foster City, CA). The sequence of MBP sense primer was 5′-ATGGCATCACAGAAGAGACC-3′ and antisense primer was 5′-CATGGGAGATCCAGAGCGGC-3′, for β-actin, the sense primer was 5′-CCACTGCCGCATCCTCTTCC-3′ and antisense primer was 5′-CTCGTTGCCAATAGTGATGACCTG-3′. Each 50 μl quantitative PCR (qPCR) consisted of 1 μl of sample DNA, 1.5 μl of 10 μm sense and antisense primers, and SYBR green PCR mix (Qiagen). Reactions were performed in optical 384-well reaction plates (ISC BioExpress, Kaysville, UT). qPCR assays were run according to the manufacturer's directions, and results were analyzed with Sequence Detection System software (Applied Biosystems). Injured spinal cord tissue samples were compared with control samples taken from the same level of the spinal cord (i.e., thoracic level 8–10). β-actin served as an internal control for total RNA levels. Fold changes were calculated as follows: ΔCt = average Ct of target gene (MBP) − average Ct of endogenous control gene (β-Actin), then we calculated ΔΔCt = ΔCt treated − ΔCt untreated, and finally the fold change = 2̂-[average ΔΔCt].
G-ratios (axon diameter/total fiber diameter) were determined for all BDA-labeled axons. We measured G-ratios for all BDA-labeled axons in randomly chosen sections within a 3 mm zone starting 200 μm caudal to the lesion. To remove bias we analyzed all positively labeled axons in a section (n = 80 axons from control mice; n = 92 axons from injured). Control measurements were performed at the T10–T11 levels, which correspond to the same location in the injured animals. Photos were taken of positively BDA-labeled axons at the following magnifications: 2950–3900× for overview, and 8900–52000× for closer examinations. Later, photographs were digitized by scanning with Epson (Long Beach, CA) Perfection 4870 Photo. Measurements of myelin sheath thickness and axon diameter were made by using NIH Image-J. A calculation of tissue shrinkage/correction factor was not made and we assumed that tissue processing effected control and injured tissue similarly. Internode lengths were measured by tracing axons from one paranode identified by CASPR staining to another using Image-J. Multiple internodes were analyzed from a total of 99 and 100 BDA-labeled axons from control and injured tissue, respectively. To ensure that CASPR and Kv1.2 labels were colocalized on Fluoro-Ruby-positive axolemma, we used Volocity software to view three-dimensional reconstructions of the tissue from confocal microscope z sections.
McIntyre et al.'s (2001) model of a mammalian myelinated axon, adapted from the database archive of the NEURON simulation environment (http://senselab.med.yale.edu/senselab/modeldb/), was implemented to estimate the effects of the observed changes in internode length, axon diameter, and myelin thickness on axonal conductance. Briefly, the axon is modeled as a multicompartmental double cable, with separate representations of the axolemma and the myelin sheath. Nodes of Ranvier, paranodal myelin attachment segments, paranodal main segments, and internodal segments are included as separate compartments with different geometry and electrical properties. Action potentials are generated at the nodes with modified Hodgkin–Huxley equations that incorporate nonlinear fast Na+, persistent Na+, slow K+ conductance, a linear leak conductance, and membrane capacitance. The internodes have no active conductance. The axolemma and myelin sheath in paranodal and internodal segments each have a passive linear conductance in parallel with membrane capacitance. Myelin capacitance is represented as a function of capacitance per unit area and myelin thickness.
Approximately 500 control and 500 injured model axons were created using the axon diameter, internode length, and myelin thickness data obtained in this study. The diameter of each Fluoro-Ruby-stained axon was used to create 8 model axons with 121 nodes and 120 internodes of variable length. The 120 lengths were taken from a uniform distribution (one SD about the mean) calculated from the measured internode lengths of axons with diameters within 0.5 μm. The myelin thickness of each of the eight axons was taken as the myelin thickness of one of the eight BDA-stained axons with the most similar diameters. Thus, data from all labeled axons from all animals were used to create the simulated axons. The electrical and other geometric parameters were taken from McIntyre et al. for simulated axons with diameters ≥3 μm, and extrapolated for smaller axons using the measurements of Berthold et al. (1983).
The model axons propagated an action potential in response to a suprathreshold, depolarizing current step delivered to the node at one end. Conduction velocity was measured as the distance between the 10th and 100th nodes, divided by the action potential conduction time between those nodes. All simulations were run in NEURON (Hines and Carnevale, 1997).
Simulations of conduction were also run using the earlier model of Brill et al. (1977) for comparison. This simpler model, available from the NEURON database, represents the axon and myelin sheath as a single conductor. Membrane dynamics in the internodes are a function of myelin conductance and capacitance, which is estimated from the relationship between the capacitance and conductance of a coaxial cable (Goldman and Albus, 1968).
All anatomical data were analyzed using two-tailed t test because we were comparing means of two independent randomly selected populations with normal distributions. The significance level was p ≤ 0.05.
Mouse contusion leads to early, widespread myelin pathology
To establish the commonality of contusion-induced myelin pathology in our mouse model with that of rat and larger species, we conducted the following experiments. First, we established the presence of demyelination at early time points after injury. The RST was traced with bilateral injections of rhodamine-dextran-amine (Fluoro-Ruby) bilaterally in the red nuclei (Fig. 1A) at the same day that the mice underwent moderate contusion injury at T9. Two and a half weeks later, we observed 9–14% of extensively labeled axons with CASPR spreading caudal to the lesion (Fig. 2Aa'). We used the staining properties of CASPR concentrated in the paranodal regions of the axon (Poliak and Peles, 2003) and voltage-gated Kv1.2 channels concentrated in the juxtaparanodal regions as an index of myelin integrity (Karimi-Abdolrezaee et al., 2004), because myelin stains themselves do not provide satisfying direct investigation (Fig. 3H). Furthermore, sections stained with toluidine blue taken 2 weeks after injury showed evidence of some demyelinated axons immediately caudal (100–300 μm) to the lesion (Fig. 2B) compared with control animals (Fig. 2C). Additional evidence of early demyelination is the presence of remyelinated dorsal column axons 12 weeks after injury (Fig. 2D). Finally, quantitative PCR analysis revealed a significant decrease in MBP expression levels both at the lesion site and 1 mm caudal at 24 h, 5 d and 2 weeks after injury (Fig. 2E). These data show that the mouse contusion lesion creates myelin pathology and decreased myelin synthesis throughout the rostrocaudal axis of the lesion similar to what has been reported in rats.
To establish whether cut axons persist with chronically demyelinated segments, we performed the following experiments. We examined myelin indices on bilaterally Fluoro-Ruby traced RST axons 12 weeks after moderate contusion injury (Schmued et al., 1990) (Fig. 1). Fluoro-Ruby extensively labeled RSTs in control mice (Figs. 1C, 3C,D) and revealed morphologically intact fibers of passage extending caudally to the lesion in the injured animals (Figs. 1B,D,E, 3B,E–G). Dextrans have been shown to rely on intact axonal transport (Glover et al., 1986; Terasaki et al., 1995); hence, we classified Fluoro-Ruby-labeled axons as actively transporting. As described previously, many axons severed rostral to the injury site ended in bulbs that were filled with Fluoro-Ruby (Li and Raisman, 1995) (Fig. 3A, inset). Such axons exhibited a variable pattern of CASPR and Kv1.2 protein staining (Fig. 3A). As described previously, internodal proteins were found to be diffused along the axolemma and in addition we observed CASPR restricted to abnormally narrow clusters. Irregular CASPR and Kv1.2 protein distribution indicative of aberrant myelination (Arroyo et al., 2002; Karimi-Abdolrezaee et al., 2004) alternated with normal-looking CASPR-labeled paranodal sites (Fig. 3A).
Spared RSTs are remyelinated with significantly shorter internodes
After confirming the aberrant myelination status of fibers cut rostrally to the lesion, we sought to determine the myelin status of spared, actively transporting fibers of passage caudal to the lesion. To do this we examined internodal proteins in isolated whole tracts or single mechanically teased fibers that extended 0.5 mm caudally and beyond the lesion epicenter (Figs. 1B, 3B,E). In addition, CASPR and Kv1.2 protein distribution on Fluoro-Ruby-labeled axons had a normal compact spatial distribution indicating the presence of healthy myelin (Fig. 3F,H). We did not observe diffuse or dispersed protein distribution on the labeled fibers. Importantly, intact axons of passage had significantly shorter internodes (mean, 217.6 μm; n = 98; p < 0.0001, two-tailed t test) (Fig. 3B,E,F) compared with control animals (mean, 389.2 μm; n = 99) (Fig. 3C,D,G), suggesting that this population had been remyelinated (Blakemore, 1974; Gledhill and McDonald, 1977) (Fig. 4). Because mean axon diameter decreased by 18%, as described previously (Blight, 1983), and was not statistically different from controls (p = 0.06, two-tailed t test), we also compared axons with small/medium (≤3 μm) and large (3–4 μm) diameters in control and injury groups to control for diameter changes. We confirmed that myelin internodes are significantly shorter within both groups. The short internodes (≤100 μm) were either randomly located on the axon interspersed with average length internodes or were found to be next to each other. In addition, we observed short internodes (≤100 μm) located 2.5 cm away from the lesion epicenter.
Spared RST axons exhibit thinner to normal myelin thickness
In addition to examining internode length in longitudinal sections, we examined the quality and thickness of myelin in spared, actively transporting axons. Bilateral injections of BDA into the red nuclei were made 8 weeks after moderate contusion injury in adult mice (Reiner et al., 2000). Three weeks after injection, spinal cords were processed for immunoelectromicroscopy and the myelin G-ratio (axon diameter/total fiber diameter) was measured for axons immunopositive for BDA (see Materials and Methods) (Fig. 5). We did not observe organelles or neurofilament compaction in positively labeled fibers normally indicative of axon degeneration (Balentine, 1978). Axons positive for BDA in injured animals (Fig. 5B) contained no evidence of demyelinated segments, but had an increased g-ratio (mean, 0.71; n = 91) compared with uninjured controls (mean, 0.68; n = 79; p = 0.06, two-tailed t test) (Fig. 5E). Because there was a trend toward smaller axon diameter in injured animals, axons were sorted into small/medium (≤2 μm in diameter) and large (>2 μm in diameter) caliber groups to further characterize the changes in G-ratio. Only small/medium axons had a significantly increased G-ratio (p = 0.025, two-tailed t test), indicating thinner myelin. Large axons did not have significantly thinner myelin as evidenced by statistically normal G-ratios (p = 0.243, two-tailed t test). Overall, although there was a trend for thinner myelin in injured mice compared with uninjured controls, the increase in G-ratio was below that reported previously for nontraced axons (Blight, 1993; Totoiu and Keirstead, 2005). In addition, G-ratio increase was positively related to internode length increase, which could indicate that thinner myelin tended to be of a longer internode (Fig. 5F). The distinct difference between the present analysis and the existing literature is the use of prospective axon labeling to delineate spared axons.
Shortened internodes lead to decreased conduction velocity
Although the changes in myelin indices are surprisingly minimal, there is a possibility that decreased internodal length and myelin thickness could combine to significantly alter the conduction properties of spared axons. To better understand the predicted relationship between shortened internodes and thinner myelin in spared axons, action potential conduction in control and injured axons was simulated using the double cable axon model of McIntyre et al. (2002). Control and injured axons were constructed using axon diameter, internode length and myelin thickness measurements from Fluoro-Ruby and BDA-stained axons. Simulated conduction velocities ranged from 6 to 48 m/s for axons with diameters from 1.2 to 7.9 μm.
The simulations showed a decrease in conduction velocity for the injured axons, but no conduction failures or changes in action potential waveform. The average conduction velocity of the axons constructed with data from the spinal injured animals was 32% slower than the velocity of the control group (15.3 vs 22.4 m/s, p < 0.01). The change in the distribution of axon diameters after injury contributes significantly to this effect (Fig. 6B). However, the majority of Fluoro-Ruby-stained axons were between 2.0 and 4.0 μm, with similar means for both control and injured groups which allowed us to control for a diameter change. Simulations of axons constructed only with data from axons with 2–4 μm diameter still showed a 21% decrease in average conduction velocity for the injured group (15.9 vs 20.2 m/s, p < 0.01) (Fig. 6A). To see the contribution of myelin thickness change to conduction velocity (CV), we repeated the simulation using the same myelin thicknesses for both the control and injury groups (i.e., only internode length differed between the control and injury axons), which produced the same result. Thus, decreased axon diameter and internode length, but not myelin thickness, are likely to contribute to significant slowing of rubrospinal axons below the level of a contusion injury.
Simulations were also performed using the single cable model of Brill et al. (1977). The decrease in conduction velocity of the simulated injured axons was not as great as with the double cable model: 14% slower than the control axons. There was no conduction failure for any simulated axons.
Passive axon labeling reveals a pathologic population with myelin deficits
As stated above, spared axons with intact axonal transport (dextran labeled) were myelinated but occasionally we observed demyelinated unlabeled axons below the lesion. To determine whether demyelination could be present on axons exhibiting a pathological transport system we used a lentiviral reporter (pCMV-GFP) to infect rubrospinal neurons 8 weeks after injury. Viral transfection resulted in axons extensively filled with green fluorescent protein (GFP) (Fig. 7A). The majority of axons (96.5%) exhibited healthy morphology whereas a small subpopulation of axons (3.5%) showed dystrophic morphology with elaborate bulbous protrusions along their axolemma (Fig. 7A,B,D) different from the controls (Fig. 7C). Axons that exhibited normal morphology also showed the same pattern of CASPR and Kv1.2 staining as Fluoro-Ruby-labeled axons (Figs. 3B–H, 7B). CASPR and Kv1.2 labeling was spatially compact and restricted to the paranodal regions; we did not observe spreading or diffusion of the paranodal proteins along the GFP filled axons with normal morphology (Fig. 7B). In dystrophic, GFP-labeled axons that demonstrated globular protrusions, CASPR and Kv1.2 proteins were not always present in a predictably periodic and compact manner. Surprisingly, we observed some regular paranodal protein staining interspersed with diffused or no staining along the axolemma (Fig. 7E). In both cases where dystrophic GFP-labeled processes could be followed longitudinally along the cord, they terminated in dystrophic end-bulbs indicating that they had retracted from their target. At this time, we cannot draw any conclusions whether healthy axons instigate remyelination or whether myelin preserves spared axons.
In this study, we describe the remyelination pattern at subacute and chronic survival periods of spared mouse RST axons that extend caudally past a moderate contusion lesion. We used a combination of classical anatomical techniques and new molecular markers to examine myelin indices around actively transporting axons and passively filled tracts. Our data confirms that mouse contusion injury results in myelin pathology including demyelinated axons during the subacute postinjury period. Importantly, our data show that by 12 weeks after injury, RST axons of passage with active axonal transport are fully remyelinated as evidenced by the presence of myelin indices similar to control but significantly shortened internodes. These observations establish a tight correlation between axon sparing and the occurrence of remyelination in the chronic period.
The concept of chronically demyelinated axons is based on anatomical, pharmacological and cell based studies. Demyelination has been widely reported after chronic spinal cord injury in animals (Gledhill et al., 1973; Blakemore, 1974; Gledhill and McDonald, 1977; Blight, 1983) and limited demyelination is observed in humans (Gensert and Goldman, 1997; Kakulas, 1999; Guest et al., 2005). Previous studies suggest that remyelination in experimental injury models is mostly incomplete and that new myelin is abnormally thin and contributes to conduction failure (Griffiths and McCulloch, 1983; Scolding and Lassmann, 1996; Nashmi and Fehlings, 2001). Abnormally thin myelin as well as shortened internodes are thought to be indicators of new myelin (Gledhill and McDonald, 1977; Blight, 1993). Combined with evidence that mRNA levels for myelin proteins MBP and PLP (Wrathall et al., 1998) are reduced in the injured spinal cord, these studies suggest myelin may be unstable or insufficiently thick to insulate spared axons. The implications are that chronic demyelination in the absence of axonal loss can contribute to dysfunction of the injured spinal cord.
Previous work on 4-aminopyridine (4-AP), a potassium channel blocker, demonstrated some neurological improvement after SCI in animal models (Bostock et al., 1981; Blight, 1989). One theory is that improvement results from blockage of potassium channels exposed on demyelinated, intact axons. However, 4-AP also potentiates synaptic transmission and skeletal muscle twitch tension (Smith et al., 2000) and human clinical trials of 4-AP in chronic SCI have been inconclusive so far where some report modest and temporary improvement on motor and sensory functions especially reduction in spasticity, whereas others do not find significant beneficial effects (Potter et al., 1998; van der Bruggen et al., 2001; Wolfe et al., 2001; DeForge et al., 2004).
In addition to potassium channel blockers, cell transplantation with the purpose of remyelinating fibers has been one of the main therapeutic approaches to improve function after SCI (Li et al., 1997; Akiyama et al., 2002; Cao et al., 2005; Cummings et al., 2005; Keirstead et al., 2005; Lepore and Fischer, 2005). Remyelination by transplanted cells of different kinds is readily observable, but only if such transplants are delivered early, namely before 3 weeks after SCI (Keirstead et al., 2005; Karimi-Abdolrezaee et al., 2006). It could be that transplanted cells outcompete endogenous oligodendrocyte progenitors that are slower to divide and differentiate than primed transplanted cells. Early remyelination could be beneficial to stabilize and preserve surviving axons because myelin provides tropic and structural support and chronically demyelinated axons have been noted to develop severe pathologies in PLP, myelin-associated glycoprotein mutants, and multiple sclerosis lesions (Griffiths et al., 1998; Yin et al., 1998; Blakemore et al., 2000; Bjartmar and Trapp, 2003).
To date, anatomical studies of demyelination have been based on correlative analyses, which do not take in to account the status of the axon (spared or severed) or its site of origin. Our findings indicate that myelin indices immediately caudal to the injury site are near control values and establish a correlation between axonal sparing and myelin status. Although myelin thickness was reduced, the G-ratios measured were within the 0.6–0.8 range, which is considered a theoretically optimal range for axon conduction (Waxman, 1980). Our results correlate with studies that observed complete remyelination after myelin was experimentally removed but the axon remained intact (Blakemore and Murray, 1981; Jeffery and Blakemore, 1995). Important caveats to the current approach include the use of the murine model where anatomical distances are small compared with larger species. For example, it may be that demyelination on spared axons is strictly limited to the injury epicenter. In the mouse, this would be a region of 0.4–0.8 mm in size, an area we were unable to examine because of the interference of intense lesion-related autofluorescence.
Because intact axons were remyelinated, we sought to label RST axons, which may be severed or lack functional transport by using a retroviral reporter (CMV-GFP) as a passive labeling technique. The purpose of this approach was to label demyelinated or thinly myelinated axons described previously to persist after spinal cord injury (Waxman, 1989; Totoiu and Keirstead, 2005). In 3.5% of axons traced in this manner caudal to the lesion, we observed some demyelination or abnormal patterns of remyelination. This subpopulation of axons exhibited a very similar pattern of demyelination or abnormal remyelination to fibers cut rostral to the lesion. The data from the passively GFP-labeled axons and axons severed rostral to the lesion demonstrate that abnormal myelin or demyelination may be confined to severed or degenerating axons. These data highlight three important findings: (1) demyelination is primarily present in severed axons in the chronic period after spinal cord injury, (2) dystrophic axons can exhibit variable regions of full and incomplete remyelination along several millimeters, and (3) complex swellings are periodic along demyelinated processes and persist in a small population for at least 12 weeks. It is important to consider that axonal degeneration may be a continuum after injury and it cannot be determined from the present data whether functionally transporting axons promote remyelination or whether remyelination leads to the return/maintenance of axonal transport.
Some virally and dextran-labeled axons in our studies resembled dystrophic axons described in head and spinal cord injury models (Li and Raisman, 1995; Povlishock et al., 1997). Previous work suggests these are dynamic structures and can be found close to the site of the lesion up to 1 week after injury, but it is not yet possible to determine if they are in a state of degeneration or persistent abortive growth (Tom et al., 2004). Future research should be done to determine whether passively labeled dystrophic axons are degenerating slowly and demyelinating or if they are stable, yet dynamic structures.
To further evaluate the implications of our myelin analyses in spared axons we adapted a mathematical model of single axonal conduction. The rationale for this was twofold. First, evidence in humans and rodents with chronic spinal cord injury suggest reduced conduction velocity, longer refractory period and reduced excitability (Waxman, 1980; Blight, 1983; Nashmi and Fehlings, 2001). Second, we theorized that a combination of shortened internodes with a marginally reduced myelin thickness could cause an increase in conduction failure or significant decrease in conduction velocity. McIntyre's et al.'s (2002) theoretical model of a single myelinated axon simulations predicted a 21% decrease in conduction velocity caused by changes in myelin for spared remyelinated RST axons with the physical parameters measured in our studies. This is a moderate reduction in conduction velocity and the effect on locomotion is likely moderate or negligible in mice. However, a 21% CV decrease might be important in larger mammals and the actual effect on nerve function will only be determined by physical measurement of CVs in the future.
In experiments where conduction velocity and compound action potential (CAP) were measured chronically after a contusion lesion similar to the injury severity modeled here, Van de Meent et al. (1996) showed a similar (12.5%) drop in conduction velocity to that predicted by our calculations. Importantly, these studies show that changes in latency are not a strong predictor of functional outcome, whereas a decrease in the CAP amplitude tightly correlates with recovery in locomotion. These data indicate that the myelin status of spared long-tract axons in the chronically contused spinal cord is less likely to contribute to functional deficits than is axonal loss.
Nonetheless, the functional conclusions that can be drawn from our simulations are limited. Our model examines only the conduction of a single action potential, but does not take into account what the effect might be when trains of action potentials are conducted along the axon (which would be more physiologically applicable). This may be particularly relevant in light of findings that remyelinated axons fatigue more readily than controls (Nashmi and Fehlings, 2001). Fatigue in combination with a drop in velocity might contribute in a significant way to loss of locomotor function. In addition, our simulations assume that membrane conductances and other biophysical properties of the axon are unchanged after remyelination. There is no quantitative data available for these parameters after spinal cord injury, and consequently there was no basis to model such pathophysiology. Direct electrophysiological studies of descending axons, including an analysis of frequency-dependent conduction, will be required to elucidate definitively the physiology of remyelinated axons after spinal cord injury.
The clinical implications of our findings are potentially important. Our data suggest that, given time, endogenous remyelination of spared axons of passage is close to that of uninjured axons. This new myelin likely contributes to the spontaneous functional recovery observed after injury (Blight, 1993). In the future, the dynamism of demyelination and remyelination needs to be studied with a careful consideration of whether an axon is intact or severed.
This work was supported by National Institutes of Health Grants NSO46724 and NS040867, the International Foundation for Research on Paraplegia, and a Royalty Research Grant (University of Washington, Seattle, WA). We thank J. Trimmer (University of California, San Diego, CA) for providing polyclonal CASPR antibody, F. Galimi (University of Sassari, Sassari, Italy) for a gift of viral plasmids, D. Possin and N. Gill (University of Washington, Seattle, WA) for technical help with electron microscopy, and J. Petruska (State University of New York, Stonybrook, NY) for comments on this manuscript. P.H. is a member of the University of Washington Institute for Stem Cell and Regenerative Medicine and the Center on Human Development and Disability.
- Correspondence should be addressed to Philip Horner, Department of Neurological Surgery, University of Washington, Seattle, WA 98195.