Myelinated axons allow the action potentials to be generated at nodes of Ranvier and signals to propagate in a saltatory pattern. The lateral borders of the myelin sheath connect with axolemma at paranodes through the axon-glial junctions,1 which segregate Na+ channels at the nodes and K+ channels at the juxtaparanodes along the axolemma.2 Specialized axo-glial junctions at the paranodes include oligodendrocyte transmembrane protein neurofascin 155 and axonal adhesion molecules Caspr-1 and contactin.3, 4 High-density Na+ channels assemble at the node through interactions with proteins including neurofascin 186 (Refs. 5 and 6) in the axolemma. K+ channels located at the juxtaparanodes are covered by a compact myelin sheath. The integrity of nodal and paranodal domains is vital to fast-action potential conduction along myelinated axons. Studies have shown that disruption of neurofascin localization precedes demyelination in tissues from multiple sclerosis (MS) patients,7 and that disruption of the axonal contactin-associated protein (Caspr)-1 may be an indicator of myelin abnormalities.8 Moreover, diffusive patterns of Kv1.2, paranodin, Caspr-2, and Nav channels were found in immunolabeled plaques from MS patients.9 Nevertheless, due to lack of control of the disease stages in these studies using MS tissues, it is not clear whether paranodal domain injury occurs at the onset of the disease. More importantly, the status of paranodal and nodal myelin at different stages of the disease was not clarified due to lack of a proper myelin observation tool.
In this paper, we address the above issues by a thorough imaging study of myelin in relapsing experimental autoimmune encephalomyelitis (R-EAE), a widely used animal model reproducing specific features of the histopathology and neurobiology of MS. Accepted as a T-cell-mediated autoimmune disease of the central nervous system,10 the EAE model has been extensively employed for investigation into the mechanisms of myelin loss11, 12 and assessment of myelin repair therapies.13, 14
Myelin is normally characterized by electron microscopy (EM) and/or histology. Although EM provides ultrastructure of myelin in resin-embedded sections, its sophisticated sample preparation and small field of view inhibit effective quantitative and/or statistical examination. Luxol fast blue staining of paraffin-embedded tissues is a widely used histological method for large-area mapping of the white matter. However, this method lacks the resolution for observation of changes in single myelin tracks, which inhibits unequivocal discrimination of remyelination from normal myelination. Clinical imaging tools, including magnetic resonance imaging and positron-emission tomography, permit noninvasive longitudinal imaging of the central nervous system and demyelinating lesions in the white matter,15, 16 but only at a low spatial resolution that is insufficient to reveal details at the single myelin track level. Currently, a correlation between myelin status and clinical scores of EAE is missing. In particular, characterization of the nature of the initial damage to myelin at disease onset remains unclear. A dearth of such knowledge has slowed the use of EAE for mechanistic studies of myelin loss, as well as therapeutic development of myelin repair strategies.
By providing contrast based on C–H bond vibration,17 coherent anti-Stokes Raman scattering (CARS) microscopy has permitted high-speed imaging of natural and chemically damaged myelin in ex vivo tissues.18, 19, 20, 21, 22 The CARS contrast arises from the CH2 symmetric vibration in myelin lipid, which comprises 70% of the dry weight of white matter. As a nonlinear optical process, CARS provides inherent 3-D submicron resolution enabling visualization of detailed myelin structures such as the paranodal myelin loops.18 In this study, we employ CARS microscopy, coupled with two-photon excited fluorescence (TPEF) imaging and confocal Raman spectral analysis,23 to distinguish the myelin sheath in naïve mice, and mice during the peak acute and remission stages of R-EAE. We show that paranodal myelin disruption corresponds to the onset of CD4+ T-cell-mediated EAE and is partially restored in the remission stage.
Materials and Methods
Induction of Relapsing Experimental Autoimmune Encephalomyelitis
Female SJL mice were purchased from the Harlan Laboratory (Indianapolis, Indiana) and maintained according to protocols approved by the Northwestern University Animal Care and Use Committee. Mice were primed subcutaneously with an emulsion containing 100 μL of incomplete Freund adjuvant (Difco, Detroit, Michigan) supplemented with 200 μg of Mycobacterium tuberculosis H37Ra (Difco, Detroit, Michigan) and 50 μg of PLP139–151 (HCLGKWLGHPDKF), which was synthesized by Genemed synthesis. Clinical scores were recorded daily using a scale of 0–5 (0, healthy; 1, limp tail or hind limb weakness; 2, limp tail and hind limb weakness; 3, partial hind limb paralysis; 4, total hind limb paralysis; and 5, moribund). All animal procedures were conducted in complete compliance with the NIH Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee of Northwestern University.
SJL mice were anesthetized with sodium pentobarbitol. During deep anesthesia, the mice were perfused transcardially with 10 mL of cold PBS (pH 7.4), and then with 10 mL of PBS containing 4% paraformaldehyde for in vivo fixation. The vertebrae were immediately excised and kept in 4% paraformaldehyde. After 24 h, <1-cm-long lumbar spinal cord was extracted and sectioned longitudinally into 100-μm slices from ventral to dorsal side using a tissue slicer (OTS-4000, Electron Microscopy Sciences, Hatfield, Pennsylvania). The slices were stored in PBS prior to imaging. For both ventral and dorsal parts, sliced slices at the same depth were used for comparison between the four groups composed of naïve mice, mice at onset of R-EAE, mice at peak acute of R-EAE, and mice at remission of R-EAE. For nuclei staining, the fixed spinal cord slices were incubated in PBS supplemented with 4 ng/mL Hoechst33342 dye for 2 h and washed with PBS prior to imaging.
Coherent Anti-Stokes Raman Scattering and Two-Photon Excited Fluorescence Imaging
Two beams at frequency ωp and ωs generated from two tightly synchronized Ti:sapphire lasers (Mira 900/Sync-lock, Coherent Inc., Santa Clara, California), with a pulse duration of 2.5 ps, were used for both CARS and TPEF imaging on the same microscope. The two beams were parallel polarized and collinearly combined. A Pockels’ cell was used to reduce the repetition rate from 78 to 7.8 MHz. The overlapping beams were directed into a laser-scanning microscope (FV300/IX70, Olympus Inc., Center Valley, Pennsylvania) and focused into a sample through a 60× water-immersion objective lens [numerical aperture (NA) = 1.2] or a 20× air-objective lens (NA = 0.75). The CARS signal was collected in the forward direction with an air-objective lens (NA = 0.5). The frequency difference between the pump and Stokes beams, ωp−ωs, was tuned to the symmetric CH2 vibration at 2840 cm−1. The same picosecond laser beams were also used for TPEF imaging of Hoechst or immunofluorescence. The TPEF signal was collected in the backward direction. Both CARS and TPEF signals were detected with the same type of photomultiplier tube (PMT, R3896, Hamamatsu, Japan). No photodamage to myelin was observed.
Confocal Raman Microspectroscopy
On the CARS microscope, a spectrometer equipped with an electron multiplying charge-coupled device (EMCCD) was attached to the side port with the slit replaced with a pinhole for confocal Raman analysis. After a CARS image was taken, the Stokes beam was blocked and a short-pass dichroic mirror directed the Raman signal into the spectrometer. The point-scan signal from the scanner externally triggered the spectrometer, which recorded a Raman spectrum at a point of interest. A detailed description can be found in elsewhere.23
Antiserum against Kv1.2 antibody (Alomone Lab, Jerusalem, Israel) was used to locate the K+ channels at the juxtaparanodes. NFC2 antibody (provided by Professor Peter J. Brophy3) was used to label Neurofascin NF155 at the paranodal myelin and neurofascin 186 at the axolemma. Antiserum against F4/80 (Invitrogen, Carlsbad, California) was used to label activated microglia and infiltrated macrophages. FITC conjugated antiserum against CD4 (Ebioscience, San Diego, California) was used to label CD4+ regulatory T cells. Platelet-derived growth factor receptor α (PDGFRα) antibody (Fitzgerald Industries International, Concord, Massachusetts) was used to label oligodendrocyte precursor cells. FITC conjugated anti-rat IgG (Sigma, St. Louis, Missouri) and Alex 488 conjugated anti-rabbit IgG (Sigma, St. Louis, Missouri) were used as secondary antibodies. The immunofluorescence images were taken by TPEF on the same laser-scanning CARS microscope.
Multimodal Coherent Anti-Stokes Raman Scattering Imaging Reveals De- and Remyelination in Relapsing Experimental Autoimmune Encephalomyelitis
R-EAE (Fig. 1) was induced by immunizing SJL mice with PLP139–151 peptide emulsified in complete Freund's adjuvant, and lumbar spinal cords were extracted at various stages during the disease. White matter in naïve mice, and mice at onset, peak acute, and remission stages of R-EAE was characterized by simultaneous CARS imaging of myelin and TPEF imaging of Hoechst-labeled cellular nuclei. The white matter from a naïve mouse (disease score = 0) showed strong CARS (red) and TPEF (green) signals [Fig. 2a], which represent ordered parallel myelin fibers and nuclei of glial cells [Fig. 2b], respectively. At the onset of EAE (defined by a disease score of 1 following a period of no clinical symptoms from the time of immunization), which manifests in mice as tail weakness, the myelin in the white matter appeared normal with fibrous structure, but the number of cell nuclei in the lumbar spinal cord was increased significantly compared to naive mice [Figs. 2c, 2d]. Immunofluorescence imaging showed a large number of CD4+ and F4/80+ immune cells in the meninges of the spinal cord (Fig. 3), indicating the infiltration of T cell and macrophage into the spinal cord during EAE onset. At the peak acute stage of EAE (score = 3 or 4), characterized by partial or complete hind-limb paralysis, demyelinating lesions [Fig. 2e] were observed near the meninges of the spinal cord and around blood vessels and were distinguished by extensive myelin debris and clustered cells [Fig. 2f]. CARS and two-photon excited immunofluorescence imaging further showed that CD4+ T cells and F4/80+ macrophage/microglia occupied these lesions (Fig. 3). During remission from acute disease (defined by a disease score of 1 following a previous peak score of 3–4), clusters of cells were observed in the meninges of the spinal cord and around blood vessels [Fig. 2g]. Adjacent to the accumulated cells, we also observed myelin fibers, small vesicles [arrows in Fig. 2h], and randomly organized cells. Combined CARS imaging of myelin and immunofluorescence imaging of labeled PDGFRα showed the existence of oligodendrocyte precursor cells among the randomly organized cells (Fig. 4), consistent with an interpretation of attempted myelin regeneration in these sites. The same types of myelin morphological changes were observed between the dorsal and ventral parts of lumbar spinal tissue extracted from the R-EAE mice. The multimodal images enabled us to quantify the density of myelin fibers, density of cell nuclei, and myelin fiber thickness in each stage of R-EAE. At disease onset, the density of cells in the white matter [e.g., on the left of the dash line in Fig. 2d] was almost twice of that in the naïve white matter. In acute lesions, the density of cells increased by 11 times, whereas no intact myelin fibers were identified. In accordance with the incomplete functional recovery in the remission stage (Fig. 1), the density of remyelinated fibers was found to be lower than that in the naïve white matter [Fig. 2i]. Moreover, axon diameter and myelin thickness were measured through the intensity profile of CARS signal from the myelin fiber. The regenerated myelin fibers were found thinner [Fig. 2j] in both dorsal and ventral parts of spinal cord and overall exhibited a significantly decreased ratio of myelin thickness to the axonal diameter (0.32 ± 0.13) compared to normal myelin fibers (0.60 ± 0.25) in the naïve white matter [Fig. 2k]. We note that the ratio of myelin thickness to the axonal diameter from normal mice presented here is not equivalent to the g ratio defined as the ratio of the inner diameter to the outer diameter of myelin fiber.
Raman Spectral Analysis Shows Degradation of Myelin Lipids
To identify conformational and/or chemical changes in naïve, degraded, and regenerated myelin, we recorded Raman spectra of myelin at the different stages of R-EAE by combination of CARS microscopy and confocal Raman microspectroscopy on the same platform.23 CARS imaging was used to locate the normal myelin in the naïve white matter [Fig. 5a], myelin debris in the demyelinating lesions [Fig. 5b], and regenerated myelin during the remission [Fig. 5c]. The Raman spectra of myelin shown in Fig. 5d were dominated by vibrational bands from lipids24, 25 due to the high lipid content in myelin. The spectral assignment and the intensity ratios between the Raman bands are summarized in Tables 1, 2, respectively. Both the C–C and C–H vibrational bands were used for determining the conformation of the hydrocarbon chains of lipids. We first studied lipid packing using the prominent bands at 2850, 2885, and 2930 cm−1, which correspond to symmetric CH2 stretching, asymmetric CH2 stretching, and CH3 stretching vibrations, respectively. As shown in Fig. 5e and Table 2, a significantly larger intensity ratio of I2930/I2885 was found in myelin debris, which reflects an increase of the intermolecular chain disorder.25 Interestingly, the regenerated myelin also showed a higher lipid-packing disorder than the normal myelin. We further studied the lipid unsaturation using the intensity ratio of I1650/I1445, which represents the ratio of C=C stretching vibration bands to H–C–H deformation bands in lipid acyl chains. We found that myelin debris gave the highest unsaturation degree [Fig. 5e]. Myelin regeneration decreased the unsaturation degree but remained higher than the normal myelin. Additionally, the intensity ratio of I1122/I1076, which is related to the ratio of the trans and cis conformation populations of C–C bonds, was employed to examine the intramolecular chain ordering in lipid hydrocarbon chains,24 but no significant difference was observed between the myelin debris, normal myelin, and regenerated myelin. In summary, Raman spectral analysis of lipid conformations shows that the degraded myelin has lower lipid-packing ordering and higher unsaturation than normal myelin, and that the lipid conformation is not fully recovered in regenerated myelin. The increase of unsaturation in myelin debris might be due to the formation of conjugated double bond systems and the isomerisation of cis to trans double bonds during lipid oxidation.26 Notably, previous measurements of lipid acyl chain composition using magic angle-spinning NMR spectroscopy also showed an increase of polyunsaturation degree in the cerebral white matter from EAE marmoset.27
Full Raman spectra and assignments of Raman peaks in myelin.
|Raman shift, cm−1||Assignment1|
|1057||Trans ν (C–C)|
|1076||Gauche ν (C–C)|
|1122||Trans ν (C–C)|
|1264||δ (=C–H) in plane cis|
|1294||δ (CH2) twisting|
|1431||δ (CH2) bending|
|1650||ν (C=C) cis|
|1675||Protein amide I region|
|2849||CH2 symmetric stretch|
|2885||CH2 asymmetric stretch|
|2930||CH3 symmetric stretch|
|3009||=C–H symmetric stretch cis|
Intensity ratios for the interested Raman peaks.
|R1 = I2930/I2885||R2 = I1650/I1445||R3 = I1122/I1076|
|Normal myelin||0.778 ± 0.024 (n = 13)||0.336 ± 0.013||0.832 ± 0.089|
|Myelin debris||0.985 ± 0.136 (n = 13)||0.788 ± 0.168||0.717 ± 0.158|
|Regenerated myelin||0.862 ± 0.031 (n = 13)||0.446 ± 0.045||0.814 ± 0.094|
R 2 represents the lipid-chain unsaturation degree. For example, R 2 = 0 for the saturated palmitic acid, R 2 = 0.58 for the monounsaturated oleic aid, and R 2 = 1.27 for the polyunsaturated linoleic acid (Ref. 23).
Paranodal Myelin Disruption Occurs at Onset of Experimental Autoimmune Encephalomyelitis and within Borders of Demyelinating Lesions
To our surprise, the density of myelin fibers [Fig. 2i] and the myelin thickness [Fig. 2j] at the onset of EAE showed no difference from those in the naïve white matter. To account for the functional deficit, we examined possible myelin degradation occurring in the paranodal domain. Normally, the paranodal myelin is held on the axolemma by the axo-glial junctions, which segregate the Na+ and K+ channels along the axolemma.2 Under the CARS microscope, we observed clear differences of paranodal myelin in the lumbar spinal tissues between naïve and R-EAE mice. In the naïve white matter, the node of Ranvier [Fig. 6a] was observed as a gap between two segments of myelin sheath. However, in the white matter of EAE spinal cord during onset and within borders of acute demyelinating lesions, a larger distance between two myelin segments was observed [Figs. 6b, 6c], which indicates a retraction of paranodal myelin toward internodes. To quantify such retraction, we define the nodal diameter as the distance of the CARS signal at the two sides across the axon and nodal length as the distance of the CARS signal at two sides along the axon [Fig. 6d]. We measured both nodal length and nodal diameter in the white matter of lumbar spinal tissue at different stages of R-EAE. Our data shown in Figs. 6e, 6f indicate that the nodal length with the same nodal diameter in the white matter of EAE spinal cord during onset, and within the lesion borders of the peak acute EAE spinal cord is much larger than that in the naïve stage.
Moreover, the extent of nodal length increase is not correlated to the nodal diameter [Fig. 6f]. The observation of elongated exposed nodes without loss of segmental myelin indicates that degradation of paranodal myelin occurs prior to the formation of myelin debris.
Paranodal Myelin Retraction is Associated with Elongated Neurofascin Distribution in Paranodal Domains
To explore molecular changes in the myelin-retracted nodes, we used NFC2 antibody to locate neurofascin155 and neurofascin186. Expression of Neurofascin186 is restricted to neurons and clusters at the nodal membrane.28 Neurofascin155 is located at the oligodendrocyte loops3 and functions as a glial receptor for the paranodin/caspr-contactin axonal complex at the axo-glial junctions.4 In the naïve white matter, the NFC2 antibody labeled both nodes and paranodes [Fig. 7a]. However, in spinal cord white matter from R-EAE mice at onset [Fig. 7b] and within acute lesion borders of peak acute spinal cord tissue [Fig. 7c], we found that NFC2 labeling extended into the internodes. The extension of NFC2 labeling to the internodes concords with the observation of retracted paranodal myelin shown in Figs. 6b, 6c. Furthermore, by measuring the length of labeled NFC2 at different disease stages, we found that with similar axonal diameter, NFC2 labeling was significantly longer in the spinal cord white matter of EAE mice during disease onset and within acute lesion borders of mice during peak acute EAE, compared to that in white matter isolated from naïve spinal cords [Fig. 7d]. The observation of elongated NFC2 labeling further confirms the retraction of paranodal myelin.
Paranodal Myelin Retraction Exposes and Displaces Juxtaparanodal Kv1.2 Channels
The K+ channels,29 especially the Kv1.1, Kv1.2, and Kvβ2 subunits, are normally located at the juxtaparanodes and protected by myelin.30, 31 To examine the effect of paranodal myelin retraction on the K+ channels, we immunolabeled Kv1.2 channels in the white matter of naïve mice, EAE mice at onset, and EAE mice at the peak acute phase, respectively. A mosaic CARS image of lumbar spinal tissue from mice at peak acute EAE is shown in Fig. 8a. Similar to that in the naïve white matter [Fig. 8b], typical paired immunolabeling of Kv1.2 channels were observed to be located at the juxtaparanodes and concealed beneath compact myelin sheath in the normal-appearing white matter (NAWM) [stars in Fig. 8c]. In contrast, exposed Kv1.2 channels were commonly observed at the borders of demyelinating lesions [arrows in Fig. 8d] and during disease onset [Fig. 8f]. The disruption of paranodal myelin also suggests consequent breakdown of axo-glial junctions between the axon and myelin at the paranodal domain. Accordingly, displacement of Kv1.2 channels into paranodal and nodal domains was observed in the acute lesion borders [Fig. 8e] and the white matter of mice with a score of 1 during disease onset [Fig. 8f]. In the lesions, no paired Kv1.2 channels were found among myelin debris [Fig. 8a]. The disruption of paranodal myelin and consequent exposure and redistribution of K+ channels would definitely impair the saltatory axonal conduction, which could account for the clinical symptoms of EAE at disease onset.
Paranodal Myelin is Partially Restored in Disease Remission
Using the R-EAE model, we further explored the status of paranodal myelin in the regeneration sites of the lumbar spinal tissue at the remission stage. In addition to the normal-appearing paranodal myelin, we also observed retracted paranodal myelin [Fig. 9a] and split paranodal myelin [Fig. 9b] with thinner internodal myelin. Consistent with CARS imaging of paranodal myelin, immunolabels of Kv1.2 channels displayed both typical pairs [Fig. 9c] and redistribution of channels [Fig. 9d]. The morphology of NFC2 labeling appeared to be normal [Figs. 9e, 9f].
Quantitative measurements of nodal length over nodal diameter revealed mixed populations of normal nodes and elongated exposed nodes [Fig. 9g]. The mean ratio of nodal length to nodal diameter was significantly reduced during remission in comparison to that at the border of lesions during peak acute, while still significantly higher than that in the naive white matter [Fig. 9h]. Moreover, quantitative analysis of NFC2 labeling showed a significant length increase in all stages of R-EAE compared to the naive white matter [Figs. 9i, 9j]. Together, these results indicate an incomplete restoration of paranodal structure in the regenerating process.
To date, it is still elusive how the myelin is degraded in multiple sclerosis.32, 33, 34, 35 There are two possible scenarios of demyelination based on previous studies. The first one is injury starting from internodal myelin, for example, thinning of internodal myelin layer by layer, as observed in lysolecithin-induced myelin swelling.19, 36 The second one initiates with paranodal domain injury.7, 9 In this study, CARS microscopy has been employed to examine these scenarios by label-free mapping of single myelin fibers in different clinical scores (0, 1, and 3) of R-EAE. With submicron spatial resolution, the CARS images allowed us to quantify the myelin thickness and determine the ratio of myelin thickness to the axonal diameter in different stages of the disease. Furthermore, by coupling CARS with two-photon immunofluorescence mapping of juxtaparanodal K+ channels, paranodal myelin retraction, and displacement of K+ channels are extensively observed at the onset of R-EAE and at the lesion border. Our results show that the loss of nodal integrity precedes the formation of myelin debris in a strict CD4+ T-cell-mediated R-EAE model of MS and that remyelination is accompanied by reestablishment of the nodal markers. Further experiments will be carried out by using in vivo imaging techniques to explore the demyelination sequences on single axons. Correlating our imaging results with the clinical scores demonstrates that alterations of the delicate nodal organization can disrupt axonal conduction in the onset stage of R-EAE. In addition, our finding that paranodal myelin is only partially restored in the regenerating sites offers one possible explanation to the incomplete functional recovery at the remission stage. We note that the relevance of this scenario of initiation of myelin destruction to that occurring in MS, which appears to be a collection of distinct histopathological subtypes,37 remains to be determined by further studies.
Understanding the mechanism underlying the disruption of paranodal myelin is important for the development of therapies aimed at delaying or preventing disease relapses. In the inflammatory lesions, infiltrating immune cells release proinflammatory cytokines, including IFN-γ, TNF-α, IL-17, GM-CSF, and LT, and numerous chemokines including RANTES, IP-10 and IL-8.11 These proinflammatory cytokines and chemokines are able to activate resident astrocytes and microglia cells,11 which then produce nitric oxide and oxygen radicals38, 39 and glutamate,40, 41 leading to white matter damage.42 Glutamate excitotoxicity has been shown to contribute to oligodendrocyte damage through glutamate receptors in the EAE model.43 Our previous work has shown that a high level of glutamate is able to inflict paranodal myelin splitting and retraction through NMDA and kainate receptor-mediated calcium influx.20 The similarity of structural changes suggests glutamate excitotoxicity as a possible mechanism contributing to the early myelin damage in the SJL R-EAE model. Indeed, the administration of glutamate receptor inhibitor before the onset of EAE has been shown to improve the neurological outcome and delay the progression of EAE.43, 44 The partial restoration of paranodal myelin shown in Fig. 9 may be related to the abnormal interactions between oligodendrocytes and axons29 during remyelination. Our result highlights the significance of not only promoting remyelination, but also recovering the paranodal domain in myelin repair therapy.
With label-free vibrational contrast, 3-D submicron resolution, multimodality, and live-animal imaging capability,45, 46 CARS microscopy is expected to have wide applications to myelin biology. Initial applications have revealed the role of glutamate excitotoxicity in paranodal myelin damage.20 Such capabilities can be extended to investigate the roles of oxidative stress and nitric oxide in myelin loss.42 Furthermore, CARS imaging can be coupled with genetic tools to explore the roles of specific myelin components, such as the galactolipid,29, 47 and axonal proteins, such as AnkyrinG,48 Caspr/paranodin,49 and neurofascin proteins50 in the organization of paranodal and nodal domains.51 Finally, CARS microscopy can be used as a reliable and efficient readout of de- and remyelination in demyelinating diseases. Recent efforts have been made to promote remyelination by using γ secretase inhibitor in the coculture system52 and EAE mice.53 With in vivo imaging capabilities,21, 45, 54 CARS microscopy would effectively allow quantitation of the outcome of the myelin repair therapies.
This work was supported by NIH Grants No. R01EB7243 (J.C.) and No. R01 NS030871 (S.D.M.), and a grant from the Myelin Repair Foundation (S.D.M.).