Introduction

The inability of adult vertebrate central nervous system (CNS) axons to regenerate is seen as a consequence of the highly inhibitory environment at the injury site [1] and of the failure in activating a cell-intrinsic program leading to the expression of regeneration-associated genes [2]. Upon CNS injury, the glial scar functions as an inhibitory barrier composed of multiple components, generally divided in three categories: myelin-associated inhibitors (MAIs), namely Nogo, myelin-associated glycoprotein (MAG), and oligodendrocyte myelin glycoprotein (OMgp) [3]; canonical axon guidance molecules, such as semaphorin 3A, ephrin B3, netrin-1, and repulsive guidance molecule A (RGMa) [4]; and chondroitin sulfate proteoglycans (CSPGs), produced by astrocytes [5].

Studies from different groups using triple knockout mice for MAG, Nogo, and OMgp produced conflicting results ranging from limited [6] to extensive [7] regeneration abilities. Although the field has largely concentrated on MAIs, other inhibitors play important roles in vivo: blocking CSPG or RGMa or deleting ephrin B3 increases axonal regeneration following spinal cord injury (SCI) [8–10]. As such, the wide variety of inhibitors present in the spinal cord milieu is thought to underlie the absence of robust axon regeneration when blocking either single or a limited combination of inhibitors. Despite the structural differences, several MAIs share receptors and an inhibitory mechanism dependent on RhoA/Rho-kinase (ROCK) activation [11], as demonstrated mainly through the use of RhoA/ROCK inhibitors [12].

Treatments of SCI remain limited and aim mainly at supportive care. Many efforts have been made for the development of new therapies that generally target the inhibitory environment of the glial scar, but so far functional recovery remains limited. To further explore the importance of myelin in the inhibition of axon regeneration, we used shiverer (shi) mice which lack myelin basic protein (MBP), a key player in myelin compaction in the CNS [13]. In the shi CNS, the absence of MBP results in a complete lack of compact myelin [14] which is accompanied by a severe phenotype comprising shivering, seizures, and early death. Here, we show that in shi mice, despite the presence of canonical protein MAIs and of axonal abnormalities generally related to decreased regeneration capacity, CNS axons have an increased ability to regenerate through the spinal cord glial scar. As lipids represent over 75 % of myelin dry weight, we further evaluated the possibility that increased axon regeneration in shi mice was due to the severely decreased abundance of myelin lipids. Corroborating this hypothesis, we provide evidence that myelin lipids, specifically cholesterol and sphingomyelin, are inhibitory for axon growth. Moreover, we show that delivery of 2-hydroxypropyl-β-cyclodextrin (HPβCD), a drug that reduces lipid levels in the SCI site, leads to increased axon regeneration of dorsal column tract axons. In summary, our work demonstrates that myelin lipids are important modulators of axon regeneration that should be regarded together with protein MAIs as critical targets to improve axon regeneration after injury.

Materials and Methods

Animals

Mice were handled according to European Union and National rules. All procedures were approved by the IBMC Ethics Committee and by the Portuguese General Veterinarian Board. Six-week-old wild-type (WT) and shi littermates of either sex, on a Swiss Webster:C3HeB/Fe5 background [15], were obtained from heterozygous breeding pairs. For drug-delivery studies 8-week-old C57BL/6 mice of either sex were used.

Electron Microscopy

The thoracic spinal cord of WT and shi littermates (11 weeks old) was embedded in Epon. Semi-thin 1-μm-thick sections were stained with p-phenylene-diamine (PPD) to visualize myelin and white matter tracts. Sixty-nanometer-thick sections of the dorsal funiculus were stained with uranyl acetate and lead citrate and examined in a JEOL JEM-1400 transmission electron microscope.

Western Blotting

Twenty five micrograms of thoracic spinal cord protein from naïve (6 weeks old, n = 4 animals per genotype) or injured (11 weeks old, n = 3 WT and n = 4 shi mice) WT and shi littermates was separated in 3–8 % Tris-acetate gels (Bio-Rad) and transferred to nitrocellulose. Similarly, 25 μg of thoracic spinal cord protein from injured animals treated with HPβCD or vehicle (12 weeks old, n = 6 for each group) was analyzed by Western blot. Antibodies used were mouse anti-MAG (a gift from Dr. Richard Quarles, NINDS, Bethesda, 1:1000), rat anti-OMgp (R&D Systems, 1:250), rabbit anti-Nogo-A (a gift from Dr. Stephen Strittmatter, Yale University, New Haven, 1:10,000), rabbit anti-Ephrin B3 (Santa Cruz Biotechnology, 1:200), rabbit anti-RGMa (Immuno-Biological Laboratories, 1:200), and mouse anti-GAPDH (Santa Cruz Biotechnology, 1:2000).

MAG and Nogo Immunohistochemistry

Paraffin-embedded spinal cords from WT and shi littermates (6 weeks old) were used for immunohistochemistry against MAG (1:500) or Nogo-A (1:5000) using ABC Vectastain and DAB (Vector Labs).

Spinal Cord Injury

Laminectomy was performed at the T9 level. Complete transection, dorsal hemisection, or left lateral hemisection were done using a micro-feather ophthalmic scalpel.

Assessment of the Lesion Area Following SCI

Five weeks after complete SCI, 10-μm sagittal spinal cord cryosections were immunostained for CSPG (Sigma-Aldrich, 1:200) and glial fibrillary acidic protein (GFAP; DAKO, 1:500). Collagen was visualized with Masson trichrome staining (Sigma-Aldrich). The injury area was measured using Photoshop CS3.

Regeneration of Dorsal Column Fibers

Animals with dorsal hemisection were allowed to recover for 4 weeks. Four days before sacrifice, 2 μL of 1 % cholera toxin B (CT-B; List Biologicals) was injected in the left sciatic nerve. Fifty-micrometer sagittal free floating sections were immunostained for CT-B (List Biologicals, 1:30,000). From each animal, all sections displaying CT-B-positive fibers being able to enter the glial scar were analyzed. Axonal regeneration was quantified by counting the number of CT-B-labeled axons within the glial scar and by measuring the length of the longest fiber found rostral to the injury border. All quantifications of axon regeneration were performed with the researcher blinded for genotype and experimental group.

Regeneration of Raphespinal Fibers

Five weeks after complete spinal cord transection, free floating sections were immunostained for 5-HT (1:20,000, ImmunoStar). Only animals where a complete injury was present in all sections were analyzed. Axon regeneration was quantified by counting 5-HT positive fibers caudally to the injury site in all spinal cord sections. Raphespinal fiber sprouting was assessed 4 weeks after lateral spinal cord hemisection by 5-HT immunostaining in cross-sections of the lumbar enlargement. 5-HT immunoreactivity was quantified in four sections/animal using FeatureJ software. Compensatory sprouting was quantified by 5-HT immunoreactivity in the ipsilateral ventral horn. For each genotype, 5-HT immunoreactivity was normalized against 5-HT immunoreactivity in the naïve spinal cord. All quantifications of axonal regeneration were performed with the researcher blinded for genotype.

Crude Membrane Isolation

Crude membranes were isolated from the spinal cord of 6-week-old WT and shi littermates, as described [16]. Briefly, tissues were homogenized in 0.32 M sucrose and nuclei removed by centrifugation at 500 × g for 30 min. Membranes were pelleted by centrifugation at 100,000 × g for 1 h and resuspended in water and protein concentration was determined.

Myelin Isolation

Myelin was isolated from the spinal cord of 16-week-old WT mice as detailed [17]. Briefly, the tissue was homogenized in 0.32 M sucrose, and after centrifugation at 900 × g, the post-nuclear supernatant was collected and carefully overlaid on an ultracentrifuge tube containing a 0.85 M sucrose solution on top of a 50 % (w/v) sucrose cushion. After centrifugation for 1 h at 37,000 × g at 4 °C in a vertical rotor, the interphase between sucrose solutions was transferred to a new ultracentrifuge tube. Two rounds of osmotic shocks were performed by adding ice-cold water and centrifugation at 20,000 × g. The final myelin pellet was stored at −80 °C until further use. Before use, myelin was sonicated and the protein concentration determined.

Lipid Isolation

Lipids were isolated from myelin and spinal cord extracts by a modified Folch two-solvent system [18]. Briefly, 150 μg of protein extract (from either myelin or total spinal cord extract) was dissolved in methanol/water mixture (1:1), and lipids were separated by liquid extraction with chloroform (Merck). The chloroform layer containing the lipids was then dried under a nitrogen stream and stored at −20 °C. Absence of proteins in the lipid fraction was confirmed by Western blotting against Nogo, MAG, Ephrin B3, and RGMa. The aqueous layer containing the protein sample was precipitated with 0.2 M perchloric acid following Folch extraction. Following centrifugation, the protein was resuspended in water and sonicated and its concentration was determined.

Lipid Extraction and Thin Layer Chromatography

Thoracic spinal cords from WT and shi littermates were collected and lipids were extracted as described [19]. The spinal cord injury site and a rostral uninjured region of WT spinal cord treated with artificial CSF (aCSF) or HPβCD were also collected for lipid extraction. Acidic and neutral lipids were further separated by reverse-phase chromatography [20–22]. All lipids were analyzed by high performance thin layer chromatography (HPTLC) as detailed [20] alongside with lipid standards. The lipids analyzed were ceramide (Cer), phosphatidylserine (PS), cholesterol (CO), sulfatide (GS), galactocerebroside (GalCer), phosphatidylcholine (PC), triglycerides (TG), sphingomyelin (SpH), cholesteryl esters (CE), phosphatidylethanolamine (PE), globotetrahexosylceramide (Gb4), lactocerebroside (Lac), phosphatidylinositol (PI), and GM1 ganglioside (GM1).

Neuron Cultures and Inhibition Assays

Primary cultures of dorsal root ganglia (DRG) neurons from P6 mice were performed as described [23]. Glass coverslips were coated with poly-l-lysine (20 μg/mL) and laminin (0.5 μg/mL) followed by either crude membranes (0.1 μg protein), myelin (0.5 μg protein), myelin protein fraction (0.5 μg), MBP (Upstate, 0.9 μg), or total protein from spinal cord extracts (3 μg). To test the effect of lipids, the myelin lipid fraction (corresponding to 0.5 μg of protein) and lipids extracted from the spinal cord of either WT or shi mice (corresponding to 2 μg of protein) were used. To test whether lipid inhibition was not restricted to neurons collected from young animals, adult DRG neurons (collected from 8-week-old mice) were cultured on top of lipids from WT and shi spinal cords as described above. Coverslips were coated with poly-l-lysine followed by lipids in chloroform/methanol/water (2:1:0.1) and laminin (0.5 μg/mL). The amount of individual lipids present in 2 μg protein from WT spinal cord lysates was determined by HPTLC and is referred to as 1× (PI = 2 ng, GalCer = 200 ng, CO = 120 ng, GS = 120 ng, Gb4 = 24 ng, SpH = 50 ng, GM1 = 0.6 ng, Lac = 20 ng, PE = 40 ng, CE = 46 ng). In the analysis of single lipids, 1×, 10-fold less (0.1×), or 10-fold more (10×) were used per coverslip. Gb4, PE, SpH, and Lac were from Matreya LLC and CO, CE, GalCer, GS, GM1, and PI were from Sigma-Aldrich. GalCer, GS, and PI were dissolved in chloroform/methanol (2:1); CO, PE, SpH, Gb4, GM1, and Lac in chloroform/methanol (1:2), and CE was dissolved in chloroform/methanol (3:1). For each lipid, the corresponding solvent was used as control. To confirm the presence of lipids following coating, lipid extraction of the coverslip was performed, and lipid content was quantified. Lipid-coating efficacy was measured by HPTLC and approximately 90 % of the coated lipids were present in the coverslips (Supplementary Fig. S1). For each condition, 5000 cells/coverslip were plated in triplicate and immunostained 12 h later against βIII-tubulin (Promega, 1:2000). In at least 100 neurons/condition, the longest neurite was traced using NeuronJ. Scholl analysis was performed in Matlab with Synapse Detector (SynD) software [24]. Using SynD, the number of processes crossing concentric circles centered at the cell body, with radiuses of consecutive multiples of 25 μm, was quantified. Mean total neurite branching was determined from the analysis of each individual neuron image. Similar experiments were performed with cortical and hippocampal neurons isolated from E17.5 WT and shi embryos as described [25, 26], respectively. For cortical neurons, 50,000 cells/coverslip were maintained for 48 h and for hippocampal neurons, and 16,500 cells/coverslip were maintained for 72 h.

Evaluation of Toxicity

DRG neurons were plated on top of either solvent, CO or SpH. Twelve hours later, cells were fixed and immunostained for βIII tubulin and activated caspase 3 (Cell Signaling, 1:400). Ten times magnification pictures were taken and the percentage of viable cells (lacking active caspase 3 staining) was determined. As a positive control, neurons were incubated with chelerythrine, a drug that induces apoptosis, at 2 μM for 12 h [27]. In a complementary assay, 12 h after plating, cells were washed and incubated with 1 μg/mL calcein (Invitrogen) for 30 min followed by 10 μg/mL propidium iodide for 5 min. Cells were then washed and ×10 magnification pictures were taken. The percentage of viable cells (cells where nonfluorescent calcein AM is converted to a green-fluorescent calcein) was determined.

Rho Inactivation Assay

WT DRG neurons were plated onto coverslips coated with laminin, myelin, solvent, CO, or SpH, as described above. Thirty minutes before plating, cells were incubated with 1 μg/mL of C3 transferase (Cytoskeleton, Inc.), a Rho inhibitor that was maintained during the entire experiment. Twelve hours later, cells were fixed and neurite outgrowth was evaluated.

ROCK Activation Assay

Fifty thousand hippocampal neurons/coverslip were plated onto coverslips coated with either solvent or SpH in the presence of 1 μg/mL of C3 transferase, or vehicle solution, as described above. A pool of 14 wells/condition was used. Seventy two hours after plating, cells were lysed in 62.5 mM Tris pH 6.8 containing 2 % SDS, 12.5 % glycerol, and 5 % β-mercaptoethanol. Lysates were then immunobloted against phospho-Ser19-myosin light chain (MLC; Cell Signaling, 1:1000), the primary phosphorylation site of MLC by ROCK [28]; total MLC (Cell Signaling, 1:500); phospho-phosphatase and tensin homologue—PTEN (Cell Signaling, 1:2000), an alternative ROCK substrate [29, 30]; and total PTEN (Cell Signaling, 1:2000) and β-actin (Sigma-Aldrich, 1:5000).

HPβCD Treatment

C57BL/6 mice (n = 24) were subjected to SCI dorsal hemisection as described above. At the time of injury, osmotic minipumps (Alzet 2006) were placed subcutaneously with a tube allowing perfusion of the injury site at a rate of 0.15 μL/hour with either 27 μg/g/day of HPβCD in aCSF (128 mM NaCl, 2.5 mM KCl, 0.95 mM CaCl2, 1.9 mM MgCl2; n = 12), or vehicle (aCSF; n = 12). Besides delivery at the injury site, 4 mg/g of HPβCD was injected subcutaneously twice a week. Five weeks following injury, the SCI site was collected and lipids were extracted and quantified as described above (n = 6, HPβCD-treated mice; n = 6, aCSF-treated mice). In n = 6 HPβCD-treated mice and n = 6 aCSF-treated mice, dorsal column fibers were labeled with CT-B, and axon regeneration was assessed as previously described.

Data Analysis

Data is shown as mean ± SEM. For single comparisons, Student’s t test was used and for multiple comparisons, one-way ANOVA was chosen followed by Tukey’s or Bonferroni’s correction. When p < 0.05, data were considered statistically significant.

Results

Axon Regeneration Inhibitors Are Present in the Shi Spinal Cord and a Standard Glial Scar Is Formed Upon SCI

In shi mice, the absence of MBP leads to an almost complete absence of compact myelin in the CNS, including the spinal cord (Fig. 1a). Despite the absence of compact myelin, Western blot of spinal cord extracts showed that all canonical inhibitors are present in the shi spinal cord (Fig. 1b). Besides a 10- and 2.5-fold decrease in the levels of MAG and OMgp, respectively, either similar (ephrin B3 and RGMa) or increased (Nogo-A) levels of inhibitors were found (Fig. 1b, c). Immunohistochemistry against MAG and Nogo-A showed a normal distribution of the inhibitors in the shi spinal cord (Fig. 1d). Following complete spinal cord transection, WT and shi mice displayed an equivalent glial scar as evaluated by Masson trichrome staining and CSPG and GFAP immunostaining (Fig. 1e, f). Moreover, following SCI, all axonal regeneration inhibitors were present in shi spinal cords (Fig. 1g), being that MAG was the only one for which decreased levels (40 % decrease) were found (Fig. 1g, h). Interestingly, following injury, increased levels of the inhibitors Nogo, OMgp, and ephrin B3 (EphB3) were present in the shi spinal cord (Fig. 1g, h). In summary, despite the absence of compact myelin, the shi spinal cord contains myelin-associated axonal regeneration inhibitors and forms a regular glial scar following SCI.

Fig. 1
figure 1

Axonal regeneration inhibitors are present in the shi spinal cord. p-phenylene-diamine (PPD) staining of thoracic spinal cords of WT and shi mice (upper panels; scale bar = 200 μm). The dorsal funiculus of WT and shi spinal cord (middle panels; higher magnification of boxed regions in upper panel; scale bar = 100 μm) was used to obtain electron microscopy images of the white matter of WT and shi mice (lower panels; higher magnification of boxed regions in middle panel; scale bar = 1 μm) (a). Western blot of Nogo, MAG, OMgp, EphB3, RGMa, and GAPDH in spinal cord extracts of naïve uninjured WT and shi littermate mice. Representative results are shown (b). Quantification of results obtained for naïve spinal cords shown in (b) (n = 4 mice/genotype) (c). Immunohistochemistry of MAG and Nogo in the dorsal funiculus of spinal cords from WT and shi littermates (scale bar = 100 μm) (d). Masson trichrome staining and immunostaining against CSPG and GFAP in the spinal cord of WT and shi littermates 5 weeks following complete spinal cord injury (scale bars = Masson trichrome 200 μm, GFAP and CSPG 50 μm). Dashed line demarks the injury borders (e). Quantification of the injury area following Masson trichrome staining and immunohistochemistry against CSPG and GFAP (n = 7 mice/genotype) (f). Western blot of Nogo, MAG, OMgp, EphB3, RGMa, and GAPDH in spinal cord extracts of WT and shi littermate mice with SCI. Representative results are shown (g). Quantification of the results obtained following SCI shown in (g) (n = 4 mice/genotype) (h). Results represented as mean ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001

Full size image

Axon Regeneration and Sprouting Are Increased in Shi Mice

Although the shi spinal cord contains all the canonical axonal regeneration inhibitors, we asked whether the absence of compact myelin could have an impact in axonal regeneration in vivo. Following dorsal hemisection, and in contrast to WT axons, shi dorsal column tract axons entered the lesion site (Fig. 2a, b) and regenerated for longer distances (Fig. 2a, c). Of note, five out of six shi mice presented regenerating dorsal column axons, compared to two out of six WT animals. In the raphespinal tract, following complete spinal cord transection, shi 5-hydroxytryptamine (5-HT)-positive fibers found caudally to the injury site were more frequent and were capable of regenerating for longer distances (Fig. 2d, e). Of note, 80 % of shi mice presented more than 100 fibers caudally to the injury site, whereas none of the WT animals analyzed was able to display this regeneration profile. The ability of contralateral unlesioned fibers to sprout in the ipsilateral injury side following lateral hemisection was used as a measure of plasticity [6]. In the shi spinal cord, a 1.7-fold increase in 5-HT positive fibers was found in the ipsilateral side (WT, 33 ± 3 %; shi, 56 ± 8 %; p < 0.05; Fig. 2f). The correlation with a functional improvement was not possible to evaluate given the severe shi phenotype. Of note, MAG was the only MAI decreased in the shi spinal cord following SCI (Fig. 1h). However, as MAG ablation is insufficient to increase axonal regeneration [31], decreased MAG levels are unlikely to underlie the increased axonal regeneration of shi mice. Combined, our data show that following SCI, shi axons have increased axonal regeneration and sprouting.

Fig. 2
figure 2

Shiverer mice present increased axonal regeneration. Cholera toxin B (CT-B) immunohistochemistry in sagittal spinal cord sections of WT and shi littermates, 4 weeks following dorsal hemisection. Arrowheads highlight regenerating CT-B-positive dorsal column axons in shi spinal cords. R rostral, C caudal, D dorsal, V ventral; dashed line demarks the injury border. Lower panels are higher magnifications of the selected boxed regions (scale bar = 100 μm) (a). Number of CT-B-positive dorsal column axons able to regenerate through the glial scar in WT and shi animals (n = 6 mice/genotype) (b). Length of the longest regenerating CT-B-positive dorsal column axon in WT and shi animals (n = 6 mice/genotype) (c). Serotonin (5-HT) immunohistochemistry in shi and WT spinal cords 5 weeks following complete spinal cord transection showing regenerating raphespinal fibers. R rostral, C caudal, D dorsal, V ventral. Lower panels are higher magnifications of the selected boxed regions (scale bar = 100 μm) (d). Quantification of regenerating raphespinal fibers (WT n = 5, shi n = 7) (e). 5-HT immunohistochemistry of the ventral white matter of spinal cord cross-sections from WT and shi littermates (WT n = 6, shi n = 7) 4 weeks following lateral hemisection to assess sprouting (scale bar = 100 μm) (f). Results represented as mean ± SEM. *p < 0.05

Full size image

Myelin Lipids Inhibit Neurite Outgrowth and Branching

Despite the alterations described in shi axons [32–34] that are generally related to decreased regeneration capacity [35], neurite outgrowth of WT and shi DRG and cortical neurons was compared. No differences in axon length were found in the case of cortical neurons (Fig. 3a). For DRG neurons, no differences were found in neurite length or in branching (Fig. 3a and Supplementary Fig. S2a–c). These data suggest that the intrinsic growth capacity of shi neurons does not underlie their increased regeneration in vivo. To evaluate whether the shi spinal cord environment might be less inhibitory than that of WT mice, and as myelin cannot be isolated from the shi CNS, we tested the effect of crude membranes prepared from spinal cords of both strains. Although neurite length (Fig. 3b and Supplementary Fig. S2d) and branching (Supplementary Fig. S2e, f) of DRG neurons were inhibited in the presence of both WT and shi membranes, the inhibition produced by shi membranes was lower, suggesting that the shi spinal cord environment presented less inhibitory cues to axonal growth. The decreased inhibitory environment of the shi spinal cord was unrelated to the lack of MBP as no effect on neurite length (Fig. 3c and Supplementary Fig. S2g) or branching (Supplementary Fig. S2h) was produced when WT DRG neurons were plated on MBP. To further confirm that the decreased inhibitory environment of the shi spinal cord was unrelated to a protein component, WT DRG neurons were grown on top of protein extract from either WT or shi spinal cords. Both extracts were equally inhibitory of neurite length and branching (Fig. 3d and Supplementary Fig. S2k–m).

Fig. 3
figure 3

Myelin lipids inhibit neurite outgrowth. Neurite outgrowth of DRG and cortical neurons from WT and shi littermates (n = 3) (a). Neurite outgrowth of WT DRG neurons grown on 0.1 μg of protein from WT and shi crude membranes (n = 3) (b). Neurite outgrowth of WT DRG neurons grown on 0.9 μg MBP, 0.5 μg WT myelin, 0.5 μg WT myelin protein, or myelin lipids corresponding to 0.5 μg of WT myelin (n = 3) (c). Neurite outgrowth of DRG neurons grown on 3 μg of total protein from either WT or shi spinal cords (n = 3) (d). Neurite outgrowth of WT DRG neurons (from P6 animals) grown on lipids extracted from either WT or shi spinal cords (n = 3) (e). Neurite outgrowth of WT DRG neurons (from 8-week-old animals) grown on lipids extracted from either WT or shi spinal cords (n = 3) (f). Representative βIII-tubulin immunocytochemistry of (e) (scale bar 25 = μm) (g). Results represented as mean ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001

Full size image

Given that lipids account for approximately 75 % of myelin dry weight, we asked whether myelin lipids would contribute to the inhibitory properties of myelin. When WT DRG neurons were grown on myelin protein or myelin lipids as substrates, a decreased neurite outgrowth and branching were observed for both, demonstrating that myelin lipids also play a role as axonal regeneration inhibitors (Fig. 3c and Supplementary Fig. S2g–j). To further demonstrate that the different lipid content of the shi spinal cord creates a more permissive environment to axonal growth, WT DRG neurons were grown on spinal cord lipids extracted from either WT or shi mice. In contrast to lipids from WT spinal cords, lipids from shi spinal cords did not inhibit neurite outgrowth or branching (Fig. 3e, g and Supplementary Fig. S2n–p). The inhibitory effect of WT lipids was not developmentally dependent as it was still observed when using adult DRG neurons (Fig. 3f and Supplementary Fig. S2q–s). These data demonstrate that neurite outgrowth of WT DRG neurons can be modulated by exposure to different lipid milieus.

Cholesterol and Sphingomyelin Inhibit Axonal Growth Through a Rho-Dependent Mechanism

To identify the lipids that might underlie the increased axonal regeneration in the shi spinal cord, we compared the lipid composition of WT and shi spinal cords. Among others, the most abundant myelin lipids were analyzed, namely cholesterol, ceramide, phosphatidylethanolamine, phosphatidylcholine, phosphatidylinositol, phosphatidylserine, sphingomyelin, GM1 ganglioside, and galactocerebroside [36]. In the shi spinal cord, we observed a decrease of most lipids (namely cholesterol, sulfatide, galactocerebroside, sphingomyelin, phosphatidylethanolamine, globotetrahexosylceramide, lactocerebroside, and GM1 ganglioside), an increase of cholesteryl esters and phosphatidylinositol, and normal amounts of ceramide + phosphatidylethanolamine, phosphatidylserine, phosphatidylcholine, and triglycerides (Fig. 4a). From all the lipids with abnormal concentrations in the shi spinal cord, only cholesterol, cholesteryl ester, and sphingomyelin inhibited DRG neurite outgrowth (Fig. 4b, c; Supplementary Fig. S3a, d, g). Cholesterol, cholesteryl ester, and sphingomyelin also evoked a decreased branching (Supplementary Fig. S3b, c, e, f, h, i). This effect was unrelated to toxicity as both under control conditions or in the presence of either cholesterol or sphingomyelin, 95 % of neurons were alive when immunostained against active caspase 3 (Fig. 4d); in addition, and using a calcein viability assay, no dead cells were found in any condition (Supplementary Fig. S3r). The effect of both cholesterol and sphingomyelin was not restricted to DRG neurons as these lipids also produced a decreased axon length in cortical neurons (Fig. 4e). A similar inhibitory effect was observed for hippocampal neurons (data not shown). For all the remaining lipids analyzed (GS, GalCer, PE, Gb4, Lac, PI, and GM1), no significant effect was found in the length of the longest neurite (Supplementary Fig. S3j), mean total length (Supplementary Fig. S3k), number of branches (Supplementary Fig. S3l), or Scholl analysis (Supplementary Fig. S3m–q).

Fig. 4
figure 4

Cholesterol and sphingomyelin inhibit neurite outgrowth through a Rho-dependent mechanism. Quantification of lipids extracted from WT and shi spinal cords analyzed by high performance thin layer chromatography (n = 3 mice/genotype) (a). Effect of individual lipids in neurite outgrowth of DRG neurons (n = 3). For cholesterol (CO) 1× = 120 ng, for sphingomyelin (SpH) 1x = 50 ng, for cholesteryl ester (CE) 1× = 46 ng. Ten times less and ten times more are represented by 0.1× and 10× represent, respectively (b). Representative βIII-tubulin immunocytochemistry of DRG neurons grown on top of solvent (control), CO, and SpH coated coverslips (scale bar = 50 μm) (c). Representative images of DRG neuron cultures grown on top of solvent, CO, or SpH, immunostained for βIII tubulin (green) and activated caspase 3 (red) (scale bar = 50 μm); as a positive control, DRG neurons plated on top of solvent were treated with chelerytrin, a drug capable of inducing apoptosis [27]. Arrowheads indicate apoptotic DRG neurons (d). Axon length of cortical neurons plated on CO and SpH (n = 3) (e). Neurite outgrowth of DRG neurons plated on top of myelin, CO, or SpH in the presence or absence of the Rho inhibitor C3 transferase (n = 3) (f). Western blot against p-myosin light chain (p-MLC), total MLC, p-PTEN, total PTEN, and β-actin of hippocampal neurons grown on top of either solvent or SpH (g). Results represent the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ns nonstatistical

Full size image

The RhoA/ROCK pathway is the major mediator of myelin inhibition [11], and the inhibitory effect of myelin can be reverted by the Rho inhibitor C3 transferase [37]. To assess whether myelin lipids share similar pathways to block axonal regeneration than those described for myelin proteins, DRG neurons were grown on the inhibitory substrates myelin, cholesterol, and sphingomyelin and in the presence of C3 transferase. As expected, C3 was able to overcome myelin inhibition (Fig. 4f; Supplementary Fig. S4a, c). Inhibition by both cholesterol and sphingomyelin was also relieved by C3 treatment (Fig. 4f; Supplementary Fig. S4a, b, d, e) suggesting that lipid inhibition of axonal growth occurs, at least in part, through a Rho-dependent mechanism. Despite that treatment with C3 improved neurite outgrowth in DRG neurons plated onto cholesterol and sphingomyelin, given that the rescue of the longest neurite was complete in neurons plated on sphingomyelin, we evaluated the Rho-mediated signaling cascade by measuring the levels of phosphorylated myosin light chain (p-MLC). Increased p-MLC is the downstream outcome of the activation of the effector molecule of Rho, ROCK [38]. In the presence of sphingomyelin, p-MLC was increased, and similarly to the neurite outgrowth effect, C3 was able to revert the activation of p-MLC (Fig. 4g), further supporting a lipid-based Rho-mediated inhibition. Of note, the effect produced by SpH was not generalized to all ROCK targets as the phosphorylation of PTEN, another ROCK substrate [29, 30] was not increased in the presence of SpH (Fig. 4g).

Reduction of Lipid Levels in the Spinal Cord Injury Site Through 2-Hydroxypropyl-β-Cyclodextrin Delivery Promotes Axonal Sprouting of Dorsal Column Tract Axons

To further confirm the inhibitory role of cholesterol and sphingomyelin in vivo, WT mice with SCI were treated with 2-hydroxypropyl-β-cyclodextrin (HPβCD), a drug capable of reducing the levels of cholesterol and sphingolipids, as already demonstrated in models of Niemann-Pick type C [39–41] and Alzheimer’s disease [42]. When comparing the changes in lipid content induced by SCI in WT mice, cholesterol and sphingomyelin remained unchanged (as well as phosphatidylcholine, ceramide + phosphatidylethanolamine, globotetrahexosylceramide + sphingomyelin, galactocerebroside), whereas cholesteryl esters, phosphatidylethanolamine, phosphatidylinositol, and sulfatide + phosphatidylserine increased at the SCI site (Fig. 5a). At the SCI site and following HPβCD delivery, we observed decreased levels of most lipids including the inhibitory lipids cholesterol (28 % decrease) and sphingomyelin (21 % decrease) (Fig. 5b). Of note, in the uninjured spinal cord, administration of HPβCD did not decrease the levels of either inhibitory lipids or of the other lipids analyzed (Fig. 5c), supporting a lack of effect of HPβCD in normal tissues. The protein levels of Nogo, MAG, EphB3, and RGMa following cyclodextrin administration were measured in the injured spinal cord by Western blot, and no differences were observed with cyclodextrin treatment (Supplementary Fig. S5a, b). In HPβCD-treated mice, improved axonal sprouting of dorsal column fibers was obtained, with a 2.5-fold increased number of axons being able to enter the lesion site (Fig. 5d, e), and a 2-fold increased length of regenerating axons (Fig. 5d, f). All the animals treated with cyclodextrin displayed regenerating axons longer than 150 μm whereas none of the untreated mice presented axons of this length. Combined, this study identified cholesterol and sphingomyelin as new myelin lipids that inhibit axonal regeneration through the glial scar and determined that HPβCD may be used to reduce the levels of inhibitory lipids in the injury site and improve axonal sprouting of dorsal column fibers.

Fig. 5
figure 5

2-Hydroxypropyl-β-cyclodextrin delivery promotes axonal regeneration following SCI. Quantification of lipids extracted from uninjured or injured spinal cord by high performance thin layer chromatography (n = 6 mice/condition) (a). Quantification of lipids in the spinal cord injury site from aCSF- and HPβCD-treated mice, 5 weeks following dorsal hemisection (n = 6 mice/condition) (b). Quantification of lipids in uninjured spinal cord from aCSF- and HPβCD-treated mice, 5 weeks following dorsal hemisection (n = 6 mice/condition) (c). CT-B immunohistochemistry in sagittal spinal cord sections of aCSF- and HPβCD-treated animals, 5 weeks following dorsal hemisection. Arrowheads highlight regenerating CT-B-positive dorsal column axons in HPβCD-treated spinal cords. R rostral, C caudal, D dorsal, V ventral; dashed line demarks the injury border. Lower panels are higher magnifications of the selected boxed regions in the upper panels (scale bar = 100 μm) (d). Number of CT-B-positive dorsal column axons able to regenerate through the glial scar in aCSF- and HPβCD-treated mice (n = 6 mice/condition) (e). Length of the longest regenerating CT-B-positive dorsal column axon in aCSF- and HPβCD-treated mice (n = 6 mice/condition) (f). Results represent the mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001

Full size image

Discussion

Our work shows that cholesterol, its esters, and sphingomyelin are myelin-associated lipid inhibitors that modulate axon regeneration, and demonstrates that HPβCD delivery reduces the levels of lipids in the SCI site promoting axonal regrowth of dorsal column axons. Lipids represent 75 % of myelin dry weight. From these, cholesterol is the most abundant, accounting for 28 % of the myelin lipid content, whereas sphingomyelin accounts for 4 %. Although cholesterol is an essential component of the plasma membrane crucial for axonal growth, cholesterol accumulation in the mammalian brain is a risk factor for neurodegenerative diseases including Alzheimer’s disease [43] and Niemann-Pick [41]. In the scenario of injury, with the presence of myelin debris within the injury site, myelin-derived cholesterol is presented to regenerating axons in a different form from that of glial-derived cholesterol. Whereas glial-derived cholesterol can stimulate axon growth by providing lipoproteins as a source of both cholesterol and apolipoprotein E to regenerating axons, free cholesterol, i.e., without the context of a lipoprotein particle, as is the case of myelin-derived cholesterol, fails to enhance axon extension [44]. In the case of sphingomyelin, its accumulation, which is the hallmark of Niemann-Pick disease, leads to changes in plasma membrane that similarly to cholesterol, also culminate in neurodegeneration [45]. As such, we propose that following CNS injury, when growing axons need to elongate through an environment filled with myelin debris, exposure to free cholesterol and to sphingomyelin contributes to axonal growth inhibition. This mechanism may also underlie, at least in part, the increased axonal regeneration in the peripheral nervous system (PNS). Despite that peripheral and central myelin have similar cholesterol and sphingomyelin content, as dedifferentiated Schwann cells and invading macrophages are able to rapidly phagocyte and clear myelin debris following PNS injury [46], regenerating axons are probably not exposed to a lipid-rich environment, in contrast to the CNS where such clearance is not as effective.

Accumulation of unesterified cholesterol and sphingomyelin is the hallmark of Niemann-Pick disease, which culminates in neurodegeneration [41, 45]. Interestingly, cultured neurons from mice lacking Niemann-Pick type C1 protein (npc1 knockout mice) displayed increased rate of growth cone collapse that was mediated by ROCK activation and reverted by ROCK inhibition [47]. In npc1 mice, administration of HPβCD delays the onset of clinical symptoms by reducing the buildup of cholesterol and sphingolipids within the nervous tissue [41, 48]. HPβCD is a well-tolerated FDA-approved drug used in animal models and in human clinical trials. Here, we show that following HPβCD delivery, decreased levels of inhibitory lipids, namely cholesterol, cholesteryl esters, and sphingomyelin, are specifically obtained in the SCI site whereas uninjured spinal cord segments have unaltered lipid composition. Upon spinal cord injury, HPβCD may exert a generalized sequestering effect allowing the removal of lipids from the injury site, and thus relieve the inhibitory level within the glial scar. Although HPβCD did not have a defined lipid specificity, we did not observe any lipid changes in the treated uninjured spinal cord, indicating that this nontoxic agent does not lead to lipid dysregulation in normal tissues. In addition, HPβCD treatment led to increased axon regeneration of dorsal column axons, supporting that delivery of compounds capable of decreasing the lipid content specifically in the injury site should be considered as a therapeutic option in the context of SCI.

At the molecular level, inhibition induced by cholesterol and sphingomyelin could be reverted by the Rho inhibitor C3 transferase. In the case of sphingomyelin, for which a stronger effect was observed with C3 transferase, we further demonstrated the downstream activation of ROCK as increased phosphorylated levels of its substrate, MLC, were generated. These data supports that both myelin proteins and lipids signal inhibition through similar pathways involving Rho activation. Cholesterol and sphingomyelin may act as ligands to induce receptor-mediated activation of the Rho signaling cascade, or alternatively they can augment receptor activation leading to increased Rho activity. Whether cholesterol and sphingomyelin engage the same receptors (e.g., Nogo receptor and PirB) as several structurally unrelated myelin proteins is to be unraveled. Specific binding sites for cholesterol have been identified in G protein-coupled receptors [49, 50]. As such, cholesterol and/or sphingomyelin at the extracellular matrix in which axons are growing may interact and engage other receptors eliciting a signaling cascade that impairs axonal growth. Sulfatide has also been identified as a lipid that specifically inhibits neurite outgrowth of retinal ganglion cells through a Rho-mediated mechanism [51], although the pathway by which this occurs remains poorly understood. The next challenge will be to further characterize the pathways by which lipids mediate the repression of neurite outgrowth.

In summary, our work shows that myelin lipids are important modulators of axon regeneration that should be considered as critical targets in strategies aiming at improving axonal growth following injury.