Aberrant Upregulation of Astroglial Ceramide Potentiates Oligodendrocyte Injury

Abstract

Oligodendroglial injury is a pathological hallmark of many human white matter diseases, including multiple sclerosis (MS) and periventricular leukomalacia (PVL). Critical regulatory mechanisms of oligodendroglia destruction, however, remain incompletely understood. Ceramide, a bioactive sphingolipid pivotal to sphingolipid metabolism pathways, regulates cell death in response to diverse stimuli and has been implicated in neurodegenerative disorders. We report here that ceramide accumulates in reactive astrocytes in active lesions of MS and PVL, as well as in animal models of demyelination. Serine palmitoyltransferase, the rate‐limiting enzyme for ceramide de novo biosynthesis, was consistently upregulated in reactive astrocytes in the cuprizone mouse model of demyelination. Mass spectrometry confirmed the upregulation of specific ceramides during demyelination, and revealed a concomitant increase of sphingosine and a suppression of sphingosine‐1‐phosphate, a potent signaling molecule with key roles in cell survival and mitogenesis. Importantly, this altered sphingolipid metabolism during demyelination was restored upon active remyelination. In culture, ceramide acted synergistically with tumor necrosis factor, leading to apoptotic death of oligodendroglia in an astrocyte‐dependent manner. Taken together, our findings implicate that disturbed sphingolipid pathways in reactive astrocytes may indirectly contribute to oligodendroglial injury in cerebral white matter disorders.

INTRODUCTION

Oligodendrocytes (OLs) are essential for the function of the vertebrate central nervous system (CNS) by forming myelin that enables saltatory conduction of nerve impulses. Myelination is a highly regulated multistep process that occurs in the human CNS primarily in late prenatal and early postnatal life 9, 36. Disruption of OL lineage progression and interaction with axons, or damage to myelin/OLs underlies many human neurological disorders, including multiple sclerosis (MS) and periventricular leukomalacia (PVL). MS is the most common inflammatory demyelinating disease of the CNS, and is pathologically characterized by inflammatory cell infiltration, OL death, demyelination, axonal damage and gliosis 39, 57. In relapses and remitting MS, certain levels of remyelination occur, but this spontaneous repair capability diminishes over time (18). PVL, on the other hand, is the principal injury of the perinatal brain that underlies cerebral palsy and cognitive deficits associated with prematurity (60). Pathologically, it is characterized by premyelinating OL injury, hypomyelination, focal periventricular necrosis, and diffuse gliosis and microglial activation in the immature cerebral white matter. In both entities, reactive astrocytes have long been recognized as a cellular feature of the pathology, but have been largely ignored on their potential contributions to OL damage. In the present study, we sought to determine if ceramides, which we show here for the first time accumulate in reactive astrocytes in both disorders, play a role in oligodendroglial injury.

Specific sphingolipid metabolites such as ceramide, sphingosine (Sph) and sphingosine‐1‐phosphate (S1P) have emerged as a new class of bioactive lipids that function as second messengers by binding to specific intracellular targets 17, 22, 37. They also function as signaling molecules capable of acting in a paracrine manner, regulating cell growth, differentiation, survival, migration and immune responses. Ceramide is at the center of interconnected metabolic pathways of sphingolipids (Supporting Information Figure S1), and is synthesized de novo from serine and palmitoyl CoA, or produced from sphingomyelinase (SMase)‐catalyzed degradation of sphingomyelin. Breakdown of ceramide via ceramidase (CDase) produces Sph, which can be phosphorylated to S1P via Sph kinases or recycled back to ceramide. Ceramide is generated in response to diverse stimuli, including proinflammatory cytokines, lipopolysaccharide and oxidative stress, and could potentially act as a shared downstream signaling molecule in cell injury common to many otherwise diverse disorders. As oligodendroglial damage is a key factor in the pathogenesis of MS and PVL, we examined in this study whether ceramide metabolism is dysregulated using post‐mortem tissues from MS and PVL patients, as well as in animal models of demyelination and remyelination. We report here that ceramide is aberrantly accumulated in reactive astrocytes and that ceramide potentiates OL damage in an astrocyte‐dependent manner. Our data suggest that reactive astrocytes may directly contribute to OL injury.

MATERIALS AND METHODS

Materials

Dulbecco's modified Eagle medium (DMEM), Hanks' balanced salt solution (HBSS) and glial fibrillary acidic protein (GFAP) antibody were obtained from Invitrogen (Carlsbad, CA, USA). Platelet‐derived growth factor (PDGF) and basic fibroblast growth factor (bFGF) were from PeproTech (Rocky Hill, NJ, USA). Antibody against Iba1 was purchased from Wako Chemicals (Richmond, VA, USA). Antibodies against Olig2 and 2′, 3′‐cyclic nucleotide 3′‐phosphodiesterase (CNPase) were from Millipore (Billerica, MA, USA). Active caspase‐3 antibody was from Cell Signaling (Danvers, MA, USA). CD68 antibody was from Dako (Glostrup, Denmark). Mouse anti‐ceramide IgM antibody (MAS0013) was obtained from Glycobiotech GmbH (Küekels, Germany), and mouse anti‐ceramide monoclonal IgM antibody (clone MID 15B4) was from Alexis Biochemicals (Plymouth Meeting, PA, USA). Antibodies against serine palmitoyltransferase (SPT), mouse acid CDase (aCDase) and NeuN were from Cayman Chemical (Ann Arbor, MI, USA). Normal mouse IgM was purchased from Jackson ImmunoResearch Lab (West Grove, PA, USA). N‐acetylsphingosine (C2‐ceramide), N‐hexanoyl‐d‐sphingosine (C6‐ceramide) and N‐oleoylethanolamine (NOE) were obtained from Sigma (St. Louis, MO, USA). C18‐ceramide was purchased from Avanti Polar Lipids (Alabaster, AL, USA). Tumor necrosis factor (TNF) was obtained from R & D Systems (Minneapolis, MN, USA). Unless specified otherwise, all other reagents were from Sigma.

Human tissues and sample preparation

Frozen and formalin‐fixed brain tissues from patients with clinically diagnosed and neuropathologically confirmed MS and from controls were obtained from the Rocky Mountain MS Center Tissue Bank (Englewood, CO, USA) and the Human Brain and Spinal Fluid Resource Center (Los Angeles, CA, USA). Formalin‐fixed human PVL and age‐related control brain specimens were from archived autopsy cases at Children's Hospital Boston, MA, USA. Frozen tissue specimens were bisected with one‐half used for tissue homogenization for sphingolipid analysis, and one‐half for histology and immunostaining examination upon post‐fixation with 4% paraformaldehyde for 1 day followed by cyroprotection with 30% sucrose in 0.1 M PBS. The fixed tissues were cryosectioned at 10 µm. Tissue homogenization was carried out in 0.05 M phosphate buffer (pH 7.0) containing 10 mM sodium pyrophosphate, 10 mM sodium fluoride, 1 mM sodium vanadate, 80 µM aprotinin, 104 µM AEBSF, 4 µM bestatin, 1.4 µM E‐64, 2 µM leupeptin, 1.5 µM pepstatin A and 100 µg/mL phenylmethylsulfonyl fluoride.

Mouse models of demyelination

Age‐matched C57BL/6 mice were used in two distinct mouse models of demyelination. The mice were housed in standard animal housing facility in accordance with NIH guidelines for Care and Use of Laboratory Animals. Food and water were provided ad libitum. Animal protocols were approved by the Texas A&M University (cuprizone model) and the Uniformed Services University of the Health Sciences [experimental autoimmune encephalomyelitis (EAE) model] Institutional Animal Care and Use Committees. Cuprizone‐induced demyelination was achieved by feeding 8‐week‐old male C57BL/6 mice cuprizone milled into normal chow at a concentration of 0.2% (Harlan‐Teklad, Madison, WI, USA) for 5 or 3 weeks when indicated. At the end of cuprizone feeding, about half of the mice were killed to examine demyelination and the other half were returned to normal diet for an additional 2 weeks to examine remyelination as described (Steelman et al, submitted). Age‐matched control mice were kept on normal diet throughout the duration of the cuprizone experiment. EAE was induced in female C57BL/6 mice between 7 and 9 weeks of age with an emulsion injected subcutaneously at two sites, one at each of the hind flanks (0.1 mL per site). The injections to the hind flanks yielded a total of 200 µg myelin oligodendrocyte glycoprotein (MOG) (NeoMPS, San Diego, CA, USA) peptide 35–55 in complete Freund's adjuvant (CFA) (DIFCO Laboratories, Detroit, MI, USA) with 500 µg Mycobacterium tuberculosis (DIFCO). Control mice (sham) were injected with CFA without the MOG peptide. Immediately following CFA injections, and 24 h later, the mice were injected intraperitoneally with 200 ng of inactivated Pertussis toxin (List Biologicals, Campbell, CA, USA). One week after induction of EAE, the mice received a booster of MOG peptide in incomplete Freund's adjuvant (IFA) and the control mice were injected with IFA without the MOG peptide. The animals were scored for clinical signs from day 10 onward and neurological signs were assessed as: 0, normal; 1, piloerection, tail weakness; 2, tail paralysis; 3, tail paralysis plus hind limb weakness/paralysis; 4, tail, hind and forelimb paralysis; and 5, moribund/dead, using increments of 0.5 points for intermediate clinical findings (32). The animals (n = 3 per group) were killed 2 weeks after EAE induction, and all received a clinical score of 2.

Primary cell cultures

Primary OLs, astrocytes and mixed glial cultures were prepared from forebrains of 1‐ to 2‐day‐old Sprague‐Dawley rat pups as described previously 40, 41. Briefly, forebrains free of meninges were dissociated and plated onto poly‐d‐lysine‐coated T75 flasks or directly into 24‐well plates for experiments using mixed glia, and fed every other day for 7–10 days. The flasks were pre‐shaken for 1 h at 200 rpm to remove lightly attached microglia followed by overnight shaking to separate OLs from the astrocyte layer. Isolated OLs were plated either by themselves or onto poly‐dl‐ornithine (PO)‐coated 24‐well plates containing established astrocytes for co‐cultures. Cells were maintained in a serum‐free basal‐defined medium (BDM) (DMEM with 0.1% bovine serum albumin, 50 µg/mL human apo‐transferrin, 50 µg/mL insulin, 30 nM sodium selenite, 10 nM d‐biotin and 10 nM hydrocortisone) containing 10 ng/mL PDGF‐AA and 10 ng/mL basic FGF for 5–9 days. In OL monocultures, contamination of astrocytes and microglia was <2%. Astrocytes were purified (>95%) from the astrocyte layer in the flask after being exposed to a specific microglia toxin l‐leucine methyl ester (1 mM) for 1 h and subcultured one to two times. For regular astrocyte and OL co‐cultures, astrocytes (2.4 × 10⁵ cells per well) were plated into 24‐well culture plates 1∼2 days before seeding OLs (6 × 10⁴ cells per well), and the co‐cultures were used within 2–3 days. For bridged, no cell contact co‐cultures of astrocytes and OLs, OLs were cultured on PO‐coated coverslips on which 1‐mm‐tall PDMS mini‐columns were attached at four corners as described (35). The coverslips were then transferred to 24 wells containing astrocytes.

Cell treatment and survival determination

Cells were treated in triplicate with C2‐, C6‐, C18‐ceramide or the CDase inhibitor NOE in the presence or absence of TNF as specified in figure legends. OL survival was analyzed 40–48 h later by immunocytochemistry and counting of O4⁺ cells with normal nuclei as described earlier (41). The total number of cells was revealed by staining all nuclei with bisbenzimide. Five to ten random consecutive fields were counted in each coverslip under 200× magnification with a total of >1000 cells counted in the control conditions. Cell survival is expressed as mean ± SEM.

Mouse tissue preparation for immunohistology analyses

Mice were anesthetized and intracardially perfused with PBS followed by 4% paraformaldehyde. Brains (cuprizone mice) and spinal cords (EAE mice) were dissected out, post‐fixed in 4% paraformaldehyde overnight at 4°C and cryoprotected in 30% sucrose until sunk. The fixed tissue was embedded in Tissue Tek OCT (Sakura, Torrance, CA, USA) and stored at −80°C. Sagittal or coronal sections of mouse brains or transverse sections of lumbar spinal cords were cut at 20 and 14 µm respectively, and mounted onto SuperFrost^® Plus slides (Fisher Scientific, Pittsburgh, PA, USA) for immunohistochemical analysis.

Mouse brain sample preparation for RNA, protein and lipid extraction

Mice were transcardially perfused with 20 mL of sterile PBS. The brain was removed, and the corpus callosum as well as portion of cingulate gyrus, external capsule and striatum was quickly but meticulously dissected using a sterile razor blade. Each piece of tissue was flash frozen in liquid nitrogen and stored at −80°C. The dissected brain regions were used for quantitative RT‐PCR (qRT‐PCR), Western blotting analysis and sphingolipid quantification.

Quantification of endogenous sphingolipids by mass spectrometry

Human autopsy tissues and dissected animal brain tissues were homogenized. An equal amount of tissue homogenate (1 mg proteins per sample) was fortified with internal standards, extracted with ethyl acetate/isopropanol/water (60:30:10, v/v/v) and analyzed by a high‐performance liquid chromatography–tandem mass spectrometry (HPLC–MS/MS) system at the Lipidomics Shared Resource Center at Medical University of South Carolina as detailed previously (8). Quantitative analysis of sphingolipids was based on calibration curves generated by spiking an artificial matrix with known amounts of target analytes and standards. Specific sphingolipid levels were normalized based on the level of total phospholipids in human samples or the amount of proteins in animal samples. Total phospholipids were determined using EnzyChrome™ Phospholipid Assay Kit (BioAssay Systems, Hayward, CA, USA).

qRT‐PCR analysis of neutral SMase (nSMase) and acid SMase (aSMase) expression

qRT‐PCR was used to determine the effects of cuprizone on nSMase and aSMase transcripts in the CNS. Briefly, RNA was extracted with TRI reagent (Sigma), quantified by NanoDrop™ 1000 spectrophotometer (ThermoFisher Scientific, Rockford, IL, USA), treated with DNase I (Invitrogen) and reverse transcribed using a Promega AMV Reverse Transcriptase Kit (Madison, WI, USA) and random oligonucleotide primers. Specific cDNAs were amplified and analyzed using a 7500 Real‐Time PCR System (Applied Biosciences, Foster City, CA, USA). Primers used were nSMase, forward‐GAGTTCCACGTCTGCTGTGA, reverse‐TTAGCACGCTGATCAAATCG; aSMase, forward‐TTCCCGAGTGCTGCTTATCT, reverse‐CTATCCAGAGCACTGAGCCC; and β‐actin, forward‐AGACTTCGAGCAGGAGATGG, reverse‐CCATCATGAAGTGTGACGTTG. All samples were run in duplicate or triplicate, and gene expression was normalized to β‐actin and fold expression was calculated using the formula 2^−ΔΔCt.

SMase activity assay

The activity of SMases was measured using the Amplex^® (10‐acetyl‐3,7‐dihydroxyphenoxazine)‐Red Sphingomyelinase Assay Kit (Molecular Probes, Carlsbad, CA, USA). Brain tissue extracts prepared as above were diluted in lysis buffer [0.05 M phosphate buffer (pH 7.0), 0.1% Triton X‐100, 80 µM aprotinin, 104 µM AEBSF, 4 µM bestatin, 1.4 µM E‐64, 2 µM leupeptin, 1.5 µM pepstatin A and 100 µg/mL phenylmethylsulfonyl fluoride] and incubated on ice for 30 minutes. After centrifugation at 17 000 × g for 15 minutes, supernatants were collected, and protein concentrations were determined using Bio‐Rad DC protein assay reagents (Hercules, CA, USA). Then, 100 µL of each sample (400 µg proteins) was added to a 100 µL assay mixture containing 100 µM Amplex Red reagent, 2 U/mL horseradish peroxidase (HRP), 0.2 U/mL choline oxidase, 8 U/mL alkaline phosphatase and 0.5 mM sphingomyelin in reaction buffer [0.1 M Tris–HCl, 10 mM MgCl₂ (pH 7.4)]. After 30 minutes incubation in dark at 37°C, the fluorescence was measured with a plate reader (FLUOstar Optima, BMG Labtech Inc., Offenburg, Germany) using excitation λ = 544 nm and emission λ = 590 nm. The values were corrected from background signal determined by omitting tissue extracts.

For aSMase activity measurement, tissue extracts were diluted in acidic buffer solution [final concentrations, 50 mM sodium acetate (pH 5.0), 0.1% Triton X‐100, 1 µg/mL aprotinin, 1 mM ethylenediaminetetraacetic acid (EDTA) and 100 µg/mL phenylmethylsulfonyl fluoride] and incubated on ice for 30 minutes. The aSMase activity was then assayed by a two‐step reaction system. First, to generate phosphocholine and ceramide, 0.5 mM sphingomyelin was added to the samples and incubated for 60 minutes at 37°C. Next, the reaction was placed on ice and 100 µL of 100 µM Amplex Red reagent, 2 U/mL HRP, 0.2 U/mL choline oxidase and 8 U/mL of alkaline phosphatase in reaction buffer (0.1 M Tris–HCl, pH 8.0) was added. The samples were further incubated for 30 minutes at 37°C and the fluorescence was determined as above.

Immunohistochemistry, terminal deoxynucleotidyl transferase‐mediated dUTP nick‐end labeling (TUNEL) and immunofluorescence microscopy

For paraffin‐imbedded tissue sections, antigen retrieval was carried out by autoclaving the tissue sections for 20 minutes in sodium citrate buffer [10 mM sodium citrate, 0.05% Tween20 (pH 6.0)] after deparaffinization. For frozen tissue sections, no antigen retrieval was needed. Tissue sections were rinsed with TBS [50 mM Tris–HCl (pH 7.4), 150 mM NaCl] and blocked with TBS‐T (0.1% Triton X‐100) containing 5% goat serum for 1 h at room temperature. Sections were then incubated overnight at 4°C with antibodies against Iba1 (1:2000), GFAP (1:1000), active caspase‐3 (1:1000), CNPase (1:100), human CD68 (1:100), mouse SPT (1:100), ceramide (monoclonal IgM, MID15B4, 1:50) or ceramide (mouse IgM, MAB0013, 1:50), or normal mouse IgM (4 µg/mL). For colorimetric detection of specific immunolabeling, the Vectastain^®ABC kit (Vector Laboratories, Burlingame, CA, USA) was used. For immunofluorescence staining, secondary antibodies conjugated with Alexa Fluor^® 488 or Alexa Fluor^® 594 (1:1000; Invitrogen) were used. Nuclei were stained with bisbenzimide as described before (41). As adult human brain tissue sections exhibited autofluorescence, before mounting the stained tissue sections with FluoroMount‐G™ (Southern Biotech, Burmingham, AL, USA), autofluorescence was quenched with 0.5% Sudan black in 70% ethanol. Additionally, because Sudan black stains neutral lipid and myelin, this step allows visualization of immunostaining and myelin in the same sections. In some cases, oil red O was used for detecting myelin. To label fragmented DNA, TUNEL was performed using an in situ cell death detection kit (Roche, Indianapolis, IN, USA). All images were captured with a fluorescence microscope equipped with an Olympus DP70 digital camera (model IX71; Olympus, Tokyo, Japan).

Negative controls in which primary antibodies were omitted did not show signals in cuprizone, EAE, PVL and MS tissue sections. Nor was there significant signal if the ceramide primary antibodies were replaced with normal mouse IgM antibodies. To semi‐quantify ceramide immunoreactivity in human brain sections, multiple fluorescence images were taken from three different regions of cerebral white matter for each case using same image acquisition settings including exposure times. For control cases, images were taken from random regions of the white matter. For MS samples, images were taken from lesion‐surrounding areas. Relative fluorescence intensity was then visually scored in a blinded fashion by two independent investigators, and the scores were consistent with that calculated using the ImageJ software (US National Institutes of Health, Bethesda, MD, USA, http://rsb.info.nih.gov/ij). Ceramide immunoreactivity was classified qualitatively and semi‐quantitatively by two independent investigators as follows: −, minimal signal; +, weak to moderate; ++, moderate to strong; and +++, strong based on fluorescence intensity and the density of ceramide‐positive cells.

Western blot analysis

Tissue homogenates were separated by 10–15% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred to PVDF membranes. The membranes were blocked with 5% nonfat milk or bovine serum albumin in TBS‐T (0.1% Tween 20) for 1 h, and then incubated with primary antibodies against SPT (1:2000) and aCDase (1:500) overnight at 4°C. Following washings, the membranes were incubated with HRP‐conjugated secondary antibody, and immunoreactive bands were visualized by chemiluminescence using SuperSignal^® detection reagents (ThermoFisher Scientific). Western blot images were acquired using Molecular Imager ChemiDoc™ XRS+ System (Bio‐Rad) and in some cases specific bands were quantified using the Quantity One^® software (Bio‐Rad) and normalized against β‐actin.

Statistical analysis

Data were expressed as mean ± SEM and analyzed by one‐way analysis of variance (ANOVA) followed by Bonferroni's post hoc test to determine statistical significance. Comparison between two experimental groups was based on two‐tailed t‐test. We considered P < 0.05 as statistically significant.

RESULTS

Ceramide is aberrantly accumulated in reactive astrocytes around/in actively demyelinating lesions in MS and in animal models of demyelination

We and others have demonstrated that premyelinating OLs and mature OLs are variably susceptible to injury induced by diverse stimuli such as proinflammatory cytokines and reactive oxygen/nitrogen species, and to excitotoxicity 10, 25. As intracellular ceramide can also be produced in response to diverse stimuli and is well documented for its capability to directly induce cell death (30), we reasoned that elevated ceramide may represent a common signal mediating OL injury. Interestingly, in actively demyelinating MS lesions, as judged by myelin breakdown and active phagocytosis of myelin debris by macrophages (Figure 1A–C), specific ceramide immunoreactivity was observed in GFAP⁺‐reactive astrocytes in and around the boarder of demyelinating plaques (Figure 1A,D,E,J), but not in normal‐appearing white matter (Figure 1K) or in the white matter from controls (not shown). The ceramide immunoreactivity was confined to astrocytes, exhibited punctuate staining patterns and did not co‐localize to activated microglia/macrophages (Figure 1F–H). Staining of adjacent serial tissue sections with an IgM isotype control antibody did not reveal any labeling (Figure 1I). Two specific anti‐ceramide antibodies that have been demonstrated by others to be specific for endogenous ceramide 14, 19, 49, 59, 61 were used. Both antibodies showed similar staining patterns in the cerebral white matter, and labeled reactive astrocytes in the lesions (Supporting Information Figure S2). Of note, ceramide immunoreactivity was also found in some blood vessels (Supporting Information Figure S2). Importantly, TUNEL⁺ cells were observed around the demyelinating lesion edge where ceramide⁺ astrocytes were also prominent in adjacent serial sections (Figure 1L). The TUNEL⁺ cells expressed OL marker CNPase, indicating damaged OLs in demyelinating lesions (Figure 1M).

Figure 1.

Aberrant accumulation of ceramide in reactive astrocytes in demyelinating lesions of human multiple sclerosis (MS). A. A representative image of actively demyelinating lesions of autopsy MS brains stained for myelin with Sudan black. Myelin debris (B) is evident at the lesion edge. C. Microglia/macrophages phagocytosing myelin debris at the edge of active MS lesions. Microglia/macrophages are identified by Iba1 staining (green). Myelin is visualized by oil red O (ORO) staining, and nuclei by bisbenzimide (Bis, blue). D–E. Double staining of the same brain section with antibodies specific for ceramide (red) and glial fibrillary acidic protein (GFAP) (green) shows strong ceramide signal in reactive astrocytes in and peri‐active MS lesions. The ceramide immunoreactivity is localized to GFAP⁺ astrocytes (F), but not to Iba1⁺ microglia/macrophages (G), and exhibits punctuate staining pattern (H). Immunostaining of serial sections with isotype control antibody is negative (I). J–K. Significant ceramide immunoreactivity in reactive astrocytes surrounding demyelinating lesions, but not in normal‐appearing white matter (NAWM) of the same cases. Representative micrographs of double immunostained sections are shown. L–M. Serial sections stained for terminal deoxynucleotidyl transferase‐mediated dUTP nick‐end labeling (TUNEL) (L, arrows) or double stained for TUNEL (brown) and oligodendrocyte (OL) marker CNPase (blue) (M), showing damaged OLs (arrows) at the border of the active MS lesion. Arrowhead indicates a TUNEL^‐/CNPase⁺ OL (M). Scale bars: A,D,E, 100 µm; B,C,H, 10 µm; F,G,I, 20 µm; J–M, 50 µm.

Immunohistological analysis of 8 control and 13 MS cases revealed ceramide immunoreactivity in reactive astrocytes in 10 MS cases (Table 1). Because of the known heterogeneity of MS neuropathology (43), we next examined the above findings using two well‐established animal models of demyelination. EAE is a T cell‐mediated disease of the CNS with inflammatory demyelinating lesions similar to type I and II MS lesions 43, 56. In contrast, cuprizone intoxication of rodents results in OL apoptosis, prominent gliosis and demyelination of specific cerebral white matter regions (eg, corpus callosum) independent of T cells 44, 46, features observed in type III and IV MS lesions 43, 49, 56. We found that ceramide immunoreactivity was indeed markedly upregulated in astrocytes, but not microglia/macrophages in demyelinating lesions in EAE mice (Figure 2A–C) and in the caudal corpus callosum of cuprizone‐fed mice (Figure 2D–H). OLs in demyelinating cerebral white matter did not appear to accumulate ceramide, as determined by double immunohistochemistry for ceramide and Olig2, a transcription factor of oligodendroglial lineage cells (Figure 2I). Other OL markers such as CC1 and CNPase were not used for double labeling with ceramide because the antibodies were all generated from mice. Five‐week cuprizone feeding caused reproducible demyelination in multiple specific brain regions in addition to the corpus callosum. In all demyelinating regions examined including cerebellum, cortex and hippocampus, but not in controls, we found ceramide accumulation in reactive astrocytes (Supporting Information Figure S3). Consistent with these observations, increased ceramide production in astrocytes was also observed in mice intoxicated for 3 weeks (Supporting Information Figure S4), when OL apoptosis peaked in the corpus callosum (46) (data not shown). Taken together, our results demonstrate aberrant upregulation of ceramide in reactive astrocytes during demyelination in MS and in animal models of MS.

Table 1.

Summary of clinicopathologic information and neuropathologic findings on multiple sclerosis (MS) and control cases.

Diagnosis	Case no.	Age/sex	Clinical diagnoses or cause of death	Microscopic classification*	Ceramide signals in astrocytes†
Control	C‐#1	86/F	nk	−	−
	C‐#2	62/F	Lewy body disease, dementia	−	+
	C‐#3	74/M	Progressive supranuclear palsy	−	−
	C‐#4	?/M	Respiratory failure, sepsis	−	−
	C‐#5	19/M	Gun shot (chest/abdomen)	−	−
	C‐#6	15/M	Normal, accidental hanging	−	−
	C‐#7	31/M	Anemia, pancytopenia, obesity	−	+
	C‐#8	29/M	Normal, motor cycle accident	−	+
MS	MS‐#1	69/M	MS	Infl.−/Demyel.−	++
	MS‐#2	44/M	MS (chronic progressive), aspiration pneumonia, grand mal seizure with status epilepticus	Infl. +/Demyel. +	−
	MS‐#3	40/F	MS (progressive, severe, unrelenting ∼10 years)	Infl. +/Demyel. +	+++
	MS‐#4	43/F	MS (chronic progressive)	Infl.−/Demyel.−	+
	MS‐#5	42/M	MS (type unavailable, ∼10 years), suicide	Infl. +/Demyel. +	++
	MS‐#6	33/M	MS (chronic progressive)	Infl. +/Demyel. +	+++
	MS‐#7	49/F	MS (type and duration unknown)	Infl. Rim +/Demyel. +	++
	MS‐#8	63/F	MS, sleep apnea, chronic urinary tract infection, respiratory failure, hypertension	Infl. Rim +/Demyel. +	−
	MS‐#9	52/F	MS (second progressive), chronic urinary tract infection, respiratory failure	Infl. Rim +/Demyel. +	+
	MS‐#10	51/F	MS (second progressive), depression	Infl. Rim +/Demyel. +	++
	MS‐#11	75/F	MS, hypertension, kidney stone surgery, goiter surgery, hypothyroidism, migraine, chronic urinary tract infection	Infl.−/Demyel.−	+++
	MS‐#12	59/M	MS (primary progressive), dementia, tobacco abuse	Infl. +/Demyel.−	−
	MS‐#13	54/F	MS (relapse/remitting)	Infl. Rim +/Demyel. +	+++

^* Microscopic classification was based on the Vienna consensus for MS lesion staging/description (56).

Infl. + (inflammatory): denote the presence of inflammatory cells (including macrophages).

Demyel. + (demyelinating): denote the presence of recognizable myelin breakdown products in macrophages.

Infl. Rim +: denote the presence of inflammatory rim (and hypocellular center).

^† Based on double immunohistochemistry. Immunofluorescence signal was analyzed by two independent investigators and the intensity of ceramide immunoreactivity in GFAP⁺ cells in relation to background signal (unaffected white matter areas of the same tissue section) was semi‐scored in scale of: −, minimum; +, weak; ++, moderate; +++, strong.

M = male; F = female; NK = not known.

Figure 2.

Accumulation of ceramide in reactive astrocytes in animal models of demyelination. A. Representative coronal sections of spinal cords from sham and experimental autoimmune encephalomyelitis (EAE) mice stained with Sudan black. Arrows indicate demyelinating white matter. B–C. Serial spinal cord sections stained for ceramide, GFAP, Iba1 or isotype control antibodies. Ceramide is localized to reactive astrocytes in the spinal cord white matter of EAE, but not control mice. Data are representative of three animals per group. D–I. Representative images of ORO‐ (D,F) and immunostained sections (E,G–I) showing that cuprizone intoxication induces extensive demyelination in the corpus callosum and marked upregulation of ceramide in reactive astrocytes (arrows in G). Ceramide immunoreactivity is not localized to microglia/macrophages (H) and appears not in Olig2⁺ OLs (I). Asterisks indicate the immunostained regions in E and G. Data are representative of three control and five cuprizone mice. Scale bars: B,C, 10 µm; D,F, 200 µm; E,G, 50 µm; H,I, 20 µm.

Accumulation of ceramide in human PVL

To determine whether astroglial accumulation of ceramide during demyelination of the adult CNS also occurs in white matter injury during development when myelin is not yet formed, we examined human PVL cases and controls related in age (Table 2). As diffuse gliosis is most pronounced surrounding foci of periventricular necrosis in PVL 24, 26, we focused on diffuse gliosis in our analysis of human PVL. Distinct and punctate ceramide staining was found within reactive astrocytes in the diffusely gliotic component of the deep white matter region (Figure 3C), whereas minimum immunolabeling was observed at the level of the subcortical white matter (Figure 3B). There appears to be a close spatial relationship between activated microglia/macrophages and ceramide‐positive astrocytes across the cerebral white matter region (Supporting Information Figure S5).

Table 2.

Summary of the clinicopathologic information and neuropathologic findings on periventricular leukomalacia (PVL) and control cases.

Diagnosis	Case #/sex	GA (weeks)*	PA (weeks)	PCA (weeks)†	Other clinicopathologic data	Microglia activation‡	Ceramide signal in AST§
Control	1/M	33	1	34	Tracheoesophageal fistula, small ventricular septal defect, respiratory distress, prematurity	−	−
	2/M	32	2	34	Bradycardia, cyanotic, prematurity, comfort care	+	−
	3/M	39	1	40	Nodular cortical dysplasia, focal pontosubicular necrosis, liver function abnormalities	++	−
PVL	PVL #1/F	31	2	33	Congenital heart defect, prematurity, sepsis, prematurity	−	−
	PVL #2/F	32	1	33	Intraventricular hemorrhage, septic shock, prematurity	+	+
	PVL #3/F	34	4	38	Congenital diaphragmatic hernia, prematurity	+	++
	PVL #4/F	35	5	40	Renal failure, diffuse sclerosis, ventilator dependent, prematurity	+++	+++
	PVL #5/M	38	2	40	Primary pulmonary hypertension, cerebral infarct, sepsis	++	+

^* GA = gestational age. The time from conception until birth. More specifically, gestational age is defined as the number of weeks from the first day of the mother's last menstrual period (LMP) until the birth of the baby. Gestational age at birth is assessed by the date of the LMP and by physical exam (Dubowitz score). Mean ± SD of GA: 34.7 ± 3.8 weeks (control) and 34.0 ± 2.7 weeks (PVL).

^† PCA = post‐conceptional age. Age since conception. Post‐conceptional age is calculated as gestational age plus postnatal age. Mean ± SD of PCA: 36.0 ± 3.5 weeks (control) and 36.8 ± 3.6 weeks (PVL).

^‡ Microglia/macrophage activation was determined by CD68⁺ cells in the cerebral white matter and diffuse component of PVL.

Labeling index: Immunoreactive intensity was defined as mild (+), moderate (++) or intense (+++), − indicates no or minimum signal.

^§ Based on immunohistochemistry. When ceramide+ cells was counted and scored, necrotic focal lesion regions were excluded and only diffuse lesions were analyzed.

NK = not known; PA = postnatal age.

Figure 3.

Ceramide is upregulated in reactive astrocytes throughout diffuse white matter lesions of human periventricular leukomalacia (PVL). A–B. Representative H&E micrograph of a typical PVL with microcysts of focally necrotic lesions (asterisks) in the deep cerebral white matter (wm) of a premature infant born at 35 gestational weeks and died at 4 postnatal weeks (case #4). Representative areas shown in B,C are depicted here. Reactive astrocytosis is prominent throughout the deep and central white matter distant from the focal lesions. B–C. Immunohistochemistry of serial sections reveals marked upregulation of ceramide in GFAP⁺ astrocytes throughout the diffuse lesion (C), but not in white matter regions with limited reactive astrocytosis and immediate below overlying cortex (ctx) (B). Representative micrographs of a Cer⁺/GFAP⁺ cell (C, arrow) at a higher magnification are shown on the bottom left. Scale bars: B,C, 20 µm; higher magnification panel in C, 5 µm.

Disrupted sphingolipid metabolism during demyelination

To confirm the above results, we then used lipid chromatography tandem mass spectrometry to determine and quantify simultaneously specific ceramide species and other sphingolipid metabolites from tissue extracts (8). C18:0 ceramide was found to be significantly elevated in MS samples when compared to controls (Figure 4A). While there appeared to be increases in C18:1, C24:0 and C24:1 ceramide levels in MS, the elevation was not statistically significant. Levels of Sph and S1P also showed variations among these human autopsy samples and were not statistically significant (Figure 4B). It should be noted that a recent study reported a decreased level of S1P and increased ratio of C16 or C18 over C24 in MS lesions, albeit the level of statistical significance was not provided (53).

Figure 4.

Specific ceramide species are elevated during active demyelination in multiple sclerosis (MS) and in animal model of demyelination. A,B. Specific sphingolipids of frozen autopsy human MS and control tissues were analyzed and quantified using high‐performance liquid chromatography–tandem mass spectrometry. Symbols indicate values from individual cases. C18:0 ceramide is statistically significantly elevated in MS (n = 11) as compared to controls (n = 8). C,D. During cuprizone‐induced demyelination, C16:0, C18:0 and C20:0 ceramides are significantly upregulated in the cerebral white matter of cuprizone‐fed mice. dhC16 and sphingosine (Sph) are also significantly elevated in cuprizone‐fed mice, whereas the level of dhSph and sphingosine‐1‐phosphate (S1P) is significantly decreased during demyelination (n = 3 animals per group). *P < 0.05; **P < 0.01; NS = not significant.

Because of the heterogeneity of human post‐mortem samples, we decided to employ the cuprizone mouse model of demyelination to investigate further alterations of sphingolipid metabolism during demyelination. We chose the cuprizone model for these analyses largely because the toxin‐induced OL death and demyelination occur reproducibly in well‐defined anatomical areas in the brain, and are independent of peripheral lymphocytes, enabling direct analysis of sphingolipid metabolism within the CNS parenchyma without the complication from activated lymphocytes. Consistent with the immunohistochemical findings, we found significantly elevated levels of C16:0, C18:0, C20:0, dihydroC16 (dhC16) ceramides and Sph, and diminished levels of dihydroS1P (dhS1P) and S1P in the white matter of cuprizone‐fed mice (Figure 4C,D). As the relative balance of ceramide and S1P is an important determinant for cell death and survival (55), this inverse relationship between ceramide and S1P/dhS1P during demyelination suggests a role for sphingolipids in regulating oligodendroglial cell demise.

SPT, the rate‐limiting enzyme for ceramide biosynthesis, is upregulated in reactive astrocytes in the corpus callosum of cuprizone‐fed mice

The level of intracellular ceramide is maintained by interconnected metabolic pathways that are localized in various subcellular compartments and are under different regulations (22). To explore whether the increased levels of astroglial ceramide in demyelinating regions were a result of increased sphingolipid metabolism, we examined enzymes involved in ceramide biosynthesis (sphingomyelin breakdown and de novo synthesis) and ceramide degradation. Expression and enzymatic activity of nSMase and aSMase were not altered in cuprizone‐fed mice when compared to controls, with aSMase accounting for the majority of SMase activity in the brain (Figure 5A,B). Similarly, aCDase expression was not significantly different between control and cuprizone‐fed mice (Figure 5C). In contrast, while SPT was essentially absent in the corpus callosum of control mice (Figure 5D), it was markedly upregulated in reactive astrocytes in the caudal region of the corpus callosum where demyelination occurs (Figure 5E,F) and where the ceramide‐positive astrocytes were observed (2, 5). Furthermore, SPT⁺ astrocytes in the caudal corpus callosum were also immunoreactive for ceramide (Figure 5H). It is noteworthy that neurons endogenously expressed high level of SPT (Figure 5G). The fact that SPT is upregulated in reactive astrocytes that also accumulate ceramide suggests a potential involvement of the de novo ceramide synthesis pathway. This is in agreement with our findings on alterations of metabolites of the de novo pathway such as dhC16 as well as dhS1P in cuprizone‐fed mice (Figure 4D).

Figure 5.

Upregulation of serine palmitoyltransferase (SPT) in reactive astrocytes during demyelination. A. Quantitative RT‐PCR (qRT‐PCR) analysis of mRNA expression of acid (aSMase) and neutral SMase (nSMase) in dissected brains of control and cuprizone‐fed mice. No significant changes in sphingomyelinase (SMase) transcripts between the two groups (n = 3 animals per group). B. Cerebral aSMase and nSMase enzymatic activities do not differ significantly between control and cuprizone‐fed mice. C. Western blotting analysis of acidic ceramidase (aCDase) expression in dissected brain tissues from control and cuprizone‐fed mice. Each lane represents one animal (n = 3 animals per group). D. Representative images of sagittal brain sections of control mice stained for myelin and SPT. SPT expression is minimal in the corpus callosum (cc, asterisks). E,F. Representative images of Sudan black‐stained sagittal brain sections of cuprizone‐fed mice, showing extensive demyelination in the caudal region of corpus callosum. Double immunostaining shows markedly increased SPT expression in reactive astrocytes in the caudal corpus callosum. Neurons express high level of SPT in both cuprizone‐fed mice (G, arrows) and controls (D). H. Representative image showing co‐localization of SPT and ceramide in the caudal corpus callosum of cuprizone‐fed mice. Scale bars: D,E, 500 µm; F, 10 µm; G,H, 20 µm.

Ceramide acts synergistically with TNF to induce apoptotic death of OLs in an astrocyte cell contact‐dependent manner

To explore the functional consequence of astroglial accumulation of ceramides to OLs, we next employed an in vitro model for cytokine‐mediated toxicity toward oligodendroglia (41). We treated mixed glial cultures with cell‐permeable C2 ceramide at concentrations not directly toxic to cells, and then examined OL survival. We found that ceramide acted synergistically with low dose of TNF, resulting in significant loss of OLs (Figure 6A,B). Interestingly, this synergism between ceramide and TNF was completely lost in OL monocultures (Figure 6C), indicating a noncell autonomous effect. Consistent with these results, in co‐cultures of astrocytes and OLs, C2, C6 and the long chain C18 ceramides all caused significant oligodendroglial death in the presence of low levels of TNF (Figure 7A). Furthermore, increasing endogenous ceramides via blocking ceramide degradation with CDase inhibitor NOE similarly potentiated TNF toxicity (Figure 7A). Under these conditions, oligodendroglial cells underwent a caspase‐dependent apoptotic cell death, as demonstrated by TUNEL of DNA fragmentation (Figure 7B,C), caspase‐3 activation in dying oligodendroglial cells in ceramide/TNF‐treated co‐cultures (Figure 7D–F) and by blockade of the cell death with zVAD‐fmk, a pan‐caspase inhibitor (Figure 7G). Interestingly, ceramide and TNF failed to induce cell death if astrocytes and OLs were co‐cultured in two separate layers distanced at 1 mm without direct cell–cell contact (Figure 7H). To further determine whether the synergism between ceramide and TNF is caused by a direct effect on astrocytes, we pretreated astrocytes with ceramide, TNF or ceramide plus TNF for 24 h followed by extensive washes before plating OLs into the pretreated astrocyte cultures. We found that astrocytes pretreated with ceramide plus TNF were subsequently toxic to added OLs (Figure 7I), albeit to a less degree as compared to those of co‐cultures (Figure 7A). Taken together, these results demonstrate that ceramide and TNF synergistically induce oligodendroglial cell death via a mechanism dependent on astrocytes. The data also agree with our immunohistological observation of co‐existence of ceramide‐laden astrocytes and TUNEL⁺ OLs around actively demyelinating lesions in some MS cases (Figure 1).

Figure 6.

Ceramide promotes oligodendroglial cell death in mixed glial cultures, but not in pure cultures. A,B. Cell‐permeable C2‐ceramide (Cer) and tumor necrosis factor (TNF) synergistically induce significant oligodendrocyte (OL) cell loss in mixed glial cultures. Mixed glia were treated with C2‐Cer (20 µM), TNF (10 ng/mL) or both as indicated for 48 h, and OL survival was quantified. C. In highly enriched OL monocultures, TNF (10–100 ng/mL) and C2‐ceramide, either alone or in combination, do not cause significant cell death. Data represent three to four independent experiments. **P < 0.01.

Figure 7.

Ceramide potentiates tumor necrosis factor (TNF)‐induced apoptotic death of oligodendrocytes (OLs) in an astrocyte‐dependent manner. A,B. C2‐, C6‐ and C18‐ceramides (20 µM) act synergistically with TNF (1–5 ng/mL) in astrocytes and OL co‐cultures, causing significant cell death. Blocking endogenous ceramide degradation with ceramidase inhibitor N‐oleoylethanolamine (NOE) also potentiates TNF toxicity. ***P < 0.001 when compared to control or corresponding drug treatment in the absence of TNF. B–E. Representative immunocytochemistry and quantification of terminal deoxynucleotidyl transferase‐mediated dUTP nick‐end labeling (TUNEL⁺) and activated caspase‐3 (Δcasp3)⁺ OLs in co‐cultures of astrocytes and OLs treated with or without C2 ceramide or TNF as indicated. ***P < 0.001 when compared to controls. F. Western blotting analysis for activated caspase‐3 in co‐cultures treated as indicated. Quantification of Δcasp3 represents two independent experiments. **P < 0.01. G. Caspase inhibitor zVAD‐fmk abolished Cer/TNF‐induced OL death in co‐cultures. H. The synergistic pro‐death effect between TNF and ceramide is abolished if astrocytes and OLs are cultured in two closely opposed layers without direct cell contact. I. Astrocytes pretreated with C2 ceramide and TNF are subsequently toxic to seeded OLs. Data represent three independent experiments. **P < 0.01; ***P < 0.001. NS = not significant. Scale bars: B,D, 10 µm.

Elevation of astroglial ceramide and depletion of S1P during CNS demyelination are reversed during remyelination

The fact that astrocytes accumulate ceramide in association with cuprizone‐induced OL damage and demyelination, and that ceramide sensitizes OLs to TNF toxicity in co‐cultures of astrocytes and OLs suggests a potential pathological role for astrocytes in demyelination. As the cuprizone model provides a unique opportunity to study signals involved in CNS remyelination processes without the complication of adaptive immune responses, we sought to determine ceramide metabolism during the remyelinating process. As mentioned, cuprizone intoxication for 3 weeks caused OL apoptosis and significant increases of C16:0/C18:0 ceramides and suppression of dhS1P/S1P (Supporting Information Figure S4D). Mice intoxicated for 5 weeks were severely demyelinated in the caudal corpus callosum. Cessation of cuprizone feeding was followed by a spontaneous repair process and remyelination (46). Two weeks after cuprizone withdrawal, more than one‐third demyelinated axons underwent active remyelination as determined by transmission electron microscopy (Steelman et al, unpublished). Increased levels of Sph, and C16:0, C18:0 and C20:0 ceramides during demyelination were either returned to control levels or significantly decreased during the active remyelinating phase (Figure 8). Concomitantly, S1P and dhS1P, which were depleted during demyelination, returned to the normal level during the remyelinating process (Figure 8). The results thus reveal a close relationship between altered bioactive sphingolipid metabolites and demyelination/remyelination, and suggest a potential of modulating the ceramide/S1P pathway in promoting myelin repair.

Figure 8.

Upregulation of ceramide and loss of sphingosine‐1‐phosphate (S1P) during demyelination are significantly reversed during active remyelination in the cuprizone animal model. Comparison of sphingolipid levels in dissected brains from control mice (n = 6), mice fed cuprizone for 5 weeks (CPZ‐5 weeks, n = 3) and mice fed cuprizone for 5 weeks followed by 2 weeks of recovery that results in efficient remyelination (CPZ‐5 + 2 weeks, n = 4). Ceramides that are significantly elevated in CPZ 5 weeks mice are significantly decreased in CPZ 5 + 2 weeks mice. Loss of S1P and dhS1P during demyelination is also recovered in 5 + 2 weeks mice. Data represent mean ± SEM. Black asterisks, when compared with control; blue asterisks, when CPZ 5 weeks compared with 5 + 2 weeks. *P < 0.05; **P < 0.01; ***P < 0.001.

DISCUSSION

Bioactive sphingolipid metabolites such as ceramide, Sph and S1P mediate multiple biological functions and have recently received increasing attention as potential therapeutic targets for diverse disease conditions ranging from cancer to atherosclerosis, inflammation and autoimmune disorders such as MS. In this study, we report that ceramides are aberrantly accumulated in reactive astrocytes in actively demyelinating MS lesions, in diffuse lesions of PVL and in animal models of demyelination. Ceramides such as C16:0, C18:0 and C20:0 are significantly increased and are associated with a concomitant decrease in S1P levels during demyelination in the cuprizone model of demyelination. In culture, ceramides act synergistically with TNF to induce apoptotic death of OLs via an astrocyte contact‐dependent mechanism. Furthermore, we found that elevated astroglial ceramide during cuprizone‐induced demyelination is restored during active remyelination in conjunction with an increase in tissue level of S1P. Our results imply that the relative levels of ceramide and S1P may determine OL cell demise.

Astrocyte dysfunction has been linked to a variety of neurological disorders 7, 28 including MS 4, 33. Overproduction of endogenous retrovirus glycoprotein syncytin or transmembrane TNF in astrocytes results in neuroinflammation and demyelination 1, 4, 5. Our current findings indicate that elevated astroglial ceramides potentiate OL injury as ceramide markedly synergizes TNF toxicity in an astrocyte contact‐dependent manner, a finding consistent with our previous reports 35, 41. It is plausible that aberrant accumulation of ceramides leads to dysfunctional astrocytes that subsequently contribute to OL injury. Alternatively, ceramides or their bioactive derivatives released from reactive astrocytes may directly impact on OLs and promote apoptotic cell death. In fact, exogenous ceramide appears to preferentially kill OLs when compared to astrocytes and neurons 11, 54. While the etiology and age predilection of MS and PVL vary considerably (eg, adults in the former vs. preterm neonates in the latter), these two disorders share pathologic features of activation of astrocytes and microglia, production of inflammatory cytokines and damage to OL lineage cells. We speculate that astroglial accumulation of ceramides may represent a stress response to pathological mediators. Indeed, increased ceramide levels have been reported in Alzheimer's disease 16, 21, HIV‐induced dementia 23, 50, amyotrophic lateral sclerosis (15) and hypoxia/ischemia (52). It is noteworthy that prominent ceramide immunoreactivity was also observed in microvessels in the cerebral white matter of PVL and MS (Supporting Information Figures S2 and S5C). Ceramide and its metabolites have been previously implicated in modulating vascular functions including vascular tones and integrity 2, 42.

Cellular levels of ceramide are tightly regulated by complex and interconnected pathways in various subcellular locations. Disruption of one particular sphingolipid metabolic pathway may compromise the other. The importance of a balanced sphingolipid metabolism is underscored by human diseases resulting from genetic defects in the metabolism of sphingolipids, including certain leukodystrophies 38, 62. Interconversion between ceramide and other bioactive sphingolipid species such as Sph and S1P allows perturbation of one potentially influencing the other. Over 2 dozen distinct enzymes can act on ceramide as a substrate or a product, and regulation of many of these enzymes is just beginning to be understood (22). Our finding of SPT upregulation in reactive astrocytes at demyelinating areas in the cuprizone animal model indicates that the de novo pathway is most likely to account for the aberrant accumulation of ceramides in these cells. In support of this idea, SPT1 and/or SPT2 have been shown to be upregulated in response to endotoxin and cytokines (47). Furthermore, lipidomics analysis reveals elevated ceramide levels in rat hippocampus after kainate‐induced excitotoxic injury (20), in part because of SPT upregulation in reactive astrocytes (27). Whether SPT is upregulated in reactive astrocytes in human MS and PVL tissues remains to be determined as we found that in all control autopsy cases, SPT was endogenously and strongly expressed in astrocytes in the cerebral white matter (Kim and Li, unpub. obs.). Thus, future studies are needed to determine the contribution of the de novo ceramide synthesis pathway to the white matter injury.

Although the current study cannot exclude the possibility that degradation of myelin sheath may account for elevations of specific ceramides during demyelination, our data do not support this as a main contributor for astroglial accumulation of ceramides. This is based on the fact that: (i) aberrant ceramide accumulation also occurs in PVL brain when myelin is not yet developed; and (ii) degradation of complex glycosphingolipids occurs primarily in late endosome and lysosomes, producing ceramides that are sequestered in lysosomes, and then converted to Sph 12, 52. Accumulation of ceramides through this salvage pathway requires ceramide synthases, key enzymes in the de novo pathway. However, we did not detect any ceramide immunoreactive microglia/macrophages in demyelinating MS lesions or in the cuprizone‐induced demyelination model, as otherwise expected for these phagocytes. Our data demonstrate that sphingolipid metabolism is disrupted during demyelination. The close relationship between elevated ceramide and depressed S1P levels found during demyelination and the restoration of the disrupted sphingolipid homeostasis during remyelination indicate the potential of modulating these sphingolipid pathways for therapeutic gains. Generally viewed as being pro‐survival and cell growth, S1P signals either intracellularly or extracellularly by engaging G protein‐coupled S1P receptors (S1P1–5) that are differentially expressed in cells and regulate diverse cellular functions. S1P released from astrocytes in response to exogenous stimuli 3, 6 could act on S1P receptors on OLs or astrocytes. For example, S1P induces process retraction, promotes OL survival (29), inhibits progenitor migration (51) and has been implicated in PDGF‐induced progenitor mitogenesis (31). Engagement of astroglial S1P receptors leads to proliferation and neurotrophic factor production 6, 63. S1P receptors 1 and 3 have recently been shown to be upregulated in reactive astrocytes in active and chronic inactive MS lesions (58). As decreased S1P tissue level during demyelination is fully restored in the remyelinating process, delineating the cellular source for the altered S1P and its interacting receptors shall provide critical insights for CNS myelin repair. In agreement with this, FTY720 (fingolimod), an analogue of S1P and a first‐in‐class S1P receptor modulator, is highly effective for relapsing and remitting MS (34) and was recently approved for the treatment of MS. Although impaired egress of immune effector cells out of lymphoid organs (45) is considered to be responsible for the beneficial effect of fingolimod in MS, FTY720 readily crosses the blood–brain barrier and can thus impact directly on CNS parenchymal cells through engagement of specific S1P receptors. In this respect, it was recently shown that FTY720 modulates glial responses and promotes remyelination after lysolecithin‐induced demyelination in cerebellar slice cultures (48). Moreover, a most recent study showed that the efficacy of FTY720 in the T‐cell‐mediated EAE demyelinating model is lost in mutant mice where S1P1 is selectively deleted in astrocytes, indicating a key role for astrocytes in the pathogenesis of EAE (13). Based upon these observations and the findings reported here, future testing of the effect of FTY720 on remyelination and sphingolipid metabolism in the cuprizone model of demyelination is warranted.

In summary, this study identifies aberrant astroglial accumulation of ceramides in human MS and PVL, and in animal models of demyelination, and demonstrates, for the first time, reversible changes of specific sphingolipid metabolites such as ceramide, Sph and S1P during active demyelination and remyelination. We further demonstrate that ceramides markedly promote oligodendroglial cell death in an astrocyte‐dependent manner and through, at least in part, acting directly on astrocytes. Our data suggest a previous unrecognized role for ceramide‐laden reactive astrocytes in promoting oligodendroglial cell death. Astrocytes could thus represent a potential drug target for modulation of sphingolipid signaling pathways in central white matter injury.

Supporting information

Figure S1. Scheme of ceramide metabolic pathways. Ceramide is generated de novo in endoplasmic reticulum through a multistep process from serine and palmitoyl CoA, or via hydrolysis of sphingomyelin by sphingomyelinases (SMases). Serine palmitoyl transferase (SPT) is the rate‐limiting enzyme for de novo biosynthesis of ceramide. Ceramide is deacylated into sphingosine (Sph) by ceramidases (CDases), and Sph can be further phosphorylated to Sph‐1‐phosphate (S1P). Ceramide is the precursor for the synthesis of complex sphingolipids through the action of sphingomyelin (SM) synthase, glucosylceramide synthase (GlcCer synthase) and UDP‐galatose : ceramide galactosyltransferase (Cer GalT). In addition, ceramide can be directly phosphorylated forming ceramide‐1‐phosphate. Many enzymes involved in ceramide metabolism are differentially regulated and compartmentalized. Bioactive sphingolipid metabolites found in this study to be increased and decreased during central nervous system (CNS) demyelination are highlighted in red and green, respectively.

Figure S2. Comparison of two different anti‐ceramide antibodies demonstrates ceramide accumulation in astrocytes, but not microglia/macrophages in MS. Two different anti‐ceramide antibodies were used: the mouse IgM antibody from Glycobiotech GmbH and the mouse monoclonal IgM 15B4. Both antibodies labeled reactive astrocytes in and around the boarder of active MS lesions. A–C. Representative micrographs of serial tissue sections (MS case #5) stained with the two anti‐ceramide antibodies, which show similar signals in GFAP⁺ cells, and with normal mouse IgM, which demonstrates minimal signals. Ceramide immunoreactivity was also observed in blood vessels. D. Immunostaining of the serial section of A–C demonstrates that ceramide immunoreactivity not localized to Iba1⁺ microglia/macrophages in the white matter. Inserts show cells at a higher magnification (A,C,D, arrows). Scale bar: 200 µm.

Figure S3. Accumulation of ceramide in reactive astrocytes in various demyelinating regions of cuprizone‐fed mice. Mice were fed cuprizone for 5 weeks and immunohistochemistry was performed. In all regions undergoing cuprizone‐induced demyelination, ceramide is markedly upregulated in hypertrophic astrocytes (arrows) as compared to the same anatomic regions of control mice (n = 3–4 animals per group). Scale bars: 20 µm.

Figure S4. Upregulation of ceramides and downregulation of S1P in the corpus callosum of mice fed cuprizone for 3 weeks. Mice were fed cuprizone for 3 weeks when oligodendrocyte apoptosis peaked and astrocytosis initiated in the corpus callosum. Immunohistochemistry and mass spectrum measurements of ceramide and S1P species were carried out as described. A–B. Double immunohistochemistry showing ceramide immunoreactivity localized to GFAP⁺ cells. C. Terminal deoxynucleotidyl transferase‐mediated dUTP nick‐end labeling (TUNEL) of apoptotic oligodendrocytes(OLs) in the corpus callosum. Myelin was also labeled by CNPase (blue). D. Mass spectrometry analysis of the corpus callosum of mice intoxicated with cuprizone for 3 weeks, showing significant increases of C16:0 and C18:0 ceramides and decrease of dhS1P and S1P levels over control mice (n = 5 animals per group). ****P < 0.001.

Figure S5. Upregulation of ceramide in the diffuse lesion of human periventricular leukomalacia (PVL) is closely associated with microgliosis. Serial sections of the PVL case shown in Figure 3 were immunostained for ceramide and CD68 (activated microglia/macrophages), and are shown here at a lower magnification. Asterisks indicate the same two focal lesions shown in Figure 3. Areas representing subcortical white matter (A), peri‐focal lesion region (B) and deep cerebral white matter (C) are shown at a higher magnification on the sides. The intensity of ceramide immunoreactivity is closely associated with microgliosis in the diffuse component of PVL.

Supporting info item

Supporting info item

Supporting info item

Supporting info item

Supporting info item