Sphingosine-1-phosphate (S1P) is a lipid-signaling molecule produced by sphingosine kinase in response to a wide number of stimuli. By acting through a family of widely expressed G protein-coupled receptors, S1P regulates diverse physiological processes. Here we examined the role of S1P signaling in neurodegeneration using a mouse model of Sandhoff disease, a prototypical neuronopathic lysosomal storage disorder. When sphingosine kinase 1 (Sphk1) was deleted in Sandhoff disease mice, a milder disease course occurred, with decreased proliferation of glial cells and less-pronounced astrogliosis. A similar result of milder disease course and reduced astroglial proliferation was obtained by deletion of the gene for the S1P3 receptor, a G protein-coupled receptor enriched in astrocytes. Our studies demonstrate a functional role of S1P synthesis and receptor expression in astrocyte proliferation leading to astrogliosis during the terminal stages of neurodegeneration in Sandhoff disease mice. Because astrocyte responses are involved in many types of neurodegeneration, the Sphk1/S1P receptor signaling axis may be generally important during the pathogenesis of neurodegenerative diseases.
The lysophospholipid sphingosine-1-phosphate (S1P) is a signaling molecule that stimulates a family of five G protein-coupled receptors, S1P1–5, to control complex physiological and pathophysiological processes (1–3). S1P receptor signaling is regulated by synthesis and secretion of S1P ligand and by differential expression and activity of S1P receptors on various cell types.
S1P is produced through phosphorylation of sphingosine by two conserved lipid kinases, sphingosine kinase (Sphk)1 and Sphk2. Sphk1 produces S1P in response to a variety of stimuli, and is controlled both transcriptionally and posttranslationally. The function and distribution of Sphk2 partially overlaps with Sphk1 (4,5). In addition to regulation by control of synthesis, S1P levels are also maintained through the activity of enzymes—an S1P-specific lyase (Sgpl1) and two phosphatases (Sgpp1,2)—that directly degrade S1P. An important S1P pool exists in the blood, having been produced mainly by red blood cells, that serves as the basis of a gradient to control the circulation of lymphocytes (6). Other S1P pools for signaling are synthesized by several cell types, including mast cells (7,8), endothelial cells (9), astrocytes (10) and neurons (11).
S1P receptors and their natural lipid ligand, S1P, are highly enriched in the central nervous system (CNS), implying that important functions may exist here for S1P signaling. Recently, we demonstrated that mice lacking the machinery for S1P receptor-mediated signaling displayed increased apoptosis and decreased proliferation of neuroblasts; these alterations led to the development of neural tube defects, suggesting that S1P signaling is required for neurogenesis during embryonic development (5). In addition to a role in CNS development, the S1P receptor signaling pathway may contribute to neurodegenerative disease, as was suggested in part by studies showing that the synthetic S1P receptor ligand FTY720 offers protection in multiple sclerosis (MS) (12).
To further explore the possible role of the canonical S1P receptor signaling pathway during a neurodegenerative process, we used the Hexb null mouse (13), a highly accurate model of Sandhoff and Tay-Sachs diseases, which are prototypic neuronopathic lysosomal storage disorders. In these disorders, an absence of lysosomal β-hexosaminidase A blocks the ganglioside degradation pathway, leading to substrate accumulation in neurons and triggering a sequence leading to neuronal cell death. During the neurodegenerative cascade, the glial cell population expands through proliferation of resident glial cells and by monocyte invasion, which accelerates the disease course (14,15). We show here that the lipid kinase that controls S1P synthesis, Sphk1, regulates a proliferative response of glial cells and associated reactive gliosis during neurodegeneration in the Sandhoff disease mouse. Furthermore, we show that the S1P3 receptor, whose expression is enriched on astrocytes, also participates in the neurodegeneration-induced glial proliferative response. Because astroglial responses can be an impediment to neuronal survival and regeneration after damage in neurodegenerative disorders (16), the results indicate that the Sphk1/S1P receptor signaling axis represents a potential therapeutic target in neurodegenerative disease.
Expression of the S1P metabolic pathway and receptors is regulated during Sandhoff disease
We determined if the expression of S1P metabolic and S1P receptor genes (Fig. 1A) was altered in Hexb−/− mouse spinal cord, which is a major site of pathogenesis and neurodegeneration during Sandhoff disease course (14,15). Gene expression was determined at 8 weeks of age, when the Hexb−/− mice were mildly symptomatic, and at 16 weeks of age, during terminal stages of the disease (Fig. 1B and C). The level of the Sphk1 mRNA was elevated at 16 weeks (Fig. 1B), while the level of S1P lyase (Sgpl1) mRNA, which specifies an enzyme that carries out the terminal degradative step of S1P, was elevated at both 8 and 16 weeks of age. Sgpp2, which encodes a S1P-specific phosphatase, was significantly decreased. The expression of S1P receptor mRNAs was not changed in the spinal cord of Hexb−/− mice at 8 weeks of age; however, at 16 weeks of age, S1P2 and S1P3 receptor mRNAs were significantly elevated (Fig. 1C).
We measured Sphk enzyme activity in spinal cord from Hexb−/− mice and wild-type mice using an assay that is relatively selective for Sphk1 activity (17), and found a significant increase in activity at 16 weeks of age in Hexb−/− mice when compared with wild-type mice (Fig. 1D). Total spinal cord S1P levels were also significantly elevated at 16 weeks of age in Hexb−/− mice when compared with wild-type mice (Fig. 1E). No increase in spinal cord Sphk activity or S1P level was noted in Hexb−/− mice at 8 weeks of age (data not shown).
Staining of spinal cord sections from 16-week-old mice using a polyclonal antibody to Sphk1 (18) resulted in positive staining of ventral horn spinal cord motor neurons in both Hexb+/+ and Hexb−/− mice (Fig. 2A and B; arrows). In Hexb−/− mice, Sphk1 staining of the motor neurons appeared to be more intensely localized to the plasma membrane than was Sphk1 staining of neurons in Hexb+/+ mice (Fig. 2D, arrows). Overall, the density of Sphk1-positive cells was higher in Hexb−/− mice than in Hexb+/+ mice (Fig. 2A and B; black arrowheads). Coimmunostaining for Sphk1 and glial fibrillary acidic protein (GFAP), a marker for astrocytes, revealed that some of the abundantly Sphk1-stained cells were astrocytes (Fig. 2C and D; arrowheads).
These results indicate that S1P metabolism is altered in the CNS in the Sandhoff disease mouse model, resulting in an increase in S1P levels together with changes in the mRNA levels of Sphk1, Sgpp2 and Sgpl1, enzymes that act directly on S1P. Both S1P2 and S1P3 receptor mRNA levels increased as the disease progressed. During the course of the disease, an increase in Sphk1-expressing cells, identified as GFAP-positive astroglia, was detected. In addition, the plasma membrane association of Sphk1 in spinal cord neurons appeared more pronounced in the disease model compared with wild-type mice. These changes appeared most prominent during the terminal stages of the disease.
Deletion of the Sphk1 gene improves neurodegenerative course of Sandhoff disease
To determine whether Sphk1 might play a functional role during the neurodegenerative process in Sandhoff disease, we generated Hexb−/− mice deficient in Sphk1, together with wild-type mice of other genotype combinations. The lifespan, body weight and motor function of each group of mice were monitored. The group of Hexb−/− mice with a single active Sphk1 allele (Hexb−/−Sphk1+/−) displayed a disease course similar to the single null Hexb−/− mice. However, the double null mice (Hexb−/−Sphk1−/−) displayed the mildest clinical symptoms (Fig. 3A) and the longest lifespan of the three groups (Fig. 3C). The double null mice continued to maintain their body weight until 17 weeks, which was the point where the single null Hexb−/− mice had lost 20% of their peak body weight (Fig. 3B). The rotorod performance of the double null mice was significantly better than single null Hexb−/− mice at all time points tested (Fig. 3D). A similar lifespan difference between genotypes was found when mice were on a mixed Sv129/C57BL/6 background (Supplementary Material, Fig. S1). These results indicate that deletion of the Sphk1 gene provides for a milder disease course in the Sandhoff model.
Sphk enzyme activity was substantially reduced in spinal cord of the double null Hexb−/−Sphk1−/− mice compared with groups with the wild-type Sphk1 alleles (Fig. 1D). Interestingly, even with this large decrease of Sphk activity, a significant increase in spinal cord S1P levels was seen in the Hexb−/−Sphk1−/− mice compared with levels in Hexb+/+Sphk1−/− mice (Fig. 1E). We then tested whether the mRNA expression of other enzymes that alter S1P levels (Sphk2, Sgpp1,2 and Sgpl1; Fig. 1A) undergo compensatory changes in the double null mice (Fig. 4A). Indeed, the expression of Sgpl1, the S1P lyase that degrades S1P (Fig. 1A), was reduced at 16 weeks of age in the double null mice compared with the single null Hexb−/− mice, providing a possible explanation for the elevation of S1P levels in Hexb−/− mice even in the absence of Sphk1.
The direct result of the gene defect in Sandhoff disease is the accumulation of glycosphingolipids GM2 and GA2, primarily in neurons (19). We determined if deletion of Sphk1, an enzyme involved in the metabolism of simple sphingolipids, might indirectly alter the storage of the complex glycosphingolipids. By quantitative thin-layer chromatography (TLC) analysis, no substantial differences in accumulation of GM2 and GA2 were observed in spinal cord of Hexb−/−Sphk1−/− mice when compared with Hexb−/−Sphk1+/+ mice at 16 weeks of age (Fig. 4B and C). These results indicate that the Sphk1 deletion does not alter GM2 and GA2 glycosphingolipid accumulation in the double null mice.
Reduced glial cell proliferation and reactive gliosis
We evaluated hematoxylin and eosin (H&E)-stained sections of spinal cord and found that stained nuclei around the margin of the ventral horn and along with the axons of motor neurons in the white matter were at a higher density in the single null Hexb−/− mice compared with double null Hexb−/−Sphk1−/− mice (Fig. 5A, arrows). Immunolabeling with antibody to the cell proliferation marker Ki67 indicated significantly more Ki67-positive cells in single null Hexb−/− spinal cord compared with double null Hexb−/Sphk1−/− spinal cord (Fig. 5A and B). In single null Hexb−/− mice, some of the Ki67-positive cells were also positive for the astrocyte marker GFAP (Fig. 5D, inset). Because most cycling glial cells express the immature cell marker vimentin (20), we examined whether these Ki67-positive cells expressed vimentin. Most Ki67-positive cells were vimentin-positive, indicating that these proliferating cells exhibited an immature astroglial phenotype (Fig. 5C, inset). Levels of vimentin and GFAP mRNA in spinal cord of the double null Hexb−/−Sphk1−/− mice were significantly decreased compared with single null Hexb−/− mice (Fig. 5C and D), consistent with lower numbers of proliferating glial cells and astrocytes in the double null mice. Further, the mRNA level of cyclin D1, a measure of cell proliferation, was also significantly lower in the double null mice compared with single null Hexb−/− mice (Fig. 5E).
Within single null Hexb−/− spinal cord, astrogliosis was demonstrated by the presence of astrocytes with hypertrophy of cell soma, increased process extensions and increased expression of GFAP (Fig. 6A and B). By comparison, astrogliosis appeared to be reduced in the ventral horn of spinal cord of double null Hexb−/−Sphk1−/− mice.
Our data indicated that the absence of Sphk1 activity resulted in a less severe neurodegenerative course that is associated with substantially less cell proliferation of glial precursors and reduced astrogliosis. Because both S1P2 and S1P3 receptor mRNAs were elevated in the disease model, we determined their cellular expression in the CNS by immunohistochemistry. In agreement with other studies (21,22), we observed that the S1P3 receptor was expressed on GFAP-positive astrocytes (Fig. 7A). Also in agreement with previous work (23), the S1P2 receptor was largely expressed on neurons, with little expression detected on GFAP-positive astrocytes (Supplementary Material, Fig. S2). Furthermore, the significant elevation of S1P3 receptor mRNA seen in the Hexb−/− mice was not detected in the double null Hexb−/−Sphk1−/− mice, indicating a link between Sphk1 and S1P3 receptor expression (Fig. 5F). We, therefore, focused our studies on exploring the possible involvement of the S1P3 receptor on the disease phenotype of Sandhoff disease mice and determining if the associated astrocyte response was altered.
Deletion of the S1P3 gene improves the course of Sandhoff disease and blunts astrocyte proliferation
To assess if the S1P3 receptor might influence the disease course of the Sandhoff model, we generated double null Hexb−/−S1P3−/− mice on the C57BL6 background. The Hexb−/−S1P3−/− double null mice had a significantly longer lifespan than the single null Hexb−/− or the Hexb−/−S1P3+/− mice (Fig. 7B). Furthermore, the motor function as determined by performance on rotorod was significantly better in the double null Hexb−/−S1P3−/− mice than in the single null Hexb−/− mice (Fig. 7C). These results demonstrate that genetic deletion of the S1P3 produces a less severe disease course in the Sandhoff model, indicating that the S1P3 receptor is functionally involved in the disease. In spinal cord of double null Hexb−/−S1P3−/− mice, the soma of astrocytes were hypertrophied (Fig. 6A and B); however, the overall pattern of GFAP immunoreactivity appeared less intense than in single null Hexb−/− mice, indicating diminished astrogliosis in the ventral horn. Correspondingly, vimentin (Fig. 5C), GFAP (Fig. 5D) and cyclin D1 (Fig. 5E) mRNA levels in the spinal cord of these double null mice were decreased compared with the single null Hexb−/− mice.
Here we show that S1P metabolism and S1P receptor expression are functionally important events during the terminal stages of a prototypical lysosomal storage disease characterized by severe neurodegeneration. The pathogenesis of storage diseases is complex and involves diverse intracellular neuronal changes (24,25), as well as profound glial responses (14,15,26). Genetic deletion of Sphk1, a key lipid kinase that produces S1P for cell signaling, resulted in a milder disease course in the Sandhoff model mouse. Glial cell proliferation and astrogliosis within spinal cord, a major site of pathogenic changes, were also significantly reduced. Similarly, deletion of the S1P3 receptor, which is enriched on astrocytes and whose expression is normally increased during the disease, produced a milder disease course, with associated neuropathologic changes compatable to those observed in Hexb−/−Sphk1−/− double null mice. Together, the results demonstrate that the Sphk1/S1P3 receptor signaling axis regulates glial proliferation and astrogliosis during the terminal stages of Sandhoff disease.
Astrocytes express mainly S1P1 and S1P3 receptors (10,22,27), with little expression of S1P2, S1P4 or S1P5 receptors. Several studies have suggested that S1P receptor stimulation is a potentially important proliferative signal for astrocytes. Direct injection of S1P into the striatum of mice induces astrocyte proliferation and astrogliosis (28). In cultured astrocytes, S1P has been shown to stimulate multiple S1P receptors, directly leading to activation of phospholipase C, ERK and ultimately cell proliferation (21,29–31). Our results using genetic deletion of the Sphk1 or S1P3 gene in the Sandhoff disease mouse model provide support for the functional relevance of the S1P-induced proliferative pathway in astrocytes.
The source of the pool of S1P generated by Sphk1 for signaling during the course of the disease has not been established. Even in the absence of Sphk1, an increase in bulk spinal cord S1P levels was detected during the disease progression, suggesting that this bulk increase alone was not sufficient for glial proliferation. This increase of S1P in Sandhoff disease brain in the absence of Sphk1 may have been due to a compensatory decrease in S1P lyase expression. Another possibility for the increase might be leakage of plasma S1P into the CNS after breakdown of the blood-brain barrier during the terminal stages of the disease.
Both astrocytes and neurons express Sphk1 and have the capacity to secrete S1P (10). Within hippocampal neurons, Sphk1 was shown to translocate to the plasma membrane and produce S1P for S1P receptor signaling (11). An intriguing possibility is that neurons, as a result of lysosomal storage or other insults, may secrete S1P locally to induce astrocytic proliferation and migration (22) via S1P receptors at the site of damage. In motor neurons of Sandhoff disease mice, Sphk1 was found to be enriched along the plasma membrane, a distribution that has been associated in several cell types with local S1P secretion (32,33). Other additional cellular S1P sources are possible, such as the astrocytes themselves, which are highly proficient at S1P secretion (10). In this case, autocrine S1P signaling in astrocytes may be induced by growth factors that stimulate Sphk1 activation and S1P release, such as basic fibroblast growth factor (30) and lysophosphatidic acid (21).
Reactive gliosis, or astrogliosis, is a typical feature of many neurodegenerative diseases (34–38). In this process, astrocytes respond to CNS damage by proliferation and hypertrophy, with increased expression of GFAP. Astrogliosis can be beneficial after injury in limiting the extent of damage. However, astrocyte responses can be detrimental in some circumstances and actually contribute to degeneration by exacerbating inflammatory pathways, secretion of toxic substances and formation of scars that are inhibitory to repair. Astrocytes have been implicated in the pathogenesis of neurodegenerative disease such as MS (34), Alzheimer disease (35), Parkinson disease (36), amyotrophic lateral sclerosis (37) and prion disease (38). Cell-cycle inhibitors that prevent astrocyte proliferation have been shown to be protective after CNS injury (39). While the results demonstrated here are focused on the terminal stages of Sandhoff disease, it is possible that the regulation of astrocyte proliferation by the Sphk/S1P receptor axis controlling may be broadly relevant to the pathogenesis of other neurodegenerative disorders. FTY720, which interacts with S1P1, S1P3 and S1P5 receptors after phosphorylation by Sphk (40,41), has been shown to significantly reduce relapse rates in MS patients (12). Recent studies have shown that astrocytes are the major CNS cell type responding to FTY720-P in mixed cultures, and that the ligand primarily transmits signals through a Gi pathway, mainly via S1P1 receptor, and promotes ERK activation and migration in astrocytes (22). These results suggest that interference with S1P receptor signaling in astrocytes might be responsible for some of the beneficial treatment effects of FTY720. In addition to astrocytes, FTY720 may have other cellular targets in MS, including oligodendroglia, neurons, leukocytes and endothelial cells, all of which are responsive to S1P signaling (27,42).
Our results demonstrate that astrocyte proliferation can be dampened by disabling the Sphk1/S1P3 receptor signaling axis in a mouse model of neurodegeneration. Because astroglial responses are common in neurodegenerative diseases, the S1P signaling pathway may be broadly relevant to many different disorders and a potential target for therapy.
MATERIALS AND METHODS
Generation of Hexb−/−, either with Sphk1−/− or S1P3−/−, double null mice
Sandhoff disease model mice Hexb−/−, Sphk1−/−, S1P2−/− and S1P3−/− were generated in our laboratory as previously described (13,43,44) and were backcrossed seven times to C57BL6 mice. Male Sphk1−/− and S1P3−/− mice were crossed to female Hexb+/− mice to generate double heterozygous Hexb+/−Sphk1+/− and Hexb+/−S1P3+/− mice, respectively, which were used to generate breeding colonies for each set of double mutant mice. A colony of Sphk/Hexb mutant mice was also generated on a mixed C57BL6/129/sv background for preliminary experiments. The genotype of the Hexb, Sphk1, S1P2 and S1P3 loci was determined for individual mice by polymerase chain reaction (PCR) from tail DNA using two pairs of primers as previously described (13,43,44). All animal procedures were approved by the National Institute of Diabetes and Digestive and Kidney Diseases and were performed in accordance with the National Institutes of Heath guidelines.
Quantitative real-time PCR assay
Total RNA from cerebral cortex and spinal cord was isolated using TRI reagent according to the manufacturer's protocol (Invitrogen). Total RNA (1–2 µg) was reverse-transcribed using a Super-Script II RT Kit (Invitrogen). Real-time PCR was performed using Assays-on-Demand (PE Applied Biosystem) for indicated genes on an ABI prism 7000 Sequence Detection System (PE Applied Biosystems). Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as an internal standard.
Measurement of Sphk activity and S1P levels
Sphk activity and S1P levels were measured in spinal cords from 8- and 16-week-old mice as described previously (45). Labeled lipids were extracted and resolved by TLC, then quantitated with a PhosphorImager (Fuji).
Measurement of GM2 and GA2 levels
The analysis of lipids followed a method described previously (13). Spinal cord lipids were isolated from 16-week-old mice with different genotypes and analyzed by TLC.
Histopathological analysis and immunohistochemistry
Mice at 8 and 16 weeks of age were transcardially perfused with ice-cold 4% paraformaldehyde. Fixed brain and spinal cord were processed for paraffin sections and stained with H&E or cut to 40 µm-thick sections with a vibratome or cryostat and processed for immunostaining as described (15). The antibodies used in immunohistochemical analysis were: rabbit anti-Sphk1 (from Tim Hla, UConn); mouse anti-vimentin (Sigma); rabbit anti-GFAP (DAKO); mouse anti-Ki67 (DAKO); rabbit anti-S1P2 and S1P3 receptor (from Suzanne Mandala, Merck). Immunoreactions were visualized using the avidin–biotin complex technique (ABC kit, Vector). For immunofluorescent staining, the primary antibodies were detected with species-specific affinity-purified FITC—and/or tetramethylrhodamine B isothiocyanate (TRITC)-conjugated IgG (Sigma-Aldrich). The sections were analyzed using a Leica DMR microscope equipped with FITC and TRITC filters.
The rotorod test was used to ascertain motor function (15). The initial speed was set to 5 rpm, and the acceleration was increased by 2 rpm/min. Performance was evaluated weekly from 10 weeks of age and was stopped when the mice could not stay on the rotorod for at least 10 s.
Data are presented as mean±standard error of the mean (SEM). Statistical analysis of the difference between two sets of data was performed using Student's t-test. To analyze the difference between more than two sets of data, one-way analysis of variance followed by Tukey's multiple comparison test was used. Differences were considered to be statistically significant at P < 0.05.
This research was supported by the Intramural Research Program of the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases.
Conflict of Interest statement. None declared.