Multi-system disorders of glycosphingolipid and ganglioside metabolism.

Glycosphingolipids (GSLs) and gangliosides are a group of bioactive glycolipids that include cerebrosides, globosides, and gangliosides. These lipids play major roles in signal transduction, cell adhesion, modulating growth factor/hormone receptor, antigen recognition, and protein trafficking. Specific genetic defects in lysosomal hydrolases disrupt normal GSL and ganglioside metabolism leading to their excess accumulation in cellular compartments, particularly in the lysosome, i.e., lysosomal storage diseases (LSDs). The storage diseases of GSLs and gangliosides affect all organ systems, but the central nervous system (CNS) is primarily involved in many. Current treatments can attenuate the visceral disease, but the management of CNS involvement remains an unmet medical need. Early interventions that alter the CNS disease have shown promise in delaying neurologic involvement in several CNS LSDs. Consequently, effective treatment for such devastating inherited diseases requires an understanding of the early developmental and pathological mechanisms of GSL and ganglioside flux (synthesis and degradation) that underlie the CNS diseases. These are the focus of this review.

The catabolism of complex GSLs also proceeds by stepwise, sequential removal of sugars by lysosomal exohydrolases to the fi nal common products, sphingosine and fatty acids ( Fig. 1 ). Individual defects in GSL hydrolases ( Fig. 3 ) result in excessive accumulation of specifi c GSLs in lysosomes leading to the various lysosomal storage diseases (LSDs) (see Table 1 ). Nonenzymatic proteins are essential to GSL degradation either by presenting lipid substrates to their cognate enzymes or by interacting with their specifi c enzyme ( 2 ). Two genes, PSAP (prosaposin) and GM2A (GM2 activator protein), encode fi ve such proteins ( Fig.  3 ) ( 2,29 ). Four saposins (A, B, C, and D) or sphingolipid activator proteins (Sap) are derived from proteolytic cleavage of a single precursor protein, prosaposin, in the late endosome and lysosome ( 30,31 ). Each of these saposins has specifi city for a particular GSL hydrolase ( Table 1 ).
N -acetylglucosaminyl-, and sialyltransferases can elongate the oligosaccharide chain of GSLs along the luminal side of Golgi ( 18,19 ), thereby defi ning the different series of GSLs. Addition of sialic acid to LacCer forms GM3 ganglioside by LacCer ␣ -2,3-sialyltransferase (GM3 synthase) that initiates synthesis of the ganglioside or sialo-GSL series ( Fig. 1 ). The type of sugars in the oligosaccharide backbone depends on the activity of glycosyltransferases in the Golgi, the cell type, and the developmental or disease stage (19)(20)(21)(22) (see below).
In addition to CERT, several other proteins participate in GSL traffi cking in the cells. GlcCer can be fl ipped into the Golgi lumen or across plasma membrane by the ATPbinding cassette transporter [also called multidrug resistance protein, MDR1, or P-glycoprotein (Pgp)]. Although this has been shown for short-chain GlcCer or neutral GlcCer, transport of naturally occurring GlcCer by Pgp has not been shown in vivo (23)(24)(25). Four-phosphate adaptor protein 2 (FAPP2) can transport newly synthesized GlcCer to the trans Golgi and back to the ER ( 26 ). It is not clear how FAPP2 transports GlcCer through the cytosol to  (16): a) conversion to galactosylceramide (GalCer) in the ER lumen, which is subsequently converted to sulfatide in the mid-Golgi ( 13,14 ), b) vesicular transport to the cytoplasmic face of the cis-Golgi where it is a precursor for GlcCer synthesis (26,316), and c) transport by ceramide transfer protein (CERT) to the mid-Golgi where sphingomyelin is formed within the lumen (319). Once formed, GlcCer also has several fates ( 27,317 ): 1) transport by FAPP2 to the ER and/or to the trans -Golgi lumen where it is converted to lactosylceramide (LacCer) ( 17,26,27 ), 2) to the cytoplasmic side of the plasma membrane by unknown mechanisms, and 3) to the extracellular matrix by exocytosis. Addition of sialic acid to LacCer initiates synthesis of gangliosides or the sialo-GSL series (318). Ceramide can also be generated through degradation of sphingomyelin in the lysosome or at the plasma membrane by aSMase and nSMase (10). Newly synthesized GSLs can exit the cell by exocytic vesicles while membrane and extracellular GSLs can be transported intracellularly via endocytosis with subsequent degradation sphingosine and free fatty acids by hydrolases in the lysosome ( 2,10 ). Abbreviations: DHCer (dihydroceramide), S1P (sphingosine 1-phosphate), SM (sphingomyelin).  Detailed mutation information for each disease can be found in the references.
The specifi c roles of GSLs in survival, proliferation, and differentiation have been demonstrated by knockout mice with embryonic lethality (GCS and SPT) ( 49,50 ) and severe neurodegenerative diseases (CGT) ( 51,52 ). Disrupted gangolioside synthesis in mice with either GM3 synthase ( 53,54 ), GM2/GD2 synthase ( 55 ), or ␣ -N-acetylneuraminide ␣ -2,8-sialyltransferase (GD3 synthase) ( 56 ) knocked out causes differential neurological impairments with normal brain formation. Mice lacking CGT are unable to synthesize galactosylceramide. They showed normal growth but had nerve conduction defi cits ( 51,52 ). A compensatory increase in GlcCer was observed. The early embryonic lethality of the GCS-defi cient mouse delineates the essential roles of complex GSLs in survival ( 50 ). Furthermore, the neurodegeneration in the GCS neuronal knockout showed the importance of GSLs in the maintenance of the CNS ( 57,58 ). To date, three human diseases are associated with mutations in enzymes involved in de novo GSL synthesis (59)(60)(61). SPT consists of three subunits and catalyzes the fi rst step in sphingolipid synthesis ( Fig. 1 ). Six missense mutations in SPT long-chain subunit 1 ( SPTLC1 ) were reported in 26 families that have autosomal dominant hereditary sensory and autonomic neuropathy type 1 ( 62,63 ). These SPT mutations caused changes in substrate specifi city that led to formation of two atypical neurotoxic postnatal day 11, GlcCer continues to decrease coincidently with axonal dendritic arborization during adulthood ( 43 ) (Sun et al., unpublished observations). Around E16 during fetal development, the central nervous system ( CNS) ganglioside distribution patterns shift sequentially from predominantly GM3 and GD3 to the a-and b-series (GD1a, GD1b, and GT1b) and GM1 ( 22 ). GalCer and sulfatide become detectable at E17 and predominate at maturity in the adult mouse brain ( 43 ). A steady increase in sulfatide content begins at about postnatal day 7 and continues into adulthood in the rat, correlating with active myelin formation. This is regional, because sulfatide levels in the spinal cord are 5-fold higher than in cerebral cortex ( 45 ).
Defects of GSL biosynthesis in mice and some human patients provided insights into the importance of GSLs and gangliosides during brain development ( Table 2 ).  permeability transition pores and release of apoptogenic factors ( 75 ). GSLs have differentially ordered domains to create selectivity in membrane transport that is important in the spatial organization of cells ( 76 ). GSLs are enriched to 30-40 mol% in some cell types, e.g., in the apical membranes of intestinal and urinary tract epithelial cells or in the myelin of axons (77)(78)(79). Complex GSLs are generally enriched on the plasma membrane, but GlcCer is mostly localized to intracellular membranes ( 80 ). GSLs are found in the vacuoles of the exocytotic and endocytotic systems with low content in the ER ( 65,73 ). Complex GSLs are not translocatable to the cytosolic surface of Golgi ( 81 ).
Rafts are ordered structures that participate in cell recognition and signaling. These microdomains or lipid rafts are composed of GSLs, sphingomyelin, and cholesterol ( 33,82 ) that provide environments for enrichment of specifi c proteins ( 83 ) in a variety of membranes. Examples include Src family kinases that are important for signal transduction ( 84 ), glycosylation-dependent adhesion/ recognition and signaling ( 85 ), and G-protein-coupled receptor-signaling ( 86 ).
Additional roles of GSLs, including ceramide, sphingosine, sphingosine-1-phosphate, lysoglycolipids, and lysogangliosides, are their inhibition or activation of apoptosis, proliferation, and stress responses ( 87 ). Cellular ceramide is an important second messenger in signal transduction that modulates a variety of these physiologic and stress responses ( For the fi eld of GSL LSDs, a major challenge remains in the unifi cation of the cell and developmental specifi city, biological functions, and pathogenic mechanisms. Given the plethora of diverse and essential cellular functions, the severity of disruptions in GSL metabolism would be anticipated to include more than just accumulation of the GSL in cells and architectural distortions of specifi c cell types. Such mechanistic understanding of the GSL storage diseases is just beginning to be delineated (See below ).

DISORDERS OF GSL AND GANGLIOSIDE DEGRADATION IN HUMANS AND MOUSE MODELS
The catabolic defects in neutral GSL and ganglioside metabolism result in LSDs with complex, multi-system progressive diseases, including neurodegeneration, that can present early in utero or childhood. All the catabolic enzymes for these diseases exist in the lysosomes ( Fig. 3 ). Later onset and more attenuated variants of each disorder have been or are anticipated to be present in humans ( Table 1 ). All these deoxyl-sphingoid bases by condensation of palmitoyl-CoA and alanine or glycine, instead of serine ( 62 ). Thus, hereditary sensory and autonomic neuropathy type 1 is caused by a gain-of-function mutation. GM3 synthase catalyzes the fi rst step of complex gangliosides synthesis ( Fig. 1 ). A lossof-function mutation in GM3 synthase ( SIAT9 ) was identifi ed in a cohort characterized by autosomal recessive infantile-onset symptomatic epilepsy syndrome ( 64 ). A nonsense mutation in exon 8 of the GM3 synthase gene produced premature termination that could abolish enzymatic activity ( 64 ). Affected individuals are homozygous for the mutation while heterozygous carriers are unaffected ( 64 ). A single case of GM2 synthase defi ciency was reported by biochemical characterization only ( 61 ). The patient died at 3 months after presenting with abnormal motor function and seizures. GM3 accumulated in the brain and liver ( 61 ).
Phenotypic analyses from GSL synthase-defi cient mice reveal that lack of all GSL is incompatible with embryonic development ( 49,50 ). Simple to complex GSLs are nonessential for early brain development but are important for brain maturation and maintenance (51)(52)(53)(54)(55)(56). Compensatory interchange between various GSLs also signifi es additional complexities in their metabolism.

Distribution and functions of GSLs
The sialic acid-containing GSLs are ubiquitously expressed in the outer leafl et of the plasma membranes, e.g., liver ( 65 ), and are most abundant in CNS and other nervous tissues ( 22,65,66 ), but there is signifi cant tissue/ regional variability in distribution. High concentrations of GD1a are present in extraneural tissues, erythrocytes, buffy coat, bone marrow, testis, spleen, and liver, whereas different GSLs are in high amounts in other tissues, e.g., GM4 in kidney, GM2 in bone marrow, GM1 in erythrocytes, and GM3 in intestine ( 19 ). GA2, GM2, and GM3 are at very low or nondetectable levels in normal brains from humans, cats, and mice ( 66 ). Cellular specifi city is exemplifi ed by GD1b localization to small neurons, whereas O-Ac-disialoganglioside localizes to large neurons ( 67 ). In liver, GM1 was detected on the canalicular and sinusoidal lining cells, but nearly absent from liver parenchymal cells; GM1 was increased in hepatic sinusoidal membranes during cholestasis ( 68 ). In skeletal muscle, the neutral GSLs and gangliosides were mostly in membrane vesicles with GlcCer predominating and only trace levels of LacCer ( 69,70 ). GalCer, sulfatide, and sphingomyelin are structural to myelin sheaths and are major lipids in oligodendrocytes and Schwann cells ( 71 ). In skin, ceramide comprises about 50% of total epidermal lipid and is generated from GlcCer in the lamellar bodies ( 72 ). These cellular microdifferences in GSLs indicate subtle, yet potentially important, metabolic roles that have not been elucidated. Cellular distribution of GSLs can be altered in pathologic states. In MDCK kidney cells, mitochondrial and peroxisomal GSLs are very low/absent ( 73 ), whereas GD1b and GD3 had high contents in mitochondria from malignant hepatomas ( 74 ). During the apoptotic process, GD3 is rapidly synthesized and relocated from the cis Golgi to mitochondria, leading to the opening of mitochondrial ease: Types A and B (NPA and B) ( Fig. 1, Table 1 ). NPA presents a neurovisceral phenotype, whereas NPB exhibits visceral manifestation ( 104 ). The ASM-defi cient mice, developed independently in two laboratories, resemble NPA or NPB phenotypes ( Table 1 ) ( 105,106 ). Neither model has detectable lysosomal ASM activity in any tissue but has reduced levels of plasma membrane-bound nSMase activity, which cleaves nonlysosomal sphingomyelin at neutral pH ( Fig. 2 ) ( 106, 107 ). The ASM gene is expressed prior to E11 during neuronal stem cell proliferation, and deficient ASM activity, due to genetic defects, can lead to loss of signaling molecules that regulate their proliferation very early on in brain development ( 108,109 ).
Krabbe disease, globoid leukodystrophy disease, is a rapidly progressive, demyelinating disease that results from insuffi cient cleavage of galactosylceramide and galactosylsphingosine by galactosylceramide-␤ -galactosidase (GALC; EC 3.2.1.46) ( Table 1 , Fig. 3 ). This enzyme has specifi city for both galactosylceramide and galactosylsphingosine, the latter being a specifi c substrate for this enzyme whereas galactosylceramide can be cleaved by at least two other lysosomal ␤ -galactosidases ( Fig. 1 ) ( 110 ), potentially accounting for the lack of large accumulations of galactosylceramide in the CNS ( 111 ). GALC has a number of nonsynonymous polymorphisms (missense or nonsense mutations) that affect the expression and function of the enzyme ( 111,112 ). There are four forms of the disease that differ in age of onset ( 113 ). Secondary pathological changes including reactive astrocytic gliosis, infi ltration of unique and often multinucleated macrophages (globoid cells), and other infl ammatory responses accelerate disease progression ( 111 ). Bone marrow/stem cell transplantation with histocompatible cells has shown some positive effects on the CNS and peripheral nervous system (PNS) phenotypes in the infantile onset disease when performed prior to the onset of signifi cant neurological involvement. However, the dementia and other aspects of the CNS and PNS disease eventually lead to death ( 114 ). GALC-deficient mice ( twitcher mouse) were discovered at the Jackson Laboratory in 1976 ( 115,116 ). This naturally occurring model is an excellent model of the infantile human disease (117)(118)(119). A transgenic globoid leukodystrophy disease Krabbe disease model was created by 'knock-in' of a missense mutation (H168C) ( 120 ). The analyses of these models demonstrated that there is a correlation between accumulation of galactosylsphingosine and the neuronal defects. The accumulation of galactosylsphingosine was detected before myelin formation ( 119 ). Bone marrow transplantation in twitcher mice stabilized galactosylsphingosine at low levels accompanied with remyelination ( 121 ). These fi ndings provide support for galactosylsphingosine as the offending agent responsible for triggering proinfl ammation, oligodendrocyte depletion, and dysmyelination in Krabbe disease.
MLD is caused by a defi ciency of arylsulfatase A (ASA; EC 3.1.6.8) that leads to the accumulation of sulfatide, a major lipid of myelin, in the CNS and PNS ( Fig. 3 ) ( 122,123 ). Currently, fi ve allelic forms of ASA defi ciency exist ( Table 1 ). Also, there are two nonallelic variants of ASA diseases are autosomal recessive except Fabry disease, which is X-linked. The phenotypic variants and the molecular causes have been variably delineated. Attempts have been made to correlate the genotype with phenotype. This is a complex and incomplete task because of the variability in presentation, nonuniform clinical delineation, and lack of complete allele characterization. However, the residual in situ hydrolase activities in patients are a major, but not unique, determinant of the age at onset and/or severity of the disease. Work by Sandhoff et al. ( 94 ) clearly established in situ residual activity as a critical correlate of phenotype with a direct, albeit nonlinear, relationship between this activity and the age at onset in Tay-Sachs disease and metachromatic leukodystrophy (MLD). Extensive clinical data show similar correlations in Gaucher disease type 1 ( 95 ). Mouse models of several LSDs have been developed ( Table  1 ). Although the mouse phenotypes are qualitatively similar to its corresponding human disease, such models can show substantial differences because of differences in GSL metabolism between humans and mice. Analyses of such mouse models of GSL synthesis and degradation have provided insight into early pathological mechanisms of GSL and ganglioside fl ux aberrations.
The disorders of ceramide, GSLs, and gangliosides degradation in humans and mice are summarized below. The phenotypes and genotypes of these disorders are summarized in Table 1 .

Disorders of ceramide degradation
Disorders of ceramide degradation are caused by acid ceramidase (AC; EC 3.5.1.23) defi ciencies that lead to Farber disease, a.k.a. Farber lipogranulomatosis ( 96,97 ). AC catalyzes the hydrolysis of ceramide to free fatty acid and sphingosine, and its function is optimized by saposin D ( Figs. 1 and 3 ) ( 98 ). However, AC can also synthesize ceramide from sphingosine and free fatty acids ( Fig. 1 ) ( 99,100 ). This reverse reaction is optimally catalyzed at pH ‫ف‬ 6.0 compared with the hydrolytic activity optimum of pH ‫ف‬ 4.5, suggesting that the two reactions likely occur in different subcellular compartments. There are seven variant forms of Farber disease that are distinguished by severity and tissue involvement ( 96 ). Although allogenic hematopoietic stem cell transplantation in Farber disease patients without CNS manifestation showed improvement in joint movement ( 101 ), there is currently no effective therapy for this disease. The AC-deficient mouse (Asah1 Ϫ / Ϫ ) undergoes apoptotic death at the 2-cell stage (E1; Fig. 4 ) because of excessive ceramide accumulation ( 39,102 ), demonstrating that AC is expressed in the embryo and that ceramide degradation is essential for survival beyond the 2-cell stage (E1) ( 102 ). In addition to increasing ceramide pools, loss of AC activity is thought to concurrently reduce pools of sphingosine and sphingosine-1-phosphate, two lipids known to promote cell growth and differentiation ( 103 ).

Disorders of GSL degradation
Mutations in the SMPD1 gene that encodes aSMase (ASM; EC 3.1.4.12) cause two types of Niemann-Pick dis-ment therapies are in clinical trials to treat type 1 and 3 diseases ( 146,147 ). There are no direct treatments available for Gaucher disease type 2. The Gaucher disease mouse models ( Table 1 ) demonstrate that accumulation of GCase substrates, GlcCer and glucosylsphingosine, begin at embryonic day 13 ( 148 ). Also, substrate storage does not affect brain formation but does lead to early and severe degeneration of the brain and disruption of the skin permeability barrier ( 149,150 ). The downstream biological consequences of GlcCer and glucosylsphingosine accumulation are currently being investigated (see section III).
GM2 gangliosidoses are a group of severe neurodegenerative disorders unifi ed by the excessive accumulation of GM2 ganglioside mainly in neuronal cells. GM2 is specifically hydrolyzed by ␤ -hexosaminidase A (EC 3.2.1.52) ( Fig.  3 ). Two genes are required for the formation of this enzyme. The HEXA and HEXB genes encode the ␣ -and ␤ -subunits, respectively ( 161,162 ). The subunits assemble in the Golgi forming ␣ ␤ 2 [ ␤ -hexosaminidase A (Hex A)] and ␤ 3 [ ␤ -hexosaminidase B (Hex B)] ( 161,163,164 ). The nonallelic GM2A gene encodes the GM2 activator protein that complexes with the ganglioside GM2 ( 165 ) (see below). Defects in any of these three genes may lead to impaired degradation of GM2 gangliosides and related substrates. Tay-Sachs disease (variant B) is the classic GM2 gangliosidosis and results from the defi ciency of the ␣ -subunit that produces the specifi c loss of Hex A, whereas Hex B is at normal or somewhat increased levels, thereby designated as variant B ( 161 ). This specifi c defi ciency in Hex A produces GM2 ganglioside accumulation in brain, because GM2 ganglioside is synthesized in appreciable amounts only in the CNS ( Fig. 3 ). Currently, there is no effective treatment for Tay-Sachs disease, although trials of "molecular chaperone" approaches are underway for the late onset variants with residual Hex A activity ( 166 ). Sandhoff disease (variant 0) is caused by mutations of the HEXB gene that leads to loss of the ␤ -subunit, resulting in defi ciency of both Hex A and Hex B activities ( Fig. 3 ). Unlike Tay-Sachs patients who store mainly GM2-ganglioside, Sandhoff patients accumulate additional N-acetyl defi ciency: 1) MLD due to saposin B defi ciency (described below) leads to inability to cleave sulfatide and globotrioaosylceramide (Gb3) in vivo as a result of the activator protein defi ciency and impaired lipid presentation to the enzyme, even though ASA and ␣ -galactosidase A are normally expressed and without mutations ( 124 ); and 2) multiple sulfatase defi ciency is a disorder that combines features of a mucopolysaccharidosis with those of MLD ( 125 ) and results in a defi ciency of all cellular sulfatases due to defects in a posttranslational enzyme essential for modifi cation of an active site cysteine that is present in all sulfatases ( 122,126 ). Details of multiple sulfatase deficiency have been reviewed ( 125,127 ). For the infantile and juvenile MLD variants, bone marrow/stem cell transplantation has been attempted with initial improvement/ stabilization followed by later progression of the CNS and PNS diseases ( 128 ). ASA-defi cient mice (Asa Ϫ / Ϫ ) have a very slowly progressive sulfatide storage resembling the early stages of human cases ( 129 ). A more severe phenotype of ASA defi ciency was achieved by overexpressing the transgene for the sulfatide-synthesizing enzyme, galactose-3-O -sulfotransferase (CST), encoded by Gal3st1 , in oligodendrocytes ( 130 ). These models have been used in preclinical trials of MLD ( 131,132 ).
Fabry disease, the only X-linked GSL storage disease, is due to a defi ciency of ␣ -galactosidase A ( 133 ), the enzyme that cleaves Gb3 and di-galactosylceramide ( Fig. 3 , Table 1 ). Uncharacteristic of most X-linked traits, both males and females have signifi cant to major clinical involvement ( 134 ). Affected patients accumulate the substrate, Gb3, in most tissues and organs ( 135 ), but the progressive deposition of Gb3 in capillary endothelial cells leads to many of the morbid manifestations ( 136 ). Increased deacylated Gb3, globotriaosylsphingosine (lyso-Gb3), in patient plasma suggests that it may be involved in pathogenesis of Fabry disease ( 137 ). The mouse model of Fabry disease (Gla Ϫ / Ϫ ) has accumulation of Gb3 in peripheral nerves and alterations of sensory nerve function that resembles neuropathic pain in Fabry disease patients ( 138 ). Gb3 accumulation in endothelium leads to vascular dysfunction, thereby providing an in vivo model to delineate the basis of cardio-and cerebrovascular complications associated with Fabry disease ( 139,140 ).
Gaucher disease results from insuffi cient cleavage of GlcCer and glucosylsphingosine by the lysosomal enzyme, acid ␤ -glucosidase or glucocerebrosidase (GCase) ( Fig. 3 ) ( 95,141 ). A homologous pseudogene ( GBA) located 16 kb downstream from the GBA gene has complicated the molecular diagnosis of Gaucher disease ( 142 ). The phenotypes are a continuum of degrees of involvement but can be divided categorically into neuronopathic (types 2 and 3) and nonneuronopathic (type 1) variants ( Table 1 ). A nonlysosomal ␤ -glucosidase (GBA2) has been identifi ed as participating in GlcCer degradation ( 143 ). Inhibition or knockout of GBA2 activity is associated with GlcCer accumulation in testis and impairment in spermatogenesis ( 144 ). However, the link of GBA2 to Gaucher disease is unclear. Enzyme replacement therapy has been successfully used to treat Gaucher disease type 1 ( 145 ), whereas new substrate synthesis inhibition and enzyme enhance-Sap B, or Sap C defi ciencies in humans are rare ( 181,(187)(188)(189). Sap D activates AC for degradation of ceramide. Heteroallelic mutations were identifi ed in the Sap D domain in a patient with an additional mutation in the Sap C domain on a different allele ( 189 ). Prosaposin and all four Sap-defi cient mice have been developed ( 98,(190)(191)(192)(193)(194). In general, individual Sap defi ciencies in mice results in slowly progressive diseases compared with total Sap-deficient mice or their cognate enzyme knockout mice ( Fig. 3 , Table 1 ). These could be due to cross-compensation in GSL degradation by multiple saposins ( Fig. 1 ). Proinfl ammation is a common pathological fi nding in the CNS from such models because of GSL accumulation. The clinical, pathological, and biochemical phenotypes of prosaposin knockout mice closely resembles those of the human disease, whereas Sap B Ϫ / Ϫ and Sap C Ϫ / Ϫ models are more slowly developing than those in humans.

Lipid traffi cking disorders
Niemann Pick disease Type C (NPC) is not a disorder directly related to the metabolism of GSLs and gangliosides but rather has major involvement in the egress of free cholesterol from the lysosome ( 195 ). NPC is characterized by the accumulation of unesterifi ed cholesterol and other lipids in endosomal/lysosomal compartments that stem from inherited defi ciencies of the NPC1 or NPC2 proteins. NPC1 is a transmembrane protein and NPC2 is a soluble protein. Both are involved in intralysosomal cholesterol traffi cking ( 196 ). Recently, NPC2 was proposed to act as a donor of cholesterol to NPC1 that facilitates cholesterol transport across the lysosomal membrane ( 196 ). Loss-offunction mutations in the NPC1 gene also lead to failure of the calcium-mediated fusion of endosomes with lysosomes, resulting in the accumulation of GlcCer, LacCer, and GM2 in late endosomes and lysosomes ( 195 ). The exact mechanism of disease causality by either cholesterol (primarily) or GSLs (secondary or primary) accumulation has not been resolved. Elevation of sphingosine was found before cholesterol and GSL accumulation ( 195 ). The importance of sphingosine in NPC1 has been addressed recently ( 197 ). NPC1-defi cient mouse models (NPC1 Ϫ / Ϫ ) show similarities with the human clinical NPC1 phenotypes ( 198 ). Lipid profi le and disease phenotype in NPC1, NPC2, and NPC1/ NPC2 double mutant mice were qualitatively and quantitatively very similar, indicating that both proteins operate cooperatively without compensation for each other in traffi cking cholesterol to the lysosomes ( 196,(198)(199)(200).
Most GSL LSDs present with neurological phenotypes that begin early in life. The analyses of the mouse models show that the lipid accumulation in the CNS starts very early on in development. Such storage may subtly alter brain and cellular functions. GSL accumulation has been postulated as a trigger to abnormal cellular responses that potentiate the pathological disease process. Although some of the mouse phenotypes differ from the analogous human disease, the biochemical profi les in the GSL synthesis or degradation pathway have provided insight into the pathological mechanisms and GSL fl ux aberrations. However, systematic analyses of GSL profi les in relation to ␤ -galactosyl terminated glycolipids and oligosaccharides in the CNS and viscera ( 167,168 ).
The Tay-Sachs disease mouse model (Hex A Ϫ / Ϫ ) displays a milder phenotype compared with Tay-Sachs patients. This is accounted for by the differences in GM2 catabolism between mouse and human ( 169 ). Two pathways have been proposed for GM2 metabolism in mice. One involves GM2 hydrolysis to GM3 mainly by Hex A and GM2 activator, which is the major pathway in humans ( Fig.  3 ) ( 161 ). The second one is specifi c to mice and involves GM2 hydrolysis by sialidase fi rst to GA2 and then to LacCer by Hex B. In Hex A Ϫ / Ϫ mice with a complete loss of Hex A, GA2 can be degraded by Hex B ( 169 ). Unlike Hex A Ϫ / Ϫ mice that have minor GM2 storage in neurons and no detectable neurological impairment, Sandhoff mice (Hex B Ϫ / Ϫ ) are severely affected. The lack of Hex A and B activities in Sandhoff mice provides an authentic model of acute human Sandhoff disease ( 169 ). Substrate reduction therapy with N-butyldeoxygalactonojirimycin (NB-DGJ), an imino sugar that inhibits GCS, has been tested in adult ( 170 ) and neonatal ( 171 ) Sandhoff mice. NB-DGJ significantly reduced total brain and GM2 ganglioside content when administered from postnatal days 2-5 in Sandhoff mice to a greater extent than in the adult mice. No adverse effects were found with NB-DGJ treatment in either age group. These results indicate that earlier intervention is an effective method for managing GSL LSDs.

Activator protein defi ciency
The physiologic importance of sphingolipid activator proteins in GSL degradation has been highlighted in the patients with mutations in the GM2 activator and the prosaposin genes; the latter encodes a precursor of four saposins (A, B, C, and D).
GM2 activator (variant AB) defi ciency is caused by mutations that lead to an inability to form GM2/GM2-activator complexes in the presence of normal amounts of Hex A and Hex B ( Fig. 3 ) ( 165,172,173 ). Mechanistically, the GM2 activator protein is a "liftase" that extracts GM2 from intra-lysosomal membranes making the GM2 terminal N-acetyl-galactosamine accessible to ␤ -hexosaminidase A for cleavage ( 174 ). The Gm2a Ϫ / Ϫ mice store mostly GM2ganglioside and low amounts of GA2 ( 175 ), similar to the Hex A Ϫ / Ϫ mice, thereby establishing its essential role in the GM2 catabolism ( 175 ).
Prosaposin/total saposin defi ciency was found in four families and was associated with multiple GSL storage ( Table 1 ) (176)(177)(178). Saposin (Sap) A is an in vivo activator for GALC ( 124 ). Sap B has lipid transfer properties and participates in degradation of sulfatide by ASA, digalactosylceramide, and Gb3 by ␣ -galactosidase A, and GM1 ganglioside and LacCer by ␤ -galactosidases ( Fig. 3 ) ( 179, 180 ). Sap C is required for GCase to achieve maximal activity in vitro and ex vivo and protects GCase from proteolysis ( 181,182 ). In comparison to Sap B, Sap C physically interacts with either phospholipid membrane or acid ␤ -glucosidase to optimize this enzyme's activity ( 183,184 ). Direct binding of Sap C to the substrate, GlcCer, has not been shown as part of this optimization ( 185,186 ). Individual Sap A, toxicity is evident in humans and mice with Gaucher disease. GSL composition studies show that glucosylsphingosine and GlcCer in Gaucher type 1 brains are substantially lower (3-10%) than that in Gaucher disease type 2 brain ( 203 ), suggesting a relationship between substrate accumulation and severity of brain disease. Galactosylsphingosine levels in twitcher mice brain progressively increase from 20 to 40 days as does the neurological phenotype ( 119,204 ). Unlike galactosylsphingosine in Krabbe disease, glucosylsphingosine has not been proven in vivo as a direct cause of CNS Gaucher disease. However, a recent glucosylsphingosine storage disease mouse model supports this concept ( 205 ). The correlations between disease progression and age-dependent GSL levels also occur in NPC disease ( 206 ), Tay-Sachs disease ( 207 ), and GM1 gangliosidosis mice ( 156 ).
GSL accumulation occurs prior to the onset of a disease phenotype, e.g., glucosylsphingosine accumulation was found at the 11 th week of pregnancy in human Gaucher disease ( 208 ). Elevated glucosylsphingosine can be detected as early as E13 during neurogenesis in Gba knockout mice ( 208 ). In V394L/SapC Ϫ / Ϫ mice, increased glucosylsphingosine is present by 14 days, i.e., before the onset of neurological signs at 30 days ( Table 1 ). Abnormal accumulation of galactosylsphingosine in twitcher mice is detected in the brain and sciatic nerves at postnatal day 7, before the onset of demyelination, and progressively increases to 100-fold normal levels, predominantly in white matter ( 209,210 ). In normal mice, GM2 is not detectable, but GM2 is elevated at postnatal day 2 in Sandhoff mice before any pathological signs ( 171 ). Such early lipid accumulation may trigger molecular alterations, e.g., proinfl ammation, as shown by global gene expression analyses in Sandhoff disease mice in which upregulation of infl ammation-related genes preceded neuronal death ( 211 ). Similar temporal analyses of CNS tissues from prosaposindefi cient mice reveal the regionally specifi c gene expression abnormalities including upregulation of numerous proinfl ammatory genes prior to neuronal degeneration ( 212 ). Although low-level excess of GSLs may not lead to early histopathological changes, they do cause cellular alterations that initiate disease processes leading to eventual end stage disease. A fi nal common pathway in the CNS for the GSL LSDs appears to be varying types and degrees of infl ammatory and proinfl ammatory reactions leading to apoptotic or other types of programmed cell death. Increased lipid levels at early embryonic stages also correlate with reduced expression of genes for neurotrophic factors [brain-derived neurotrophic factor (BDNF) and nerve growth factor (NGF)] and MAPK pathway genes (ERK) in the cortex, brainstem, and cerebellum of neuronopathic Gaucher disease mice at E17.5, E19.5, and P1 ( 213 ). The defi ciencies in neurotrophic factors (BDNF and NGF) cause sensory neuron degeneration in mice ( 214,215 ). These early abnormalities in lipid and neurotrophic factors correlated with substantial reductions in cellular density, neurodegeneration, and microglial infi ltration in Gba knockout mouse brains at E13-14 ( 149,150,208 ). Injection of the GCase inhibitor (cyclopellitol) early brain development ( Fig. 4 ) and the early molecular events have yet to be elucidated. Such understanding should be useful in defi ning the initiating and propagating mechanisms that would provide a road map for designing therapeutic strategies.

PATHOLOGICAL CONSEQUENCES OF DISORDERS IN GSL METABOLISM
GSL disorders present complex disease phenotypes and diverse pathological changes. The defects in GSL metabolism affect cellular and molecular functions and lead to abnormalities of the CNS and visceral organs ( Fig. 5 ). The mechanisms leading to this pathology are poorly understood. Are these disorders triggered by accumulated lipid or insults from secondary molecular defects in the cells? Does infl ammation potentiate the disease? What causes neuronal drop out? Could lesions result from lack of products as well as from storage of substrate in the GSL pathway?
The following section summarizes some of the pathological and molecular fi ndings, which may be preludes to understanding the signature pathways in the GSL diseases.

Evidence of GSL accumulation correlating with disease progression
Disruption of GSL fl ux is thought to initiate and potentiate LSD phenotypes. The slow accumulation of sulfatide in Asa Ϫ / Ϫ mice does not disrupt myelin formation, which occurs in the fi rst 3 weeks postnatally in the mice but does destabilize myelin integrity in adults ( 201 ). Although lysosomal accumulation of Gb3 begins in utero, the signs and symptoms of Fabry disease may not be apparent until childhood or adolescence ( 202 ). The association of GlcCer, and, especially glucosylsphingosine, with neuronal into wild-type mice led to a 17-to 31-fold increase of glucosylsphingosine in brain, liver, and spleen with concomitant neurological abnormalities ( 216 ). These fi ndings provide further evidence for the relationship of abnormal substrate levels in GCase defi ciencies and neuronopathic phenotypes. These data suggest that a toxic stimulus is present long before birth or at least early in postnatal life, and that therapy begun during infancy may not prevent progressive neurologic damage. In this regard, the conditional experiments of Xu et al. ( 150 ) are instructive. A chemically induced model of CNS Gaucher disease was produced by administering a covalent inhibitor, conduritol B epoxide (CBE), of GCase to neonatal mice ( 150 ). Wild-type and GCase point mutated mice were given daily injections of CBE from postnatal day 5 to day 11. About 1 day after CBE treatment was stopped, the mice developed neuropathic signs. No mice survived if CBE was continued through day 13. Because the wild-type and mutant GCases have t 1/2 ‫ف‬ 60 h in cells ( 217 ), the neuropathic signs in wild-type mice or those with D409H homozygosity stabilized for the next several months in the absence of CBE. However, with a lower activity mutant, D409V/null, the neuropathic signs progressed to very severe within 2 months; unlike the wild-type and D409H homozygotes, the glucosylceramide levels in brain did not decrease, and there was continuing evidence of neuroinfl ammation. These data support the concept that specifi c reconstitution levels of in situ activity must occur at the appropriate developmental time frame to avoid propagation of disease. Furthermore, once established, the neuronopathic disease does not resolve but can be stabilized.

Secondary storage or defi ciency of GSLs
Alterations of GSL fl ux can directly affect membrane composition and disrupt cell homeostasis that may cause secondary storage ( 38 ) and potentiate disease processes. Such secondary accumulations of GSLs and gangliosides are common features associated with neuropathology in several LSDs, e.g., Niemann-Pick diseases with GM2 and other GSL accumulation (218)(219)(220). Pronounced increases in gangliosides, GlcCer, and LacCer occur in prosaposin defi ciency ( 178,190 ) and increases in GM2 and GM3 are evident in the cathepsin D-defi cient mouse ( 221 ), as they are in the CNS of several mucopolysaccharidoses, MLD, Tay-Sachs disease, and Gaucher disease ( 222 ). Importantly, in neurons with secondary gangliosides accumulation, the pathological changes were essentially identical to those in the gangliosidosis ( 223 ). Because GM2 and GM3 are precursors of many gangliosides and are the fi nal common metabolites in the ganglioside degradative pathway ( 224 ), the discovery of secondary ganglioside accumulation indicates the importance of the endosomal/lysosomal system in the overall regulation of ganglioside expression in neurons.
In the NPC 1 mice, sphingosine accumulation reduces the lysosomal calcium pool and may trigger secondary GSL storage ( 195 ). GCase mutations, even in the heterozygous state, are predisposing factors for Parkinson disease ( 225,226 ). It is unknown whether the GSL or other cellular abnormalities determine the association between these two diseases, but increased ␣ -synuclein deposition and Lewy bodies occur in Gaucher disease type 1 brains ( 227 ).
A clear connection has yet to be established between the primary lysosomal defect and the accumulation or defi ciency of secondary compounds that become involved in the storage process. Signal transduction derangements and altered traffi cking of GSLs in the endosomal/lysosomal system would be possible causes. Regardless of the mechanism, secondary GSL storage compounds appear to be actively involved in disease pathogenesis.

Infl ammation
Infl ammation, constituting a local response to cellular insult, occurs in visceral and CNS tissues as a general pathological event in GSL and ganglioside LSDs. The excess lipid is causally linked to a proinfl ammatory response, including activation of the microglial/macrophage system to produce infl ammatory cytokines leading to CNS damage. In Gaucher disease type 1, GlcCer accumulates mainly in cells of mononuclear phagocyte origin. Serum levels of macrophage-derived cytokines that are pro and antiinfl ammatory mediators have been variously elevated ( 228 ). They include IL-1 ␣ , IL-1 receptor antagonist, IL-6, TNF ␣ , pulmonary and activation-regulated chemokine, and soluble IL-2 receptor (229)(230)(231)(232). Such mediators clearly could play a role in disease progression (233)(234)(235). In ASM knockout mice, the majority of genes with elevated expression in lung and brain were those involved in the immune/ infl ammatory response ( 236 ). Activation of microglial cells and astrocytes in CNS and PNS parallels such cytokine increases. Ex vivo, gangliosides activate microglia to produce the proinfl ammatory mediators nitric oxide and TNF ␣ ( 237 ). Alternatively, gangliosides also suppress toll-like receptor-induced proinfl ammatory cytokine expression ( 238 ). In Sandhoff disease mice and humans, macrophage/microglial activation and subsequent cytokine expression result in neuronal dysfunction ( 211 ). Similarly, progressive CNS infl ammatory changes were coincident with the onset of clinical signs in Tay-Sachs, Sandhoff, and GM1 gangliosidosis mice ( 239 ).
Use of antiinfl ammatory drugs in several GSL LSD models slowed disease progression ( 240,241 ). This indicates that multiple infl ammatory pathways participate. Nonsteroidal cyclooxygenase-dependent antiinfl ammatory drugs (e.g., indomethacin, aspirin, and ibuprofen) extended the lifespans of Sandhoff disease and NPC1 mice ( 240,242 ). In Sap A Ϫ / Ϫ mice, such treatment suppressed only early, but not later, proinfl ammation, indicating involvement of the cyclooxygenase pathway, but it is not the major infl ammatory pathogenic pathway (Barnes et al., unpublished observations). Strikingly, continuous high levels of estrogen during the pregnancy period of Sap A Ϫ / Ϫ mice protected against proinfl ammation and demyelination even though galactosylceramide and galactosylsphingosine continued to accumulate ( 191,243 ). These studies implicate the proinfl ammatory reactions in the propagation and potentially initiation of GSL-mediated CNS degeneration.
The underlying molecular mechanisms of GSL-or ganglioside-induced infl ammation are not fully elucidated. Several disparate results have suggested signaling pathways. Macrophage inducible nitric oxide synthase can be regulated by GSL-mediated intracellular calcium elevation ( 244 ). AMP-activated protein kinase plays a role in regulating astrocytic galactosylsphingosine-mediated infl ammatory responses ( 245 ). Similarly, sulfatide induces infl ammatory cytokine production in microglia and astrocytes in addition to stimulating phosphorylation of p38, ERK, and JNK ( 246 ). Whole genome transcriptome analyses of prosaposin-deficient mice implicated CEBP ␦ as a common mediator in the STAT3 proinfl ammatory pathway ( 212 ). GlcCer-derived ceramide may play an antiinfl ammatory role by mediating p38 ␦ MAPK activation, p38 ␦ phosphorylation, and regulation of IL-6 ( 247 ); the partial defi ciency of GlcCer-derived ceramide in Gaucher disease may release, via p38 ␦ MAPK, suppression of IL-6 and the consequent downstream infl ammatory effects ( 247 ). Greater insights into these basic mechanisms could facilitate developing adjunctive approaches to altering the progression of CNS disease in the GSL LSDs.
Several GSLs have roles as immune modulators through invariant natural killer T (iNKT) cell selection that relies on interaction of glycolipid antigens through the MHC class I-related glycoprotein CD1d in late endosome/lysosomes ( 248,249 ). In addition to ␣ -galactosylceramide, the mammalian lipid, isoglobotrihexosylceramide, was proposed as an endogenous glycolipid antigen ( 249 ). This has been questioned since the isoglobotrihexosylceramide synthase knockout mice do not have iNKT cell abnormalities ( 250 ). Defi ciency of iNKT cells occurs in prosaposin defi ciency, Sandhoff disease, and NPC1 disease ( 248,249 ), indicating that several GSLs, gangliosides, and lysophospholipids may participate in iNKT cell selection ( 251 ), but the role of these cells and the molecular connection remain unclear.

Neuronal degeneration
The majority of GSL LSDs with CNS involvement result from processes that lead to neuronal death ( Table 1 , Fig.  5 ). CNS neuronal cell loss and gliosis in Gaucher disease types 2 and 3 correlate with GlcCer and, particularly, glucosylsphingosine levels ( 252,253 ). Neurological deterioration in human GM1 gangliosidosis can span a period from birth up to 30 years; apoptosis is the major mechanism of neuronal death, but this is not universal, because several other cell death mechanisms have been implicated in different LSDs. Apoptosis is well understood and is linked to an unfolded protein response (UPR) that includes upregulation of BiP and CHOP, and activation of JNK2 and caspase-12 ( 254 ). GM1 loading of wild-type neurospheres causes depletion of ER calcium stores and activation of the apoptotic pathway, implicating ER GM1 accumulation as an inducer of UPR ( 254 ). However, the UPR is not general for GSL LSDs, because UPR is not evident in the Gaucher disease brain ( 255 ). Apoptotic cell death does occur in mouse and human Sandhoff and Tay-Sachs diseases that have CNS accumulation of GM2 and GA2 ( 169,256 ). How-ever, no Hex A or Hex B protein is made in the respective models and, thus, malfolding of a mutant protein cannot be implicated. Apoptotic neuronal death is not evident in several GSL disorders, including prosaposin knockout, Gaucher disease, and individual saposin defi cient models. Recently, alterations of autophagy pathways in several LSDs, e.g., NPC1 and Batten disease ( 257 ), suggest that impairment of the autophagosome system that can lead to neuron death by as-yet-undetermined mechanisms. Altered calcium homeostasis was described in NPC1 disease ( 195 ), ASM knockout mice ( 102,109 ), Sandhoff mice ( 258 ) and GM1 gangliosidosis mice ( 254,259 ). A mechanistic link among GSL accumulation, Ca 2+ homeostasis, and neuronal cell death may be of signifi cance for delineating the neuronal degeneration in GSL disorder diseases.
Purkinje cells are a unique population of neurons that are very vulnerable to pathogenic insults, and their degeneration occurs in several GSL storage disease mice ( 98,192,193,260,261 ). The losses of Purkinje cells in these mice progress chronologically from lobule I to X . In ASM knockout mice, progressive Purkinje cell loss begins at ‫ف‬ 7 weeks, coincident with the appearance of defi cits in motor function ( 261 ). Purkinje cell death is detected in 60 day old NPC1 mice that store cholesterol and multiple GSLs (GM2, GM3, GlcCer, and LacCer) ( 260 ). Decreased Purkinje cell numbers occur in human Sap C defi ciency ( 262 ) and Sap C Ϫ / Ϫ mice ( 193 ). Purkinje cell loss is also observed in Sap D-deficient mice ( 98 ). LC/MS analyses of cerebella from Sap C Ϫ / Ϫ or Sap D Ϫ / Ϫ mice revealed small increases in lactosylsphingosine (LacSph) and LacCer or hydroxy ceramide, respectively ( 98,193 ). By comparison, Sap A Ϫ / Ϫ and Sap B Ϫ / Ϫ mice do not exhibit Purkinje cell death, probably due to their selective effects on myelin development ( 191,212 ). Although GM2 storage is detected in Purkinje cells of Sandhoff disease and GM2 activator knockout mice, Purkinje cell death was not reported ( 169,175 ). Thus, some GSLs, e.g., ceramide, GlcCer, and LacCer, or LacSph, may cause Purkinje cell loss, but others may not. Several mouse models unrelated to GSL disorders also present with Purkinje cell loss (263)(264)(265), thereby implicating other pathways in degeneration of these cells. The mechanism(s) by which Purkinje cells degenerate are unknown, but negative TUNEL assays in Sap C Ϫ / Ϫ and Sap D Ϫ / Ϫ mouse brains suggest that apoptosis is not the major mechanism ( 243,266 ). Recently, an imbalance was proposed between induction and fl ux through the autophagic pathway in contributing to cell stress and neuronal loss in NPC ( 267,268 ). Neurochemical studies in NPC1 mice demonstrate serotonin and its main metabolite, 5-hydroxyindoleacetic acid, are altered in the cerebellum and cortex of these mice. The levels of the inhibitory amino acid glycine also were 3-fold higher in the cerebellum of NPC1 and increased L-DOPA was detected in the vermis of the cerebellum. These results implicate changes in neurotransmitters that could be a molecular basis of cerebellar neuropathology leading to Purkinje cell death ( 269 ).

Axonal degeneration
Axonal degeneration can precede and lead to neuronal death by the infl ammatory, myelination defects, and ab-normal lipid and protein metabolism ( 270 ) that occur in GSL and ganglioside LSDs. In GM1 and GM2 gangliosidosis, axonal spheroids and dendritogenesis correlate with progressive ganglioside accumulation ( 271 ). In humans and mice with Sap C defi ciency, the soma of neurons and Purkinje cells have normal initial ultrastructure, but inclusion bodies and focal swellings of dendrites develop ( 193,262 ). Dysregulation of autophagosomes in axonal termini has been implicated in axon degeneration ( 272 ) but has not been explored in GSL LSDs. The pathological roles of axonal spheroids in several GSL disorders, and possible links to neuronal degeneration, have been discussed in a recent review ( 273 ).
Abnormal morphology in the sensory nervous system is a common pathology in the GSL LSDs. Foamy storage material occurs in the dorsal root ganglion of GM1 gangliosidosis, Sandhoff disease, GM2 activator defi ciency, Sap B , prosaposin knockout mice, and NPC1 mice, as well as GM2/GD2 synthase knockouts that lack complex gangliosides ( 274 ). The storage of GSLs in dorsal root ganglion causes uncoordinated body movement in those models. GM2/GD2 synthase knockout, twitcher , and Sap A Ϫ / Ϫ mice exhibit degenerative changes in sciatic nerve, including myelin fragmentation and myelin layer separation with resultant synaptic rearrangement ( 118,191,274 ). GalCer or sulfatide are components of myelin GSL, and their excess accumulation causes demyelination in ASA, Krabbe disease, Sap B-defi cient patients, and mouse models ( 124 ) ( 130,275 ). In primary neuronal and neuroblastoma cultures, lysosomal accumulation of sulfatide mediates increases in endosomal ceramide that leads to apoptosis ( 276 ), thereby implicating GSL fl ux alterations due to sulfatide storage in neuronal death. The psychosine hypothesis proposed that lysosphingolipids are responsible for the neuronal disease pathology and cell death mediated through aberrant modulation of PKC pathways ( 277 ). The fi ndings that other lyso-lipids, lyso-sulfatide, LacSph, lyso-GM1, lyso-Gb3, and glucosylsphingosine are present in the GSL disorders extended this hypothesis to the lyso-sphingolipid hypothesis ( 278 , 279 ). Experimental evidence to support this hypothesis is growing. Galactosylsphinosine (psychosine) is undetectable in normal tissues but is at high levels in the brains of Krabbe (globoid cell leukodystrophy) patients and mouse models ( twitcher ) ( 280 ). The pathology of infantile Krabbe disease includes severe dys/demyelination caused by apoptotic loss of myelin-forming oligodendrocytes and Schwann cells thought to be linked to accumulation of the highly toxic GALC substrate, galactosylsphingosine ( 277,281,282 ). Importantly, this was the fi rst disease in which a lysoglycolipid was implicated as a major pathogen and led to the postulation that lysoglycolipids were the toxic agents that were responsible for the disease progression in many GSL storage diseases. In the twitcher mouse brain, galactosylsphingosine accumulation leads to alteration in lipid raft structure, which may cause abnormal cellular function in oligodendrocyte and Schwann cells and lead to retrograde axonal degeneration ( 283 ). A similar mechanism may be operative in neuronopathic Gaucher disease mice exhibiting increased CNS glucosylsphingosine levels ( 205,208 ). Axonal degeneration has been correlated with elevated GlcCer and glucosylsphingosine levels ( 205 ), and glucosylsphingosine toxicity has been demonstrated in neurons by suppression of neural outgrowth ( 284 ). Both GlcCer and glucosylsphingosine modulate Ca 2+ fl ux in brain microsomes ( 285 ). Defective Ca 2+ homeostasis, especially increases of Ca 2+ release, presumably from the ER, has been suggested to be responsible for neuronal cell death ( 286 ). However, the direct effect of in vivo GlcCer and glucosylsphingosine levels on neuron and axon degeneration remains to be clarifi ed. Several other lyso-glycolipids have been detected in GSL LSDs and their role and/or mechanism of toxicity provide fertile ground for developing support for this universal lyso-sphingolipid hypothesis.

Electrophysiologic studies in GSL disorders
The application of long-term potentiation (LTP) analyses has provided insight into the pathogenesis and levels of neural involvement of the GSL and lysosomal disorders ( 193,212,287 ). For such analyses, a constant level of presynaptic stimulation is converted to a large postsynaptic output, i.e., LTP, a leading experiment to study learning and memory ( 288 ). Sap C Ϫ / Ϫ and V394L/SapC Ϫ / Ϫ mice have been tested and exhibit reductions of LTP, together with marked accumulation of GlcCer and glucosylsphingosine, suggesting abnormal synaptic plasticity ( 193,205 ). Sap B Ϫ / Ϫ hippocampi have sulfatide accumulation and normal LTP, showing that although excess sulfatide induced myelin dysfunction, it does not affect LTP ( 194 ). Although the molecular basis for LTP alteration association with GSL levels has not been elucidated, calcium dysregulation has been implicated as having a direct effect on LTP ( 289 ), which is mediated through glutamate effects on Na + and Ca 2+ fl ux. Indeed, disrupted calcium homeostasis was found in isolated brain microsomes from a Gaucher disease type 2 patient, the Sandhoff disease mouse ( 285,290 ), and the ASM knockout mouse ( 291 ). Thus, selected abnormal GSL levels could alter LTP by disruption of calcium-mediated vesicular release of neurotransmitters or related receptor functions, as suggested recently for galactosylsphingosine interference of lipid raft assembly ( 283 ). The alteration of LTP in selected GSL disorders also could result from secondary pathology, such as proinfl ammation and neurotrophin defect ( 292,293 ). Neurotrophins have been suggested to modulate LTP ( 292 ). BDNF, a positive effector, defi ciency was found in NPC1 mice ( 294 ), whereas NGF defi ciency occurs in Gaucher disease and Sandhoff disease ( 213 ). Such defi cits in the brain could lead to impairment of LTP and form a basis of more global cognitive and learning defi cits in patients.

THERAPY
Treatments for the GSL LSDs ( Table 3 ) focus on two primary approaches to decrease the excess accumulated/ accumulating GSLs by increasing their degradation rates or decreasing their synthesis rates. The former centers on increasing the hydrolytic capacity of the GSL enzymes by cells to increase its activity in situ by improving traffi cking to the lysosome, enhancing its stability, improving its catalytic rate constant, reestablishing proteostasis by diminishing the UPR that a mutant enzyme creates, or a combination of all these mechanisms. This approach has been called molecular chaperone therapy, but is more appropriately termed enzyme enhancement therapy or EET. The current studies and trials with EET use potent competitive inhibitors of selected lysosomal enzymes as the chaperone agents.

Treatments by increasing GSL degradation
The success of ERT in ameliorating many, but not all, of the visceral manifestations of Gaucher disease type 1 has reconstituting target cells to above-threshold amounts of the specifi c enzyme function via several different approaches: 1) directly supplying the needed enzyme (so called enzyme replacement therapy or ERT) that can be taken up into target cells and delivered as active enzyme to the lysosome by receptor-mediated endocytosis; 2) indirectly providing enzyme by surrogate factories, transplantation of unaffected cells (e.g., hematopoietic stem cells or induced pluripotent stem cells) or addition of genes to cells that can produce wild-type or greater in vivo levels of the needed enzyme for secretion and internalization by the target defi cient cells, so called, metabolic cross correction by cell or gene therapy; and 3) use of small molecules that interact directly with the specifi c mutant enzyme in spurred the development of this approach for other LSDs with major visceral involvement, such as Fabry disease ( 295 ). This restriction to visceral disease derives mainly from the inability of intravenously administered enzymes to cross the blood brain barrier in therapeutically significant amounts. Data are now emerging that ligands, other than the typical sugars, i.e., mannose-6-phosphate or mannose, may allow transit of proteins from the circulation across the blood brain barrier, particularly at very high plasma concentrations, as shown in the mouse models of ␤ -glucuronidase or ASA defi ciencies ( 132,296 ). Such effects were enhanced by using an amino acid tag to increase the delivery of the enzyme to the brain substance ( 297 ). In addition, the use of intraventricular or intrathecal administration of appropriate enzymes could provide for significant therapeutic effects in the CNS. Similar considerations also apply to postneonatal delivery of AAV viral vectors for genetic transduction following intravenous administration. This limitation appears to have been partially overcome by the inclusion of disease-specifi c peptide ligands into AAV vectors that would then bind to endothelial cells of affected animals and express the needed enzyme, thereby facilitating uptake into affected CNS cells including neurons ( 298 ). Such an approach led to widespread distribution of the expressed enzyme in the brain following intravenous administration with resultant biochemical and phenotypic improvement. Transplantation of hematopoietic, embryonic, neural progenitor, or induced pluripotent stem cells with or without transduced additional genes is being used to deliver the needed therapeutic enzymes (299)(300)(301)(302)(303). Such cells can either add to existing cell populations in the brain, i.e., glia or neurons, and/or act as metabolic factories for the enzymes that can then be secreted and taken up by enzyme defi cient cells in the CNS and other organs to varying degrees. Such approaches have had varying success in altering the CNS phenotypes because of the timing and nature of the underlying CNS pathology in the GSL LSDs. The degree of reversibility is determined by the time of treatment and/or the overall distribution of the therapeutic cells and/or enzyme to different regions of the brain (see below).

Treatments by decreasing GSL synthesis
The approach of inhibiting the synthesis of GSLs to treat the corresponding diseases was fi rst suggested by Norman Radin some 30 years ago . The general concept is to inhibit GCS, because GlcCer is the precursor of the lipids in the majority of GSL LSDs, but more specifi c enzymes for the synthesis of higher GSLs and gangliosides could also be targeted. Because the complete defi ciency of GCS is lethal in utero, partial inhibition of GlcCer synthesis is the goal with the intent to reestablish a balance between synthesis and degradation. For this approach to work, either residual, albeit low, enzyme activity or an alternative pathway for GSL disposal is needed. This approach has been termed Substrate Reduction Therapy. However, this is a misnomer, beccause the goal of all the therapeutic approaches is to reduce substrate accumulation; a better term would be substrate synthesis inhibition therapy or SSIT.
Oral SSIT with N-butyl-deoxynojirimycin (NB-DNJ) (miglustat or Zavesca; Actelion, Basal Switzerland) has been approved by the EMEA and FDA for treatment of Gaucher disease type 1 patients who cannot tolerate ERT ( 147,304 ). This drug is in Phase 3 trials for Sandhoff disease ( 158 ) and for NPC1; miglustat has been approved by EMEA ( 305 ). The therapeutic effects of NB-DNJ are less than achieved with intravenous ERT using mannoseterminated GCase ( 304 ). Use of NB-DNJ in Gaucher disease type 3 patients with established CNS disease failed to show signifi cant improvement in neurological signs, specifi cally abnormal eye movements ( 147 ). However, sufficient neurological effects were found in NPC1 disease for EMEA approval for that condition ( 306 ) (http://nnpdf. org; http://www.emea.europa.eu/pdfs/human/opinion/ Zavesca_33596508en.pdf). More recently, a more potent GCS inhibitor (Genz-112638) based on the parent compound, 1-phenyl-2-decanoylamino-3-morpholino-1-propanol, has shown major clinical and biochemical effects in visceral organs from humans and mice with Gaucher disease type 1 variants ( 307 ). Importantly, the lungs were substantially cleared of GlcCer storage in the mice, an organ that has poor biochemical response to ERT ( 307 ). Genz-112638 is a Pgp substrate and, although it rapidly crossed the blood brain barrier, signifi cant levels of the drug are not achieved due to rapid removal by Pgp ( 307 ). Phase II data of Genz-112638 from the human trial show similar safety and initial effi cacy at 6-12 months to ERT with imiglucerase ( 308 ).
A clear advantage of drugs used for SSIT and EET is their ability to cross the blood brain barrier in potentially therapeutically signifi cant amounts. Furthermore, the EET approach to re-engineer the mutant enzyme in situ for better function has additional advantages of oral administration and avoiding immunological reactions and the necessity of specifi c ligands for receptor mediated endocytosis. The fi rst generation of these agents for the GSL storage diseases are primarily iminosugars that are either 5-or 1-deoxy or 3,4-dideoxy glycoside derivatives of glucose or galactose, and they are potent reversible competitive inhibitors of specifi c lysosomal hydrolases. The current agents, isofagomine [3,4-dideoxy-5-(hydroxymethyl) piperidine] and DGJ (1-deoxy galactonojirimycin), have shown signifi cant ability to enhance the activity of a great variety of different mutant GCases or ␣ -galactosidases A, respectively, in tissue culture and in peripheral blood leukocytes of affected patients with Gaucher or Fabry disease, respectively (309)(310)(311). Clinical trials in Fabry disease are ongoing with DGJ, but the development of isofagomine has been stopped due to the lack of clinical effect in all but one patient with Gaucher disease type 1 (http://fi les. shareholder.com/downloads/AMTX/763312895 × 0 × 322616/e986f511-4f40-4412-972d-068e6d7dd058/FOLD_ News_2009_10_2_General.pdf). Clearly, alternative drugs with different targets on the mutant enzymes might have greater clinical effects.
For CNS LSDs, early therapeutic interventions have been shown to improve diseases outcome in several LSD mouse models. NPC1 mice treated with cyclodextrin at P7 versus P21 showed longer lifespans ( 312 ). NB-DGJ treatment started at P2, when GM2 accumulation was detected, significantly reduced total gangliosides in Sandhoff disease mice ( 171 ). Isofagomine treatment on V394/SapC Ϫ / Ϫ mice started in utero (E15) extended lifespan versus treatment started at P21 (Sun et al., unpublished observations). The dosage used in the early intervention at embryonic stage is critical, because they may be toxic in utero effects. Understanding the GSL profi le and function during brain development ( Figs. 4 and 5 ) will be important to design the therapy and to defi ne the timing for effective treatment of CNS LSDs without disrupting normal brain development.
The disease stage and pathogenic mechanisms of the CNS GSL storage diseases are of major importance to the development of therapies. Many of the studies of ERT, cell transplantation, and gene therapies have shown promising results in animal models in which the disease processes may be reversible, e.g., GSL storage is present in selected CNS cells, but the neurons remain viable. In such cases, one can envision the reestablishment of GSL fl ux in such cells might lead to recovery of their function and a positive effect on phenotype. In comparison, if the disease process has neuronal loss as a major pathogenic mechanism, recovery or even stabilization might be impossible; e.g., in neuronopathic Gaucher disease mice, neither complete enzyme reconstitution within neurons nor provision of microglia with wild-type GCase activity led to restoration or rescue of the CNS disease ( 150 ). Neuronal death is predominant in these models. Much more detail of the pathogenic mechanisms will be required for development of disease-specifi c appropriate therapeutic approaches.

CONCLUSIONS AND PERSPECTIVES
Advances in understanding the pathophysiology and treatment of the GSL diseases hinge upon molecular insights into the basic and initial mechanisms of disease causation. Elucidation of the earliest molecular events caused by the subclinical accumulation of GSLs could provide initial targets that are fundamental to interfering with the propagation of the neuronal death pathways. Such insights are critically important as the primary neuronal disease, cell death versus GSL storage, would require highly different intervention strategies and timing of initiation. Recent proposals that retrograde axonal degeneration due to lipid raft aberrations provide additional pathways and targets for therapy and for the understanding of the role of GSLs in cell development and physiology. These rare diseases continue to provide platforms for the general understanding of GSL biochemistry and therapy for many diverse and more common diseases.