Glycosphingolipids in signaling and development: From liposomes to model organisms

Authors


Abstract

Since the launch of the raft hypothesis in 1997, data generated in liposomes and cultured cells have highlighted the role of glycosphingolipids (GSLs) in the dynamic organization of biological membranes and the activity of signaling complexes. In parallel studies, genetic analysis of the GSL synthetic pathway has begun to reveal some of the specific roles of GSLs in vivo. Here, we review the role of GSLs in signaling in the context of a refined raft hypothesis. Recent genetic studies in worms, flies, and mice give us the opportunity to integrate these in vivo data with earlier in vitro liposome studies. Developmental Dynamics 241:92–106, 2012. © 2011 Wiley Periodicals, Inc.

INTRODUCTION

Since their discovery more than 100 years ago, the status of glycosphingolipids (GSLs) has very much changed from a biochemical peculiarity of the nervous system to a key lipid of the plasma membrane involved in infectious diseases, lysosomal degradation disorder pathologies (or GSL storage diseases), neurodegenerative diseases, multiple sclerosis, cancer, and diabetes (Ginzburg et al., 2004; Xu et al., 2010). The work of many laboratories over the last two decades has demonstrated how the chemical properties of GSLs underlie their activities in biological membrane (Sahl et al., 2010; Brown and Rose, 1992; Kenworthy and Edidin, 1998; Sharma et al., 2004; Douglass and Vale, 2005; Goswami et al., 2008; Eggeling et al., 2009). Since the proposal of the fluid mosaic model of Singer and Nicholson (Singer and Nicolson, 1972), whereby proteins and lipids are freely diffusing within a relatively homogenous membrane lipid bilayer, our representation of the membrane organization has drastically changed (Kusumi et al., 2010; Simons and Ikonen, 1997; Anderson and Jacobson, 2002; Hakomori, 2002a,b; Mayor and Rao, 2004). In current models, the diversity and the biophysical properties of the membrane lipid components, including glycerophospholipid, cholesterol, and sphingolipid (ceramide, sphingomyelin, phosphoethanolamine ceramide, and GSLs) molecules, are driving the formation of a mosaic of lipid nanodomains of 5–10 nm in diameter (Fig. 1; Sahl et al., 2010; Sharma et al., 2004; Pinaud et al., 2009). The interplay between these lipids, the membrane-associated or -embedded proteins, and the underneath cortical actin network, regulates the clustering and thereby the mobility of nanodomains into larger proteolipidic domains (100–300 nm) (Douglass and Vale, 2005; Goswami et al., 2008) that are believed to mediate the stability of signaling complexes and to control their signaling efficiency (Ianoul et al., 2005; Gales et al., 2006; Pontier et al., 2008).

Figure 1.

GSL and sphingomyelin nanodomains. Simplified model of the plasma membrane showing that cholesterol, glycerophospholipids, sphingomyelin, and GSLs form different types of ordered nanodomains within the external leaflet of the membrane. Transmembrane proteins (e.g., Receptor tyrosine kinases [RTK] or G-protein coupled receptors [GPCR]) and membrane-associated proteins (GPI-anchored proteins or Heterotrimeric G proteins) localize differentially depending on their relative affinity for the different ordered lipid environments. For instance, GPI-anchored proteins exhibit a high affinity for simple and complex GSL nanodomains (Pinaud et al., 2009). Considering the average size of nanodomains (5–20 nm), proteins are most likely present in only one or three or four copies per nanodomain. Dynamic rearrangement of these nanodomains upon, for example, ligand activation, can lead to the formation of larger proteolipidic platforms (100–300 nm) (Kusumi et al., 2010).

While the important molecular principles of the influence of GSLs on membrane architecture and the function of signaling molecules are partly understood, the transfer of this knowledge to the level of developing and aging organisms remains challenging. The genetic analysis of GSL metabolic pathways in worms, flies, or mice is a first step towards addressing this complex issue (Furukawa et al., 2004; Stolz et al., 2008; Marza et al., 2009; Pizette et al., 2009; Hamel et al., 2010).

In this review, we aim at giving an update on the role of GSLs in vivo. We attempt to integrate our current knowledge on the properties of GSLs from in vitro studies with studies on the role of GSLs in signal transduction from cultured cells and model organisms.

SYNTHESIS AND BIOPHYSICAL PROPERTIES

Structure and Synthesis

GSLs share with all sphingolipids a sphingoid backbone that results from the condensation of a serine with an acyl chain. This reaction constitutes the limiting step of the sphingolipid synthesis and is under the control of complex metabolic regulations (Breslow and Weissman, 2010). These sphingosine molecules are further acylated in the ER by ceramide synthases to generate ceramide (Acharya and Acharya, 2005; Pewzner-Jung et al., 2006). Ceramide can then be modified by various sphingomyelin synthases to generate phosphocholine ceramide (also called sphingomyelin) or phosphoethanolamine ceramide (Acharya and Acharya, 2005; Ternes et al., 2009), converted into galactosylceramide molecules or translocated to the cis-golgi apparatus to be converted into glucosylceramide (Fig. 2) (Degroote et al., 2004; Xu et al.2010). Both galactosyl- and glucosylceramide molecules initiate the formation of GSLs. Galactosylceramide are subsequently modified into sulfatides while glucosylceramide molecules give rise to all the other types of GSLs that will all be synthesized in the Golgi apparatus (Fig. 2). The complexity of the oligosaccharidic groups added to the glucosylceramide module varies greatly among GSLs ranging from a two sugar molecule (Lac-Cer or Mac-Cer) to complex oligosaccharide chains covalently bound to neuraminic acid molecules in mammals or phosphoetanolamine groups in insects (Fig. 2) (Chen et al., 2007).

Figure 2.

GSL synthetic pathways. The glycosyltransferases and GSLs discussed in this review are in blue and red, respectively.

In invertebrates, a single type of GSLs has been identified so far, namely the arthroserie (Chen et al., 2007), that is characterized by the presence of a mannose residue in place of the galactose moiety found in LacCer modules (Fig. 2). In mammals, several types of modifications can occur at the level of the sugar chain and, accordingly, different series of GSLs have been characterized: globo-, lacto/neolacto-, sulfo-, or phosphoglycosphingolipids, as well as a series of GSLs containing some neuraminic acid moieties, namely the gangliosides (Fig. 2). For an extensive classification of the various types of sphingolipids, the reader is referred to the work of Fahy et al. (2005).

The Glycosphingolipid Synthases

Glycosphingolipid synthases are evolutionarily conserved from insects to mammals (Schwientek et al., 2002; Griffitts et al., 2003, 2005; Wandall et al., 2003, 2005; Johswich et al., 2009). They preferentially localize in the Golgi apparatus according to a cis-trans organization (Wandall et al., 2005; Johswich et al., 2009) with the enzymes catalyzing the first reactions in the biosynthetic pathway being localized closer to the ER–Golgi interface than the enzymes catalyzing subsequent and more complex modifications (Maccioni et al., 2011). In addition, sequential reactions appear to be catalyzed by multi-enzymatic molecular complexes (Giraudo et al., 2001). Also, the first reaction, i.e., the addition of a glucose to a ceramide, takes place on the cytosolic leaflet of the Golgi membrane whereas all subsequent additions of sugar moieties are catalyzed on the lumenal side of the Golgi (Maccioni et al., 2011). Consequently, complex GSLs are mainly asymmetrically localized in the external leaflet of the membranes. While the bulk of GSL neosynthesis occurs in Golgi membranes, the presence of glycosyl hydrolases and transferases in the plasma membrane indicates that dynamic regulation of GSL composition can also occur at the cell surface (Sonnino et al., 2010).

Biophysical Properties and Phase Separation in Biological Membranes

In comparison to average membrane glycerophospholipids, the ceramide acyl chains of the 400 species of GSLs found in mammals are longer and highly saturated. These chemical specificities result on average in a higher order (Lo) to disorder (Ld) phase transition temperature (Tm) (Sonnino and Prinetti, 2010; Prinetti et al., 2009). While the Tm for a typical unsaturated glycerophospholipid is far below 0°C, saturated lipids, such as sphingomyelin or simple glucosylceramide, have in contrast a Tm of 45° and 80°C, respectively. Hence, GSL have a strong tendency to acquire gel-like ordered structures (Lo phase) and laterally segregate from lipids with lower Tm that are exhibiting a disorder organization. In addition to conferring high Tm, the length and the saturation level of the GSL acyl chains will determine the ability of each GSL molecule to create local interdigitations at the interface of the two membrane leaflets. These local interdigitations rigidify the corresponding region of the membrane (Roux et al., 2005). In contrast to mammals, insect-derived GSLs have shorter acyl chain (C14 vs. C20) (Rietveld et al., 1999; Acharya and Acharya, 2005). The average Tm of the arthroserie GSLs in insects should hence be reduced, accordingly. These differences in acyl chain length and associated Tm between arthro-GSLs and vertebrate GSLs may relate to the fact that invertebrates are allotherm and have to adapt the fluidity of their membranes to lower temperature.

Complex GSL headgroups increase the lateral volume and the amount of hydrogen bonds formed with the membrane surrounding water, thereby affecting their Tm characteristics. Bulkiness and hydrogen bonds are also believed to drive and stabilize phase separation and correlate with positive membrane curvature (Westerlund and Slotte, 2009). For similar acyl chains, complex GSL lipids pack more tightly than sphingomyelin. This organization has been shown to attenuate the affinity of complex GSLs for cholesterol, which can only mix with complex GSLs in the presence of sphingomyelin or simple GSLs (Sonnino and Prinetti, 2010; Prinetti et al., 2009).

The specific biophysical properties of the distinct GSL species may impact on the organization of biological membranes. Illustrating this point, the GM1 and GM3 gangliosides form independent clusters in biological membranes. These clusters exhibit a differential sensitivity to cholesterol depletion or actin depolymerization pharmacological treatments (Fujita et al., 2007, 2009). Thus, GSLs contribute to the formation of a heterogeneous mosaic of nanodomains in biological membranes.

Association of Membrane Proteins to GSL-Containing Nanodomains

The molecular basis underlying the localization of specific membrane proteins within a specific ordered nanodomain environment has been intensively studied. Post-translational modifications with glycosylphosphoinosityl (GPI) or palmytic acid groups have been shown to mediate the preferential partitioning of membrane proteins within raft nanodomains (Brown and Rose, 1992; Zacharias et al., 2002; Levental et al., 2010). GPI or palmitoyl groups contain saturated hydrocarbon chains conferring an affinity for the sphingolipid ceramide backbone. Proteins can also interact directly with specific GSLs through their glycosylated headgroups in membrane (Griffitts et al., 2005; Romer et al., 2007). Cholera toxin binds specifically to GM1 (Fishman, 1982) whereas the shiga-toxin and the HIV envelop protein gp120 bind the globoside trihexosylceramide (Gb3) (Mahfoud et al., 2002; Romer et al., 2007). Detailed structure function analysis performed on HIV gp120 protein has led to the identification of peptidic motifs with specific GSL-binding properties (Mahfoud et al., 2002). Such GSL Binding Motifs (GBM) have now been identified in various eucaryotic proteins, including the Prion protein (PrP) (Mahfoud et al., 2002), the alzheimer-associated beta-amyloid (Mahfoud et al., 2002), the Parkinson-associated α-synuclein (Fantini and Yahi, 2011), TNF receptor family members (Chakrabandhu et al., 2008), the Notch receptor ligands Serrate and Delta (Hamel et al., 2010), and the prominin/CD133 stem cell marker (Taieb et al., 2009). Interestingly, the specific binding of α-synuclein to GM3 favors its correct folding (Di Pasquale et al., 2010). Thus, protein conformation can be modulated by GSL-protein interaction. This activity of GSL likely underlies some of its in vivo roles in the regulation of signaling complexes.

REGULATORY FUNCTIONS

Regulation of Protein Conformation and Activation

While many reports have invoked a role for nanodomains in signal transduction, only a few studies have precisely investigated the mechanistic role of GSLs (Yoon et al., 2006; Coskun et al., 2011). A recent study by Coskun and colleagues on the role of GM3 in the regulation of the Epidermal Growth Factor Receptor (EGFR) is a landmark exception concerning the direct impact of GSLs on signaling proteins (Coskun et al., 2011). By embedding a purified form of the receptor in proteoliposomes of defined lipid composition, these authors have clearly demonstrated that GM3 molecules limit receptor dimerization and autophosphorylation. This GM3-mediated inhibition can be overcome by the EGF ligand and relies on the direct interaction between the GM3 neuraminic acid group with an extracellular lysine localized proximally to the EGFR transmembrane domain. Mutation of this lysine residue abrogates inhibition of EGFR dimerization by GM3 and results in its constitutive autophosphorylation and activation. These results establish a direct link between EGFR/GSL interaction and the receptor oligomerization and function. In biological membranes, EGFR has been shown to display both low (nM range) and high (pM range) affinity binding sites for its ligand (Sigismund et al., 2005; Coskun et al., 2011) and the proteoliposomes used in this study reconstitute only the low-affinity EGF-binding site. Nevertheless, these data confirm previous results concerning the role of GM3 in the EGFR activity regulation in membranes (Yoon et al., 2006) and correlate with the existence of an EGF-induced monomeric-dimeric dynamic transition of the receptor in membranes of living cells (Zhang et al., 2009).

The lysine residue implicated in the GM3-mediated EGFR inhibition appears to be conserved in other tyrosine kinase receptors, such as the insulin receptor (Kabayama et al., 2007). Again, the interaction between the lysine and the GM3 molecules are believed to attenuate the constitutive receptor activity (Kabayama et al., 2007). Very interestingly, this molecular scenario is entirely consistent with the observation that the insulin receptor is constitutively phosphorylated in GM3-deficient mice (Yamashita et al., 2003). The interaction of the insulin receptor with GM3 was proposed to competitively restrain the presence of the receptor in caveolae domains, hence its signal transduction (Kabayama et al., 2007). Whether GM3 regulates the activity and the localization of the receptor via a conformational-induced change similar to the one observed for EGFR remains to be determined.

The GM1 molecule has also been implicated in the regulation of several receptors, among them the PDGF and the TRK receptors. Modulation of GM1 has dramatic impact on the membrane localization of the PDGF receptor (Veracini et al., 2008) and is believed to regulate the TRK receptor function by promoting the autocrine stimulation of its own ligand, the neurotrophin molecules (Rabin et al., 2002; Mallei et al., 2004). Whether GM1 interacts directly with the PDGF receptor and/or neurotrophin to modulate their conformation, as shown for GM3 and the EGFR receptor, is not known.

Regulation of Endocytosis

The biophysical properties of GSLs confer them with the ability to modulate the membrane curvature. Recent studies on Shiga toxin internalization have revealed how the binding of Gb3 with the homo-pentameric subunit B of the Shiga toxin drives the formation of a local Lo liquid ordered phase (Windschiegl et al., 2009) contributing to the membrane curvature (Sorre et al., 2009) and the formation of membrane tubules (Romer et al., 2007). The interplay between GSLs and the underneath cortical actin seems to be required for the scission of tubules and vesicles, the actin directing appropriate cholesterol-dependent membrane rearrangements allowing for the scission (Romer et al., 2007).

While these data suggest that GSL identity may control membrane dynamics, the actual implication of nanodomains or GSLs in the regulation of protein trafficking is experimentally difficult to establish (Sigismund et al., 2008). Part of the difficulty comes from the diversity in the molecular machineries supporting cholesterol-regulated clathrin-independent internalization events (Mayor and Pagano, 2007). We, therefore, limit our discussion below to two examples of clathrin-independent internalization that likely involve GSL-based nanodomains since the signaling activity of the internalized receptors is known to be modulated by raft nanodomains.

The first example concerns the internalization of the EGFR. While the relationship between EGFR and clathrin-independent internalization has been known for many years (Chinkers et al., 1979), how this trafficking modulates EGFR function was only clarified in 2005 when Sigismund et al. demonstrated that a clathrin-independent mechanism was most likely involved in the attenuation of EGFR signaling (Fig. 3; Sigismund et al., 2005). The mono- or polymonoubiquitination of EGFR was shown to drive its internalization through a clathrin-independent and epsin-dependent mechanism. The binding of epsin to the ubiquitin moiety appeared to inhibit the epsin engagement by the clathrin protein and segregate the receptor/ubiquitin/epsin complex away from the clathrin-coated pits (Chen and De Camilli, 2005). This event occurred preferentially at high concentrations of ligand and appeared to result in EGFR degradation (Sigismund et al., 2005, 2008; Huang et al., 2007). In contrast, clathrin-mediated internalization occurred at a low concentration of EGF (high-affinity binding site) and is required for receptor coupling to its classical signaling pathways PI3K and/or MAPK pathways (Goh et al., 2010; Sigismund et al., 2005, 2008).

A related model may apply to the TGF-β receptors that play many essential roles during development, in particular for stem cell maintenance (Di Guglielmo et al., 2003). TGF-β superfamily members signal through heterodimeric complexes of Ser-Thr kinase receptors. Activated TGFβ receptor phosphorylates the proteins Smad2 and Smad3 that then bind the Smad4 effector to modulate gene expression. Smad7, a negative regulator, eventually associates with the receptor and recruits the E3-ubiquitin ligases Smurf-1 and Smurf-2 to promote receptor down-regulation and degradation. In a seminal paper, Di Guglielmo et al. (2003) showed that this differential coupling of the receptor to Smad2/Smad3 or Smad7/Smurf1/Smurf2 was correlated with distinct trafficking routes. Specifically, the Smad2/Smad3 complex was endocytosed through a clathrin-dependent mechanism while the Smad7/Smurf-1/Smurf-2 complex was trafficked through specific nanodomains known as caveolae (Di Guglielmo et al., 2003).

This regulatory mechanism may be important for stem cell maintenance and differentiation (Xia et al, 2010; Yamazaki et al., 2009). In Drosophila, Smurf 2 down-regulates TGFβ signaling in the daughter cells of the germline stem cells that undergo differentiation (Xia et al., 2011). It will be of interest to examine whether the function of Smurf2 in Drosophila involves nanodomain-mediated or clathrin-independent internalization of the receptor. In the mouse, a similar regulatory mechanism appears to modulate hematopoiesis (Yamazaki et al., 2009). In this context, inhibition of nanodomain clustering (which may reflect the ability of nanodomains to dynamically internalize) was shown to depend on TGFβ receptor activity and to be essential for stem cell maintenance (Yamazaki et al., 2006). Also, down-regulation of the TGFβ signaling in differentiating cells correlates with their de novo ability to cluster nanodomains. Thus, regulation of nanodomain clustering by TGFβ receptor signaling appears to be important to inhibit differentiation. It will be of interest to decipher the molecular mechanisms underlying this regulation. One simple and exciting possibility is that TGFβ receptor signaling regulates the expression of genes involved in the GSL biosynthetic pathway.

CONTROL OF GSL SYNTHESIS DURING DEVELOPMENT

The extent to which the lipid composition of the membrane varies during development, correlates, and/or directs differentiation is largely unknown. This may in part be due to the lack of appropriate molecular tools to probe for quantitative and cell-specific changes in developing organisms. Nevertheless, changes in the expression of ceramide and GSL synthase enzymes and the presence of their corresponding lipid products have been reported during the differentiation of human embryonic stem cells (Liang et al., 2010; Park et al., 2010). Multipotent Oct4-positive stem cells have mainly GSLs with a globoserie signature whereas more differentiated cells exhibit an increase in GM3 and GD2 gangliosides. While changes in GSL content have been previously observed during development as well as in various pathologies (Hakomori, 2002a,b; Xu et al., 2010), this is, to our knowledge, the first report indicating that stem cell differentiation correlates with a change in lipid composition of the membrane. It will be important to investigate whether and how this change impacts on differentiation.

Remarkably, several of the stem cell markers that are routinely used to FACS purify stem cells, including prominin/CD133, CD44, Sca-1, Thy, β1- and β4-integrin proteins, are bona fide nanodomain-associated proteins. These proteins are highly expressed at the surface of stem cells from blood, mammary gland, muscles, fat body, pancreas, intestine, and prostate (Liu et al.2011; Prestegarden and Enger, 2011; Wojakowski et al., 2011; Zeki et al., 2011; Pontier and Muller, 2009). It would, thus, be of interest to examine whether changes in GSL composition correlate with both the dynamic removal of these stem cell markers from the surface during differentiation and the transcriptional regulation of GSL synthase enzymes (Liang et al.2010).

GENETIC ANALYSIS OF GSL FUNCTIONS

Genetic Studies in Model Organisms

GSLs are essential for the development of multicellular organisms. Mutation and/or silencing of the gene encoding the GlcT enzyme, a unique enzyme that regulates the production of glucosylceramide, is lethal in C. elegans, D. melanogaster, and M. musculus (Yamashita et al., 1999; Marza et al., 2009; Kohyama-Koganeya et al., 2004) (Tables 1 and 2; Fig. 2). In particular, mouse embryos undergo a massive apoptosis following their gastrulation, possibly due to the accumulation of ceramide that is predicted to occur in the absence of glucosyltransferase (Hannun and Obeid, 2008). Along this idea, accumulation of ceramide molecules upon mutation of the neutral ceramidase gene leads to the degeneration of Drosophila photoreceptors (Acharya et al., 2008).

Table 1. List of the Different Genetically Engineered Mice for GSL Synthases
Affected GSL productEnzyme namePhenotypeReference
Glucosylceramide (GlcCer)Glucosylceramide synthaseFollowing gastrulation, embryos die and exhibit a massive apoptosis.Yamashita et al. (1999)
Lactosylceramide (LacCer)β4Gal-T6 or β4Gal-T5Embryos die early (E10), most likely because of a defect in the development of extra- embryonic tissue. The mutation also affects O- and N-glycoprotein synthesis.Kumagai et al. (2010); Nishie et al. (2010)
GM3GM3 synthase (Siat9 enzyme) Sialyl transferaseMice exhibit hearing loss and a hypersensitization of the insulin pathway, which confer to them a resistance to an induced high-fat diet and the development of a subsequent diabetic phenotype. Mice survive more than 1 year.Yamashita et al. (2003); Yoshikawa et al. (2009)
Gb3 GlobosideGb3 synthaseMice seem to develop normally and have, in addition, a resistance to invasion by the Shiga toxin.Okuda et al. (2006)
GA2 or GM2 or GD2N-acetyl-galactosyl transferase or GM2 synthase (Galgt1)Mice are viable and survive beyond 1 year of age, displaying slow axonal degeneration and locomotor deficits. Males are sterile.Takamiya et al. (1996); Sheikh et al. (1999); Liu et al. (1999)
GalCerGalactosylceramide synthaseMice develop almost normally but quickly start to show hind limb weakness, ataxia, and neurodegeneration. Males are sterile, most likely because of the lack of specific testis lipids, the seminolipids.Coetzee et al. (1996); Honke et al. (2002)
GlcNAcGal-Glc- ceramideß3GnT-5Mice are viable and fertile; 80% of them die between 5–15 months with splenomegaly, enlarged lymph nodes, fur loss, and obesity phenotypes. B cell function is also affected.Togayachi et al. (2010)
GD3GD3 synthaseMice are viable and fertile.Okada et al. (2002)
GD3/GM2/GD2Double KO combinationMice develop and live without a major problem until they suddenly die around 12 weeks of age with skin and epithelium lesions as well as CNS and PNS degeneration. It has been proposed that ganglioside deficiency may activate ectopic complement activation that would favor the induction of neurodegeneration.Inoue et al. (2002); Ohmi et al. (2009)
GM3/GM2/GD2Double KO combinationMice develop according to Mendelian frequency but start to develop ataxia, hind limb weakness, and uncoordination, which increase with age. A minimal percentage of mice survive beyond 5 months. Mice have been shown to display a high level of lactosylceramide and lactosylceramide-3- sulfate, which have been correlated with catastrophic neurodegeneration associated to vacuolar cytosol observed especially in oligodendrocytes. The phenotype is reminiscent of GalCer synthase mutation.Yamashita et al. (2005)
Table 2. Phenotypic Characteristics of GSL Synthase Mutants in Invertebrates
SpeciesGlycosphingolipid productEnzyme namePhenotypeReference
D. melanogasterGlucosyl ceramideGlcTEmbryos die, exhibiting a massive apoptosis upon gene silencing (upon dsRNA injection).Kohyama-Koganeya et al. (2004)
Mactosylceramide/ Man-Glc-Cer (MacCer)EghZygotic mutants develop until late larval/pupal stages and exhibit perturbation in photoreceptor axogenesis. Maternal mutants are mildly neurogenic and lack their ventral ectoderm. Hypomorphic mutations allowing adult viability are associated with perturbation of the courtship behavior.Ellis and Carney (2011); Fan et al. (2005); Goode et al. (1996a); Rubsam et al. (1998); Soller et al. (2006); Wandalll et al. (2003, 2005).
GlcNAcβ1-3Manβ1- 4Glcβ1-Cer (Ap3Cer)BrnZygotic mutants develop until late larval/pupal stages. The development of follicular tissue is affected and this phenotype has been associated with perturbations in the gradient formation of one of the EGFR ligands in Drosohila as well as Notch ligand loss of function. Maternal embryonic mutants exhibit a mild hyperplasia of the nervous system and a hypoplasia of their epiderm. Brnfs107 hypomorphic alleles have a locomotion defect similar to β4GalNAcTA mutants (see below).Chen et al. (2007); Goode et al. (1992, 1996a, 1996b); Wandall et al. (2005); Schwientek et al. (2002); Pizette et al. (2009)
Complex GSLsβ4GalNAcTA β4GalNAc TBDouble mutants are viable. β4GalNAcTA defective flies are uncoordinated and exhibit locomotion defects, most likely associated with impairment of their neuromuscular junction formation. β4GalNAcTB loss of function affects specifically follicular epithelium morphogenesis leading to the fusion of embryonic dorsal appendages.Chen et al. (2007); Haines and Irvine (2005); Johswich et al. (2009); Stolz et al. (2008).
Complex GSLsα4GalNAcT1/ α4GalNAcT2Double KO are viable and fertile.Hamel et al. (2010)
C. elegansGlucosylceramideGlcTMutation is lethal and rescued by expressing the protein in a subset of intestinal cells.Marza et al. (2009)
Mactosylceramide- MacCerBre 3No developmental defect but mutation rescues a Notch gain of function.Griffitts et al. (2005)
GlcNAcβ1-3Manβ1- 4Glcβ1-Cer- Cer or Ap3CerBre 5No developmental defect but mutation rescues a Notch gain of function.Katic et al. (2005); Griffitts et al. (2005)
Complex GSLBre 4No developmental defect noticed.Griffitts et al. (2005)

Genetic analysis of subsequent enzymes in the GSL synthesis pathway has provided interesting insights about the in vivo role of GSLs. Deletion of the genes involved in the synthesis of MacCer in invertebrates (corresponding functionally to LacCer in vertebrates) (Wandall et al., 2005) is compatible with an apparent normal life in C. elegans (Katic et al., 2005) but not in Drosophila (Goode et al., 1992, 1996b) (Table 2). In mice (Fig. 2 and Table 1), genetic redundancy and functional diversification may complicate the issue. Indeed, two enzymes appear to regulate the production of LacCer. Additionally, these enzymes are also involved in both O- and N-protein glycosylation (Yamashita et al., 1999). Hence, it is difficult to evaluate the specific impact of LacCer deletion in vertebrate membranes. Simple model organisms, therefore, constitute a precious source of information concerning the role of MacCer and LacCer molecules in development. In Drosophila, MacCer is synthesized by a glycosyltransferase known as Egghead (Egh) (Fig. 2 and Table 2; Wandall et al., 2003). While mutations in the egh gene result in the accumulation of GlcCer lipids, mutations in the gene encoding the enzyme responsible for the glycosylation of MacCer into GlcNAcMacCer, known as Brainiac (Brn) in flies (Fig. 2 and Table 2), lead to the accumulation of both MacCer and GlcCer lipids (Wandall et al., 2005; Pizette et al., 2009). Interestingly, egh and brn mutants exhibit very similar phenotypes that may result from either the accumulation of GlcCer lipids (or any product upstream of GlcCer in the sphingolipid synthesis pathway) or the depletion of complex GSLs (i.e., any product downstream of MacCer in the GSL synthesis pathway). While individuals carrying strong egh and brn alleles die at late pupal stages with no obvious anatomical perturbations, hypomorphic mutant flies are viable but female sterile. Genetic mosaic analyses have revealed that the synthesis of complex GSLs is required in the female germ-line for proper oogenesis (Goode et al., 1992, 1996a, b). In particular, egh and brn mutant oocytes fail to induce the correct proliferation and epithelial morphogenesis of the wild-type somatic follicular cells that surround the oocyte, resulting in defects similar to those seen upon decreased Notch and/or EGFR signaling (Goode et al.1992, 1996a, b; Pizette et al., 2009). Interestingly, the worm homologues of both egh and brn also modulate cell–cell communication mediated by Notch (Katic et al., 2005). We discuss below how egh and brn mutations might affect such essential signaling networks at the molecular level.

LacCer molecules can be further modified in mice by four different enzymes (Fig. 2 and Table 1) to generate, either GM3 (the precursor of all gangliosides), Gb3 (the first GSL of the globoserie), GA2 (the precursor of all asialo-serie GSL), or GSLs from the lacto/neolacto-serie. Interestingly, these four enzymes are developmentally regulated (Xu et al., 2010). Of note, the product of Brn may be either considered as a neutral GSL similar to Gb3 or as an acidic GSL similar to GM3 when covalently linked to phospho-ethanolamine (PE) molecules (Chen et al., 2007). While brn is an essential gene in flies (but not in worms), none of the four enzymes is strictly required for mouse embryonic development, at least individually (Table 1) (Togayachi et al.2010; Takamiya et al., 1996; Sheikh et al., 1999; Yamashita et al., 2003; Okuda et al., 2006; Yoshikawa et al., 2009). This may suggest that the GM3/GA2, Gb3, and Lacto- GSLs are to a certain extent interchangeable during the developmental period as well as after birth (Tables 1 and 2). Mutant mice, however, exhibit specific deleterious phenotypes, suggesting that compensation between the distinct GSL series is not complete (Table 1). Of interest, the GM3 synthase mutant mice exhibit a specific degeneration of the Conti organ and are believed to become deaf early after birth (Yoshikawa et al., 2009).

Finally, we note that the loss of a GSL species may result in the increase of other GSL species from alternate branches (Yamashita et al., 2005; Furukawa et al., 2008) More generally, lipidomic studies in yeast and flies have indicated that compensatory changes maintain cellular homeostasis upon loss of specific lipid species (Guan et al., 2009; Carvalho et al., 2010). Thus, deducing the role of a given GSL species only from loss-of-function studies may be misleading and, obviously, genetic data are best interpreted in the context of complementary molecular data. Also, future lipidomic studies of GSL mutants assessing expression of the various enzymes involved in sphingolipid/GSL metabolism and the relative amount of their molecular products should provide important insights into in vivo sphingolipid homeostasis.

GSLs in Neurological Diseases

This impact of GM3 loss on mouse hearing illustrates the importance of gangliosides in nervous system development. In flies, egh and brn mutant embryos display a mild hyperplasia of the central nervous system and of the chordotonal neurons in the peripheral nervous system (Goode et al., 1996a; S.P. and F.S., unpublished results). Interestingly, chordotonal organs in flies and Conti organ neurons in mammals are specified by a conserved proneural transcription factor called Atonal in flies and Math1 in mice (Ben-Arie et al., 2000; Wang et al., 2002). The lineage specificity seen in flies may constitute an important advantage to determine the identity of the GSL-sensitive factors in Atonal/Math-derived neurons in mammals.

Neurodegenerative phenotypes reflecting the essential role of complex GSLs and gangliosides in the maintenance of neuron homeostasis have been observed in various GSL mutant genetic backgrounds. Mice deleted either for GM2 and Gd3 or GM2 and GM3 synthases develop normally but start to display hind limb weakness, ataxia, and a global neurodegeneration between 3 and 12 weeks of age (Inoue et al., 2002; Yamashita et al., 2005; Ohmi et al., 2009). Knocked-out mice for both GM3 and GM2 synthases do not form gangliosides and display the strongest phenotype. In the majority of the cases, the mice die around 2 months of age with unusually high levels of lactosylceramide and lactosylceramide sulfate (Yamashita et al., 2005). Of interest, the accumulation of MacCer products observed in hypomorphic brn flies (Pizette et al., 2009, Wandall et al., 2005) is accompanied by locomotion defects (Chen et al., 2007). Similarly, mutant flies for the β4GalNAcTA enzyme that catalyzes the production of more complex GSLs (Fig. 2) also present locomotion defects and abnormal development of their neuromuscular junctions (Haines and Irvine, 2005; Chen et al., 2007; Haines and Stewart, 2007; Table 2). Whether or not these various mutants display neurodegenerative features will be important to establish.

In any case, these studies illustrate how animal models may further our understanding of neurodegenerative diseases observed in GSL Storage Disorders as well as in a number of neurodegenerative diseases, such as Parkinson or Alzheimer diseases during which GSL metabolism is also perturbed. The role of GSLs in these pathologies has already been reviewed (Ginzburg et al., 2004; Xu et al., 2010) and we refer the reader to these reviews for a detailed analysis. Here, we simply wish to emphasize that GSL genetic models may constitute powerful tools in order to (1) clarify whether cellular degeneration may be hampered by reducing GSL accumulation; (2) define putative genetic and/or pharmacological targets allowing to compensate for such accumulation and help neurons to reacquire their normal homeostasis (Ginzburg et al., 2004); and (3) set regenerative strategies. This late idea is exemplified by the improved regeneration of hearing neurons in deaf cats observed following GM1 adjunction (Leake et al., 2007), which is believed to increase the autocrine expression of neurotrophic factors and lead to TRK receptor stimulation (Rabin et al., 2002; Mallei et al., 2004).

DEVELOPMENTAL ROLES OF GSLS IN SIGNALING

Our current understanding of the influence of GSLs in development is very limited and largely relies on studies carried out in Drosophila. These have suggested that GSLs affect primarily the activity of signaling ligands. We, therefore, discuss here how alterations in GSL synthesis in mutant flies may affect ligand diffusion, trafficking, and activity.

Complex GSLs Favor the Optimal Diffusion of the EGFR Ligand, Gurken

During oogenesis, the formation of dorsal follicular appendages requires the activation of the EGFR pathway in somatic follicular cells. The EGFR ligand, Gurken, is synthesized and secreted by the germ-line and stimulates EGFR activation in the follicular cells during this process (Pizette et al., 2009). As described above, defective GSL synthesis in the germ-line, but not in the follicular cells, affects both cell proliferation and epithelium organization in the somatic follicular cells surrounding the egg. GSL mutant germ-line clones are also associated to defects in dorsal follicular appendage morphogenesis reminiscent of EGFR mutant phenotypes. Therefore, it was suggested that egh, brn, and β4GalNacTB regulate Gurken function (Goode et al., 1992, 1996a, b; Chen et al., 2007; Pizette et al., 2009) without affecting the function of its receptor. Importantly, brn mutants were shown to affect the proper extracellular diffusion of Gurken ligand at the germ/follicular cell interface (Pizette et al., 2009). The binding of Gurken to GSLs at the surface of the oocyte has thus been proposed to promote its proper diffusion and its availability for interaction with EGFR on the opposite membrane of the follicular cells. Whether GSLs in the oocyte membrane interact directly with Gurken and influence its conformation, as proposed for other GSL-interacting proteins (Fantini and Yahi, 2011), remains to be investigated.

All together, in vitro data demonstrating the influence of GSLs on the conformation and activity of the EGFR itself (Coskun et al., 2011) contrast strikingly with the influence of egh or brn mutations on the EGFR ligand function of Gurken, which is observed during fly oogenesis (Pizette et al., 2009). Additionally, we note that the worm homologues of egh and brn have been functionally linked to Notch but not EGFR in C. elegans. These differences may be interpreted to suggest that in vivo, GSLs act in a context-dependent and/or species-specific manner. Consistent with this interpretation, GM3 synthase mutant mice exhibit an increased phosphorylation of the insulin receptor in skeletal muscle cells but not in adipocytes (Yamashita et al., 2003). Future studies characterizing the molecular signaling impact of GSL in vivo should aim at understanding the molecular basis underlying context-dependent GSL specificity, which will certainly help to fill up the current gap between in vitro and in vivo results.

GSLs Modulate the Activity of Notch Ligands

Notch receptors are activated by transmembrane ligands of the Delta (Dl) and Serrate (Ser)/Jagged families. In the fly embryo, one function of Notch is to limit the number of neuroepithelial cells that become neural precursor cells; in Notch mutant embryos, the nervous system forms at the expense of the epidermis. Interestingly, a similar phenotype has been described upon loss of complex GSLs in embryos produced from egh and brn germ-line clones (Goode et al., 1996a). During oogenesis, Notch signaling is required in part, at stage 6, in cells of the follicular epithelium, to precisely regulate their transition from a mitotic cell cycle to an endocycle (Deng et al., 2001). In this context, Dl signals from the oocyte to activate Notch in the follicular cells surrounding the oocyte. Decreased Notch activation in follicular cells produces defects similar to those seen in the absence of Egh or Brn in the germ-line (Goode et al., 1996a; Rubsam et al., 1998), suggesting that GSLs may positively regulate the activity of Dl in the oocyte.

The binding of Dl (or Ser) to Notch appears to be insufficient for receptor activation. Indeed, ubiquitination of the ligands by the E3 ubiquitin ligases Mindbomb or Neuralized is essential for receptor activation (Le Borgne et al., 2005a). One current model proposes that ubiquitinated ligands are endocytosed and “activated” by an epsin-dependent mechanism before being recycled to the plasma membrane (Fig. 4) (Wang and Struhl, 2004, 2005). Another model proposes that endocytosis of the ubiquitinated ligands is more directly implicated in Notch receptor activation with endocytosis generating a mechanical force pulling on Notch, which would result in exposing a cleavage site on the extracellular part of the receptor (Le Borgne et al., 2005a; Banks et al., 2010). At stage 6 of oogenesis, endocytosis of Dl by an epsin- and dynamin-dependent (but clathrin-independent) mechanism is required in the oocyte for inducing Notch receptor activation in follicular cells (Windler and Bilder, 2010). This signaling activity of Dl may, however, not require its subsequent recycling since the activities of rab11 and auxillin, two proteins implicated in membrane recycling, are not required for Notch receptor activation in follicular cells (Banks et al., 2010; Windler and Bilder, 2010). Together, these data may suggest that complex GSLs facilitate the Dl-mediated pulling of Notch receptor, hence its activation, by modulating the membrane lipid environment of Dl.3

Figure 3.

EGF-induced EGFR trafficking. Models of EGF-induced EGFR trafficking as proposed by Sigismund et al. (2005, 2008) and Chen et al. (2005). In the absence of EGF (1 in A and B), EGFR, Src, and H-Ras localize putatively into distinct nanodomains. In the presence of a low EGF concentration (A; pM range), EGFR in a high-affinity state engage signaling molecules such as Src and H-Ras (2) and cluster into clathrin-coated pits (3). These events are associated with the EGFR-mediated activation of MAPK (via Ras) and PI3K (via Src). At a higher concentration of EGF (B; nM range), the E3 ubiquitin ligase Cbl is recruited and induces the mono- or polymonoubiquitination of the EGFR (2). These ubiquitin moieties recruit Eps15 to the receptor, thereby inhibiting the recruitment of EGFR into clathrin-coated pits (Chen and De Camilli, 2005). Ubiquitinated-EGFR/Eps15 complexes are then endocytosed in a clathrin-independent manner (3). This endocytosis is sensitive to the levels of membrane cholesterol, possibly because of its role in the clustering of the GM3-rich nanodomains that contain EGFR. It will be important in the future to precisely determine how GSL composition and levels modulate clathrin-dependent (A) versus clathrin-independent (B) endocytosis.

Figure 4.

Endocytosis of the Notch ligands. Model illustrating two possible modes of internalization of Serrate (and Delta). Ubiquitination by the E3 ubiquitin ligases, Mib and Neur, is required for the endocytosis and signaling activity of Serrate (and Delta). Ubiquitinated ligands can be internalized in a clathrin-, epsin-, and auxillin-dependent manner (1). Clathrin-independent endocytosis (Windler and Bilder, 2010) of ubiquitinated ligands may involve complex GSL, possibly through the direct interaction of Serrate (and Delta) to complex GSL via their GBM motif (Hamel et al., 2010) (2). Whether nanodomains enriched in sphingomyelin/simple GSL and ceramides that are present in MultiVesicular Bodies mediate the sorting and recycling of Serrate (and Delta) is not known.

Data obtained recently in Drosophila wing imaginal discs are consistent with such a model. Overexpression of α4GalNAcT1, a GSL synthase acting after the β4GalNAcTA/TB step (Table 2) and leading to the formation of complex arthroGSLs, was shown to compensate for the defect in Notch activation observed following the silencing of mindbomb (Hamel et al., 2010). The accumulation of Ser ligand at the plasma membrane seen upon mindbomb silencing (Le Borgne et al., 2005b) could also be overcome by the overexpression of α4GalNAcT1 (Hamel et al., 2010). Whether these results reflect a role of complex GSLs in Ser endocytosis and/or recycling, or in the direct Ser/Notch binding/pulling event, remains an open question.

Interestingly, the impact of membrane organization on Notch ligand activity may have been conserved through evolution since mutation of the worm homolog of brn reduces the ligand-dependent activity of an activated form of a Notch receptor (Katic et al., 2005). In addition, the modulation of membrane cholesterol content in mammal cells has also been proposed to affect Notch ligand functions (Heuss et al., 2008). The presence of evolutionarily conserved GSL-binding motifs in the extracellular domains of both Dl and Ser suggest that these ligands can directly interact with specific GSLs at the plasma membrane (Hamel et al., 2010). These GSL-binding motifs are similar to those found in PrP and HIV gp120 (Mahfoud et al., 2002). PrP and HIV gp120 proteins have both been shown to affect membrane organization and appear capable of hijacking membrane trafficking machinery such as the one implicated in membrane nanotube or cytoneme formation (Sowinski et al., 2008; Gousset et al., 2009). It will be very interesting in the future to test whether the direct binding of Dl and Ser to membrane GSLs can affect their ability to drive their own trafficking and localization into membrane domains involved in signaling.

Finally, we note that studies in Drosophila and in cultured cells point toward a role of GSLs in modulating the activity of an epsin/ubiquitin-based clathrin-independent endocytosis of Dl (Goode et al., 1996a; Banks et al., 2010; Windler and Bilder, 2010; Hamel et al., 2010) and EGFR (Chen and De Camilli, 2005; Sigismund et al., 2005, 2008). Future studies should address whether and how GSL-based nanodomains actually regulate the epsin/ubiquitin machinery.

PERSPECTIVES

The data reviewed here suggest that GSLs have at least two essential functions in development, a first one in the global sphingolipid homeostasis (Hannun and Obeid, 2008) and a second one in the regulation of key signaling pathways essential to development. Two important future goals are to first understand how perturbations in the synthesis of GSLs affect sphingolipid metabolism and, second, to mechanistically link the biological activities of specific GSLs with their particular biophysical properties. There are at least 400 species of GSLs in mammals. Characterizing the biophysical properties of these various GSLs and understanding the functional significance of this diversity is thus a challenging task. In line with this idea, a better understanding of the physical influence on the membrane organization of specific GSLs will certainly allow us to understand more clearly how these lipids affect the membrane properties of stem cells, differentiated cells, and/or aging cells. To reach these goals, the community will need to expand the molecular probes (such as antibodies or GSL-binding peptides) to detect specific GSLs in their native cellular environments using photonic live microscopy and gain access to their ultrastructure organization using electron or atomic force microscopy. Future collaborative work between geneticists, cell biologists, and physicists should unveil new exciting facets of the role of membrane components in differentiation and aging.

Acknowledgements

The authors warmly thank their past and present colleagues for fruitful and intellectually challenging discussions and especially F.-X. Campbell-Valois, D. Del Alamo, J. Fantini, and H. Rouault. S.P. received a post-doctoral fellowship from the Association pour la Recherche contre le Cancer (ARC) and the European Molecular Biology Organization (EMB0-ALTF553-2009). S.P. also obtained a Marie Curie Action/International Reintegration Grant (IRG-277152) from the European commission.

Ancillary