Glycosphingolipid metabolism in cell fate specification

Cellular membranes serve as both barriers and interfaces between topologically distinct biological spaces. The lipid composition of these membranes varies at different cellular locations. For example, the plasma membrane (PM) is rich in sphingolipids compared to intracellular membranes, which results in the PM having distinct biophysical properties (Holthuis and Menon, 2014). Sphingolipids contain a hydrophobic ceramide (Cer) backbone that is composed of a saturated fatty acid and sphingoid base. This allows sphingolipids to establish lateral interactions (both homotypic and with sterols) to yield a tightly packed and thick membrane structure (Hannun and Obeid, 2018; Holthuis et al., 2001). Owing to this lipid composition, the PM is less permeable to ions and peptides compared to intracellular membranes, which matches with its ‘barrier’ function towards the extracellular environment (Holthuis and Menon, 2014). Sphingolipids also show incomplete miscibility with phospholipids, which results in lateral phase partitioning of the membrane and thus in the formation of membrane microdomains (Simons and Ikonen, 1997). Such microdomains have different affinities for proteins depending on the length and composition of their transmembrane domains, or on their lipid-based membrane anchoring. Specifically at the PM, sphingolipids participate in signaling events by recruiting signaling molecules to, or sequestering them at, membrane microdomains for the modulation of their activities and for their processing into the endocytic cycle (Holthuis and Menon, 2014; Holthuis et al., 2001; Simons and Ikonen, 1997). Given these properties, sphingolipids are proposed to function as fundamental membrane organizers and to make up the fabric of eukaryotic PMs in order to influence the interaction with the extracellular environment (Hannun and Obeid, 2018; Holthuis et al., 2001).

Interestingly, different cell types exhibit a specific sphingolipid array at their PMs (Hakomori, 2003; Ngamukote et al., 2007) (Table S1). Indeed, sphingolipids are subjected to remarkable structural variations that lead to the production of hundreds of different species (Hannun and Obeid, 2008, 2018). A substantial part of this variability derives from the heterogeneous elongation of glycan chains that are covalently linked to the sphingolipid backbone in the synthesis of the class of compounds known as glycosphingolipids (GSLs). GSL-associated glycans range have between one and more than 20 sugar residues, with 11 different monosaccharide types being used in vertebrates (D'Angelo et al., 2013a). Importantly, the elongation of glycans in GSLs is not driven by a template; instead, it entirely depends on the relative expression and organization of their specific synthetic enzymes (Bieberich et al., 2002; Giraudo and Maccioni, 2003). Still, GSL production is tightly controlled during differentiation programs; as a result, specific GSLs are used as differentiation stage or cell-type-specific markers (D'Angelo et al., 2013a). In addition, GSL composition can substantially vary among single cells in syngeneic cell populations (Majoul et al., 2002; Russo et al., 2018; Snijder et al., 2009). Furthermore, specific GSL glycans appear to organize interactions with receptors that are located at the PM in order to modulate their activity (Bremer and Hakomori, 1982; Bremer et al., 1984; Coskun et al., 2011; Farooqui et al., 1999; Liu et al., 2008; Mirkin et al., 2002; Mutoh et al., 1995; Park et al., 2012; Toledo et al., 2004). This occurs, for instance, in the case of the GM3-dependent inhibition of epidermal growth factor receptor (EGFR) signaling, which maintains EGFR in an inactive state in the absence of its ligand (Coskun et al., 2011). By contrast, GD1a and GM1 enhance EGFR activation (Li et al., 2001, 2000; Liu et al., 2004). Thus, two otherwise identical cells can react differently to the same stimulus owing to their different composition in GSLs.

Whereas the role of cell-to-cell variability in GSL composition in differentiated cells remains to be understood, non-genetic heterogeneity has been proposed to contribute to cell-type diversification in developmental processes (Huang, 2009). Specifically, non-genetic heterogeneity provides cells with transitory ‘states’ to potentially orient their fates towards diverging directions (Huang, 2009). Given the role of GSLs in modulating cell responses to environmental cues, along with their extensive structural variation, cell-to-cell heterogeneity in GSL composition might therefore help in generating identity patterns during tissue morphogenesis. In this Review, we discuss the role of GSLs as cell-fate determinants, focusing on (1) how GSL diversity is generated, (2) what GSL changes occur when cells differentiate toward alternative fates, and (3) how the GSL metabolism is controlled by differentiation programs. Finally, we will speculate on how GSLs can contribute to tissue patterning and morphogenesis.

GSL synthesis is initiated at the cytosolic membrane leaflet of the endoplasmic reticulum (ER), where Cer is produced from its precursor sphinganine by the consecutive action of enzymes that catalyze its acylation and desaturation (Mullen et al., 2012). Cer can then be converted into several compounds that include sphingosine, Cer-1-phosphate, acyl-Cers, sphingomyelin (SM) and GSLs (Hannun and Obeid, 2018; Holthuis et al., 2001; Merrill et al., 2005). SM and GSLs are synthesized at the interface between the ER and Golgi complex and constitute the major sphingolipids at the PM. For its metabolic conversions, Cer can be galactosylated in the ER to produce galactosylceramide (GalCer), extracted from ER membranes by the lipid-transfer protein ceramide transfer protein (CERT) and delivered to the trans-Golgi where SM is synthesized (Hanada et al., 2003), or transported in vesicles to the cis-Golgi where it is glucosylated to produce glucosylceramide (GlcCer) (Funakoshi et al., 2000) (Fig. 1). Whereas SM cannot be further processed in an anabolic direction, GalCer is the precursor of GSLs from the gala-series, also known as sulfatides, which include sulfo-GalCer, (α2-3)-sialylated GalCer (GM4), di-GalCer (i.e. Gal-GalCer) and di-sulfo-GalCer, which are produced at the Golgi complex where the enzymes for GalCer processing reside (Merrill, 2011) (Fig. 1).

Apart from the gala-series GSLs, all other GSLs have GlcCer as a precursor (D'Angelo et al., 2013a; Merrill, 2011). GlcCer is converted into lactosylceramide (LacCer; Gal-GlcCer) (Kumagai et al., 2010; Nishie et al., 2010), which is the metabolic branching point for the formation of all remaining GSLs. They are categorized into four classes (i.e. the globo, lacto, ganglio and asialo series) and their cumulative number exceeds 400 GSLs (D'Angelo et al., 2013a; Merrill, 2011; van Meer et al., 2008). Thus, LacCer is the substrate of (1) GA2 synthase (GA2S) for the synthesis of GA2 (GalNAc-LacCer) and of asialo-series GSLs (Nagata et al., 1992), (2) of GM3 synthase (GM3S) for the synthesis of GM3 (NeuAc-LacCer) and of ganglio-series GSLs (Ishii et al., 1998), (3) Gb3 synthase (Gb3S) for the synthesis of Gb3 (Gal-LacCer) and of globo-series GSLs (Kojima et al., 2000), (4) Lc3 synthase (Lc3S) for the synthesis of Lc3 (GlcNAc-LacCer) and of lacto-series GSLs (Biellmann et al., 2008) (Fig. 1).

After being conveyed to one of these four major metabolic directions, GSLs are processed in glycosylation pathways. There, GSLs are often substrates of multiple possible reactions that lead to further diverging metabolic directions or to the formation of branched glycan structures (Fig. 1) (D'Angelo et al., 2013a; Merrill, 2011). The elongation of glycan residues in GSLs is indeed the result of the ordered action of glycosyltransferases; their relative levels, topological organization within the Golgi stack and presence in multi-enzymatic complexes are key factors in the determination of the metabolic outcome (Maccioni et al., 2011). Along with these parameters, another factor that influences glycan elongation in GSLs is substrate availability. GlcCer, the common precursor of most GSLs, can be delivered to specific sub-Golgi regions by different transport mechanisms [i.e. vesicular, or non-vesicular through the action of the lipid transfer protein 4-phosphate adapter protein 2 (FAPP2, also known as PLEKHA8); D'Angelo et al., 2007], where each of these transport routes feeds a distinct glycosylation pathway (D'Angelo et al., 2013b). However, in spite of the non-deterministic nature of the GSL synthetic system, when the database of GSL structures was analyzed (Sud et al., 2007), they were found to be assembled according to regular patterns; this suggests that structural heterogeneity in GSL structures is not the result of a random process and points to them having a biological function.

A major limitation in our understanding of the structural and functional features of GSLs derives from technical difficulties: determining the GSL composition of a biological sample remains an analytical challenge. GSL composition is specific to the species, cell type and condition (Hakomori, 2008). Moreover, GSLs largely differ in their abundance, chemical stability and biophysical properties, which makes their uniform extraction from biological samples difficult. In addition, the monosaccharide units in GSL chains have very similar chemical structures, which, together with heterogeneous positioning and the anomery of the sugar–sugar bonds and glycan chain branching, complicate GSL analysis (Merrill, 2011). However, the accuracy in resolving GSL composition has improved as technologies have improved. Thus, whereas orcinol-sulfuric acid staining and radioactive labeling with ³H- or ¹⁴C-labeled monosaccharides coupled to chromatographic separation are still valuable procedures for a rapid and inexpensive assessment of GSL composition (Schnaar and Kinoshita, 2015), detection with specific lectins or antibodies and mass spectrometry-based methods now represent the golden standards for GSL profiling (Wuhrer, 2013). Thanks to these advancements, it is now possible to evaluate GSL changes in biological samples with good accuracy, although an absolute quantification is often not possible owing to lack of complete reference standard samples (Farwanah and Kolter, 2012).

In the following sections, we will discuss the changes GSLs undergo during cellular differentiation in developmental processes, as well as during oncogenic transformation of tissues.

Numerous studies have reported that the composition of GSLs in the membrane is remodeled during embryonic development (Cochran et al., 1982; Handa and Hakomori, 2017; Kannagi et al., 1983; Ngamukote et al., 2007; Yamashita et al., 1999). These compositional changes have been evaluated during the three major developmental stages in mice {i.e. preimplantation [embryonic day (E) 0.5–6.5], gastrulation (E6.5–E10.5) and organogenesis (E10.5–E17.5)} (Handa and Hakomori, 2017) (Fig. 2). The preimplantation phase is dominated by GSLs of the lacto series [i.e. stage-specific embryonic antigen 1 (SSEA-1) and Le^y] and globo series (i.e. Forssman antigen, Gb4, SSEA-3 and SSEA-4) (Handa and Hakomori, 2017; Sato et al., 2007). During gastrulation, production of the ganglio-series GSLs is induced in both neuronal and glial cell precursors (Goldman et al., 1984), whereas SSEA-3, Forssman antigen and Gb4 globosides are restricted to visceral mesoderm cells and to the inner cell mass of the growing blastocysts (Handa and Hakomori, 2017). Finally, during the organogenesis phase, the GSLs that are most prominently synthetized are gangliosides; their relative amounts change in the nervous system from post gastrulation (E8) to adult ages. Thus, GM3, GD3 and GD2 are expressed at day E8, whereas GM1, GD1a, GD1b, GT1b and GQ1b are induced starting from E14 (Ngamukote et al., 2007) (Fig. 2).

Changes in GSL expression have also been measured during in vitro differentiation of pluripotent cells into the three germ layers (i.e. ectoderm, mesoderm and endoderm) (Liang et al., 2010, 2011; Russo et al., 2018) (Table S1). Pluripotent stem cells express GSLs of the globo and lacto series (Breimer et al., 2017; Liang et al., 2010, 2011; Russo et al., 2018), including Gb3, Gb4, Gb5 (SSEA-3), α1-2 fucosylated-Gb5 (Globo H), sialyl-Gb5 (SSEA-4) and disialyl-Gb5, globo-A, Lc3, Lc4, SSEA-1 and fucosyl-Lc4 (Breimer et al., 2017) (Fig. 3). The levels of globo- and lacto-series GSLs decrease upon differentiation of pluripotent stem cells to neuronal progenitors, which is followed by the increase in the synthesis of GD3, GM3, GM1 and GD1 (Kwak et al., 2006; Liang et al., 2011; Marconi et al., 2005; Russo et al., 2018). In contrast, when embryonic stem cells differentiate into definitive endoderm, the major GSL that is expressed is Gb4 (Liang et al., 2011) (Fig. 3). GSL composition dynamically changes during the differentiation of mesenchymal stem cells (MSCs) from adult bone marrow into multiple cell lineages. Indeed, MSCs express SSEA-4 (Bergante et al., 2014; Gang et al., 2007) along with GD1a and GD2 gangliosides (Bergante et al., 2014), whereas in MSC-derived adipocytes, the major GSLs are GM3 and GD1a (Kojima et al., 2015), and GM3 and GD3 are expressed in MSC-derived chondrocytes (David et al., 1993). Moreover, lacto-series GSLs and GM3 are expressed in pre-B-cells, whereas mature and activated B cells express GM3 and the globo-series GSLs Gb3 and Gb4 (Taga et al., 1995; Wiels et al., 1991; Wipfler et al., 2011) (Fig. 3).

These data indicate that developmental programs are accompanied by the reprograming of GSL metabolism.

Importantly, active GSL synthesis is required for embryonic development: both GlcCer synthase (GLS, encoded by UGCG) and LacCer synthase (B4GALT5) (Fig. 1) knockout (KO) mice, which are unable to synthesize GSLs through the GlcCer precursor, die by E10.5 (Allende and Proia, 2014; Nishie et al., 2010; Yamashita et al., 2002, 1999). In both cases, the embryo is able to progress through pre-implantation phase, but not beyond formation of the three germ layers (Allende and Proia, 2014; D'Angelo et al., 2013a; Yamashita et al., 1999). Further analyses of mice that harbor defects in the pathways of GSL synthesis support the idea that there is a tissue-specific role for the GSL subclasses (Allende and Proia, 2014; D'Angelo et al., 2013a). Knockout of B3GNT5 – the gene encoding the first enzyme involved in the synthesis of lacto-series GSLs (i.e. Lc3 synthase) (Fig. 1) – results in either preimplantation lethality or multiple postnatal defects (Biellmann et al., 2008). Conversely, the genetic disruption of globo or ganglio series GSL production yields a wide range of immune and neurological phenotypes, respectively (Allende and Proia, 2014). In addition, loss-of-function mutations in three genes that encode enzymes involved in the synthesis of ganglio-series GSLs cause neuronal disease in humans (Boccuto et al., 2014; Boukhris et al., 2013; Fragaki et al., 2013; Harlalka et al., 2013; Simpson et al., 2004).

Altogether, this evidence highlights that (1) GSL cell composition is remodeled when cells differentiate, and (2) that GSL synthesis has a role in differentiation and development. Along these lines, aberrant changes in GSL metabolism are coupled to altered cell differentiation and malignant cell transformation, as discussed in the following section.

Aberrations in GSL metabolism have also been linked to cancer (Gouaze-Andersson and Cabot, 2006; Morad and Cabot, 2013; Ogretmen, 2018). In fact, similar to the events during normal embryonic development and tissue lineage differentiation, cells rearrange their GSL composition during oncogenic transformation (Hakomori, 1998; Hakomori and Zhang, 1997). This rearrangement has been suggested to contribute to cellular transformation, metastasization and the emergence of multi-drug resistance (Gouaze-Andersson and Cabot, 2006; Hakomori and Zhang, 1997; Jacob et al., 2014; Kovbasnjuk et al., 2005). A recent study on mammalian target of rapamycin (mTOR)-induced liver cancer showed that hyperactive mTOR signaling results in increased GSL synthesis (Guri et al., 2017), and that GSL production is strictly required for mTOR-dependent cancer development (Guri et al., 2017), but how exactly do GSLs contribute to the different aspects of oncogenesis?

Signal transducers, adhesion molecules and growth factor receptors that participate in malignant transformation and development of drug resistance are often GSL targets. For instance, in breast cancer, increased GD3 and GD2 synthesis favors stem cell proliferation by fostering the activation of growth factor receptors on the PM (Liang et al., 2013) and promoting resistance to treatment with Gefitinib, a tyrosine kinase inhibitor that targets EGFR (Liang et al., 2017). Cisplatin is a chemotherapeutic agent that is used for the treatment of a number of cancers, such as non-small cell lung cancer (NSCLC) and malignant pleural mesothelioma (MPM). It induces Cer production, leading to cell cycle arrest and apoptosis (Dasari and Tchounwou, 2014; Nowak, 2012). Drug-resistant cancer cells escape apoptosis by increasing GSL synthesis at the expense of an accumulation of Cer, which also leads to increased expression of the multidrug resistance-associated protein 1 (MRP1), which stimulates drug efflux (Tyler et al., 2015).

GSL reprograming has a role in the epithelial-to-mesenchymal transition (EMT), which is the process that enables metastatic cellular invasion in the context of cancer progression. The induction of EMT in vitro by transforming growth factor β (TGFβ) treatment is accompanied by a reduction in the levels of asialo-GSLs GM1 and GM2, whereas complex gangliosides are, in turn, induced during this process (Guan et al., 2009; Mathow et al., 2015). Interestingly, a subpopulation of cells that express low levels of epithelial markers has been identified in prostate tumors. This subpopulation expresses high levels of SSEA-4 and spontaneously escapes from adhesive colonies and forms invadopodia-like migratory structures. This supports the idea that SSEA-4 is a marker for metastasizing cells that have acquired a mesenchymal nature (Sivasubramaniyan et al., 2015).

Moreover, for a number of tumors, the overproduction of a specific GSL has been reported. These GSLs can be used as tumor-associated antigens (TAAs) for the definition of the tumor type and stage (Table 1 and references therein). Importantly, GSLs that serve as TAAs have been exploited to develop vaccine strategies to elicit a specific cytotoxic and/or humoral immune response against tumor cells (Dobrenkov and Cheung, 2014). GD2-targeted immunotherapy of neuroblastoma has become the first GSL-targeting immunotherapy to obtain food and drug administration (FDA) approval for medical care (Dobrenkov and Cheung, 2014). Moreover, innovative strategies to target GSL-TAAs also imply that toxins that use these GSLs as the natural receptors in their target cells could be used for cancer treatment; this is the case for a Shigella toxin, which recognizes Gb3 that is overexpressed in gastric adenocarcinomas (Geyer et al., 2016).

Thus, metabolic alterations of GSLs are inherent components of cancerogenesis as they (1) originate from the malignant transformation process, (2) contribute to cancer-relevant phenotypes and (3) define cancer-specific cell states.

The aforementioned metabolic changes in GSLs, both in cancerogenesis and developmental contexts, are often the consequence of a reprograming in the expression of genes that encode the enzymes that synthesize GSLs. During neural differentiation, for instance, the expression of genes encoding enzymes for the synthesis of globo- and lacto-series GSLs (i.e. A4GALT, encoding Gb3 synthase, and B3GNT5, encoding Lc3 synthase) decreases; at the same time, expression of genes encoding enzymes of the ganglio series synthesis pathway (i.e. ST3GAL5, encoding GM3 synthase, and B4GALNT1, encoding GA2/GM2 synthase) increases (Liang et al., 2010, 2011; Russo et al., 2018). Similarly, during EMT, the production of GSLs is switched from the asialo to ganglio series (Fig. 4A) owing to the induction of ST3GAL5 and ST8SIA1 (encoding GD3 synthase) and to the repression of B3GALT4 (encoding GA1/GM1 synthase) (Mathow et al., 2015). These data suggest that dedicated regulatory circuits exist to reorient the GSL pathways.

During TGFβ-induced EMT, the mothers against decapentaplegic homolog 3 and 4 (Smad3–Smad4) complex represses B3GALT4 by binding to its promoter (Guo et al., 2015), whereas zinc finger E-box-binding homeobox 1 (Zeb1) (a transcriptional target of Smad3–Smad4) binds to and activates the promoters of both ST3GAL5 (Mathow et al., 2015) and ST8SIA1 (Dae et al., 2009). Importantly, exogenous provision of GA1 (one of the products of B3GALT4/GA1 synthase) inhibits TGFβ-induced EMT, which suggests that GSLs are both targets and regulators of the same signaling pathway (Guan et al., 2009; Guo et al., 2015) (Fig. 4A). Moreover, we recently demonstrated that, during differentiation of stem cells into neural cells, the decrease in globo-series GSLs (owing to A4GALT/Gb3 synthase repression) triggers the expression of the chromatin-remodeling factor autism susceptibility gene 2 protein (AUTS2), which, in turn, binds to the ST3GAL5 promoter where it stimulates local histone acetylation and transcriptional activation of ganglio-series GSLs (Russo et al., 2018). Similar to what is seen during EMT, the addition of GSLs that are repressed in differentiated cells (globo-series GSLs) counteracts both the differentiation process and metabolic reprograming (Russo et al., 2018) (Fig. 4B). Interestingly, ganglio-series GSLs (i.e. GM1) have been found to stimulate neuronal differentiation and to sustain the expression of enzymes that synthesize ganglio-series GSLs (i.e. GM2S) by promoting histone acetylation at their promoters (Tsai et al., 2016; Tsai and Yu, 2014), which ultimately leads to maturation of the neuronal population (Fig. 4C). An increase in histone acetylation at the GM2S (B4GALNT1) promoter was indeed observed in developing mouse brains, where it correlates with GM2S mRNA expression (Suzuki et al., 2011). This evidence reveals the existence of a two-way relationship between GSL metabolism and transcriptional programs that affect cell fate determination.

How exactly do GSLs influence gene expression? One fundamental feature of GSLs is to organize protein–carbohydrate or carbohydrate–carbohydrate interactions with structural proteins and receptors at the PM (Chakrabandhu et al., 2008; Coskun et al., 2011; Kawashima et al., 2009; Liang et al., 2017; Liu et al., 2004; Mutoh et al., 1995; Park et al., 2012; Russo et al., 2016). The GSL-glycan moiety can, indeed, interact directly with a specific amino acid residue (Coskun et al., 2011) within a protein domain, or with a glycan portion (Heuss et al., 2013) of PM proteins. In doing so, GSLs regulate PM proteins through (1) their conformation, (2) their accessibility to ligands, (3) their oligomerisation state and/or (4) their partitioning into membrane microdomains (Russo et al., 2016).

Thus, GSLs can modify cell signaling in response to specific stimuli. Examples for this are the interactions of GSLs with EGFR (Hofman et al., 2008; Park et al., 2017; Coskun et al., 2011) and the notch ligand delta-like 1 (Dll1) (Heuss et al., 2013). Here, their signaling – and as a consequence, the downstream transcriptional responses – are influenced by their interaction with GSLs at the PM. GSLs also affect signal transduction and gene expression by regulating endocytosis (Lakshminarayan et al., 2014). GSLs interact with the secreted carbohydrate-binding protein galectin-3, which in turn triggers GSL-dependent biogenesis of specific cargo-laden endocytic carriers (Lakshminarayan et al., 2014). Through this mechanism, GSLs regulate the exposure of specific receptors and proteins at the PM. Another level of action for GSLs on gene expression is directly within the nucleus, where GSLs have been found at the nuclear lamina where they directly contact chromatin and influence the activity of promoters (Tsai et al., 2016).

Thus, the impact of GSLs on signal transduction pathways and on transcriptional programs upon their activation (Regina Todeschini and Hakomori, 2008). As outlined above, the transcriptional responses to changes in cellular GSL content often involve enzymes of the GSL synthetic pathway itself, thus resulting in self-contained regulatory loops that lead to metabolic switches. The final output of these transcriptional metabolic interplays influences cell fate decisions and differentiation programs.

Glycans of GSLs protrude out of the PM towards the extracellular space. This peculiar position makes GSLs specifically suited to interact with glycans and proteins that are present either on the same PM or on the PM of adjacent cells. Through these interactions, GSLs influence signaling, receptor trafficking, cell–cell contacts and adhesion, and, thus, ultimately gene expression and cell fate determination. The structural diversity in GSLs and their tissue-specific production suggest that distinct GSLs influence cell fate decisions towards differentiation trajectories.

During development and tissue pattering, individual progenitor cells are subjected to specific differentiation programs in order to achieve the formation of functional anatomical structures through morphogenesis (Basson, 2012). Whereas specific hormone gradients sustain morphogenetic processes, the events that initiate morphogenesis usually happen in a uniform context, that is, among undifferentiated and genetically identical cells that are exposed to a homogeneous environment (Tabata and Takei, 2004).

Cell-to-cell variability in gene expression (either stochastic or dependent on the cellular microenvironment) has been proposed to drive these early morphogenetic events by changing cell differentiation potential (Huang, 2009). This provides otherwise identical cells with the capability to break symmetry within the population (Huang, 2009) (Fig. 5). According to this concept, progenitor cells can follow alternative differentiation trajectories to achieve one of multiple stable states; this eventually leads to cell fate decisions that depend on the oscillating expression of a key factor (Huang, 2009). Besides proteins and nucleic acids, small molecules, which include lipids and, specifically, GSLs, are able to influence these cell differentiation programs. Moreover, GSL composition varies among cells in a syngeneic cell population owing to cell cycle phase (Majoul et al., 2002), the local microenvironment (Snijder et al., 2009) or to metabolic circuits (Russo et al., 2018). Whether this variability is involved in symmetry-breaking events in morphogenesis remains to be addressed (Fig. 5).

Research devoted to the dissection of the role for GSLs in regulatory gene expression circuits at single-cell resolution in advanced models of development and morphogenesis (i.e. in organoids) is probably the missing and required step to attain a sufficient body of knowledge on the role of GSLs in development, morphogenesis and tissue patterning.