Vertebrate protein glycosylation: diversity, synthesis and function

Abstract

Protein glycosylation is a ubiquitous post-translational modification found in all domains of life. Despite their significant complexity in animal systems, glycan structures have crucial biological and physiological roles, from contributions in protein folding and quality control to involvement in a large number of biological recognition events. As a result, they impart an additional level of ‘information content’ to underlying polypeptide structures. Improvements in analytical methodologies for dissecting glycan structural diversity, along with recent developments in biochemical and genetic approaches for studying glycan biosynthesis and catabolism, have provided a greater understanding of the biological contributions of these complex structures in vertebrates.

Glycans associated with cell surface and intracellular proteins and lipids contribute to numerous biological functions in animal systems¹. Oligosaccharide structures at the cell surface influence interactions with the extracellular environment by providing ligands for cell adhesion, macromolecule interactions and pathogen invasion^1,2. The glycans associated with cell surface receptors and proteins also directly modulate protein function and signalling, as well as altering the dynamics of glycoprotein endocytosis and cell surface half-life through binding to multivalent lectins³. Glycan structures on newly synthesized glycoproteins are crucial for protein secretion, as they influence protein folding, provide ligands for lectin chaperones, contribute to quality control surveillance in the endoplasmic reticulum (ER) and mediate transit and selective protein targeting throughout the secretory pathway^4,5. More general roles of glycan structures include contributions to the regulation of cytosolic and nuclear functions, immune surveillance, inflammatory reactions, auto-immunity, hormone action and tumour metastasis^1,6. It is clear that these post-translational modifications confer an additional information content at the molecular and cellular level that extends beyond the simple information flow from genome-derived transcripts to encoded protein functions⁷.

Protein glycosylation is broadly used by cell biologists to monitor protein transit through the secretory pathway (for example, endoglycosidase H-resistant glycan structures are used to monitor protein translocation in the Golgi⁸). Moreover, glycans detected by specific antibodies⁹ and ectopically expressed fluorescently tagged glycosylation enzymes¹⁰ are used as markers for intracellular compartments. Glycan structures also distinguish cell types in developing and mature animal tissues². Despite this limited use of mammalian glycan structures or glycosylation machinery as tools for cell biology studies, few cell biologists have truly embraced the diversity and complexity of glycan modifications for their contributions and roles in biological systems. This results from an intimidating collection of challenges for studying the functions of glycan structures at the molecular, cellular and organismal level.

It has been estimated that approximately 700 proteins are required to generate the full diversity of mammalian glycans (estimated to be ≥7,000 structures), which are assembled from only ten monosaccharides: fucose (Fuc), galactose (Gal), glucose (Glc), N-acetylgalactosamine (GalNAc), N-acetylglucosamine (GlcNAc), glucuronic acid (GlcA), iduronic acid (IdoA), mannose (Man), sialic acid (SA) and xylose (Xyl)^7,11,12. Among these proteins, ~200 are glycosyltransferases, which are enzymes that extend acceptor glycan structures using nucleotide or lipid-linked sugars as activated donor substrates. Competition between glycosyltransferases that possess overlapping glycan acceptor preferences but different donor specificities can strongly influence the relative abundance of glycan structural features in the total glycome of a cell or tissue. Thus, the glycan biosynthetic potential of a given cell is determined by multiple factors, including relative enzyme abundance and localization, abundance and trafficking of glycoprotein substrates, and the availability of activated sugar donors in appropriate secretory compartments.

Glycans can be attached to polypeptide structures through amide linkages to Asn side chains (N-glycosylation), through glycosidic linkages (O-glycosylation) to side chains of Ser/Thr, hydroxylysine (collagen) or Tyr (glycogenin), or through C-C linkages to the C2 position of Trp (C-mannosylation). Alternatively, they be can attached as a linker structure bridging glycosylphosphatidylinositol anchors to protein backbones¹³ (TABLE 1). Even among the O-linked structures to Ser/Thr, at least six distinct glycan types are known in animal systems (BOX 1; TABLE 1), with most found on extracellular, cell surface or intracellular membrane compartments. An O-linked monosaccharide addition (GlcNAc-β-Ser/Thr) is also found on a large number of cytosolic and nuclear proteins, where it contributes to the regulation of numerous cellular functions and can compete with Ser/Thr phosphorylation⁶. Monosaccharides and complex extended glycan structures are also found attached to many other macromolecule classes in animal systems, from glycolipids to secondary metabolites, but these latter structures are beyond the scope of this Review.

Table 1.

Consensus motifs and enzymes responsible for various glycosylation reactions^*

Type	Linkage	Enzyme	Consensus sequence	Domain	Examples
N-glycosylation	GlcNAc-β-Asn	OST	N-X-(S/T)‡ (standard sequons)		Nascent polypeptides
			N-X-C‡,§, N-G§, N-X-V‡,§ (non-standard sequons)
O-glycosylation	GalNAc-α-Ser/Thr	ppGALNTs	Isoform specific||		Mucins
	GlcNAc-β-Ser/Thr	OGT	No set consensus¶		Cytosolic, nuclear proteins
	GlcNAc-β-Ser/Thr	EOGT	Unknown	EGF	Notch, Dumpy
	Xyl-β-Ser	XYLT1, XYLT2	a-a-a-a-G-S-G-a-(a/G)-a (‘a’ represents Asp or Glu)#		Heparin, proteoglycan core proteins
	Fuc-α-Ser/Thr	POFUT2	C-X-X-(S/T)-C-X-X-G	TSR	Thrombospondin, ADAMTS family
		POFUT1	C2-X-X-X-X-(S/T)-C3	EGF	Notch, clotting factors
	Glc-β-Ser	POGLUT1	C1-X-S-X-(A/P)-C2	EGF	Notch, clotting factors
	Man-α-Ser/Thr	POMT1, POMT2	I-X-P-T-(P/X)-T-X-P-X-X-X-X-P-T-X-(T/X)-X-X**		α-dystroglycan
	Gal-β-Hyl	GLT25D1, GLT25D2	X-Hyl-Gly (collagen repeats)‡‡		Collagen, adiponectin
	Glc-α-Tyr	GYG	Tyr194 of GYG§§		GYG (autoglycosylation during glycogen formation)
C-mannosylation	Man-α-Trp	Unknown	W-X-X-W||||	TSR	Thrombospondin, ADAMTS family
Glypiation	Pr-C(O)EthN-6-P-Man	Transamidase	No set consensus¶¶		THY1, NCAM1

^*References provided in the text except where noted.

^‡X cannot be Pro.

^§Determined by high-throughput proteomic analysis of lectin-enriched glycopeptides²¹. Other non-standard sequons have been observed for recombinant immunoglobulin G2 antibodies expressed in Chinese hamster ovary cells²³ or mouse laminin²², but these structures have not been confirmed more widely in animal systems.

^||Variation in ppGALNT specificities was determined empirically¹⁵⁴.

^¶Prediction software available from the YinOYang WWW server¹⁵⁵.

^#The specificity for the initiation of proteoglycan core synthesis is defined empirically¹⁵⁶.

^**Proposed consensus¹⁵⁷ with variable support from mass spectrometry analysis¹²³.

^‡‡Collagen domains are modified at Lys residues to form hydroxylysine (Hyl) and galactosylated, then they are extended with an α1,2-Glc residue prior to triple helix formation¹⁴⁴.

^§§Autoglucosylation and extension of the initial glycogen polymer on the glycogenin backbone occurs on the Tyr194 hydroxyl group¹⁴³.

^||||Mannose liked to the C2 of the indole ring of the tryptophan residue based on the corresponding consensus sequence¹¹¹.

^¶¶Prediction of consensus sequence has been defined based on hidden Markov model and prediction software¹⁵⁸. ADAMTS, a disintegrin and metalloproteinase with thrombospondin motifs; EOGT, EGF domain-specific O-linked GlcNAc transferase; GLT25D1/2, glycosyltransferase 25 family member 1/2; GYG, glycogenin; NCAM1, neural cell adhesion molecule 1; OGT, O-linked GlcNAc transferase; OST, oligosaccharyltransferase; POFUT, protein O-fucosyltransferase; POGLUT, protein O-glucosyltransferase; POMT, protein O mannosyltransferase; ppGALNTs, polypeptide GalNAc transferases; TSR, thrombospondin type 1 repeat; XYLT, xylosyltransferase.

Box 1. Major classes of vertebrate glycan structures.

Most glycans on membrane and secreted proteins are found in N-linkage to Asn or in O-linkage to Ser/Thr. N-linked glycans that have undergone minimal mannosidase processing are called ‘high-mannose’ glycans. Addition of N-acetylglucosamine (GlcNAc) to the α3 arm of an acceptor containing five Man residues produces a hybrid glycan^14,18. Hybrid glycans can be extended on the α3 arm with galactose (Gal), N-acetylgalactosamine (GalNAc), fucose (Fuc) and sialic acid (SA). Further processing of hybrid glycans produces complex glycans, which can also be decorated with Gal, GalNAc, Fuc and SA.

Glycosaminoglycans (GAGs) are O-linked glycans initiated by a conserved tetrasaccharide (GlcA-β1,3-Gal-β1,3-Gal-β1,4 Xyl-β) and classified by the composition of their disaccharide repeat. GAG chains can be post-synthetically modified by sulphation and epimerization (GlcA conversion to IdoA), producing substantial heterogeneity¹⁴⁰. Glycoproteins carrying one or more GAG chains are called proteoglycans, and can be secreted or transmembrane or glycosylphosphatidiylinositol (GPI)-anchored. Hyaluronic acid, a GAG-like polymer of the extracellular matrix, is the only glycan that is not linked to a protein or lipid¹⁴¹.

Other O-linked glycans are classified by their initiating monosaccharide. Addition of GalNAc to Ser/Thr initiates mucin-type O-linked glycans. Extension with Gal, GlcNAc or GalNAc produces eight different core structures¹⁴². For example, addition of a branching GlcNAc to a core 1 disaccharide (Gal β1,3-GalNAc) produces core 2. Man initiates another class of O-linked glycan (O-Man glycans). A subset of O-Man glycans are extended with a repeating disaccharide — (-3-Xyl-α1,3-GlcA-β-)_n — in phosphodiester linkage to an incompletely defined bridging moiety (X)^121,127. Two types of O-linked Fuc glycan and one type of O-linked Glc glycan can be added to specific Cys-rich domains, as well as C-Man residues on Trp side chains^94,118. O-linked GlcNAc is found on the extracellular domains of some proteins and on numerous cytosolic and nuclear proteins, but different enzymes mediate extracellular and nucleocytoplasmic O-GlcNAcylation⁶. O-linked glycan structures attached to other amino acids include Gal on hydroxylysine of collagen domains that is extended by the addition of an α1,2-Glc residue and the addition of a Glc on Tyr of glycogenin that is extended and branched with additional Glc residues to form glycogen^143,144. In addition to proteins, sphingolipids can be modified by glycosylation. Eukaryotic cell surfaces are enriched in glycosphingolipids, which are ceramide-linked glycans that can be capped and branched with Fuc and SA¹⁴⁵.

Glycan structures attached to proteins can be highly complex, with numerous possibilities for branching and anomeric linkage, and accordingly they have much greater structural diversity than linear nucleic acid or polypeptide structures⁷. Additional complexity and diversity arises from the fact that mammalian glycans are synthesized in intricate biosynthetic pathways, and the resulting products depend on potential competition among multiple enzymes for the same glycan substrates^12,14. This competition, when combined with the relative inefficiency of the enzymatic reactions, often results in a diverse ensemble of branched glycan structures (termed ‘microheterogeneity’), even at the same glycosylation site of otherwise identical proteins¹⁵. The diversity of glycan modification can also produce complex pleiotropy, where unique modifications on one glycosylation site may alter function or recognition within a specific cellular context, but may cause other effects or be functionally silent in other contexts. Finally, the diversity of glycan structures found in a typical cell or tissue sample presents considerable challenges for structural elucidation, even when using the most contemporary analytical approaches¹⁶. Thus, the goal of determining the roles and contributions of individual oligosaccharides in various biological contexts is equivalent to hunting for a handful of discrete glycan functions within a forest of cellular carbohydrate structures. This complexity commonly dissuades non-glycobiologists from exploring the functional information content embedded within these glycan modifications.

To facilitate broader exploration by non-glycobiologists of the functional information content embedded within glycan structures, this Review article focuses on four key areas of contemporary glycobiology research that have a great impact within the cell biology community: the roles of glycans in quality control; the mechanisms of glycoprotein transport and modification through the Golgi complex; protein domain-specific glycosylation; and high-throughput analytical approaches for glycan structural profiling. The overall goals are to highlight the latest developments in the glycobiology field while emphasizing the unique information content and functions that glycans have in animal systems.

N-glycan synthesis and quality control

Transfer of glycan precursor to nascent polypeptides

N-glycosylation is the most highly studied form of protein glycosylation in eukaryotic organisms. It has been estimated that approximately half of all human proteins are glycoproteins, and most of those contain N-glycan structures¹⁷. N-glycans are initially synthesized as a lipid-linked oligosaccharide (LLO) precursor, and the glycans are transferred from LLO to a nascent polypeptide chain during translation. Although the general details of co-translational glycan transfer have been known for decades¹⁸, the mechanistic details of oligosaccharide transfer have only recently been revealed¹⁹. Eukaryotic organisms generally use a multi-subunit oligosaccharyltransferase (OST) on the lumenal face of the ER membrane to catalyse glycan transfer to acceptor peptide sequons²⁰, which are comprised of an Asn-X-(Ser/Thr) tripeptide (and less frequently of Asn-X-Cys²¹ and other non-standard sequons^22,23), where X can be any amino acid except for Pro (FIG. 1a; TABLE 1). One OST subunit, STT3, contains the catalytic site of the enzyme²⁴ and the mechanism of its interaction with the tripeptide sequon has long been hypothesized on the basis of theoretical and biochemical studies²⁵.

Figure 1. Protein N-glycosylation and quality control of protein folding.

a | During glycoprotein biosynthesis, the translation of nascent polypeptides is followed by their translocation through the SEC61 pore and the simultaneous transfer of a glycan from a lipid-linked intermediate to peptide acceptor sequons by the oligosaccharyltransferase (OST). One cleft in the STT3 subunit of OST scans for Asn-X-Ser/Thr acceptor sequons, while an adjacent cleft binds the glycan donor. b | Glycan trimming through Glc removal occurs immediately after transfer by α-glucosidase I (GIsI) and the α-glucosidase II α–β heterodimer (GIsIIα/β). Folding intermediates containing Glc₁Man₉GlcNAc₂ structures are ligands for the lectins calnexin or calreticulin, which function in complex with ERp57. Dissociation from the lectins is followed by further Glc cleavage. Additional chaperone assistance is provided by the ATP-driven chaperone BiP (also known as GRP78). Correctly folded glycoproteins are packaged for transport to the Golgi. c | Incompletely folded glycoproteins are recognized by the folding sensor UDP-Glc:glycoprotein glucosyltransferase (UGGT1). They are then re-glucosylated through the addition of a Glc residue back to the glycan structure and are reintegrated into the calnexin cycle. d | Terminally misfolded glycoproteins are subjected to endoplasmic reticulum (ER) disposal by Man trimming (through the activity of ER α-mannosidase I (ERManI) or GolgiManIA, GolgiManIB and GolgiManIC (not shown), followed by the activity of ER degradation-enhancing α-mannosidase-like 1 (EDEM1), EDEM2 and EDEM3 (which are homologues of Htm1 (also known as Mnl1) in yeast)). The trimmed glycans bind the ER lectins OS9 or XTP3B (not shown) and are translocated into the cytosol via a complex of derlin 1 (DER1), DER2 and DER3 (the SEL1L complex; Hrd3 in yeast) using the driving force of the cytosolic ubiquitin binding protein and ATPase functions of valocin-containing protein (VCP; also known as TER ATPase; known as Cdc48, Ufd1 or Npl4 in yeast). The peptide is deglycosylated by a cytosolic PNGase and degraded by the proteasome¹⁴⁹.

Recent work has identified a homologous form of N-glycosylation on similar acceptor sequons of periplasmic and secreted glycoproteins in a restricted set of Gram-negative bacteria^26–28. Bacterial glycosylation also relies on the transfer of glycans from an LLO precursor to nascent polypeptide chains, but the enzyme responsible is a single polypeptide, PglB, with high sequence similarity to eukaryotic STT3 (REFS 27–29). The structure of Campylobacter lari PglB with a bound peptide acceptor was recently solved¹⁹, and the co-complex provides interesting insights into the mechanistic details of sugar transfer for both bacteria and eukaryotes. PglB is anchored in the bacterial membrane via 13 transmembrane helical segments while a large globular domain faces into the periplasm. Two clefts are found at the junction between the transmembrane and globular domains, and the acceptor peptide was found to be positioned across one of the clefts with its Asn side chain extending through a hole into the adjacent cleft, which is presumed to bind the LLO donor. The nature of the interactions between the peptide acceptor and the active site cleft, the indirect catalytic role of a protein-bound divalent cation, and the conformational change in a loop region required for product release all provide novel insights into the mechanism of glycan transfer²⁵. The additional subunits required for eukaryotic OST activity presumably act to facilitate coupling to the Sec61 protein translocation pore and to select full-length Glc₃Man₉GlcNAc₂ LLO glycan substrates, as truncated glycans are transferred with greatly reduced efficiency in animal systems²⁰.

N-glycosylation in quality control of protein folding

Following transfer of Glc₃Man₉GlcNAc₂ glycans to nascent polypeptide chains on the lumenal face of the ER in eukaryotes, trimming of two Glc residues results in the formation of a GlcMan₉GlcNAc₂ structure that acts as a ligand for the membrane-bound or soluble lectin chaperones, calnexin and calreticulin³⁰ (FIG. 1b). These lectins are tethered to ERp57 (also known as protein disulphide-isomerase A3 (PDIA3)), and they protect nascent polypeptides from hydrophobic aggregation and non-productive disulphide bonding during their folding itinerary³¹. Iterative cycles of Glc removal by glucosidase II and Glc re-addition by UDP-Glc:glycoprotein glucosyltransferase (UGGT1), followed by re-binding to the lectin chaperones, help to facilitate efficient folding of newly synthesized glycoproteins in the ER lumen^30,31 (FIG. 1c). Glucosidase II is comprised of an α (catalytic) and a β (lectin) subunit that facilitate substrate interactions, but neither glucosidase subunits nor the ER lectins calnexin and calreticulin appear to recognize the polypeptide backbone of the folding intermediate. However, UGGT1 clearly recognizes glycan structures in proximity to incompletely folded ‘molten globule’ peptide structures, and it acts as a folding sensor that drives re-entry into the ‘calnexin cycle’^30,32.

Control of misfolded glycoprotein disposal from the ER

In addition to recruiting chaperones during protein folding, glycan structures define the ER residence time for the downstream quality control of newly synthesized glycoproteins. Glycoproteins with slow folding kinetics owing to mutation or incomplete oligomeric assembly of subunits are targeted for degradation by the activity of Man trimming enzymes, which is the basis for the ‘mannose timer’ model of ER quality control³⁰. Glycan trimming is followed by recognition of the trimmed structures by components of a multiprotein complex at the ER membrane that facilitates ‘retro-translocation’ and subsequent proteasomal degradation in a process termed ER-associated degradation (ERAD)^30,31. Recent studies on ERAD in Saccharomyces cerevisiae demonstrated that two related ER mannosidases (Mns1 and Htm1 (also known as Mnl1)) are required for sequential trimming of terminal Man residues³³, with Htm1 working in a complex with Pdi1 (REF. 34). The resulting trimmed glycan is recognized by the ERAD lectin Yos9 (OS9 in humans along with its homologue XTP3B (also known as ERLEC1)), which mediates transfer to the retrotranslocon complex (FIG. 1d).

In mammals, the components involved in glycan trimming and recognition are more complex. Seven enzyme isoforms related to the yeast trimming man-nosidases are found in mammalian cells, where they have roles in both ERAD and glycan maturation to complex type structures in the secretory pathway³⁵. Mannosidase inhibitor and overexpression studies indicate that Man trimming is the proximal and rate-limiting step in ERAD and is tightly coupled to components of the calnexin cycle to assure effective quality control surveillance and clearance of misfolded proteins from the lumen of the ER³⁶. However, the subcellular location of the ERAD compartment, the substrate specificities and the localizations of the trimming enzymes, as well as their relative contributions in ERAD versus glycan maturation, remain controversial^36,37.

Neither the trimming mannosidases nor the Yos9 homologues are believed to recognize the folding state of the underlying polypeptide, and only some glycans on ERAD substrates are required for effective disposal^38,39. An intriguing model has recently been developed suggesting that some protein glycosylation sites mark secondary structure elements that are sensitive to the overall folding state of the protein³⁸. A bipartite signal that combines trimmed glycans and underlying protein conformation would allow the selection of ERAD substrates while sparing fully folded intermediates and ER-resident proteins from disposal. The identity of the ERAD components that couple the glycan and peptide recognition components still remains unclear³⁸. Thus, protein glycosylation has key roles in assisting in protein folding as well as in defining the timing for protein disposal as a part of the quality control function in the early secretory pathway.

Golgi organization and glycomic diversity

Glycan processing in the post-ER secretory pathway

After leaving the ER, glycoproteins traverse the Golgi apparatus, where their glycans are susceptible to further processing. Generally, few ‘high-mannose’ glycans reach the cell surface of most differentiated vertebrate cell types⁴⁰. Almost all glycoprotein glycans are subject to trimming and extension as they transit the Golgi. Localization studies have demonstrated that glycan processing enzymes are distributed across the Golgi in a cis-to-trans topography that correlates successive glycan maturation steps with vectorial traffic through the organelle^41,42 (BOX 2). However, this generic Golgi architecture, where glycan processing enzymes serve as defining markers for specific Golgi subcompartments, is an oversimplification. Overlapping or altered enzyme distributions are common, and different cell types or organisms frequently present unique Golgi organizations^43–47. Thus, Golgi heterogeneity, coupled with an incomplete understanding of the mechanisms that regulate membrane trafficking and sub-Golgi enzyme targeting, complicate the standard definitions of Golgi compartmentalization.

Box 2. Distribution of glycan processing enzymes across a generic Golgi apparatus.

Transitional endoplasmic reticulum (ER) and cis-Golgi compartments contain enzymes that initiate O-linked glycosylation and mannosidases for trimming high-mannose N-linked glycans. Medial compartments contain enzymes that branch O-linked glycans and that initiate complex N-linked glycan formation, branching and core fucosylation^9,74. The N-linked glycans on glycoproteins that achieve this level of processing become resistant to removal by endoglycosidase H, and studies of trafficking frequently rely on this characteristic to report the transit of a protein through the Golgi. Trans compartments elaborate additional branching and capping reactions (galactosylation, sialylation, sulphation and external fucosylation) on complex N-linked and O-linked antennae⁴². Beyond the trans-Golgi, capping reactions are continued in the trans-Golgi network (TGN), a tubulovesicular compartment that stages glycoprotein cargoes for secretion or sorting to specific cell surface or subcellular membrane domains. Bulk flow of Golgi contents from cis-to-trans is counterbalanced by retrograde transport that returns resident Golgi proteins to appropriate cisterna⁶¹. Glycoprotein substrates carrying appropriate glycan acceptors, glycosyltransferases possessing appropriate enzymatic activities, and nucleotide sugar transporters of the appropriate specificities must all be brought together within the same functional unit of the Golgi apparatus in order to achieve glycan maturation.

It is easy to envision that alterations in the glycan assembly line could affect glycan expression. However, it is not always possible to predict the precise effects of altered Golgi morphology on glycan production. Pharmacologic interventions that disperse the Golgi stack produce glycosylation changes that range from relatively minor glycan modifications (after low doses of nocodozole) to extreme under-processing of N-linked glycans (after brefeldin A administration)^48–53. More subtle changes to the Golgi architecture — such as changes that cells might routinely use to modulate their glycan expression — have not been investigated in any significant detail in terms of their impact on the cellular glycome.

Current models of Golgi trafficking offer many intriguing candidates for proteins that may influence cellular glycan expression. Many excellent reviews have presented the evolution of evidence supporting and contradicting the three major models for Golgi trafficking — cisternal maturation, vesicular transport and rapid partitioning — and have also discussed several alternative models^54–61. Each of these models makes specific predictions regarding the mechanisms by which glycan biosynthetic enzymes are brought into contact with their substrates (BOX 3). From a glycomic viewpoint, the most important consideration is whether these models can account for the diversity of cellular glycans and whether they accommodate the known biochemical and localization characteristics of the glycan processing machinery.

Box 3. Models of Golgi trafficking in animal cells.

The original cisternal maturation model proposed that glycoproteins exit the endoplasmic reticulum (ER) in vesicles or vesiculotubular structures that coalesce into cis-cisternae, which subsequently receive enzymes that function in the early stages of glycan processing. Cisternal maturation would then be achieved by the sequential delivery of enzymes responsible for successively terminal glycan modifications. Ongoing delivery of new ER cargo vesicles would drive forward the translocation of maturing cisternae, which would eventually bud transport vesicles at the trans-Golgi¹⁴⁶ (part a of the figure).

The vesicular transport model supposes the existence of stable Golgi cisternae that possess characteristic sets of glycan processing enzymes⁵⁸. Glycoprotein cargos are moved by targeted anterograde vesicular transport from early to late cisternae (part b of the figure). However, the vesicular transport model is inconsistent with the enrichment of glycosyltransferases in transport vesicles and with the trafficking of macromolecular complexes that are too large to fit within vesicles^54,60. Subsequent reassessments of the cisternal maturation model introduced retrograde transport of glycan processing enzymes as a mechanism to retrieve processing capacity into appropriate cisternae Both the cisternal maturation and vesicular transport models share the assumption that transported proteins progress through the Golgi as a wavefront. However, photobleaching measurements of glycoprotein progression through the Golgi in live cells have recently challenged this conveyer belt assumption by demonstrating that product disappears uniformly from all cisternae and that new protein arriving at the Golgi quickly accesses all cisternae⁶³ (part c of the figure). These results have led to the formulation of the rapid partitioning model, which proposes a major role for Golgi lipid composition in establishing and regulating the polarity of the organelle and the distribution of its resident proteins^147,148. The rapid partitioning model proposes the coexistence of two distinct lipid domains in each stack of the Golgi, one enriched in sphingolipid and the other enriched in phosphoglycerolipid. The phosphoglycerolipid domains carry out transport functions. The sphingolipid domains are home to Golgi resident proteins, such as glycan processing enzymes that may partition into these domains based on their sphingolipid composition^62,64. Glycoprotein processing and transport occur as substrate proteins partition between these two domains. Although the central tenants of the major models are mutually exclusive, many of the proposed mechanisms are not, and the most useful conceptualization of Golgi trafficking allows for the simultaneous existence of aspects of each model (part d of the figure).

Compartmentalization of Golgi glycosylation enzymes

Several mechanisms have been proposed for the targeting and capturing of specific glycosylation enzymes within Golgi sub-compartments that ensures orderly glycan processing. The biophysical characteristics of the transmembrane domains of some processing enzymes may stabilize their distribution within membrane domains that possess specific ratios of sphingolipids, glycosphingolipids and phosphoglycerolipids^62–64. It is not currently clear whether sufficient resolution exists in the matching of transmembrane domains with lipid micro-environments to account for specific cisternal sorting or whether this matching serves primarily as a means of excluding enzymes from transport to the cell surface. Essentially all Golgi glycosyltransferases are type 2 transmembrane proteins with very small NH₂-terminal cytoplasmic domains. In yeast, a consensus sequence in the NH₂-terminal cytoplasmic tail that confers binding of vacuolar protein sorting 74 (Vps74) has been identified in a subset of glycosyltransferases; in turn, this sequence interacts with coatomer protein complex I (COPI), suggesting a mechanism for segregating transferases into transport vesicles^65,66. Golgi phosphoprotein 3 (GOLPH3) has been identified as the animal homologue of Vps74, but the yeast consensus sequence does not appear in the cytoplasmic tails of any human glycosyltransferase⁶⁶, and little evidence supports a dominant role for the cytoplasmic tail in localization of animal glycosyltransferases. However, biochemical, mutagenesis and domain-swapping studies have demonstrated that amino acid sequences within the cytoplasmic, transmembrane and stalk domains of individual transferases, as well as disulphide bond formation between the lumenal domains of glycosyltransferase monomers, can contribute to targeting^67–72.

In the absence of a single clear retention or sorting mechanism for glycosyltransferases, it is reasonable to assume that these enzymes probably use a combination of multiple, weak sorting signals for localization^73,74. The kin recognition and oligomerization model, which was originally based on co-immunoprecipitation and mis-localization studies, proposed that glycosyltransferases form homo-oligomers with each other and hetero-oligomers with functionally related enzymes⁷⁵. As larger aggregates are formed, they reach a size at which they are excluded from transport vesicles, thus stabilizing their localization within a specific cisterna. In the context of the vesicular transport model, the aggregation hypothesis was useful. However, in the cisternal maturation and rapid diffusion models, localization by aggregation is problematic, as it does not allow for dynamic changes in cisternal enzyme content. Nonetheless, elegant recent studies using bi-molecular fluorescence complementation have demonstrated that specific transferases form homodimers and heterodimers in living cells⁷⁶. The organization of transferases into biosynthetic complexes has also been described for N-linked glycan precursor biosynthesis, yeast mannan extension and glycosaminoglycan and glycosphingolipid processing, but not for mucin-type O-linked glycosylation^74,77–79. Oligomeric organization is thought to confer increased efficiency, as the product of one enzyme is immediately accessible as substrate for the next. Further work will be required to determine the extent to which glycosyltransferase localization and oligomerization are useful mechanisms for regulating cellular glycomic composition.

Golgi trafficking sculpts glycan profiles

Extensive testing of the three major models of Golgi organizational dynamics has driven substantial progress towards identifying molecular components of the Golgi trafficking machinery. Key proteins involved in intra-Golgi trafficking include vesicle COPI, tethering factors of the Golgi matrix (golgins and multiprotein complexes), small GTPases (the RAB family, ADP-ribosylation factor (ARF) and ARF-like GTPases, and associated GTPase activating proteins (GAPs) and guanine exchange factors (GEFs)) and the membrane fusion machinery (NSF (N-ethylmaleimide-sensitive factor), SNAP (soluble NSF attachment protein) and vesicle- and target-SNAREs (SNAP receptors)) (FIG. 2a–c). Among these components, the conserved oligomeric Golgi complex (COG complex) has been directly implicated in the regulation of cellular glycan profiles (FIG. 2d). The COG complex is composed of eight highly conserved protein subunits. To date, mutations in three human COG subunits (COG1, COG7 and COG8) have been identified, and each causes altered protein glycosylation and severe multi-systemic developmental defects, forming a subgroup of a growing class of inherited diseases known as congenital disorders of glycosylation (CDGs)⁸⁰. Fibroblasts from these patients with CDG have proven to be useful experimental tools, shedding considerable light on the organization and function of the COG complex^61,81–84. Knockdown of COG components in wild-type cells identified a set of proteins comprised of Golgi glycosylation enzymes and other related proteins (Golgi mannosidase II, four glycosyltransferases, the Golgi SNARE protein syntaxin 5 and several golgins). These proteins are collectively termed GEARs, and their stability and localization requires an intact COG complex^10,61,85. Experimentally validated interactions between these GEARs and COG subunits support the hypothesis that Golgi vesicle tethering may occur in two stages: initial capture by golgins followed by COG-mediated stabilization that enhances vesicle association with the fusion machinery⁶¹. Thus, the COG complex may drive retrograde transport steps to completion and may also provide an opportunity for a final quality check before the commitment to fusion. Reduced function of the COG complex would be predicted to decrease retrograde transport, leading to mislocalization and degradation of the GEARs that mediate glycan processing.

Figure 2. Control points for Golgi trafficking and cellular glycomic diversity.

Intra-Golgi transport mechanisms provide regulatory points for glycome modulation. Two idealized Golgi compartments are depicted. Guanine nucleotide exchange cycles acting on small GTPases regulate trafficking steps important for glycosylation. a | ADP ribosylation factor 1 (ARF1) activation recruits coatomer protein complex I (COPI) coats to nascent transport vesicles. b,c | Golgin interactions with cisternal and vesicular GTPases influence vesicle capture^150,151. At least 15 small GTPases of the RAB family and three of the ARF and ARF-like (ARL) family are associated with specific cisternae, vesicles and tethering factors^56,152. Golgin tethering factors exist as transmembrane or peripheral membrane proteins that are retained at Golgi cisternae through binding interactions with myristoylated Golgi reassembly-stacking protein of 55 kDa (GRASP55; also known as GRS2) or GRASP65 (also known as GRS1) or small GTPases (prenylated RABs, acylated ARF1 or ARL1). d | The conserved oligomeric Golgi (COG) complex, a multiprotein-tethering complex, interacts with a subset of proteins called GEARs (including golgins), facilitating the fusion of appropriately tethered transport vesicles and modulating the stability of processing enzymes. e,f,g | Sphingolipid and cholesterol content, as well as glycosphingolipid complexity (that is, glycan size and charge) increase from early to late Golgi^148,153. Glycosyltransferases catalysing sequential reaction sets may congregate into selective Golgi processing domains on the basis of preferences for specific lipid content imposed by the biophysical characteristics of their transmembrane domains. Therefore, altered lipid biosynthesis may affect glycosylation by contracting or expanding facilitative lipid domains within Golgi cisternae. t SNARE, target soluble N-ethylmaleimide-sensitive factor attachment protein receptor; v-SNARE, vesicle SNARE.

The major models for Golgi trafficking universally require a concerted process for retention and partitioning of glycosylation enzymes across the Golgi stacks in the face of considerable anterograde flow. The glycan structures delivered to the cell surface should reflect the length of time necessary for maturation within a given cisterna, which, in turn, should be coupled to antero-grade and retrograde transport efficiency. However, there is currently no reason to suspect that these transport events occur at constant rates across the organelle. Differential control of intra-organellar transport may provide cells with regulatory points for expanding, contracting or otherwise shifting their glycome. These regulatory points may include COPI coat dynamics, COG complex availability, RAB protein cycling, golgin expression and lipid biogenesis (FIG. 2a–g). Furthermore, a handful of paradigms have been described in which cellular responses to external stimuli affect Golgi form and function, presumably by altering trafficking. For example, administration of epithelial growth factor (EGF) and platelet-derived growth factor (PDGF) to HeLa cells activates a Golgi-localized SRC kinase that induces a redistribution of the enzymes responsible for initiating O-linked glycosylation from the cis-Golgi to the ER, where they drive increased mucin glycosylation⁸⁶. Other Tyr and Ser/Thr kinases have recently been implicated as regulators of Golgi morphology and, in some cases, protein glycosylation^87–90. These emerging paradigms, in which growth factors, cytokines, chemokines and other signalling molecules stimulate changes in Golgi dynamics, promise to reveal stronger mechanistic links between cell signalling and glycomic diversity.

Protein domain-specific glycosylation

Ubiquitous forms of N- and mucin-type peptide-O-GalNAc modifications have been extensively studied for decades. In addition, the selective modification of N-glycans on restricted families of proteins has been recognized for many years (for example, selective polysialylation of N-glycans on neural cell adhesion molecule 1 (NCAM1)⁹¹ and a handful of other proteins, or GlcNAc-1-PO₄ addition to terminal Man residues on lysosomal enzymes⁵). Unconventional glycan additions have also been described for a small number of proteins^21–23. An emerging field of glycobiology has recently focused on the selective glycosylation of distinct protein domains with specific classes of glycan structures (TABLE 1; BOX 1). These novel forms of glycosylation have been shown to have crucial roles in modulating biological recognition events in development and physiology.

The first of these domain-specific modifications was identified on the Cys-rich EGF domains of the cell surface transmembrane signalling proteins of the Notch family, where the glycans were found to modulate Notch–ligand interactions in developmental cell fate decisions⁹². Subsequent studies identified other protein domain-specific modifications, including modifications of thrombospondin type 1 repeats (TSRs) and other selective protein modifications such as those found on the mucin-type domains of α-dystroglycan^93–97 (TABLE 1; BOX 1). In each case, initial peptide modification occurs in the ER, but further extension of the glycan structure continues in the Golgi complex.

EGF domain-specific glycosylation

EGF repeats are Cys-rich domains of ~30–40 amino acids that are found in many cell surface and secreted proteins and are frequently involved in protein–protein interactions⁹⁸. The Notch family of single-pass, transmembrane receptors has served as a paradigm for studying glycosylation of EGF repeats, but modification of other EGF domain-containing proteins has not been well studied. In contrast with classical O-GalNAc glycosylation, Ser and Thr residues between the second and third Cys residues within EGF domains are modified by O-Fuc addition by the enzyme protein O-fucosyltransferase 1 (POFUT1; also known as OFUT in Drosophila melanogaster)^99,100 (TABLE 1). The Fuc residue can be extended by additional glycosyltransferases to form the tetrasaccharide SA-α2,6-Gal-β1,4-GlcNAc-β1,3-Fuc-α1-O-Ser. Pofut1 is essential in mice¹⁰¹, as is Ofut1 in D. melanogaster¹⁰², and conditional knockouts of Pofut1 in adult mice revealed roles of O-Fuc modification in haematopoietic homeostasis¹⁰³. The Fringe family of β3-GlcNAc transferases initiate elongation of glycans on O-Fuc residues and are modulators of Notch activity⁹².

EGF repeats of Notch have also been found to contain modifications with O-Glc residues added to Ser side chains before the second Cys residue of the EGF repeat by the action of protein O-glucosyltransferase 1 (POGLUT1; also known as O-glucosyltransferase Rumi in D. melanogaster¹⁰⁴) (TABLE 1). This glycan can also be extended by the addition of two Xyl residues^105,106. Most of the O-Glc consensus sites in mouse NOTCH1 appear to be occupied, but variability exists among cell types and depending on motif locations within the protein¹⁰⁷. Poglut1-knockout mice do not survive past day 9.5, whereas heterozygous Poglut1^+/− mice develop normally, suggesting an essential role for O-glucosylation in embryonic development¹⁰⁸.

A third type of novel glycosylation of EGF repeats was recently described in D. melanogaster, where a single O-GlcNAc residue was found to be added to a Ser/Thr residue following the fifth Cys residue in an EGF domain of the extracellular matrix protein Dumpy¹⁰⁹ and some EGF domains of NOTCH1 (REF. 110) (TABLE 1). Extension of the O-GlcNAc residue has not been detected. Previous studies had identified O-GlcNAc linkages exclusively on a large collection of cytosolic and nuclear proteins⁶, but the detection of EGF domain-specific O-GlcNAc modification was the first example of such a linkage on an extra-cellular protein¹⁰⁹. O-GlcNAc addition to extracellular proteins is catalysed through the action of a recently identified protein called EGF domain-specific O-linked GlcNAc transferase (EOGT)¹⁰⁹. D. melanogaster Eogt mutants exhibit aberrant cell–matrix interactions and are embryonic lethal in a Notch-independent manner¹⁰⁹. Mammalian EOGT orthologues have also been identified. In summary, EGF domains can act as acceptors of at least three unique types of glycosylation that all have crucial roles in metazoan development.

TSR domain-specific glycosylation

Similar to EGF repeats, TSR domains are found in numerous cell surface and secreted proteins as tandem repeats of 50–60 amino acids with disulphide bonds formed by six conserved Cys residues. Two forms of domain-specific glycosylation are found in TSRs that are distinct from those found in EGF domains (TABLE 1). Similar to EGF domains, O-linked Fuc can be added to Ser/Thr residues of TSRs by the action of POFUT2 (REFS 111,112), which is a distantly related homologue of the EGF repeat-modifying POFUT1 protein¹¹³, and the O-Fuc is extended by the addition of a Glc residue through the enzymatic action of β-3-glycosyltransferase-like (B3GALTL)^114,115. RNA interference knockdown of Pofut2 transcripts or mutagenesis of the acceptor sequence in a model TSR-containing protein resulted in reduced secretion, suggesting a functional role for this modification¹¹⁶. Recently, mouse knockouts in the Pofut2 locus revealed several developmental defects in epithelial to mesenchymal transition (EMT), as well as defects in cell localization and patterning¹¹⁷.

The second type of unique TSR glycosylation is the addition of Man to a Trp residue through a C-C bond (C-mannosylation)¹¹¹. The C-Man residue is added to the TSR co-translationally, using dolichol-phosphate-Man as the sugar donor, before protein folding^118,119. Mutagenesis to replace Trp residues within a model TSR-containing protein suggests a role for C-mannosylation in the control of protein secretion¹²⁰, but the mannosyltransferase that creates this structure has yet to be identified.

Protein-selective glycosylation of dystroglycan

Finally, a distinct class of protein domain-specific glycosylation that contributes to cell–matrix interactions has been found predominantly on the cell surface protein α-dystroglycan (αDG), although other substrates may exist. Dystroglycan consists of two subunits, α and β, which are part of the membrane-spanning dystrophin glycoprotein complex that links the extracellular matrix with the cellular cytoskeleton. The α subunit is comprised of globular N- and C-terminal domains containing N-glycosylation sites and an extended central domain rich in classical O-GalNAc mucin-like structures, as well as glycans initiated with an O-Man linkage¹²¹ that are essential for binding of dystroglycan with its ligands (laminin, agrin, perlecan and neurexin) or to viruses¹²¹. Defects in O-Man addition and extension on αDG lead to several forms of congenital muscular dystrophy collectively known as dystroglycanopathies⁹⁵.

The initial synthesis of the O-Man linkage is catalysed by protein O-mannosyltransferase 1 (POMT1) and POMT2, and the structure is extended by protein O-linked-Man β-1,2-N-acetylglucosaminyltransferase 1 (POMGNT1), which adds a β1,2-GlcNAc onto the O-Man core (TABLE 1). The most predominant O-Man structures of rabbit skeletal muscle and mouse brain αDG are linear tetrasaccharides containing terminal Gal and SA residues^122,123, and defects in Pomt1, Pomt2 or Pomgnt1 cause loss of laminin binding and subsequent dystrogly-canopathies. However, the linear tetrasaccharide structure is not necessary for αDG ligand binding¹²⁴.

The products of several additional genes, including LARGE1, LARGE2 (also known as GYLTL1B), fukutin (FKTN) and fukutin-related protein (FKRP), contribute to O-Man glycan modifications on αDG. Defects in these genes result in underglycosylation of αDG, disrupted ligand interactions, reduced stability of muscle fibres, disrupted basement membrane, altered neuronal migration and characteristic dystroglycanopathies^121,125. However, the exact functions of these gene products remain unclear. Several lines of evidence indicate that functional O-Man structures are comprised of large, branched glycans that have resisted complete structural characterization (BOX 1). Mass spectrometry and NMR have detected a phosphodiester linkage to an unknown substituent at the 6-hydroxyl position of the core O-Man residue on αDG¹²⁶, but the identity of the responsible phosphotransferase or kinase is currently unknown. In addition, LARGE1 has recently been shown to contain two glycosyltransferase domains that catalyse the extension of a novel (-3-Xyl-α1,3-GlcA-β1-) homopolymer in vitro¹²⁷. Although the nature of the bridging moiety that might link the proteoglycan-like homopolymer and the O-Man core is still unclear, the size of this product of the glycosyltransferase activity of LARGE1 is consistent with the high molecular weight of the laminin binding forms of αDG isolated from natural sources. Recent work with αDG-null cells and tissues also revealed that there is an approximately equal abundance of O-Man oligosaccharides in wild-type and αDG-null animals^122, and overexpression of LARGE1 in these cells was shown to restore laminin binding¹²⁸. Therefore, other proteins besides αDG also carry O-Man structures, providing additional targets for understanding the role of LARGE1 in laminin binding.

Protein domain-specific glycosylation is an emerging field of study with great potential for elucidating important biological recognition functions. As many glycosyltransferase-related coding regions in animal systems remain uncharacterized, we anticipate that additional classes of protein glycosylation will be characterized in the future that have key regulatory roles in biological systems.

Glycan profiling and global regulation

The diversity of glycan structures in animal systems presents a huge analytical challenge for determining detailed glycan structural profiles in complex biological systems¹⁶. As glycan structures are commonly altered in response to biological signalling pathways, there is a major effort to develop broad analytical strategies that identify glycan signatures for cellular transitions¹²⁹ or biomarkers for disease states^130,131. In addition, detailed snapshots of glycan structural diversity can provide insights into potential biological recognition events and regulatory mechanisms for glycan biosynthesis¹². Limitations in structural analysis depend on the glycan class, the complexities of the sample workup and the analytical tools that are available for structural inquiry¹³².

Numerous approaches are now available for highly sensitive, broad N-glycan analysis¹⁶. These approaches include mass spectrometry analysis directly after enzymatic release of the glycan structures; methods that employ front-end chromatographic separations before mass spectrometry analysis; methods that employ chromatographic separation and N-glycan identification on the basis of known elution positions of standards; and the analysis of glycan structures on the basis of lectin or antibody binding. Similar approaches are also available for O-glycan and glycolipid analyses, but chemical release rather than enzymatic release methods are generally used before analysis¹³³. Recent advances have also provided direct analysis of intact glycolipid structures in complex samples. However, few methods have been developed for low sample consumption and for high-throughput analysis that can tease apart complex mixtures of structural isomers. Moreover, although numerous data sets are now available for glycan structures from various mammalian cell and tissue sources, challenges still exist in linking glycan structures to the corresponding peptide backbones.

A complementary strategy for glycan structural analysis is the use of lectin- and antibody-based micro-arrays¹³⁴. In this analytical format, target proteins are captured with protein-specific antibodies and then their glycan structures are analysed using lectins or antibodies. This strategy has a particular benefit, as it enables high-throughput diagnostic profiling of glycans on distinct protein backbones. Multiple platforms for capture of target glycoproteins and glycan detection are available^135,136. However, the benefits of this high-throughput approach are tempered by the limited ability of lectins and antibodies to distinguish fine structural details on glycans.

The availability of high-throughput platforms for glycan profiling has provided unique insights into regulated glycan expression in vertebrate organisms and in human disease. Numerous efforts are underway to identify glycans or glycoproteins as cancer biomarkers^131,137. Profiling of a broad range of glycan structures in a high-throughput platform is also being performed in combination with human genome-wide association studies to examine the regulation of glycan expression in human populations^138,139. Correlations between glycan profiles and gene expression data are also emerging as an approach to assess mechanisms of glycan regulation in animal systems¹¹; in these studies, complex sets of glycan and transcript data are being mined to examine the relative contribution of transcriptional and post-transcriptional control of glycosyltransferases¹². Thus, the dynamic changes in glycomic diversity are slowly yielding to structural analysis, providing opportunities to delineate the fine details of glycan structural changes that occur in complex biological systems.

Conclusions

Contemporary studies of protein glycosylation are revealing numerous examples where these post-translational modifications have essential roles in biological recognition events. Glycans have crucial functions throughout the cell, from the cytosol and secretory compartments to the cell surface and the extracellular space. Conserved contributions of N-glycan structures in chaperone interactions and protein quality control are initiated co-translationally, and glycan processing continues throughout the dynamic collection of secretory pathway compartments that lead to the cell surface. The complex assortment of glycosylation enzymes comprises an intricate assembly line for glycan maturation from the ER through the Golgi. Notably, the localization, dynamics, interactions and regulation of glycosylation enzymes within these compartments, as well as how glycosylation enzymes compete for the same substrates, remain active areas of study. Of the many roles that glycans have at the cell surface, emerging paradigms have highlighted the importance of protein domain-specific glycosylation in facilitating or modulating biological recognition events.

The diversity of glycan structures clearly provides an additional level of information content in biological systems, but the challenge for the future lies in identifying the critical contexts in which glycan functions contribute to biological regulation within the bewildering array of heterogeneous glycan structures. An essential requirement for future glycosylation studies is the ability to obtain detailed profiles of glycan diversity in various biological contexts. Recent advances in analytical strategies are beginning to provide the necessary breadth, depth and sensitivity of analysis to define the full spectrum of glycan complexity. Continued high-throughput adaptations of these approaches will generate structural data sets that will be essential for revealing the mechanisms that regulate glycan biosynthetic pathways, defining unique glycan signatures for disease states and providing correlations with biological functions, thereby increasing our ability to decode the many functions of glycoprotein glycans in complex biological systems.

Glossary

Lectins

Carbohydrate binding proteins that are involved in various biological recognition phenomena

Sialic acid

(SA: also known as neuraminic acid (Neu)). SAs are a family of nine-carbon monosaccharides with a carboxylic acid at the C1 position and a glycerol side chain at C7-C9. Two SAs predominate in vertebrates, with either an N-acetyl group (NeuAc) or an N-glycolyl group (NeuGc) at C5. Humans do not make NeuGc, but can obtain it from their diet and incorporate it into glycoconjugates

Lipid-linked oligosaccharide

(LLO). Extended carbohydrate structure attached through a phosphodiester linkage to a polyisoprenoid lipid, usually dolichol. N-linked glycan precursor structures are commonly assembled as a lipid-linked intermediate before transfer to a polypeptide side chain

Oligosaccharyltransferase

(OST). A multi-enzyme complex (or single subunit in bacteria) in the endoplasmic reticulum membrane that transfers glycan structures from a lipid-linked oligosaccharide precursor to acceptor sequences on nascent polypeptides

Acceptor peptide sequons

Short amino acid sequences on glycan-acceptor polypeptide chains that are recognized by glycosyltransferases prior to glycosylation

Calnexin

Integral membrane endoplasmic reticulum lectin that recognizes GlcMan₉GlcNAc₂ glycans on early glycoprotein folding intermediates and, in collaboration with an associated thiol oxidoreductase, ERp57, aids in protein folding as a part of endoplasmic reticulum quality control

Calreticulin

A soluble lectin in the lumen of the endoplasmic reticulum that contains a K-D-E-L endoplasmic reticulum retrieval sequence and, in a similar way to calnexin, binds to GlcMan₉GlcNAc₂ glycoprotein folding intermediates to facilitate protein folding and quality control

ER-associated degradation

(ERAD). A protein quality control pathway in which misfolded lumenal or integral membrane proteins are recognized (often through trimmed glycan structures) for disposal by translocation into the cytosol for proteasomal degradation

Nocodozole

A pharmacological agent that blocks mitosis by binding to tubulin and inhibiting microtubule polymerization. Inhibition of microtubule polymerization also causes the dispersal of well-organized Golgi stacks into smaller units distributed throughout the cytoplasm

Brefeldin A

A fungally derived antibiotic that inhibits the guanine nucleotide exchange factor (GEF) responsible for activating ARF1 GTPase. Activation of ARF1 recruits coatomer protein complex I (COPI) to Golgi membranes to form vesicles. In the absence of Golgi vesicle formation, cis and medial cisternae fuse with the endoplasmic reticulum and the trans and trans-Golgi network components disperse into a drug-induced entity called the brefeldin compartment

Cisternal maturation

An early model of Golgi trafficking that proposed that Golgi cisternae formed from the fusion of vesicles leaving the endoplasmic reticulum and were subsequently matured by the import of appropriate processing enzymes as the cisternae were pushed forward towards the trans face. The original cisternal maturation model had its groundings in extensive electron microscopic observations of Golgi morphology

Vesicular transport

A model for Golgi trafficking, which proposed that cisternae are stable structures and that cargo proteins move between cisternae in transport vesicles that are targeted to specific Golgi domains

Rapid partitioning

A model for Golgi trafficking, which proposes that cargo proteins can exit the Golgi in vesicles arising from all cisternae and that new protein arriving at the Golgi rapidly gains access to the entire apparatus. Distinct transport and processing domains within each Golgi cistern are defined by their lipid content, which varies systematically from cis-to-trans and may provide favourable environments for cisternae-specific subsets of glycan processing enzymes

Glycosphingolipids

Glycoconjugates comprised of a ceramide lipid (Cer) carrying a glycan headgroup. The glycan is usually initiated by Glc, although GalCer is an important component of some cells. GlcCer is elongated to generate four major types of neutral cores in mammalian tissues (ganglio-, globo-, lacto- and neolacto-series), each of which can be capped and branched to produce additional structural diversity

Type 2 transmembrane proteins

Single pass transmembrane proteins with their amino terminus oriented towards the cytosol and their carboxyl terminus facing the lumen of the secretory pathway or cell exterior

Vacuolar protein sorting 74

(Vps74). A yeast protein identified as having a vacuolar protein sorting function. Yeast vps74 mutants are deficient in intra-Golgi transport and exhibit mis-localized glycosyltransferases. Golgi phosphoprotein 3 (GOLPH3) is the mammalian homologue of Vps74

Coatomer protein complex I

(COPI). A protein complex that is recruited to membranes by ARF (ADP-ribosylation factor) GTPases and that mediates intra-Golgi and Golgi-to- endoplasmic reticulum retrograde transport

Mucin-type O-linked glycosylation

A form of protein glycosylation initiated by the addition of a GalNAc residue to protein Ser or Thr side chains and extended and branched with other complex termini. It is commonly found on highly glycosylated mucin glycoproteins, but also on many non-mucin polypeptides

Golgins

A family of Golgi tethering factors characterized by their extended coiled-coil domains. Individual family members exhibit specific sub-Golgi localizations, GTPase interactions and effector activities. The golgins form long filaments that emanate from the cytoplasmic face of Golgi cisternae, providing access to transport vesicles in the near environment. Some golgins possess multiple binding sites for RAB proteins, suggesting that they capture transport vesicles carrying specific RAB proteins

Conserved oligomeric Golgi complex

(COG complex). A multiprotein tethering factor comprised of eight protein subunits. The complex facilitates Golgi retrograde transport and glyco-syltransferase localization

Congenital disorders of glycosylation

(CDGs). A growing group of recessive human disorders characterized by altered protein glycosylation. In type 1 CDGs, the number of N-linked glycans added to a protein is affected. In type 2 CDGs, the nature of the glycan on a glycoprotein is changed. Most CDGs arise from partial loss-of-function mutations and are diagnosed clinically by altered serum protein glycosylation

Retrograde transport

The movement of vesicles containing Golgi-resident proteins (glycosyltransferases and other processing enzymes) from late to early cisternae. Retrograde transport provides a mechanism for retrieving enzymes that are displaced forward by anterograde flow or cisternal maturation

Anterograde flow

The bulk movement of cisternal contents and cargo proteins from early to late Golgi compartments

EGF domains

Protein motifs comprised of ~30–40 amino acids, including six Cys residues forming three characteristic disulphide bonds, and a mainly β-sheet structure, found in all ERBB-binding growth factors and many other cell surface and extracellular matrix proteins

Thrombospondin type 1 repeats

(TSRs). Protein domains of ~50 amino acids. They are rich in Cys residues and are comprised of three antiparallel β-strands along with regions without a secondary structure. TSRs are found in matrix and transmembrane proteins with functions in matrix organization, cell–cell interactions and cell guidance

Dystroglycanopathies

A clinically heterogeneous collection of muscular dystrophies that are associated with aberrant glycosylation of α-dystroglycan, which is a component of the multiprotein dystrophin glycoprotein complex that bridges the extracellular matrix, plasma membrane and cytoskeleton