N-glycan-dependent protein maturation and quality control in the ER

Abstract

The vast majority of proteins that traverse the mammalian secretory pathway become N-glycosylated in the endoplasmic reticulum (ER). The bulky glycan protein modifications, which are conserved in fungi and humans, act as maturation and quality-control tags. In this Review, we discuss findings published in the past decade that have rapidly expanded our understanding of the transfer and processing of N-glycans, as well as their role in protein maturation, quality control and trafficking in the ER, facilitated by structural insights into the addition of N-glycans by the oligosaccharyltransferases A and B (OST-A and OST-B). These findings suggest that N-glycans serve as reporters of the folding status of secretory proteins as they traverse the ER, enabling the lectin chaperones to guide their maturation. We also explore how the emergence of co-translational glycosylation and the expansion of the glycoproteostasis network in metazoans has expanded the role of N-glycans in early protein-maturation events and quality control.

We are very sad to report that the original corresponding author on this Review, Daniel N. Hebert, passed away on 8 December 2024 as the article was close to submission. We would like to dedicate this Review to the 20 plus years of Daniel’s work within the glycobiology field. We are also grateful for editing assistance from our colleague Lila Gierasch as the article reached completion.

Introduction

Most proteins that pass through the mammalian secretory pathway are modified by N-linked glycosylation — the attachment of a sugar moiety to the nitrogen atom of an asparagine (Asn) residue¹. N-glycoproteins include proteins that reside in the membranes or lumens of the organelles that make up the secretory and endocytic pathways of the cell, as well as proteins that are integrated into the plasma membrane or secreted into the extracellular space (Fig. 1). N-glycoproteins serve a wide range of functions in protein folding, trafficking, signal transduction, cell adhesion, cell–cell communication and immunity². N-glycans consist of strings of monosaccharides that can influence the structure, stability, localization and function of proteins^3–6. It has only become clear in recent years how important N-glycans are in the proper folding, maturation, quality control and trafficking of secretory proteins, a process referred to as glycoproteostasis^6–10.

Fig. 1 |. The secretory pathway in eukaryotic cells.

Proteins containing an N-terminal signal sequence are directed to the endoplasmic reticulum (ER), where proteins are translocated through the ribosome–Sec61 complex, folded, and appended with N-linked glycans. Once folded, proteins are trafficked to the ER–Golgi intermediate compartment (ERGIC) and the Golgi apparatus, where the glycans are extensively modified by a series of transferases and glycosidases. After traversing the Golgi, proteins are trafficked to their final destination, such as the extracellular space, plasma membrane, endosomal or lysosomal compartments. Red shapes indicate glycans with multiple mannose residues, mostly found on proteins in the ER and the ERGIC. Blue shapes represent glycans that are modified in the Golgi, becoming hybrid or complex glycoforms.

N-glycans are assembled and added to proteins in the endoplasmic reticulum (ER), the entry portal to the secretory pathway of the cell^11,12. The core oligosaccharide is sequentially built at the ER membrane through a process that is highly conserved from fungi to humans and involves over thirty proteins¹³. This process commences on the cytoplasmic face of the ER membrane, where a lipid-linked Man₅GlcNAc₂ is produced, which is then flipped to the inner ER leaflet, where another four mannoses and three glucoses are transferred to create the completed lipid-linked precursor of Glc₃Man₉GlcNAc₂ (where Glc is glucose, Man is mannose and GlcNAc is N-acetylglucosamine). Large heteromeric protein complexes called oligosaccharyltransferases (OSTs) then transfer the oligosaccharides all together to nascent chain Asn residues found in the consensus sequence (sequon) Asn–X–Ser/Thr/Cys (in which X is any amino acid except for proline, and the last amino acid is serine, threonine or cysteine) to create the modified glycoprotein^12,14.

In the ER, the composition of glycans is dynamic as they are modified depending on the folded state of the nascent chain to which they are attached. Glycosidases and glycosyltransferases match the folded state of the glycoprotein to the glycan composition. Critical to the glycoproteostasis network in the ER are the lectins calnexin and calreticulin, which serve as carbohydrate-binding molecular chaperones. Lectins recognize specific glycan compositions and oversee the maturation and the proper trafficking of nascent chains^15,16. In this manner, properly folded and assembled proteins are efficiently targeted to their downstream locations, whereas irrevocably misfolded proteins are directed to degradation.

N-glycosylation has diverse connections to health and disease. Glycoproteins decorate the surface of cells, so their accessibility makes them frequent therapeutic targets for a range of diseases. Glycoprotein misfolding accounts for dozens of loss-of-function disease states in which mutations in the glycoprotein cause misfolding and lead to the production of proteins that are marked for ER retention and subsequent degradation^17,18. Protein misfolding can also create toxic aggregates or gain-of-toxicity diseases^19,20. Mutations may also occur in the components of the N-glycosylation proteostasis network, leading to a growing number of over 70 different congenital diseases of glycosylation²¹.

In this Review, we discuss the N-glycosylation processes in the ER and how the differential mechanisms of N-glycan transfer dictate the role of glycans in assisting folding and quality control in the ER. Importantly, a better understanding of these mechanisms will shed light on the role of any given N-glycan in the maturation and quality control of a specific protein. Moreover, understanding N-glycan-mediated processes promises to elucidate how organisms have evolved to handle the challenges encountered by an increasingly large and complex secreted proteome (the secretome). Finally, we also compare glycoproteostasis networks across species to explore how their components have diversified along with the complexities of their N-glycoproteomes.

Mechanisms of N-glycosylation

The first step of the N-glycosylation process is carried out by OSTs that reside on the luminal face of ER membranes of eukaryotic cells. OSTs transfer preassembled glycans (Glc₃Man₉GlcNAc₂) from membrane-embedded lipids to nascent polypeptide chains. The two major OSTs in mammalian cells are specific to either co- or post-translational modification of secretory proteins entering the ER²² (Fig. 2). Insights from recent studies have shown that there are fundamental connections between OST-mediated N-glycosylation and the mechanisms of initial folding in the ER.

Fig. 2 |. N-linked glycosylation in human cells.

N-linked glycans are added to a specific sequence motif (also known as a sequon): Asn–X–Thr/Ser/Cys, where X is any amino acid except proline, and the third amino acid of the sequon can be threonine, serine or cysteine. Glycans can either be added post-translationally by the oligosaccharyltransferase B (OST-B) complex or co-translationally by the OST-A complex. a, For post-translational glycosylation, acceptor sites must remain accessible, and cannot be buried by folding, for the OST-B complex to append the carbohydrate to the protein. GlcNAc is N-acetylglucosamine. b, For co-translational glycosylation, the OST-A complex binds directly to the Sec61 translocon to glycosylate the protein as it enters the endoplasmic reticulum (ER).

Complexity of oligosaccharyltransferases in N-glycosylation

OSTs are heteromultimeric membrane protein complexes²³. Each OST complex consists of a catalytic subunit called STT3, which is part of an enzymatic subcomplex containing six additional subunits (DAD1, OST4, OST48, ribophorin I (RPN1), RPN2 and TMEM285 in humans). These components are found in all OSTs and are conserved across single-cell and multicellular eukaryotes^14,22,23.

In yeast and Caenorhabditis elegans, the enzyme-specific subcomplexes of OSTs contain a singular form of Stt3 along with an additional subunit of either Ost3 (dolichyl-diphosphooligosaccharide–protein glycosyltransferase subunit 3) or its paralogue Ost6²⁴. In more complex organisms such as plants, insects, fish and mammals, the OST machineries have undergone gene duplication, leading to two distinct isoforms termed OST-A and OST-B. In humans, the corresponding catalytic subunits are termed STT3A and STT3B, respectively²⁵. Stt3 in yeast and C. elegans share more similarity with STT3B in terms of sequence homology and mechanism of modification¹⁴. Both OST-B and the yeast OST modify proteins post-translationally^26,27 (Fig. 2a), limiting modification by these enzymes to consensus sites that are not hidden by early protein folding and thus remain accessible post-translationally. In multicellular eukaryotes, either the subunit MAGT1 or its paralogue TUSC3 form a subcomplex with STT3B to aid in scanning and recruiting acceptor sites^28,29. Both human MAGT1 and TUSC3 contain thioredoxin domains, and their oxidoreductase active sites promote the recognition of sites possessing proximal Cys residues by forming intermolecular disulfides with these substrates^28,30. MAGT1 and TUSC3 have also been shown to use a cytosolic reductive pathway to help reduce oxidized sequences and support glycosylation by STT3B³¹. Yeast Ost3 and Ost6 also possess luminal thioredoxin domains, and their active-site Cys residues are required for the efficient glycosylation of a subset of consensus sites³².

The emergence of STT3A and OST-A in plants, insects, fish and mammals suggests an adaptive strategy to assist the increasingly complex glycoproteome of eukaryotic organisms. The oligosaccharyltransferase complex subunits OSTC (also known as DC2) and KCP2 bind to STT3A to help localize OST-A to the Sec61 translocon³³. Cryo-electron microscopy (cryo-EM) structures of the OST-A complex demonstrate that OSTC forms a direct interaction between Sec61 and the OST-A complex, positioning STT3A to act co-translationally on substrates as they emerge into the ER lumen^34–36 (Fig. 2b). This arrangement permits efficient access to the acceptor site on the nascent amino acid chain before protein folding occurs. The active site in STT3A is located about 15 Å from the membrane surface^23,37. This distance is in agreement with earlier results showing that an acceptor site needs to be about 11 amino acids away from the membrane to be modified³⁸. Furthermore, the acceptor peptide must form a 180° loop to place the modified Asn into the STT3 active site^12,23. This conformation also positions the Thr, Ser or Cys residue in the +2 position such that it can form a hydrogen bond with the conserved Trp–Trp–Asp of STT3 (ref. 23). These findings provide evidence that the OST-mediated enzymatic reaction leads to selective pressure for glycosylation sites to be located in flexible or unstructured regions, which in turn are expected to reside on the surface of glycoproteins after folding and assembly.

Over the past several years, high-resolution, single-particle cryo-EM-derived structures of isolated OST complexes^34–36 and cryo-electron tomography (cryo-ET) images of ER membranes³⁷ combined with glycoproteomics studies of CRISPR-engineered cells³⁹ and the biochemical analysis of protein biogenesis⁴⁰ have greatly advanced our understanding of the mechanism of N-glycosylation by OSTs. Because OST-A acts co-translationally in multicellular eukaryotes, it has early access to acceptor sites and is responsible for the modifications of approximately 70% of the sites³⁹. In multicellular eukaryotes, OST-A and OST-B work together to ensure that all accessible acceptor sites are efficiently recognized. Importantly, their differential timing of modification has a notable role in the ability of the N-glycan to support early maturation and quality-control events while the glycoprotein resides in the ER. In yeast and in C. elegans, where only the OST-B type OST is present, the role of N-glycans in protein maturation is expected to be reduced.

Co- and post-translational N-glycosylation

Studies using cells in which either STT3A or STT3B were knocked out showed that most glycosylation sites are modified co-translationally by STT3A (the OST-A complex)³⁹. Consistent with this observation, the unfolded protein response pathway is activated in cells deficient in STT3A but not in STT3B^14,26. As OST-B acts post-translationally, it modifies acceptor sites skipped by the earlier-acting OST-A complex. Although OST-A is positioned to scan most sites co-translationally, acceptor sites near the N and C termini of the mature protein are frequently missed by OST-A and then N-glycosylated by OST-B.

Most secretory proteins contain N-terminal signal sequences and are translocated into the ER through SEC61 translocons⁴¹. The hydrophobic signal peptides are embedded into the membrane of the ER, which anchors the protein in the membrane and constrains the movement of the mature N terminus until the signal sequence is cleaved. This geometry renders acceptor sites within ten amino acids of the N-terminal signal sequence cleavage site unable to reach the STT3A active site and consequently inaccessible for efficient co-translational glycosylation by OST-A until the signal sequence is cleaved (Fig. 3a). Glycosylation sites in these locations were found to be modified by OST-B based on studies using STT3A and STT3B knockdown cells²⁶. An earlier study found that N-glycosylation of influenza haemagglutinin at position Asn8 took place after cleavage of the signal sequence by the signal peptidase complex and thus was probably mediated by OST-B⁴².

Fig. 3 |. Cases of post-translational glycosylation by the OST-B complex.

a, Post-translational glycosylation through the oligosaccharyltransferase B (OST-B) complex occurs if the N-linked glycosylation motif is found within the first ten amino acids (AA) of the mature N terminus, because the signal peptide is embedded in the endoplasmic reticulum (ER) membrane and is unable to reach the OST-A complex for glycosylation owing to steric hindrance, until the nascent chain is liberated by the signal peptidase complex (SPC). b, Posttranslational glycosylation also occurs if the motif is located within the last 65 amino acids from the C terminus of the nascent chain, because this type of motif cannot be reached by the OST-A complex owing to the rapid translocation of the nascent chain into the ER. c,d, The OST-B complex also glycosylates sites that are missed by the OST-A complex, such as those located close to each other within the sequence (c), or those containing a Cys within the +1 or +2 positions of the sequon (d). e, Post-translational glycosylation can also occur for multipass transmembrane proteins as the protein associated with translocon (PAT) complex, which is required for inserting the transmembrane sections of a protein into the membrane of the ER, displaces the OST-A complex from binding to the Sec61 translocon.

During translation and translocation, approximately 35 amino acids are protected inside the ribosome, 20 are within the ER membrane or the Sec61 translocon, and another ten or so amino acids are needed on the luminal side of the ER to reach the OST-A active site. As translation is terminated once a stop codon is reached, the last 65 amino acids from the C terminus are rapidly released into the lumen of the ER (Fig. 3b). This quick entrance into the ER hinders efficient STT3A recognition. Hence, acceptor sites that are within the 65 C-terminal amino acids of a glycoprotein are frequently OST-B-dependent⁴³. However, rapid folding of the C terminus can hinder OST-B post-translational modification. Consistent with these general principles, neuroserpin, which bears a potential site of N-linked glycosylation at Asn385, nine amino acids from the C terminus, can be efficiently modified post-translationally by OST-B but only when a nearby disease-associated mutation (Gly376Glu) disrupts early C-terminal folding^44,45. A similar case was observed for transthyretin, which has a potential acceptor site 29 amino acids from the C terminus that is modified by OST-B only when folding is disrupted by a mutation⁴⁶. In general, OST-B glycosylation has been shown to be inversely related to the stability of the protein⁴⁷.

Efficient co-translational glycosylation requires that OST-A properly load glycans, scans nascent polypeptides from the N- to the C-terminal direction and glycosylates the motif. Sequons containing serine instead of threonine as the acceptor hydroxyls are often missed owing to the reduced efficiency of OST-A to append N-glycans⁴⁸. In addition, closely spaced sequons often pose stereochemical and kinetic challenges that prevent OST-A from working efficiently, resulting in sites being skipped in the middle of the array. Once OST-A has glycosylated the first acceptor site of the array, the stereochemistry of the bulky glycan makes it difficult for STT3A to gain access to subsequent proximal sites⁴⁸. Knocking down STT3A or STT3B in HeLa cells revealed that closely spaced acceptor sites containing serine in the +2 position, such as Asn–X–Ser–Asn–X–Thr or Asn–X–Ser–Asn–X–Ser, were hypoglycosylated because sequons containing Ser interact more weakly with STT3A than do those containing Thr at the +2 position⁴⁹ (Fig. 3c). Long tandem repeats of Asn–X–Thr sequons were also shown to be hypoglycosylated by the OST-A complex. Therefore, closely spaced acceptor sites are often hypoglycosylated by OST-A and rely on OST-B for post-translational modification⁴⁹.

As nascent polypeptides enter the ER, the oxidizing environment supports the formation of disulfide bonds. Given the constraints in accessing the STT3A active site, N-glycosylation sites proximal to disulfide bonds are refractory towards glycosylation²⁸. Specific sequons, either containing Cys in the consensus motif (Asn–Cys–Thr/Ser) or having proximal Cys residues, are frequently skipped by OST-A²⁸ (Fig. 3d). It is likely that the potential of the Cys to form disulfide bonds makes these sites unavailable for glycosylation. A subset of these sequons rely on OST-B to modify these protected sites because the OST-B complex contains oxidoreductase activity (mediated by the subunits MAGT1 or TUSC3, see above, which are absent in OST-A)^28,30. MAGT1 or TUSC3 can reduce disulfide bonds, providing increased flexibility and accessibility to the STT3B active site²⁸. Alternatively, Cys-rich domains may be modified early on by OST-A before the disulfides form³⁹. Thus, OST-A and OST-B use different strategies to optimize the modification of problematic Cys-containing regions.

N-glycosylation of membrane proteins

Ribosomes and Sec61 form ribosome–translocon complexes that co-translationally translocate newly synthesized polypeptides into the ER⁴¹. Recent work has provided important details about the mechanisms of ER targeting, translocation and biogenesis of multipass membrane proteins (also known as polytopic membrane proteins)^40,50. Multipass proteins with long N-terminal tails are targeted to the ER by N-terminal signal sequences and translocated into the ER lumen by Sec61³⁹. Because the OST-A complex interacts directly with the Sec61 translocon, it can glycosylate consensus sites on long N-terminal tails of polytopic membrane proteins co-translationally, which is consistent with results from glycoproteomics^39,40. As translation progresses and transmembrane regions are inserted into the ER membrane, Sec61 acts as a ribosome receptor and recruits additional translocation machinery such as the protein associated with translocon (PAT) complex to support membrane insertion of the newly synthesized transmembrane domains⁴⁰. The PAT complex associates with Sec61 on the same side of the translocon as OST-A, displacing OST-A from the Sec61 complex (Fig. 3e). With OST-A absent from the Sec61–PAT complex, downstream glycosylation sites within short luminal loops of multipass transmembrane proteins rely on OST-B for their post-translational glycosylation³⁹. Interestingly, cryo-ET of ER-derived microsomes using human embryonic kidney (HEK) cells showed that 69% of membrane-bound ribosomes contained Sec61–OST-A complexes suitable for co-translational glycosylation, whereas 26% contained Sec61–PAT complexes³⁷. It will be of interest to determine how these ribosome–translocon fractions correlate with cell lines that carry different secretory protein loads.

Alternatively, proteins with short N-terminal tails can use their first transmembrane segment to initiate translocation by engaging the ER membrane protein complex⁵¹. There is thus far no evidence that the ER membrane protein complex is associated with OST-A. Therefore, any glycosylation sites positioned in short luminal N-terminal tails of multipass proteins would be expected to be modified post-translationally by OST-B, although this remains to be tested.

The emerging mechanism for multipass membrane proteins is that recruitment of the ER translocon machinery occurs while they are being translated and is based on the properties of the individual client, and that this is a dynamic process with associated factors changing as the process progresses⁵¹. Only N-glycans positioned in a long N-terminal luminal tail of a multipass membrane protein are expected to utilize OST-A for co-translational glycosylation³⁹. Any additional downstream sites of glycosylation will probably be modified by OST-B. Therefore, these sites are unlikely to play a part in early protein-maturation events.

Lectins are ER chaperones for N-glycosylated proteins

Molecular chaperones transiently interact with nascent and newly synthesized chains to assist with their maturation⁵². Classical chaperones bind directly to the polypeptide, most often in an adenine nucleotide-regulated manner⁵³. Although the ER possesses members of the heat-shock protein 70 (Hsp70) and Hsp90 chaperone families (BiP and GRP94 (also known as endoplasmin), respectively), it relies on lectin chaperones, which bind carbohydrate modifications on proteins^6,54, to mediate glycoproteostasis and facilitate proper folding and quality control of secretory proteins.

N-glycosylation processing triggers lectin binding

Instead of recognizing their clients through exposed hydrophobic patches like other chaperones, lectin chaperones bind exposed hydrophilic N-linked glycan modifications^15,55. Their binding can thus be regulated by the composition of the carbohydrate on the nascent chain, which is in turn controlled by ER glycosidases and glycosyltransferases.

Upon entry of a protein into the ER and glycosylation by OST-A, the transferred Glc₃Man₉GlcNAc₂ glycoform is rapidly trimmed by ER resident glucosidases (step 1 in Fig. 4). Glucosidase I (GlsI), a type II transmembrane protein, cleaves the outermost glucose to create diglucosylated glycoforms^13,56. High-resolution native electrophoresis coupled to mass spectrometry of mouse fibroblasts revealed that GlsI can associate with Sec61 and the OST-A complexes and is thus positioned to act on the nascent amino acid chain as it enters the ER and is glycosylated⁵⁷.

Fig. 4 |. Comparison of lectin chaperone cycles in yeast and humans.

a, Saccharomyces cerevisiae lack the co-translational glycosylation machinery because they encode only one OST paralogue, which is closely related to OST-B. Therefore, glycosylation can only occur post-translationally, and the acceptor site must remain exposed after protein folding to be recognized by the OST-B machinery (step 1). Once glycosylated, the glycan is trimmed by glucosidase I and II (GlsI and GlsII) into a monoglucosylated state, allowing the protein to bind calnexin for lectin chaperone-mediated folding assistance (step 2). Once the remaining glucose is trimmed by GlsII the protein can continue to fold (step 3) and can either be trafficked out of the endoplasmic reticulum (ER) if it reaches its native form (step 4) or be targeted for degradation (step 5). b, In humans, proteins can be glycosylated co-translationally by OST-A as they enter the ER. Once they are glycosylated, the glycan is rapidly trimmed into a monoglucosylated state (step 1), allowing the protein to bind the lectin chaperones calreticulin or calnexin for lectin chaperone assistance (step 2). Associated folding factors (AFFs) bound to calnexin or calreticulin provide further folding assistance by aiding in the formation of disulfide bonds or the isomerization of prolines. The protein is released from these chaperones upon removal of the terminal glucose by GlsII, and the client protein can continue to fold until it reaches its native form (step 3). If the client protein reaches its native state, it can be exported from the ER (step 4), or, if it needs further folding assistance, it is recognized and re-glucosylated by the UDP-glucose:glycoprotein glucosyltransferases 1 or 2 (UGGT1 or UGGT2), which allows the protein to re-enter the lectin chaperone cycle for further folding assistance (step 5). If the protein is terminally misfolded, or persistently engaging lectin chaperones, it can be targeted for degradation (step 6).

Malectin is a membrane-associated ER-resident lectin that binds proteins harbouring diglucosylated glycoforms, but its role in the ER is unclear⁵⁸. ER stress upregulates malectin, and its binding to glycans could potentially delay subsequent processing and thus influence the residence time of glycoproteins in the ER⁵⁹. Malectin was discovered to be associated with both OST complexes, and its interactions were found to be stronger with OST-B than OST-A²³. In the same study it was observed, using cryo-electron microscopy maps of the OST complexes, that malectin was close to the general OST components of the OST complexes. Malectin may utilize its binding to diglucosylated glycan-modified proteins to recruit them to the OST complexes. This role could be especially important for OST-B given that the substrate must be positioned near the membrane to gain access to this OST complex. The membrane-associated malectin may hold the substrate in the vicinity of the membrane to aid with further modification.

The diglucosylated glycan is further processed to a monoglucosylated and eventually to a non-glucosylated glycan by α-glucosidase II (GlsII), a soluble non-covalently linked heterodimeric protein⁶⁰. The larger α-subunit of GlsII possesses catalytic activity, whereas the β-subunit contains a C-terminal ER-retrieval sequence and has been shown to be necessary for activity, solubility and localization of the enzyme^61,62.

The sequential action of GlsI and GlsII creates monoglucosylated glycans that then bind to the lectin chaperones calnexin, a type-I transmembrane protein, and calreticulin, its soluble paralogue^63,64 (step 2 in Fig. 4). The lectin chaperones bind specifically to monoglucosylated glycoforms^15,56 and consist of an N-terminal globular domain, a C-terminal domain and a proline-rich region domain (P-domain)⁶⁵. The N-terminal domain contains a single carbohydrate-binding site⁶⁶. Isothermal titration calorimetry of purified calreticulin indicates that the lectin-binding domain binds monoglucosylated glycans with submicromolar affinity⁶⁷. Substrate binding to both calnexin and calreticulin appears to be strongly influenced by calcium binding, because treatment with EGTA (ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid) in vitro reduced the affinity of both lectins to bind radiolabelled Glc₁Man₉GlcNAc₂ carbohydrates⁶⁵. Indeed, the luminal domain of calnexin and calreticulin has a highly conserved, high-affinity calcium-binding site⁶⁶.

Scaffolding role of lectin chaperones

Various biochemical techniques have demonstrated that the P-domain of calnexin and calreticulin lectins serves as a scaffold for associated folding factors (AFFs) including the protein disulfide isomerase A3 (PDIA3, also known as ERp57), cyclophilin B (CyB), and endoplasmic reticulum resident protein 29 (ERp29)⁵⁵. This binding positions AFFs to act on their substrates^68,69.

ERp57, which contains four thioredoxin domains, is recruited to nascent polypeptide chains by its association with the lectin chaperones and accelerates the formation of disulfides during protein maturation or the isomerization of mispaired Cys^70,71. It was shown in a purified system that the action of ERp57 on RNase B, which contains four interwoven disulfides, was enhanced when this oxidoreductase was in complex with calnexin and calreticulin^72,73. Additionally, the knockdown of ERp57 in liver hepatocytes caused a substantial decrease in the secretion of glycoproteins⁷⁴ and its knockout in the semi-permeabilized human epithelial cell line HT1080 compromised the isomerization of non-native disulfides within glycoproteins⁷¹.

The roles of the six ER-resident peptidyl-prolyl cis–trans isomerases in protein maturation are poorly understood⁷⁵. CyB is a calnexin- or calreticulin-associated peptidyl-prolyl cis–trans isomerase that is upregulated by ER stress, and its overexpression can suppress apoptosis⁷⁶. Treatment of cells with cyclosporin A, a high-affinity inhibitor of CyB, demonstrated that this AFF aids in the maturation of collagen, transferrin and the C_H1 domain of human IgG^77,78. These findings are consistent with a role for CyB in the protein folding of glycoproteins containing multiple cis-proline residues.

Another AFF, ERp29, presents a puzzle. In most organisms, ERp29 contains one N-terminal thioredoxin-like domain, but is thought to lack thioredoxin enzyme activity because it does not contain a Cys–X–X–Cys motif. The Cys–X–X–Cys motif, found within many oxidoreductases, influences the reduction potential of the protein and can facilitate disulfide bond formation or isomerization^79,80. Surprisingly, ERp29 from Drosophila melanogaster does contain a Cys–X–X–Cys motif⁸¹. How this apparent loss of activity has influenced the chaperone function of ERp29 has yet to be uncovered, but it has been shown to be involved in the maturation of thyroglobulin and collagen^82,83.

As ERp57, CyB and ERp29 bind to the same site on calnexin and calreticulin, only one of these AFFs can bind to each lectin chaperone at a time⁸⁴. This leads to interesting questions. Are the AFFs associated with the lectin chaperones before substrate binding or after? How does the appropriate AFF get recruited to the calnexin–client complex or the calreticulin–client complex at the correct time? Can multiple AFFs sequentially assist during one round of lectin chaperone binding? Thus, the ability of calnexin and calreticulin to bind a variety of AFs allows their clients to overcome a diverse set of folding bottlenecks, such as disulfide bond formation and disulfide or proline isomerization, supporting an increase in the rate and efficiency of maturation of their clients.

Lectin chaperones act as site-specific holdases

Chaperone binding can slow the folding and maturation of nascent chains. Treatment with glucosidase inhibitors that abolish lectin chaperone binding accelerates the co-translational folding and oligomerization of the viral fusion glycoprotein haemagglutinin; however, the efficiency of maturation was diminished^15,85,86. Furthermore, haemagglutinin folding was arrested when haemagglutinin was trapped on lectin chaperones⁸⁶.

Importantly, lectin chaperone binding can specifically restrict the bound region of the substrate, thereby directing the folding of a protein by allowing unbound regions to fold first⁵³. In this manner, lectin chaperones can direct the folding trajectory by acting as site-specific holdases. Consequently, glycan positioning can have an important role in directing the folding process. N-terminal glycans on haemagglutinin and the serpin antithrombin (ATIII) help to protect the N terminus and delay its folding through lectin chaperone binding, thus allowing the medial and C-terminal domains, respectively, to fold first^42,45,87. It has been proposed that the N-terminal 50 amino acids (following the signal-sequence cleavage site) constitute the N-terminal ‘chaperone selection zone’^45,88 (Fig. 5). N-glycans in this region mediate initial binding by the lectin chaperones, whereas their absence can allow for early binding of the nascent chain to BiP. More C-terminal glycans then support the handover to the lectin chaperone network.

Fig. 5 |. Factors determining early ER chaperone binding.

The positioning of the glycan within the early N terminus (around 1–50 amino acids, AA) can determine early chaperone binding. If an N-glycan site is present within this window, the protein will primarily interact with lectin chaperones, but if no N-glycan site is present in this sequence region, the protein may instead interact with the endoplasmic reticulum (ER) heat-shock protein 70 (HSP70) chaperone BiP. OST-A, oligosaccharyltransferase A.

The binding of lectin chaperones to substrates appears to retard the removal of the remaining glucose from the Glc₁Man₉GlcNAc₂ glycan as they block glucose hydrolysis by GlsII^60,89. How then are substrates released from lectin chaperones for GlsII trimming? One hypothesis proposes that as clients fold into their native form, the interaction between substrates and lectin chaperones weakens, allowing them to dissociate⁹⁰. Another model posits that the remaining glucose is trimmed by GlsII during transient glycoprotein release from the lectin chaperones thanks to the low binding affinity of the carbohydrate to the chaperone⁹¹. At this stage the protein can continue to fold (step 3 in Fig. 4), and if it reaches its native state, be trafficked out of the ER (step 4 in Fig. 4).

Recognition and modification of misfolded substrates

If a protein remains improperly folded upon release from the lectin chaperones, it can be recognized by the ‘gatekeeper’ of the secretory pathway — the UDP-glucose:glycoprotein glucosyltransferase (UGGT), which recognizes misfolded substrates and adds a glucose to select glycans, thus allowing the substrate to rebind calnexin and calreticulin for further rounds of chaperone assistance^15,56 (step 5 in Fig. 4b).

The gatekeeper UGGT enables lectin chaperone rebinding

Humans have two paralogues of UGGT (UGGT1 and UGGT2) that share 53.7% amino acid identity, and both have been shown to possess glucosyl transferase activity in cells^92–94. Both UGGTs are soluble, ER-resident proteins, contain an ER-retrieval sequence and have a similar domain organization^95,96. The architecture of UGGTs has been revealed through a crystal structure of the UGGT homologue from Chaetomium thermophilum (CtUGGT)⁹⁶, an earlier, lower-resolution structure of D. melanogaster UGGT⁹⁷, and analysis of the AlphaFold 2 prediction of human UGGT1 (ref. 98). All homologues adopt a U-shaped structure formed by their four thioredoxin like-domains (TRXL1–TRXL4), all of which lack the canonical Cys–X–X–Cys motif found on redox proteins such as protein disulfide isomerase family members⁹⁹. The C-terminal glucosyltransferase domain of UGGTs belongs to the GT24 family^96,99; it binds UDP-glucose and transfers a glucose onto the Man₉GlcNAc₂ glycan on a client protein.

Binding of a substrate to UGGT occurs through its ‘sensory domain’, which consists of the four TRXL domains. The structure of CtUGGT revealed hydrophobic patches believed to be involved in discrimination of non-native substrates^96,97,99. TRXL1–TRXL4 probably participate in substrate recognition similar to other non-catalytically active thioredoxin domains, such as those found in protein disulfide isomerase, that have been shown to be involved in substrate binding¹⁰⁰. Additionally, in vitro glycosylation studies of human UGGT1 showed that truncation of the sensory domain reduced its ability to transfer glucose to a synthetic Man₉GlcNAc₂ glycan¹⁰¹.

With a wide range of glycosylated proteins traversing the ER, how does UGGT recognize the array of clients requiring further lectin chaperone assistance? Most of our current understanding derives from in vitro studies using purified components^101–105. From these studies, it was concluded that UGGT recognizes non-native or near-native model substrates and peptides containing exposed hydrophobic residues. Additionally, the sequence surrounding the glycan was also shown to affect the ability of UGGTs to recognize and modify model peptides¹⁰³. Although these studies with model substrates are valuable to understand how UGGT may recognize specific substrates, they do not address how UGGT can modify clients found within the ER.

Cell-based studies of UGGT have often relied on the overexpression of model substrates and viral proteins that may be non-native to the cell type used^106–108. The use of non-native cell types can be problematic because secretion of proteins is dependent on cell type¹⁰⁹. To overcome these issues, endogenous cellular substrates of UGGT have been identified using cells that lack the active dolichyl pyrophosphate Man₉GlcNAc₂ α1,3-glucosyltransferase (ALG6) protein. ALG6 adds the first glucose to the immature glycan as it is being synthesized. Genetic deletion of the ALG6 enzyme therefore promotes the transfer of the Man₉GlcNAc₂ glycoform by the OST rather than of the normal Glc₃Man₉GlcNAc₂ form^94,107. Thus, in these cells, clients can only have glucose added to their N-glycans by UGGT. The first cellular substrate of UGGT identified in Chinese hamster ovary cells was the lysosomal protein prosaposin¹⁰⁷. Pulse-chase analysis showed that the presence of UGGT resulted in persistent binding of prosaposin to lectin chaperones. Additionally, UGGT1 was necessary for proper maturation of prosaposin, because deletion of UGGT1 resulted in its aggregation and trafficking defects. Furthermore, it has been demonstrated that UGGT is directly involved in the maturation of MHC class I molecules and efficient peptide loading¹¹⁰. UGGT reglucosylates molecules containing peptides that are not bound tightly to the MHC class I protein, resulting in reglucosylation, rebinding to calreticulin and the peptide loading complex, so that a new, more immune-stimulating peptide can be loaded into the MHC class I proteins. This process ensures that the best peptide is paired with the MHC class I molecule prior to secretion from the ER^110,111.

Mechanism of UGGT substrate recognition

A study using glycoproteomics coupled with tandem mass tags has expanded the cellular substrates of UGGT in HEK cells⁹⁴. UGGT1 and UGGT2 were found to recognize a diverse set of cellular proteins. Substrates with differing preference for UGGT1 or UGGT2 could be classified according to several characteristics: UGGT1 preferred larger, single-pass transmembrane proteins destined for the plasma membrane, whereas UGGT2 preferentially modified smaller, soluble proteins destined for the lysosome⁹⁴. This difference may arise from lower sequence conservation in the sensory domains between UGGT1 and UGGT2⁹².

Identifying which N-glycans in a substrate are modified by UGGTs is an important step in understanding the cellular mechanism for recognition, as N-glycan choice identifies regions or domains requiring further folding assistance. Recently, a modified filter-aided sample preparation method for mass spectrometry was used to identify the sites where glycans were modified by UGGT on the serpins alpha-1 antitrypsin and antithrombin III^45,112. Key to the assay was the use of wild-type calreticulin immobilized to beads to capture serpins modified by UGGT⁴⁵. The most highly reglucosylated sites were the most C-terminal glycans for both alpha-1 antitrypsin and antithrombin III. This suggests a mechanism for quality control: if the C terminus of the serpin does not fold correctly early, which is essential for it to adopt its metastable state, the C-terminal N-glycan of the protein will be reglucosylated.

The ability of UGGT to recognize a wide range of clients may be due to the flexibility of the protein. Analysis of the CtUGGT structure from X-ray crystallography showed flexibility especially in the TRXL2 and TRXL3 regions⁹⁶. A bending, clamping and twisting motion of these domains was predicted that potentially explains how UGGT can recognize different sizes or folded states of proteins¹¹³. It has been speculated that the site of misfolding cannot be more than 70–80 Å away from the N-glycan that is modified by UGGT.

The recognition of clients by UGGT may be aided by selenoprotein F (SEP15), a soluble, seleno-Cys-containing 15-kDa protein. SEP15 has been found exclusively in complex with UGGT in mammalian cells, whereas the UGGTs are not always bound to SEP15 (ref. 114). Whether SEP15 contributes to substrate recognition by UGGT remains unclear. In humans and other organisms, SEP15 contains a selenocysteine instead of a cysteine at position 65, suggestive of a redox function¹¹⁵. The redox potential of D. melanogaster Sep15 indicates it may act as a reductase or a cysteine isomerase¹¹⁵. Hydrogen–deuterium exchange analysis of the D. melanogaster UGGT–Sep15 complex has suggested a putative binding site for SEP15 within residues 262–295 of UGGT⁹⁷. Modelling of the complex of human UGGT1–SEP15 using AlphaFold 2 led to the prediction that there is a critical interface in the complex and that two residues (Phe243 and Lys262 in human UGGT1) were necessary anchors for SEP15 binding⁹⁸. In the predicted human UGGT1–SEP15 structure, the selenocysteine of SEP15 is positioned near the active site of UGGT1. Although the role of SEP15 remains to be determined, the geometry of the UGGT1–SEP15 complex suggests that SEP15 may assist in re-glycosylation by selecting and delivering proteins containing mispaired disulfides to UGGT for modification or by reducing the disulfides of clients that are already interacting with UGGT and thus allowing the client to either refold or be targeted for dislocation and degradation if the folding process has reached a dead end.

UGGT as a therapeutic target

The deletion of Uggt1 in mice is embryonically lethal¹¹⁶, whereas mice lacking Uggt2 are viable but have impaired glucose metabolism¹¹⁷. The role of UGGT as a gatekeeper for glycoprotein secretion and trafficking was initially established in Arabidopsis thaliana¹¹⁸. Here, a mutant of the brassinosteroid receptor was retained in the ER by lectin chaperone binding. ER retention was suppressed after deletion of the Arabidopsis thaliana homologue of UGGT, enabling the export of the receptor.

UGGT acts on disease-causing protein mutants more efficiently than on their wild-type forms. Using an assay to measure glycosylation in cells, it was demonstrated that disease-causing variants of the serpins alpha-1 antitrypsin and antithrombin III were found to be glucosylated at higher levels than their wild-type counterparts^45,119. Similarly, a mutated form of tumour-associated calcium signal transducer 2 (TROP-2) associated with gelatinous drop-like corneal dystrophy was modified at a higher level than wild-type TROP2, and the deletion of UGGT1 enabled the cell-surface appearance of the disease variant, which is otherwise retained in the ER¹²⁰. Together, this suggests that UGGT recognizes and modifies disease variant proteins that are off-pathway. These findings point to UGGT as a novel target for therapeutic development. A reduction in UGGT activity may allow some responsive mutants (disease variants that have slight protein-folding defects but still possess some activity) to exit the ER and resume trafficking.

UGGT is effectively inhibited by its product UDP, by UDP-2-deoxy2-fluoro-d-glucose (U2F), by synthetic analogues of N-glycans or by truncated N-glycan variants that act by mimicking the full glycan^121–123. Squaryl-derivatives of UDP have also been shown to inhibit UGGT activity, although their specificity to UGGT is questionable¹²⁴. A fragment-based lead discovery approach was recently undertaken to find novel inhibitors of UGGT¹²⁵. The catalytic domain of CtUGGT was soaked with different chemical compounds and crystallized to determine binders. From this screen, the ligand 5-[(morpholin-4-yl)methyl]quinolin-8-ol (5M-8OH-Q) was identified and shown to interfere with N-glycan binding. Using a cellular assay to measure glycosylation, the molecule was demonstrated to inhibit both UGGT1 and UGGT2 activity in cells, but with low affinity^107,125. Further drug screens or optimization of existing candidates could lead to more potent and specific UGGT1 or UGGT2 inhibitors. With the preference of UGGT2 to re-glucosylate lysosomal proteins, and the tolerance to knockout of UGGT2 in mice, this paralogue may be an attractive target for developing inhibitors, given that many lysosomal storage diseases result from trafficking defects^94,117,126.

ER exit of glycosylated proteins

UGGT recognition ceases when the substrate acquires its native structure and therefore the substrate is not subject to re-glycosylation and no longer binds the lectin chaperones, enabling folded clients to freely exit the ER through bulk flow. Glycans may also be remodelled by ER-resident mannosidases, which remove mannoses and thus expose glycan compositions that are recognized by downstream lectin cargo receptors such as protein ERGIC-53, vesicular integral-membrane protein 36 (VIP36) and VIP36-like protein (VIPL)¹²⁷. This prepares them for packaging into COPII vesicles and can contribute to the selective anterograde trafficking to the Golgi^128,129.

The extraction of persistently non-native substrates from the lectin chaperone-binding cycle is also mediated by a series of additional ER exo-mannosidases that target substrates for degradation¹²⁷. A model has been proposed in which a glycoprotein that is not folded properly in a timely manner is subjected to mannose residue trimming by ER degradation-enhancing alpha-mannosidase-like protein 2 (EDEM2, in complex with the oxidoreductase thioredoxin TXNDC11) to Man8B, an isomer of Man₈GlcNAc₂^130,131 (step 5 in Fig. 4a and step 6 in Fig. 4b). It is possible that the trimming by EDEM2, and to an extent by its paralogue EDEM1, is accelerated by the unfolded state of the protein and the exposure of hydrophobic regions, indicating that the protein is non-native and requires degradation^132,133. This may explain why only misfolded proteins are degraded, whereas properly folded ER proteins are not. This step is followed by additional mannose trimming by EDEM3 and EDEM1 to expose α1–6-bonded mannoses on the clients^130,134,135. α1–6-bonded mannoses are recognized by the lectins OS9 or XTP3B (also known as ERLEC1) for delivery to the HRD1SEL1L complex in the ER membrane, promoting cytosolic retrotranslocation and subsequent ER-associated degradation (ERAD)¹³⁶ (Fig. 6).

Fig. 6 |. Role of mannosidases in protein trafficking or degradation.

The ER degradation-enhancing alpha-mannosidase-like proteins EDEM1–EDEM3 modify N-linked glycans on substrates, targeting a subset for degradation by ER-associated degradation (ERAD)¹²⁵. Substrates are extracted from the lectin chaperone cycle either by trimming of the glycan A-branch, which makes such trimmed glycoproteins a poor substrate for UGGT1 re-glycosylation, or by interacting with components of the ERAD machinery. One hypothesis posits that EDEM2 initiates ERAD by trimming the outermost mannose sugar on the B-branch. Once exposed, the outer mannose is removed from the B-branch and the glycan becomes a substrate and is modified by EDEM3 and EDEM1 until the α1–6 mannose on the C-branch is accessible. Exposure of the α1–6 mannose on the C-branch allows the substrate to bind to protein OS9 or XTP3B, which directs the substrate to the retrotranslocon and ultimately degradation by the proteosome.

Glycoproteins can also be degraded by a process termed ER-to-lysosome-associated degradation (ERLAD)^127,137. In this pathway, ERAD-resistant protein mutants are delivered to the lysosome for degradation through their persistent interaction with calnexin¹³⁸. For example, high-molecular-mass polymers of an alpha-1 antitrypsin Z (AAT-Z) mutant formed in the ER are too large for retrotranslocation through the HRD1 complex^139,140. It has been proposed that persistent lectin chaperone engagement is necessary for delivery of AAT-Z polymers to the lysosomal compartments for degradation. Knockout of calnexin impaired the delivery of mutant AAT-Z to the lysosome, and UGGT1 was necessary for efficient delivery to the endosomal compartments¹³⁸. ERLAD provides an attractive model for the degradation of large glycoprotein oligomers, as it overcomes many of the hurdles of translocating large protein aggregates through the ER membrane¹⁴¹.

The level of N-glycosylation of a client protein and consequently the extent to which the protein is directed out of the ER for degradation can have a role in regulating the cellular concentration of the client and in ensuring that misfolded clients are destroyed. A compelling example of this has been reported for one of the most abundant, glycosylated ER chaperones GRP94 (also known as HSP90B1)¹⁴². The predominant functional form of GRP94 carries one co-translationally added glycan. During ER stress, GRP94 loses its structural integrity and exposes an additional N-glycan sequon, leading to its hyperglycosylation^39,143,144. These modifications support GRP94 destabilization and degradation. The hyperglycosylation of cryptic N-glycosylation acceptor sites and consequent shunting of substrates to degradation can therefore be used to regulate protein levels, functions and disposal, even for a long-lived ER-resident chaperone such as GRP94.

Species-dependent differences of the glycoproteostasis network

The role of N-glycans as protein maturation and quality control tags has evolved as the proteomes have diversified from single-cell eukaryotes to humans. An interesting comparison can be drawn between the N-glycan machineries of humans and Saccharomyces cerevisiae. Many of the components found in the human lectin chaperone cycle are absent in yeast (Fig. 4). S. cerevisiae lack the ability to facilitate protein binding to lectin chaperones early, because the machinery is absent to append glycans co-translationally^14,22. Proteins produced in yeast can only interact with lectins once they have been recognized and modified post-translationally by the OST-B complex (Fig. 4a). Interestingly, both Schizosaccharomyces pombe and S. cerevisiae contain a single lectin chaperone family member, calnexin, but lack calreticulin^145,146. Although the yeast homologue of calnexin binds monoglucosylated proteins in the same manner as calnexins from more complex organisms, there is a fundamental difference between the two: the yeast homologue lacks the C-terminal cytoplasmic tail of mammalian calnexin, which has been shown to be palmitoylated and phosphorylated and has been implicated in a variety of secondary functions, such as ribosomal interaction, calcium signalling and localization of calnexin itself^147–149. Furthermore, the C-terminal tail contains a di-acidic motif thought to be necessary for ER retention of calnexin in mammals. Given that yeast calnexin does not have this motif, there must be an additional signal to ensure that it remains localized in the ER¹⁵⁰.

Beyond the limited interaction of proteins with lectin chaperones in yeast, another key difference is the absence of any homologue of UGGT in S. cerevisiae (Fig. 4). Lacking this key quality control enzyme means that proteins produced in S. cerevisiae only have one opportunity to interact with lectin chaperones for folding assistance. Once released upon glucose trimming by GlsII, S. cerevisiae proteins are unable to rebind calnexin because no UGGT-catalysed re-glycosylation is possible (Fig. 4a). Once the glycan is modified by GlsII in S. cerevisiae the protein client must complete its folding and be trafficked out of the ER or be targeted for degradation (steps 4 and 5 in Fig. 4a). In humans, and many other organisms, the presence of UGGT ensures that proteins requiring further folding assistance can re-enter the lectin chaperone cycle to obtain their native form (step 5 in Fig. 4b). Interestingly, S. pombe does contain a homologue of UGGT, suggesting an early evolutionary need for glycan-dependent protein quality control¹⁵¹.

Taken together, the above comparisons point to several key and fundamental differences in both lectin chaperone assistance and protein quality control between humans and S. cerevisiae. We propose that these changes correlate with the nature and expression demands of the respective proteomes of these organisms. The proteome sizes of H. sapiens and S. cerevisiae are vastly different, and the fraction of proteins that traverse the secretory pathway is greatly increased in the multicellular system compared to the unicellular yeast: 34.0% H. sapiens versus 22.2% S. cerevisiae. Moreover, 21.3% (4,351) of the proteins in H. sapiens are N-glycosylated compared to just 6.1% (411) in S. cerevisiae (unpublished data based on an analysis of Uniprot proteomes for each organism).

Importantly, the N-glycan-dependent protein quality-control system of S. cerevisiae has evolved to handle the endogenous clientele produced in this organism adequately for its survival. However, comparison with multicellular organisms suggests that the ability of the yeast quality-control system to facilitate maturation of exogenously expressed proteins, especially those originating from other species, is likely to be suboptimal. Without early lectin chaperone binding and the quality control provided by UGGT, the use of yeast to produce human proteins may fail because many substrates require these processes for efficient maturation and trafficking^45,152,153. As an example, the efficient production of human serpin proteins in yeast, such as alpha-1 antitrypsin and antithrombin, is likely to encounter challenges, given that these proteins have been shown to benefit from lectin chaperone intervention⁴⁵.

Conclusions and perspectives

The role of N-glycosylation in the folding and quality control of secretory proteins has until recently been underappreciated. N-glycosylation is usually grouped with post-translational modifications. This is misleading, as we have discussed, because the vast majority of N-glycosylation in metazoans occurs co-translationally as the nascent polypeptide enters the ER lumen³⁹. The co-translational addition of N-glycans in multicellular eukaryotes that possess an OST-A complex positions these modifications to aid with early maturation and quality-control events. An important future avenue of study is to relate the positioning of N-glycans to the architecture and folding mechanisms of secretory proteins in multicellular organisms.

Comparisons between the glycoproteostasis network in multicellular organisms with those in yeast and C. elegans are fascinating, given that these latter organisms have only OST-B components and thus add glycans only post-translationally. In these organisms, glycans are expected to have a limited impact on early folding and maturation events. S. cerevisiae have an even more streamlined N-glycan-mediated proteostasis system in their ER, lacking UGGT and thus supporting only a single round of calnexin binding to glycans that were added post-translationally. They also lack the soluble lectin chaperone calreticulin, as well as many associated factors that assist with folding. The implications of the differences in glycoproteostasis between different organisms will be a fruitful area of inquiry in the future. Practically, these differences affect how well exogenous proteins can be successfully expressed and produced. In addition, it will be exciting to elucidate the relationship between the folding mechanisms (managing proper folding, maturation and quality control) of secretory proteins in organisms with fundamentally different components.

As we consider the implications of such insights into N-glycanmediated proteostasis of secretory proteins, it becomes clear that mutations in the proteostasis machinery itself can have broad impacts on several levels, resulting in the onset of human diseases. For example, recent analyses of patient genotypes have led to the discovery of mutations in UGGT that appear to be causative in devastating diseases¹⁵⁴. Defects in this ‘gatekeeper’ in glycoproteostasis will affect the performance of ER glycoproteostasis and thus the folding of a large fraction of the secretome. Therapeutic approaches to UGGT-related diseases must target the function of this very important molecular machine within the ER.

Of the greatest importance for future studies is a better understanding of the multiple roles of N-glycosylation in proteostasis. Consequently, the concept of a chaperone must be expanded to encompass recognition of carbohydrate modifications, specifically including how their remodelling reports on stages in protein maturation in the ER. Until recently, the concept of a chaperone has been focused on protein-based molecular recognition as the means of identifying incompletely or misfolded clients. Now we can add to this paradigm the use of N-glycan tags and their addition to substrates as a major form of proteostasis.

Glossary

Bulk flow

The passive movement of secretory proteins through the ER without a specific targeting signal but instead via concentration gradients.

ER-associated degradation

(ERAD). Mechanisms by which proteins undergo retrotranslocation from the ER for degradation by the proteosome.

ER-derived microsomes

Small vesicles derived from the ER used to study resident proteins or processes within the ER, such as degradation and protein folding.

ER membrane protein complex

A complex that facilitates integration of multi-span transmembrane proteins into the membrane of the ER as the protein is translocated into the ER.

ER-retrieval sequence

Specific sequence at the C terminus of a substrate that is recognized by proteins in the Golgi apparatus and results in the retrograde trafficking of the protein to the ER.

ER-to-lysosome-associated degradation

(ERLAD). Mechanisms by which proteins that are too large for retrotranslocation, and thus not suited for ER-associated degradation, are captured and exported for degradation by endocytic or lysosomal vesicles.

Fragment-based lead discovery

A drug-discovery method in which small low-affinity binders are discovered and subsequently optimized with modification side groups.

Influenza haemagglutinin

A glycosylated membrane protein found on the surface of the influenza virus.

Lysosomal storage diseases

A group of metabolic diseases caused by mutations or deficiencies in certain lysosomal proteins, resulting in the build-up of toxic compounds within the lysosome.

Serpins

A superfamily of structurally similar proteins that inhibit serine protease.

Translocon

A protein complex that facilitates protein translocation into the ER.

Unfolded protein response

A cellular pathway activated in the ER in response to an accumulation of misfolded proteins aimed at restoring homeostasis.