Structural and mechanistic studies of the N-glycosylation machinery: from lipid-linked oligosaccharide biosynthesis to glycan transfer

Abstract

N-linked protein glycosylation is a post-translational modification that exists in all domains of life. It involves two consecutive steps: (i) biosynthesis of a lipid-linked oligosaccharide (LLO), and (ii) glycan transfer from the LLO to asparagine residues in secretory proteins, which is catalyzed by the integral membrane enzyme oligosaccharyltransferase (OST). In the last decade, structural and functional studies of the N-glycosylation machinery have increased our mechanistic understanding of the pathway. The structures of bacterial and eukaryotic glycosyltransferases involved in LLO elongation provided an insight into the mechanism of LLO biosynthesis, whereas structures of OST enzymes revealed the molecular basis of sequon recognition and catalysis. In this review, we will discuss approaches used and insight obtained from these studies with a special emphasis on the design and preparation of substrate analogs.

1 Introduction

N-glycosylation is one of the most common post-translational modifications found in nature. N-glycans play essential roles in protein folding, trafficking, and cell–cell communication (Nothaft and Szymanski 2010; Breitling and Aebi 2013; Cherepanova et al. 2016; Varki 2017). The process starts with the multi-step biosynthesis of a lipid-linked oligosaccharide (LLO) donor, which is subsequently transferred to asparagine residues in secreted proteins containing the glycosylation sequon N-X-S/T, where X can be any amino acid except proline (Aebi 2013; Cherepanova et al. 2016). In bacteria, an acidic residue is required in position −2 (if zero position denotes the acceptor asparagine), resulting in the extended sequon D/E-X-N-X-S/T (Nothaft and Szymanski 2010; Aebi 2013; Cherepanova et al. 2016).

LLO biosynthesis involves the sequential addition of monosaccharide units to a polyisoprenyl carrier coupled to a phosphate (archaea) or pyrophosphate (bacteria and eukaryotes) group. In bacteria, LLO assembly takes place at the cytoplasmic side of the plasma membrane and requires the action of several membrane-associated glycosyltransferases (GTs) that utilize nucleotide-activated sugars as donor substrates (Fig. 1A). In the human pathogen Campylobacter jejuni, LLO biosynthesis starts with the transfer of N,N′-diacetylbacillosamine-phosphate (Bac-1-P) to undecaprenyl-phosphate (Und-P), catalyzed by the phosphoglucosyltransferase PglC (Glover et al. 2006; Ray et al. 2018). The resulting Und-PP-Bac is elongated by PglA and PglJ, which catalyze the addition of a single GalNAc unit each, yielding α-1,3 and α-1,4 glycosidic linkages, respectively (Weerapana et al. 2005). Subsequently, the processive GT PglH adds three α-1,4 GalNAc units (Troutman and Imperiali 2009; Ramírez et al. 2018). Finally, PglI adds a branching β-1,3 glucose to the third GalNAc, yielding Und-PP-Bac-GalNAc₂ (Glc)GalNAc₃. The fully assembled LLO is then recognized and flipped to the periplasm by the ATP-binding cassette transporter, PglK (Alaimo et al. 2006; Perez et al. 2015). In the periplasm, the LLO serves as a donor substrate for the single-subunit OST, PglB (Fig. 1A).

Fig. 1.

N-glycosylation pathway. (A) Schematic of the N-glycosylation pathway in Campylobacter jejuni. (B) Schematic of the N-glycosylation pathway in Saccharomyces cerevisiae. The branches of the completed glycan are labeled (A), (B), and (C). Monosaccharide symbols follow the SNFG (Symbol Nomenclature for Glycans) system.

In higher eukaryotes, the LLO consists of a dolichyl-pyrophosphate carrier (Dol-PP) linked to an oligosaccharide composed of fourteen glycan units arranged in three branches termed A, B, and C (Fig. 1B). The first part of eukaryotic LLO biosynthesis takes place at the cytoplasmic side of the endoplasmic reticulum (ER) membrane. It requires the action of membrane-associated and integral membrane GTs that utilize nucleotide-activated sugars as donor substrates. In contrast, the subsequent transfer reactions in the ER lumen are catalyzed exclusively by integral membrane GTs that utilize dolichyl-phosphate activated sugars (Fig. 1B). LLO biosynthesis starts with the transfer of N-acetylglucosamine-phosphate (GlcNAc-1-P) to dolichyl-phosphate (Dol-P), catalyzed by ALG7 (Kukuruzinska et al. 1994). This is followed by the addition of a β-1,4 GlcNAc catalyzed by the presumed complex ALG13/ALG14 (Gao et al. 2008; Lu et al. 2012). Subsequently, five mannose (Man) units are added by the mannosyltransferases ALG1, ALG2, and ALG11 (Burda et al. 1999). ALG1 adds a β-1,4 mannose (Couto et al. 1984), which serves as an acceptor of the next two mannose units transferred by ALG2, yielding α-1,3 and α-1,6 glycosidic linkages and creating the first branching point of the oligosaccharide (Yamazaki et al. 1998; O’Reilly et al. 2006) (Fig. 1B). The α-1,3 Man (A-branch) is then elongated by ALG11, which transfers two α-1,2 mannose units, yielding the intermediate Dol-PP-GlcNAc₂-Man₅ (Cipollo et al. 2001; Gao et al. 2004). This intermediate is flipped to the ER lumen by an unknown mechanism possibly involving the presumed flippase RFT1 (Helenius et al. 2002). In the ER lumen, the mannosyltransferases ALG3, ALG9, and ALG12 transfer four mannoses from Dol-P-Man onto the growing LLO (Aebi et al. 1996; Burda and Aebi 1999; Frank and Aebi 2005; Manthri et al. 2008). First, ALG3 and ALG9 complete the B-branch by adding one mannose each, yielding α-1,3 and α-1,2 glycosidic linkages (Aebi et al. 1996; Frank and Aebi 2005; Manthri et al. 2008). Subsequently, ALG12 and ALG9 complete the C-branch, adding a mannose unit each, yielding α-1,6 and α-1,2 glycosidic linkages (Burda and Aebi 1999; Frank and Aebi 2005). In the final steps, the glucosyltransferases ALG6 and ALG8 elongate the A-branch with the addition of one α-1,3 glucose unit each (Stagljar et al. 1994; Reiss et al. 1996), and ALG10 catalyzes the addition of an α-1,2 glucose, resulting in the fully-assembled LLO Dol-PP-GlcNAc₂-Man₉-Glc₃ (Burda and Aebi 1998). This molecule is subsequently recognized as a donor substrate by eukaryotic OST (Fig. 1B).

In the past 10 years, high-resolution structures and in vitro studies have provided mechanistic insight into several of the reactions described above. This has in part been made possible by chemical and chemo-enzymatic synthesis of LLO donor and acceptor substrate analogs and by the structure determination of several enzymes of the N-glycosylation machinery by X-ray crystallography and single particle cryo-EM (Table 1.).

Table 1.

Experimentally determined structures of enzymes involved in N-glycosylation.

Protein	Activity	PDB ID	Reference
Bacteria
PglC	Phosphoglycosyltransferase	5W7L	(Ray et al. 2018)
PglH	Glycosyltransferase	6EJI, 6EJK, 6EJK	(Ramírez et al. 2018)
PglK	LLO flippase	5C73, 5C76, 5C78, 5NBD, 6HRC	(Perez et al. 2015, 2017, 2019)
PglB	Oligosaccharyltransferase	3RCE, 5OGL, 6GXC	(Lizak et al. 2011; Napiórkowska et al. 2017, 2018)
Archaea
DPMS	Glycosyltransferase	5MM1	(Gandini et al. 2017)
AglB	Oligosaccharyltransferase	3WAJ, 3WAK, 6GMY, 7E9S	(Matsumoto et al. 2013, 2017; Taguchi et al. 2021)
Eukaryotes
ALG7/PGT	Phosphoglycosyltransferase	6FM9, 5O5E, 5LEV, 6FWZ, 6BW6, 6BW5	(Dong et al. 2018; Yoo et al. 2018)
ALG6	Glycosyltransferase	6SNH, 6SNI	(Bloch et al. 2020)
OST	Oligosaccharyltransferase complex	Yeast: 6EZN, 6C26, 7OCI 8AGB, 8AGC, 8AGE	(Bai et al. 2018; Wild et al. 2018; Neuhaus et al. 2021; Ramírez et al. 2022)
		Human: 6S7O, 6S7T	(Ramírez et al. 2019)

2 Preparation of lipid-linked oligosaccharide analogs

Structural and mechanistic investigation of enzymes requires access to high-purity substrates or substrate analogs. For enzymes involved in the processing of LLO molecules, this requirement is a challenge due to the low water-solubility of the polyisoprenyl carriers undecaprenyl and dolichyl (Fig. 2A). Particularly, the extraction yields of eukaryotic LLOs from native sources are low, especially for LLO intermediates, which are less abundant in the cell. In this section, we will discuss approaches used for the extraction of native LLOs and the preparation of LLO analogs containing either the native polyprenyl carriers or shorter lipid moieties.

Fig. 2.

Chemo-enzymatic synthesis of soluble LLO analogs. (A) Structures of native LLO in C. jejuni and of higher eukaryotes. (B) Pipeline for chemo-enzymatic synthesis of LLO analogs. (1) Preparation of synthetic LLO precursors: chemical structures of the synthetic C20-PP-GlcNAc (bacterial precursor) and of Dol20-PP-GlcNAc₂ (eukaryotic precursor). (2) Enzymatic glycan elongation catalyzed by bacterial Pgl or eukaryotic ALG enzymes. (3) Analysis: resulting LLO analogs are used as substrates for in vitro glycosylation of fluorescently labeled peptides using PglB or eukaryotic OST enzymes. The resulting glycopeptides are analyzed by SDS-PAGE to confirm glycan elongation. (C) Chemical structure of synthetic, non-reactive phosphonate glycosyl LLO analogs for bacterial (left) and eukaryotic (right) OST enzymes. Monosaccharide symbols follow the SNFG system.

2.1 Preparation of LLO molecules containing native polyisoprenyl carriers

Initial approaches to obtain LLO molecules containing the native polyisoprenyl carriers Und-PP (bacteria) or Dol-PP (eukaryotes) included the extraction from native sources or the in vitro reconstitution of the LLO biosynthesis pathway using recombinantly expressed and purified enzymes (Kelleher et al. 2001; Glover et al. 2005; Olivier et al. 2006). A method successfully used to produce bacterial LLO involves the transformation of E. coli cells with a plasmid carrying the pgl cluster, which encodes the complete N-glycosylation machinery of C. jejuni (Kowarik et al. 2006), but a defective pglb gene, preventing glycan transfer and leading to LLO accumulation. The E. coli strain SCM6 is used in this experiment because it is defective in O-antigen biosynthesis. As a result, all cellular Und-P is used for the biosynthesis of the desired LLO, which can be extracted with organic solvents and enriched by reversed-phase chromatography. This method also allowed the preparation of biosynthetic LLO intermediates by inactivation of specific GT genes in the pgl cluster (Lizak et al. 2014; Perez et al. 2015; Ramírez et al. 2018). As an alternative to the production of bacterial LLOs in E. coli, chemo-enzymatic methods can be used. For example, enzymatic UDP-Bac synthesis was combined with Bac-1-P transfer to Und-P and subsequent glycan elongation using recombinantly expressed and purified Pgl enzymes (Tai and Imperiali 2001; Glover et al. 2005; Weerapana et al. 2005; Olivier et al. 2006).

For eukaryotic LLOs, extraction protocols from porcine pancreas and yeast cultures have been described (Das and Heath 1980; Kelleher et al. 1992, 2001). These approaches led to fractions enriched in the mature Dol-PP-GlcNAc₂-Man₉-Glc₃ but containing other LLO intermediates as impurities (Das and Heath 1980; Kelleher et al. 1992, 2001; Karaoglu et al. 2001). Alternative approaches including chemical synthesis of high-mannose glycans were described (Boltje et al. 2009; Li et al. 2015; Shivatare et al. 2016). However, their linkage to polyisoprenyl carriers has not been tested.

Although the preparation and extraction of LLO analogs carrying the native polyisoprenyl carriers allowed initial studies of N-glycosylation enzymes, these approaches usually require high concentrations of detergent for solubilization, which is incompatible with structural studies and with biotechnological approaches aimed to high-throughput production of glycoproteins. In the next section, we will discuss chemo-enzymatic strategies developed to overcome these difficulties yielding sufficient amounts of LLO analogs for various in vitro applications.

2.2 Chemo-enzymatic synthesis of soluble LLO analogs

A generalized protocol for the preparation of soluble LLO analogs carrying glycan moieties of different size includes the (i) preparation of synthetic LLO precursors, (ii) enzymatic glycan elongation and (3) analysis of the resulting LLO analogs (Fig. 2B).

2.2.1 Preparation of synthetic LLO precursors

Chemical synthesis of LLO precursors has focused on using short lipid moieties that offer increased solubility (Musumeci et al. 2013; Liu et al. 2014; Napiórkowska et al. 2017; Ramírez et al. 2017a; Boilevin and Reymond 2018). For bacterial LLOs, polyprenyl tails of different length and stereochemistry coupled to a pyrophosphate moiety and a single GlcNAc unit could serve as donor substrates of PglB (Liu et al. 2014; Napiórkowska et al. 2018). In contrast, eukaryotic OST enzymes were unable to transfer a single GlcNAc, but could transfer a GlcNAc₂ (GlcNAc-β1-4GlcNAc) moiety, which is present at the reducing end of all eukaryotic LLOs (Ramírez et al. 2017a; Eyring et al. 2021; Neuhaus et al. 2021). The highest turnover rates were observed with lipid tails containing four isoprenyl units featuring the stereochemistry of their native counterparts undecaprenyl or dolichyl (C20-PP-GlcNAc for PglB and Dol20-PP-GlcNAc₂ for eukaryotic OSTs, Fig. 2B) (Napiórkowska et al. 2017; Eyring et al. 2021). Furthermore, the LLO analog phytanyl-PP-GlcNAc₂, which carries five saturated isoprenyl units, was accepted as a donor substrate of the single-subunit OST from T. brucei STT3A (Rexer et al. 2017).

2.2.2 Enzymatic glycan elongation

Synthetic LLO precursors could be elongated in vitro using recombinantly expressed and purified GTs (Fig. 2B). In this way, C20-PP-GlcNAc was extended to C20-PP-GlcNAc-GalNAc₂, using purified PglA and PglJ from C. jejuni. The resulting LLO analog allowed the structural and mechanistic characterization of the GT PglH from C. jejuni (Ramírez et al. 2018), which will be discussed in Section 3.2. Elongation of eukaryotic LLO analogs was achieved using purified ALG enzymes. First, ALG1, ALG2, and ALG11 were used to generate Dol20/Dol25-PP-GlcNAc₂Man₅ (Ramírez et al. 2017b). Further glycan elongation required the expression and purification of integral membrane GTs, which use Dol-P-Man or Dol-P-Glc as donor substrates. Therefore, analogs of dolichyl-phosphate activated mannose and glucose were required for in vitro glycosylation reactions (Bloch et al. 2020; Ramírez et al. 2022). We will discuss the preparation of those analogs in Section 2.3. An alternative approach using membrane fractions of E. coli cells expressing ALG enzymes allowed the elongation of the LLO analog Phytanyl-PP-GlcNAc₂ to LLOs carrying high mannose-glycans (Li et al. 2018, 2019).

2.2.3 Analysis of synthetic LLO analogs

Chemo-enzymatically generated LLOs have to be evaluated for their ability to act as enzymatic substrates. A common approach is to use them as glycan donors for OST-catalyzed in vitro glycosylation of fluorescently labeled acceptor peptides (Fig. 2B). The resulting glycopeptides are analyzed by SDS-PAGE and mass spectrometry to confirm the composition of the attached glycans (Ramírez et al. 2017b, 2022; Bloch et al. 2020).

2.3 Chemical synthesis of dolichyl-linked sugar donors for luminal ALG enzymes

To produce dolichyl-phosphate-sugar analogs, two strategies were explored, both using shortened polyprenyl tails: chemical synthesis of the substrate analogs Dol25-P-Man and Dol25-P-Glc (Bloch et al. 2020), and the enzymatic synthesis of phytanyl-P-Man from phytanyl-P using recombinantly expressed and purified dolichylphosphate-mannose synthase (Dpm1) from Saccharomyces cerevisiae (Li et al. 2019). Both strategies proved successful for generating Dol-P-Man analogues that can be utilized by the mannosyltransferases ALG3, ALG9 and ALG12. The synthesis of a Dol-P-Glc analog has so far only been achieved by chemical synthesis. Synthetic Dol25-P-Glc was successfully used as a donor substrate by the glucosyltransferases ALG6, ALG8 and ALG10 in vitro (Bloch et al. 2020; Ramírez et al. 2022).

2.4 Chemical synthesis of non-reactive donor substrates

Non-reactive substrates are powerful tools for structural and biochemical studies of GTs. Because they act as competitive inhibitors, they are commonly used to capture pseudo-Michaelis complexes of enzyme-catalyzed reactions. Non-reactive analogs of nucleotide-, lipid phosphate- or lipid pyrophosphate-activated sugars have been generated replacing the phosphate group connected to the anomeric C1 atom of the reducing-end monosaccharide with a thiophosphate, a phosphonate, or a hydroxyphosphonate group. Such modifications result in poor leaving groups (Garneau et al. 2004; Chang et al. 2006; Hsu et al. 2014; Napiórkowska et al. 2017; Ramírez et al. 2018). For example, glycosyl phosphonates were successfully used as competitive inhibitors of bacterial and eukaryotic OST enzymes (Fig. 2C) (Napiórkowska et al. 2017; Ramírez et al. 2017a). The bacterial OST PglB of Campylobacter lari was crystallized in complex with a phosphonate LLO analog and an acceptor peptide, resulting in a ternary complex (Napiórkowska et al. 2017). Similarly, a crystal structure of C. jejuni PglH was determined bound to a phosphonate analog of UDP-GalNAc and an acceptor LLO analog (Ramírez et al. 2018).

3 Structural insight into LLO biosynthesis

3.1 Structures of phosphoglycosyltransferases initiating LLO biosynthesis

The first step in the biosynthesis of various glycolipids (including the LLO substrates for N-glycosylation, O-antigens, teichoic acids and peptidoglycan) is the attachment of a sugar-phosphate group to a polyprenyl-phosphate carrier (Denapaite et al. 2012; Ray et al. 2018; Liu et al. 2020). This reaction is catalyzed by phosphoglycosyltransferases (PGTs), which are divided into two families based on their architecture. Monotopic PGTs, such as the bacterial PGT PglC, associate with one leaflet of the lipid bilayer (Entova et al. 2018; Ray et al. 2018; Allen et al. 2019). In contrast, polytopic PGTs including the bacterial WecA, TarO, and MraY enzymes are integral membrane proteins (Lehrman 1994).

In Campylobacter, the PGT PglC catalyzes the transfer of Bac-1-P from UDP-Bac to Und-P. The crystal structure of Campylobacter concisus PglC revealed that it associates with the plasma membrane via an amphipathic helix (Fig. 3A) (Entova et al. 2018; Ray et al. 2018). This allows Und-P to interact with PglC through electrostatic interactions between the phosphate group and a cluster of positive residues in the predicted active site. The catalytic mechanism proposed for PglC involves the nucleophilic attack of the catalytic base Asp93 on the α-phosphate of UDP-Bac, resulting in a covalently bound intermediate. A nucleophilic attack of Und-P on the covalent intermediate, then yields Und-PP-Bac and UMP.

Fig. 3.

Structural insight into LLO biosynthesis. (A) PGTs initiating LLO biosynthesis. Monotopic PGT: X-ray structure of Campylobacter concisus PglC (PDB: 5W7L). Polytopic PGT: X-ray structure of DPAGT1 in complex with bound tunicamycin (B) LLO-elongating GTs. Membrane-associated GT: X-ray structure of C.jejuni PglH in complex with LLO analog and non-reactive UDP-CH₂-GalNAc (PDB: 6EJK). Integral membrane GT: cryo-EM structure of S. cerevisiae ALG6 in complex with donor substrate analog (PDB: 6SNH). (C) Left, schematic of bacterial LLO flipping mediated by PglK. Middle, X-ray structure of C. jejuni PglK in outward-occluded conformation (PDB: 6HRC). In the inset, the proposed recognition site for the undecaprenyl moiety. Monosaccharide symbols follow the SNFG system.

In humans, the transfer of GlcNAc-1-P to Dol-P is catalyzed by a polytopic PGT, DPAGT1 (equivalent to ALG7 in yeast). DPAGT1 is inhibited by tunicamycin, an off-target effect that prevents its use as an antibiotic for treating bacterial infections (Tamura et al. 1976). Tunicamycin is a complex molecule containing uracil, tunicamine, GlcNAc and an aliphatic tail. Its mode of action is to mimic the Michaelis complex formed during PGT-catalyzed phosphoglycosyl transfer reactions (Tamura et al. 1976; Keller et al. 1979). Therefore, it acts as a competitive inhibitor of MraY, an essential PGT involved in the cell wall biosynthesis of many bacterial pathogens. DPAGT1 has been shown to exist predominantly as a dimer with each protomer composed of 10 transmembrane (TM) helices, and the active site located on the cytoplasmic side (Dong et al. 2018; Li et al. 2018; Yoo et al. 2018). The nucleotide moiety of the donor substrate is bound through hydrophobic interactions involving the uracil ring and multiple hydrogen bonds to DPAGT1 (Dong et al. 2018). Dol-P is proposed to bind in a hydrophobic groove where the aliphatic tail of tunicamycin was bound (Fig. 3A). Structural insight into substrate recognition by DPAGT1 has been used for designing tunicamycin derivatives with higher specificity for bacterial MraY and low cross-reactivity with DPAGT1 (Dong et al. 2018).

3.2 Structures of LLO-elongating GTs

GTs are essential for the biosynthesis of lipid-linked oligosaccharides. They catalyze the transfer of a glycan unit from either nucleotide- or lipid phosphate-activated sugars to the growing oligosaccharide. Since the acceptor substrate, and in some cases also the donor substrate, is lipidic, these enzymes are either associated with the membrane or integral membrane proteins.

In C. jejuni, the membrane-associated GT PglH transfers exactly three α-1,4 GalNAc units to the growing LLO (Fig. 3B) (Troutman and Imperiali 2009). PglH is a GT-B and consists of two Rossman-like domains connected by a linker region. Two X-ray crystal structures of PglH were solved containing bound C20-PP-GlcNAc-GalNAc₂, which had been generated as described in Section 2.2. PglH contains an amphipathic helix named “ruler helix” that plays a dual role in membrane anchoring and glycan counting (Ramírez et al. 2018). The “ruler helix” contains three positively charged residues that interact consecutively with the pyrophosphate group of the LLO as the glycan is elongated (Fig. 3B). After the third GalNAc transfer, the pyrophosphate group cannot bind to PglH anymore, and the product Und-PP-Bac-GalNAc₄ is released (Ramírez et al. 2018). Other membrane-associated members of the GT-B family exhibit a hydrophobic helix at the same position as the “ruler helix” of PglH, suggesting that its role in mediating association with the membrane is conserved. However, the “ruler helix” in PglH is longer than the equivalent helices of other GTs that catalyze single sugar transfers, suggesting that it might be a unique feature of processive membrane-associated GTs of the B superfamily.

In eukaryotes, the glucosyltransferase ALG6 catalyzes the transfer of the first glucose unit from Dol-P-Glc to the LLO intermediate Dol-PP-GlcNAc₂-Man₉. Structures of ALG6 in distinct states were determined using single-particle cryo-EM (Bloch et al. 2020). ALG6 exhibits the characteristic fold of the C superfamily of GTs (GT-C) (Alexander and Locher 2023). The structure of ALG6 with bound donor substrate analog Dol25-P-Glc revealed that the lipid moiety was accommodated in a hydrophobic groove formed by TM6, TM7, and TM8, whereas the glucose was located in the active site (Fig. 3B). Mutational studies revealed that residues Asp99 and His378 play an essential role in ALG6 activity (Bloch et al. 2020). Further studies are needed to understand the specificity exhibited by ALG enzymes toward the LLO acceptor substrates.

3.3 L‌LO flipping

In bacteria, LLO translocation is catalyzed by the flippase PglK, a homodimeric ATP binding cassette (ABC) transporter (Alaimo et al. 2006). Based on X-ray crystal structures of PglK in distinct states, molecular dynamics (MD) simulations, and biochemical studies, a mechanism of ATP-driven LLO translocation was proposed (Perez et al. 2015, 2017, 2019). It postulates that the LLO molecule is recognized on the surface of PglK. An external helical (EH) segment located on the periplasmic side of the membrane forms a hydrophobic groove where the undecaprenyl moiety is proposed to bind (Fig. 3C), and a cluster of positively charged residues facing the cytoplasm participates in the recognition of the pyrophosphate group (Perez et al. 2019, 2015). During flipping, the headgroup of the LLO molecule is proposed to enter the outward-facing translocation cavity, and ATP hydrolysis is thought to squeeze out the LLO headgroup, which is released into the periplasm (Fig. 3C).

In eukaryotes, the essential membrane protein RFT1 has been postulated as the flippase of the LLO intermediate Dol-PP-GlcNAc₂-Man₅, which is transported from the cytoplasmic to the luminal side of the ER membrane (Frank et al. 2008; Sanyal et al. 2008; Haeuptle and Hennet 2009; Rush et al. 2009; Jelk et al. 2013). This proposal was supported by the observation that depletion of RFT1 in yeast leads to accumulation of Dol-PP-GlcNAc₂Man₅, and strongly reduced levels of fully assembled LLO (Helenius et al. 2002). The same phenotype was observed in patient-derived fibroblasts that carried the R67C mutation in the human RFT1 ortholog (Haeuptle et al. 2008). However, this view was challenged by another study that reported Dol-PP-GlcNAc₂Man₅ flipping activity in vesicles derived from crude membrane extracts independently of the presence of RFT1 (Frank et al. 2008). Subsequent studies suggested that RFT1 is not directly responsible for the flipping of Dol-PP-GlcNAc₂-Man₅, but it might instead act as a chaperone controlling access to the LLO (Jelk et al. 2013). The issue is currently unresolved because no direct experimental evidence of flippase activity of RFT1 is available.

4 Structural studies of OSTs

OST enzymes exhibit different complexity in their architectures (Kelleher and Gilmore 2006; Shrimal and Gilmore 2018). In bacteria and archaea, OST is a single-subunit enzyme (Dell et al. 2010; Jarrell et al. 2014; Nothaft and Szymanski 2013, 2010), whereas eukaryotic organisms display architectures ranging from single-subunit enzymes (e.g. Trypanosoma brucei and Leishmania major), to complexes composed of four (e.g. Plasmodium falciparum), six (e.g. Cryptosporidium parvum), seven (insects and plants), or eight subunits (mammals and yeast). In multimeric complexes, the catalytic subunit is named STT3 and is a structural homolog of the bacterial PglB protein (Fig. 4A, B).

Fig. 4.

Architecture of OST enzymes. (A) Structure of the single subunit OST PglB from Campylobacter lari (PDB:5OGL). (B) Structure of the octameric OST complex from S. cerevisiae (PDB: 6EZN). (C) Cryo-EM map of human OST-A (EMD-10110) with subunits colored as in (B). DC2 mediates the interaction with the translocation channel SEC61 and ribophorin-I interacts with the translating ribosome. (D) Comparison of the cryo-EM maps of human OST-A and OST-B. The C-terminal domain of ribophorin I forms a helix bundle in OST-A, but not in OST-B.

4.1 Architecture of single-subunit and multimeric OST enzymes

Initial insight into the architecture of OST enzymes was obtained from crystal structures of the bacterial and archaeal proteins, PglB and AglB (Lizak et al. 2011; Matsumoto et al. 2013, 2017; Napiórkowska et al. 2018; Taguchi et al. 2021). These enzymes belong to the GT-C superfamily (Alexander and Locher 2023). They contain an N-terminal transmembrane domain composed of 13 TM helices and a C-terminal periplasmic domain (Fig. 4A). The catalytic site is located at the external membrane boundary, which allows OST to access both the membrane-embedded donor substrate (LLO) and the soluble acceptor polypeptide.

The first structures of a multimeric OST are from S. cerevisiae and were determined by single-particle cryo-EM (Bai et al. 2018; Wild et al. 2018). They revealed that OST subunits are arranged in three subcomplexes that had been previously identified as intermediates of OST assembly in vivo (Mueller et al. 2015). Subcomplex I consists of OST1 and OST5; subcomplex II is formed by OST4, the catalytic subunit STT3, and an oxidoreductase chaperone (OST3 or OST6); subcomplex III consists of OST2, WBP1, and SWP1 (Fig. 4B). In mammalian cells, two functionally distinct OST complexes co-exist, termed OST-A and OST-B (Kelleher et al. 2003; Mueller et al. 2015; Shrimal et al. 2015). They contain distinct catalytic subunits (STT3A or STT3B) and perform distinct cellular functions, which can be rationalized based on their structures. OST-A catalyzes co-translational glycosylation as it forms a complex with the translocation channel SEC61 and an ER-membrane-associated ribosome. The interaction between OST-A and SEC61 is mediated by DC2, an OST-A-specific subunit that binds STT3A but not STT3B (Shrimal et al. 2017;Braunger et al. 2018 ; Ramírez et al. 2019). In contrast, OST-B does not associate with the translocon but contains an oxidoreductase (MAGT1 or TUSC3), suggesting that OST-B is more similar to yeast OST (Cherepanova et al. 2014). Binding of OST-A to the translating ribosome is mediated by the C-terminal domain of ribophorin-I (OST1 in yeast) (Yu et al. 1990; Braunger et al. 2018). Although ribophorin-I is present in both OST complexes, it forms a helical bundle contacting the translating ribosome only in OST-A (Fig. 4C, D). In OST-B, a distinct structural arrangement that includes an extended N-terminal segment of STT3B precludes an interaction with the ribosome (Fig. 4C, D) (Braunger et al. 2018; Ramírez et al. 2019). As a result of this structural organization, OST-A is part of a super-complex that includes the translocon and the translating ribosome, which provides access to the nascent, unfolded polypeptide chain of secretory proteins as they reach the ER lumen (Ruiz-Canada et al. 2009; Shrimal et al. 2013a; Cherepanova et al. 2019). Consistent with this, OST-A glycosylates the majority of N-X-S/T sequons. In contrast, OST-B glycosylates sequons skipped by OST-A, especially those located near the C-terminus of proteins, near disulfide bridges, or in loops connecting transmembrane helices of integral membrane proteins (Shrimal et al. 2013b; Cherepanova et al. 2014, 2019). The joint activity of OST-A and OST-B allows for efficient glycosylation and ensures high glycan occupancy in secreted proteins (Shrimal et al. 2015; Cherepanova et al. 2019).

4.2 Binding and recognition of the N-glycosylation sequon

The binding site of the N-glycosylation sequon N-X-S/T is located at the interface of the transmembrane and luminal domain of STT3 proteins. In order to place the acceptor asparagine in the active site of STT3, acceptor peptides have to form a 180° loop, explaining why glycosylation sequons are located in flexible or partially unfolded regions, or why the modification by OST precedes glycoprotein folding (Fig. 5A). Binding of serine or threonine residues at the +2 position of the glycosylation sequon is mediated by hydrogen bonding to the side chains of the “WWD motif” (Trp-Trp-Asp sequence), which is strictly conserved in STT3 proteins (Fig. 5B). In bacteria, where an extended glycosylation sequon D/E-X-N-X-S/T is recognized, the acidic residue at the −2 position forms a salt bridge with the residue Arg331 of PglB (Lizak et al. 2011). Eukaryotic OSTs do not require an acidic residue at this position, and do not have a residue equivalent to Arg331. Instead, a cavity exists on the surface of STT3 that can accommodate larger side chains (Wild et al. 2018; Ramírez et al. 2019, 2022). Binding of acceptor peptide to OST requires a divalent metal ion (e.g. Mg²⁺ or Mn²⁺) bound to the active site of OST (Lizak et al. 2011; Gerber et al. 2013). The M²⁺ ion is coordinated by a cluster of acidic residues that simultaneously interact with the acceptor amide group of the asparagine (Fig. 5B).

Fig. 5.

Substrate recognition and catalysis by OST enzymes. (A) Structure of C. lari PglB in complex with acceptor peptide and non-reactive LLO analog (PDB: 5OGL). Bound substrates are shown as spheres. (B) Left, sequon recognition by C. lari PglB. (PDB: 3RCE). Right, active site of S. cerevisiae OST as observed in the structure of a ternary complex containing non-acceptor peptide and reactive LLO (PDB: 8AGC) (adapted from Ramírez et al. 2022).

4.3 Binding of LLO

Molecular insight into LLO binding has been obtained from structural studies of OST enzymes bound to either reactive (Section 2.2) or non-reactive (Section 2.4) LLO analogs (Napiórkowska et al. 2017, 2018; Ramírez et al. 2022). These studies revealed that the binding pocket for the polyprenyl-pyrophosphate moiety is highly conserved in STT3 proteins. The lipid tail is accommodated in a hydrophobic groove formed by TM6, TM8, and TM11, whereas the pyrophosphate group is coordinated by a cluster of positively charged residues and the divalent metal ion bound in the active site. Structural studies of PglB revealed that the reducing end GlcNAc of a bound LLO analog interacts via hydrogen bonds with polar residues in the vicinity of the active site. Of note, the hydroxyl group of a strictly conserved tyrosine residue (Tyr468 in PglB) interacts with the N-acetyl group at C2, which is present in the bacillosamine unit of bacterial LLO and GlcNAc of eukaryotic LLO.

The cryo-EM structure of S. cerevisiae OST bound to the LLO analog Dol20-PP-GlcNAc₂-Man₉-Glc₃ revealed how the fully assembled LLO is recognized (Ramírez et al. 2022). This is important for ensuring the attachment of the complete oligosaccharide to newly synthesized proteins, which is a pre-requisite of correct glycoprotein folding and intracellular trafficking. The recognition involves mainly two parts of branch A of the LLO, the GlcNAc₂ moiety at the reducing end and the terminal Glc₃ at the non-reducing end. These two parts represent the initiating and the terminating steps of LLO biosynthesis. Recognition of the GlcNAc₂ moiety exclusively involves residues of the catalytic subunit STT3, and the binding site is strictly conserved in all eukaryotic organisms. In contrast, binding of the terminal Glc₃ moiety occurs in a pocket formed by the non-catalytic subunits WBP1 and OST2. This pocket is not conserved in all eukaryotes, which correlates with the observation that only OST systems with a complete subcomplex-III (WBP1, OST2, and SWP1) transfer Glc₃-containing oligosaccharides. These findings suggest a coordinated evolution of the N-glycosylation pathway, resulting in a correlation between OST architecture and structure of the transferred oligosaccharide (Ramírez et al. 2022).

4.4 Catalysis and reaction cycle

In order to visualize pseudo-Michaelis complexes of OST enzymes, structures of ternary complexes with analogs of both substrates were determined. This was achieved by using either non-reactive LLO analogs (including glycosyl phosphonates described in Section 2.4) or non-reactive peptides, where the acceptor asparagine is replaced by 2,4-diaminobutanoic acid (Dab) (Napiórkowska et al. 2017, 2018; Ramírez et al. 2022). The structures obtained with those approaches revealed that the simultaneous binding of the substrate analogs facilitates the ordering of transmembrane helix TM9 and the engagement of the external loop connecting TM9 and TM10 (EL5). The latter is relevant because it completes the active site of the catalytic STT3 subunit, as EL5 contains conserved catalytic residues. As a result of the conformational changes, a tunnel is formed that connects the binding sites of LLO and peptide (Fig. 5). The side chain of the acceptor asparagine is lodged in this tunnel and approaches the C1 of the reducing end unit of the LLO (Fig. 5B). Two distinct mechanisms have been proposed for the activation of the amide group during the OST-catalyzed transfer reaction. (1) Acid/base activation by a catalytic base in the active site of STT3 proteins or (2) Activation by twisting of the amide group induced by the immobilization of the acceptor asparagine via hydrogen bond interactions with acidic residues at the active site (Lizak et al. 2013). Given that the structures of STT3 enzymes have not revealed a catalytic base that could deprotonate the acceptor amide, the second hypothesis may be more likely. However, this has not been demonstrated experimentally. After glycan transfer, positively charged residues interacting with the pyrophosphate group of Dol-PP likely stabilize the leaving group.

5 Conclusion and future perspectives

Recent structures have advanced our mechanistic understanding of the N-glycosylation machinery, including LLO biosynthesis and OST-catalyzed glycan transfer. Given that the strict order of eukaryotic glycan assembly is likely determined by the high substrate specificity of the ALG enzymes, future studies aimed at revealing the molecular basis of their specific substrate recognition are needed. Future studies into OST function might focus on understanding the regulation of concerted translocation, N-glycosylation and folding of secreted proteins at the ER membrane. Furthermore, structural insight into the N-glycosylation machinery will open new avenues for the design of specific N-glycosylation inhibitors that could have potential uses as helpers in chemotherapeutic treatments and as antiviral drugs.