Protein N-glycosylation and O-mannosylation are catalyzed by two evolutionarily related GT-C glycosyltransferases

Abstract

The structural folds of glycosyltransferases are categorized into three superfamilies: GT-A, GT-B, and GT-C. Few structures of GT-C fold existed in the Protein Data Bank prior to the recent advent of high-resolution cryo-EM, because the glycosyltransferases are large membrane proteins that are difficult to crystallize. The use of cryo-EM has resulted in the structures of several key GT-C glycosyltransferases. Here we summarize the latest structural features of and mechanistic insights into these membrane enzyme complexes.

INTRODUCTION

Glycosylation of biological macromolecules – proteins, lipids, and nucleic acids – occurs in all domains of life. In fact, glycosylation is the dominant post-translational modification of proteins, playing important roles in diverse biological processes such as protein folding and stability, cell–cell recognition, protein–protein or protein–ligand interactions, and the regulation of enzyme activity [1]. In protein glycosylation, a glycosyltransferase catalyzes the transfer of a simple sugar or an oligosaccharide from a phosphate-containing carrier molecule to either a nitrogen or an oxygen of an acceptor amino acid. The acceptor can be an asparagine or arginine in protein N-glycosylation, or a threonine, serine, or tyrosine in protein O-glycosylation. The modifying sugars can include fucose, glucose, galactose, mannose, N-acetylglucosamine (GlcNAc), and N-acetylgalactosamine (GalNAc). The formation of the glycosidic bond between these sugars and their acceptors either retains or inverts the stereochemistry of the anomeric carbon atom of the transferred sugar [2]. Protein glycosylation occurs primarily in the endoplasmic reticulum (ER) and the Golgi apparatus, where glycosyltransferases reside [3].

The Carbohydrate Active enZyme database (CAZy)(http://www.cazy.org) now lists 111 families of glycosyltransferases (GT1 – GT111), with a total of over 70,000 putative glycosyltransferases from all species. Structurally, three basic folds have been identified: GT-A, GT-B, and GT-C, although many glycosyltransferases appear not to have any of the three folds. The GT-A fold contains a Rossmann domain with several α-helices flanking a β-sheet core on both sides. The GT-A enzymes generally feature a DXD motif and a divalent metal ion such as a Mn²⁺ at the catalytic site. The GT-B fold contains two Rossmann-like domains, and the catalytic site is at the interface between the two domains. The GT-B enzymes don’t have a DXD motif and are independent of metal ions. Both GT-A and GT-B enzymes use only a nucleotide sugar as the donor substrate. The GT-C fold enzymes are membrane proteins with a dozen or so transmembrane helices and one or more soluble domains in the ER lumen or bacterial periplasm. The catalytic site of a GT-C enzyme is located between the transmembrane domain and a soluble lumenal domain. The GT-C enzymes do not use nucleotide sugar as donors; they exclusively use lipid-linked sugar donors, and they usually catalyze inverting reactions with the simple S_N2 mechanism [2].

There are only a few GT-C glycosyltransferase structures in the Protein Data Base, due largely to the difficulty in crystallizing these large membrane proteins. The advent of high-resolution cryo-EM is rapidly improving this situation [4,5]. Cryo-EM requires a smaller amount of purified sample and tolerates a higher degree of sample heterogeneity, so it is ideally suited to studying the GT-C enzymes. Recently, several cryo-EM structures have been reported for glycosyltransferases belonging to four GT families. The yeast protein O-mannosyltransferase Pmt1–Pmt2 complex is a GT39 family member [6]. The yeast and human oligosaccharyltransferase (OST) complexes belong to the GT66 family [7–10]. The yeast ALG6 catalyzes one step of the biosynthesis of a dolichol-linked oligosaccharide and belongs to the GT57 family [11]. The arabinofuranosyltransferase (AftD) and arabinosyltransferases (EmbA, EmbB and EmbC) belong to the GT53 family and are involved in mycobacterial cell-wall glycolipid synthesis [12–14]. This review focuses on the structures and our mechanistic understanding of PMT, OST, and ALG6 enzymes that are involved in protein glycosylation, and it ends with a brief description of the mycobacterial AftD and Emb enzyme structures.

Both protein N-glycosylation and protein O-mannosylation occur in the ER lumen

Secreted or membrane proteins can undergo N-glycosylation and O-mannosylation in the ER co-translationally, as mediated by two translocon-associated membrane enzyme complexes, the OST complex and the protein mannosyltransferases (Fig. 1).

Figure 1. N-glycosylation and O-mannosylation pathways in the ER of Saccharomyces cerevisiae.

Both OST and PMT complexes can directly associate with the Sec61 translocon to catalyze co-translational protein glycosylation. The OST donor substrate dolichol-PP-OS is sequentially synthesized by 14 enzymes (ALG1–14). The Pmt1/2 donor substrate dolichol-P-man is synthesized in the cytosol and flipped into the ER lumen for protein O-mannosylation. Post-translational protein glycosylation also occurs in eukaryotes but is not illustrated here for simplicity.

In protein N-glycosylation, the OST catalyzes the transfer of a preassembled 14-sugar oligosaccharide (OS; Glc₃Man₉GlcNAc₂) from the donor, dolichol diphosphate (Dol-PP-OS), to the NxS/T peptide motif in acceptor proteins [15]. The dolichol-PP-OS is synthesized from dolichol-P by sequentially adding a phosphate and fourteen sugars. Each step of sugar addition is catalyzed by a specific glycosyltransferase. The first seven sugars (two GlcNAc and five mannoses) are individually transferred from nucleotide sugar donors to dolichol-PP in the cytosolic side of the ER, catalyzed by ALG7, ALG13/14, ALG1, ALG2, ALG2, ALG11, and ALG11, respectively [16]. The dolichol-PP-Man₅GlcNAc₂ is then flipped from the cytosolic side to the luminal side by an ABC-type lipid flippase. Then, additional seven sugars (four mannoses and three glucoses) are individually transferred from dolicol-P-sugar donor in the luminal side of the ER to form the final OST donor substrate, dolichol-PP-OS, as catalyzed by ALG3, ALG9, ALG12, ALG9, ALG6, ALG8, and ALG10, respectively [16].

In protein O-mannosylation, dolichol-phosphate mannosyltransferase (DPMS) first converts dolichol-P to dolichol-P-Man at the cytosolic face of the ER membrane [17]. Then, dolichol-P-Man is flipped to the luminal face of the ER by the ATP-driven flippase in order to serve as the mannose donor for O-mannosylation, as well as for the synthesis of the OST substrate dolichol-PP-OS [18]. Finally, the protein O-mannosyltransferases (POMTs in human, PMTs in yeast) catalyze the transfer of a mannose to a serine or threonine of the protein [19].

Protein N-glycosylation is catalyzed by the eight-subunit OST membrane complex

In bacteria and archaea, the OST is a single-subunit enzyme named PglB and AglB respectively. Their crystal structures provided the first mechanistic insights into N-glycosylation [20–23]. The eukaryotic OST, however, is an eight-subunit membrane protein complex [24,25]. In Saccharomyces cerevisiae, the OST comprises Ost1, 2, 4, and 5; Stt3; Wbp1; Swp1; and either Ost3 or Ost6 [26,27]. Ost3 and Ost6 are paralogs and associate with the Sec61 and Ssh1 translocons, respectively [26]. Ost3 and Ost6 have a thioredoxin-like fold and form a mixed disulfide bridge with the nascent polypeptide substrate to facilitate N-glycosylation [28]. Stt3 is the catalytic subunit that is homologous to PglB and AglB. In contrast, mammals have two forms of the catalytic subunit, STT3A and STT3B [29], leading to the assembly of two functionally distinct OST complexes, OST-A and OST-B. Those two complexes share six subunits (ribophorin-I, ribophorin-II, OST4, OST48, DAD1, and TMEM258). OST-A catalyzes co-translational glycosylation and binds the translocon via the adaptor protein DC2 [29–31]; OST-B catalyzes post-translocational glycosylation, and it modifies the NxS/T motifs near folded protein elements or disulfide bridges. OST-B does not have DC2, but instead contains either MAGT1 or TUSC3, which are redox chaperones paralogous to the yeast Ost3 and Ost6.

Earlier structures of the eukaryotic OST were low resolution and revealed only the spherical shape of the complex [32–35]. Since the cryo-EM resolution revolution, several higher-resolution OST structures have been reported, including the yeast OST complex determined in detergent and nanodiscs at 3.5 Å and 3.3 Å, respectively; the mammalian OST-Sec61-ribosome complexes at resolutions of 3.5–5.5 Å; and, most recently, the human OST-A and OST-B complexes at 3.5 Å resolution [8] (Fig. 2A–D).

Figure 2. The cryo-EM structures of the yeast and human OSTs.

(A) Cryo-EM 3D map of the yeast OST complex (EMD-7336) in a frontside and a backside view (viewed within the membrane plane). (B) Cryo-EM 3D map of the human OST-A complex (EMD-10110). (C) Cryo-EM 3D map of the human OST-B complex (EMD-10112). (D) Cryo-EM 3D map of a mammalian OST–translocon–ribosome complex at 4.7 Å overall resolution (EMD-4317).

These structures have demonstrated the conserved architecture of all eukaryotic OST complexes [36]. The catalytic subunit Stt3 is located in the center of the complex, surrounded by the other seven subunits [37]. The eight subunits of the yeast OST can be structurally grouped into three subcomplexes: Ost1-Ost5, Ost2-Swp1-Wbp1, and Stt3-Ost4-Ost3, the interfaces of which are mediated by phospholipid molecules in the transmembrane region. Interestingly, the yeast Stt3 has a fold and substrate binding site similar to those of PglB, containing a highly conserved DDX motif that coordinates a divalent metal ion, Mn²⁺ [7,9]. Also, the peptide acceptor and sugar donor captured in the human STT3B structure are arranged similarly to those in the PglB structure.

The OST structures are consistent with an inverting S_N2 reaction mechanism. Therefore, the catalytic mechanism of N-glycosyltransferases is conserved from Bacteria and Archaea to Eukarya. The amino group of the acceptor asparagine needs to be activated in order to launch the nucleophilic attack on the anomeric carbon of the donor OS, but the detailed activation mechanism is unsettled and requires further investigation [22,23]. The structures support a direct interaction between the redox chaperone subunit and the translocon, which is Ost3 or Ost6 in the yeast OST and DC2 in the mammalian OST-A. The structural studies satisfactorily explain why OST-A functions co-translationally while OST-B only functions post-translationally. In the OST-A structure, the C-terminal region of RPN1 has a cytosolic helix bundle domain that interacts with a translating ribosome, explaining its role in co-translational glycosylation [8]. The corresponding RPN1 domain in the OST-B structure is disordered, likely due to the presence of a long cytoplasmic N-terminal segment in STT3B that seems to interfere with the folding of the helix bundle. The disordered RPN1 domain is incapable of interacting with the ribosome, leading to the OST-B function in post-translational glycosylation [8].

Protein O-mannosylation in yeast is mediated by the Pmt1–Pmt2 membrane complex

Protein O-mannosylation is essential for unfolded protein O-mannosylation and protein quality control [38]. As mentioned above, protein O-mannosyltransferases (PMTs) are located in the ER membrane and transfer a mannose from the donor substrate dolichol-P-Man to a serine or threonine residue at the lumenal side of the acceptor proteins [19]. All PMT enzymes characterized so far assemble as a dimer. In yeast, there are two major PMT complexes: the Pmt1–Pmt2 heterodimer and the Pmt4–Pmt4 homodimer. In animals, however, there is only one PMT complex, the POMT1–POMT2 dimer [19,39]. Recently, the atomic-resolution cryo-EM structures of the yeast Pmt1–Pmt2 heterodimer in complex with a peptide acceptor and a lipid-linked sugar donor were reported (Fig. 3A–B) [6].

Figure 3. Cryo-EM structure of the yeast Pmt1-Pmt2 complex.

(A) Cryo-EM 3D map of yeast Pmt1–Pmt2 complex (EMD-20236) viewed from the side (left, from within the membrane plane) and from the top (right, from ER lumen). (B) The Pmt1–Pmt2 structure (PDB code 6P25) is shown in cartoons; colors are the same as in panel A.

Pmt1 and Pmt2 each contain 11 transmembrane helices (TMHs) and a luminal β-trefoil fold termed the MIR domain that is found between TMH7 and TMH8. Both the transmembrane regions and the MIR domains of Pmt1 and Pmt2 are highly similar [6]. The interface between the Pmt1 and Pmt2 subunits is confined to the cytosolic and luminal regions. This unique interface results in a large cavity between the transmembrane regions of Pmt1 and Pmt2. This cavity appears to allow diffusion of the lipid-linked sugar donor into enzyme active site in each subunit [6].

The dimer interface on the luminal side is asymmetric. Only the MIR domain of Pmt1 interacts with Pmt2; the MIR domain of Pmt2 is disordered, forming an open gate on that side [6]. The substrate binding in the Pmt1–Pmt2 complex structure is also asymmetric, with the acceptor peptide binding to Pmt2 and the donor dolichol-P (minus the mannose) being associated with Pmt1. Therefore, both Pmt1 and Pmt2 are functional transferases, and they likely function alternatively. In addition, the Pmt1/2 structures align reasonably well with the Stt3 structures in the transmembrane domains, despite Stt3 having two additional TMHs [6], which reveals a close evolutionary relationship between protein N-glycosylation and protein O-mannosylation.

The acceptor peptide and sugar donor in the catalytic sites of the yeast Pmt1–Pmt2 complex are coordinated by conserved residues and they align well with the donor and acceptor of the bacterial oligosaccharyltransferase PglB [6]. There are no divalent metal ions in the yeast Pmt1–Pmt2 structure, and it is unclear whether the enzyme requires a metal ion. However, a highly conserved DE motif is present in both Pmt1 and Pmt2 at the N-terminal end of the external loop horizontal helix 1 (EL1-HH1). The conserved aspartate (D77 in Pmt1 and D92 in Pmt2) likely participates in activating the hydroxyl group of Ser/Thr in the acceptor peptide. This observation suggests that the PMT enzymes catalyzing protein O-mannosylation are conserved with the N-glycosylation enzymes not only in their overall architecture, but also in their catalytic mechanisms.

ALG6 is involved in synthesis of the donor substrate of the OST complex

ALG6 transfers the first of three glucose moieties onto the preassembled OS precursor glycan, Man₉GlcNAc₂ (Fig. 1). Recently, the cryo-EM structures of ALG6 alone and in complex with a sugar donor analogue were determined at a resolution of 3.0 Å and 3.9 Å, respectively [11] (Fig. 4A–C). The mass of ALG6, at about 60 kDa, is small and difficult for cryo-EM. To overcome this problem, a conformation binder Fab fragment was developed and used to form an immunocomplex with ALG6. The increased mass facilitated particle alignment and 3D reconstruction of the cryo-EM images.

Figure 4. Cryo-EM structure of the yeast ALG6 and its structural alignment with other GT-C glycosyltransferases.

(A) Cryo-EM 3D map of the yeast ALG6-Fab complex at 3.0 Å resolution (EMD-10258). (B) Cryo-EM 3D map of the yeast ALG6-Fab complex bound to a sugar donor at 3.9 Å average resolution (EMD-10257). (C) A side (left) and a top view (right) of the ALG6 structure (PDB code 6SNH), shown in cartoons and colored in rainbow from N-terminal blue to C-terminal red. (D) A side view (left) and a lumenal view (right) of the structural alignment of ALG6 (PDB code 6SNH) with yeast Stt3 (PDB code 6C26), human STT3A (PDB code 6S7O), STT3B (PDB code 6S7T), and yeast Pmt1 and Pmt2 (PDB code 6P25). The names of enzymes whose structures contain a sugar donor are underlined.

ALG6 contains 14 transmembrane helices and two insertion loops in the ER lumen (EL1 and EL4) (Fig. 4C). The overall fold of ALG6 is again similar to those of Stt3s and PMTs, especially for the first seven TMHs (including the luminal helices EL1-h1 and EL1-h2) and TMH9–10 (Fig. 4D). These regions form the typical, conserved GT-C glycosyltransferase fold. The ALG6 structure contains neither a divalent metal ion nor the DXD motif. However, the donor sugar binding site of ALG6 is almost identical to those of PMTs and Stt3. The donor substrate resides in a membrane-exposed groove formed by TMH6, TMH7, and TMH8. The dolichylphosphate-linked sugar moiety is not present in the ALG6 structure; its position may be on the opposite side of the donor binding pocket in ALG6. Five acidic residues line the substrate binding pocket, but only Asp69 is essential. This residue is likely the general base that activates the C3 hydroxyl group of the terminal A branch mannose of the acceptor dolichol-PP-GlcNAc₂Man₉, leading to the nucleophilic attack on the anomeric C1 carbon of the donor glucose. Therefore, the ALG6 structure underscores the idea that either an aspartate or a glutamate at the N-terminal end of EL1-h1 may be a key catalytic feature in all GT-C enzymes.

The GT-C glycosyltransferases are involved in mycobacterial cell-wall synthesis

The mycobacterial cell wall is composed of covalently linked three-tiered mycolic acid–arabinogalactan–peptidoglycan complexes. The mycobacterial embCAB operon encodes three membrane-embedded arabinosyltransferases, EmbA, EmbB, and EmbC, that transfer an arabinose from the donor decaprenyl-phosphate-arabinose to an acceptor arabinogalactan or lipoarabinomannan. These proteins assemble either a heterodimer, EmbA–EmbB, or two homodimers, EmbB₂ and EmbC₂, with each enzyme further associating with an acyl carrier protein AcpM. These enzymes are targeted by the antibiotic ethambutol, which is used to treat TB. Cryo-EM structures have recently been reported for the EmbA–EmbB–AcpM₂, EmbB₂–AcpM₂, and EmbC₂–AcpM₂ complexes [12,40], as well as a monomeric EmbB [14], confirming the overall GT-C fold for all three enzymes. EmbA, B, and C have similar structures, each containing 15 TMHs and two periplasmic domains. Both periplasmic domains contain a carbohydrate-binding jellyroll fold. The highly conserved catalytic DDX motif precedes a short periplasmic horizontal helix (HH1), similar to the OST and PMT structures. AcpM is a small helix bundle that binds to the cytoplasmic surface of each glycosyltransferase. Ethambutol can occupy the enzyme active site, mimicking the two native substrate arabinoses, one from the donor substrate and the other from the acceptor substrate.

The mycobacterial arabinofuranosyltransferase AftD transfers an arabinofuranose from the donor decaprenyl-phosphate-arabinofuranose to either arabinogalactan or lipoarabinomannan. The cryo-EM structure of a mycobacterial AftD showed that AcpM binds to the cytoplasmic surface of the transferase and that it regulates the enzyme’s function [13]. AftD contains 16 TMHs (with the first 11 forming the GT-C fold) and a large periplasmic soluble module composed of five domains, including three carbohydrate-binding domains. The catalytic site is located between the transmembrane domain and the soluble module and is lined with five charged, highly conserved residues. Although AftD, EmbA-C, and the earlier reported aminoarabinose transferse ArnT [41] all transfer sugars to acceptor substrates other than a protein (e.g., a lipid or a glycan), their core structures and active sites nevertheless resemble those of protein glycosyltransferases such as PMT and OST. This observation suggests that all GT-C enzymes may share a similar core structure and catalytic pocket.

Perspective

The advent of modern cryo-EM has greatly enriched our understanding of the GT-C family of protein glycosyltransferases. The first batch of structures revealed the general conservation of the GT-C fold and the basic catalytic mechanism, yet it is clear that individual enzymes have evolved unique structural features to accommodate their respective donor and acceptor substrates, which can vary a lot in size and properties. Because the traditionally difficult membrane enzymes are no longer difficult for cryo-EM, we expect to see many more cryo-EM structures of GT-C glycosyltransferases in the near future. For example, it will be very interesting to obtain structures of the recently identified metazoan-specific O-mannosyltransferases, TMTC1–4; the DPY19 enzymes that catalyze protein C-mannosylation; and the ALG3/8/9/10/12 proteins that synthesize OS precursors for N-glycosylation.

Even more exciting is the second wave of the so-called resolution revolution that came quietly in the midst of the Covid-19 pandemic, promising true atomic resolution by cryo-EM [42–44]. The structures of large membrane proteins such as the GABA receptor can now be obtained at better than 2-Å resolution. This unprecedented resolution capability will help to address several key issues about the GT-C enzymes, including their detailed reaction mechanism and how the nascent substrate polypeptide is directed to the catalytic site and then funneled away after glycosylation. The new capability may also lead to atomic-resolution structures of mycobacterial GT-C enzymes such as AftD, EmbA, EmbB, and EmbC, which would facilitate the development of small-molecule inhibitors targeting these enzymes.

Highlights.

• The GT-C glycosyltransferases are membrane-associated enzyme complexes that use dolichol-linked donor substrates.

• Eukaryotic OST complexes mediate protein N-glycosylation.

• The yeast Pmt1–Pmt2 complex catalyzes protein O-mannosylation.

• The yeast Alg6 is involved in the synthesis of oligosaccharide donor substrates for OST.

• The glycosyltransferases EmbA-C and AftD are involved in mycobacterial cell-wall synthesis.