Keywords

1 Introduction

Asparagine-linked (N-linked) glycosylation is an essential protein modification, affecting a number of basic cellular processes such as protein folding, its half-life, trafficking and immunogenicity as well as its interactions between cells, cells and extracellular matrix components or pathogens (Varki and Gagneux 2017). In all eukaryotic cells, N-glycans are synthesized in two specialized organelles, the endoplasmic reticulum (ER) and the Golgi apparatus. Together, these organelles harbor dozens of functionally distinct glycosyltransferases and glycosidases that sequentially modify the growing oligosaccharide chain (Kornfeld and Kornfeld 1985; Dunphy 1985; Spiro 2002; Rabouille et al. 1995). Yet, it is much less clear how this sequence of enzymatic reactions is orchestrated to guarantee faithful synthesis of N-glycans, considering that enzymes do not use any template, can compete with each other for the same substrate and/or the acceptor, and even localize in the same Golgi compartment. Another puzzling issue is an intrinsic microheterogeneity of glycans made by the cell. For example, an N-glycan attached to a specific asparagine of a given protein can be different from an N-glycan attached to the same site in another protein molecule. Distinguishing this “background noise” from dynamic changes that are functionally important, e.g., during embryonic development, cell differentiation, and aging can sometimes be problematic. Nevertheless, unlike other polymerization events in the cell, glycosylation apparently need not be a high-fidelity system, since cells normally tolerate such microheterogeneity without facing problems in cell survival or proliferation.

The other side of the coin is that this variation can sometimes lead to a devastating disease. Congenital disorders of glycosylation (CDGs) are a rare, yet diverse group of serious, often multiorgan diseases characterized by defects in glycosylation (Freeze et al. 2014; Francisco et al. 2020). More than 140 different CDG syndromes are known as of today, the severity of which varies from prenatal death to survival into adulthood with a relatively normal life span. Disturbed N-glycosylation forms the largest group of the CDGs. It is divided into two groups (Type I and Type II) based primarily on the genetic defect and the site it is affecting. Type I CDGs are characterized by defects in the synthesis of N-glycans in the endoplasmic reticulum (ER), while the Type II CDGs have problems in their processing in the Golgi apparatus. In addition to CDGs, glycosylation changes play an important role in many other human diseases including autoimmune diseases, inflammation, tumorigenesis, and its progression (Reily et al. 2019). Yet, the underlying mechanistic details that cause these changes are incompletely understood, as are also the reasons why some changes lead to disease and some do not. Partly, this is due to the dynamic and variable nature of the glycan themselves, their cell- and tissue-specific expression (Medzihradszky et al. 2015) as well as the lack of tools that would allow glycan editing at will in a specified glycosylation site or protein itself.

2 Biosynthesis of N-Glycans in the Endoplasmic Reticulum

2.1 Building Blocks for N-Glycan Synthesis

The early steps in N-glycan biosynthesis in the endoplasmic reticulum (ER) are conserved in all three domains of life (Dell et al. 2010), whereas their processing and maturation differ markedly. All N-glycans share a common core structure (asn-GlcNAc2Man3-) which is further elongated in a species- and tissue-specific manner (Medzihradszky et al. 2015) by adding a few other subterminal or terminal sugar residues to the core structure. Depending on the sugar residue and the linkage type used, these additions can significantly influence the structure of the N-glycan (Medzihradszky et al. 2015). The main sugar residues utilized as building blocks are N-acetylglucosamine (GlcNAc), mannose (Man), galactose (Gal), fucose (Fuc), and sialic acid (N-acetylneuraminic acid (Neu5Ac) being the predominant form), of which the latter two act as chain-capping residues. In some instances, N-acetylgalactosamine (GalNAc) residues can be used to construct an N-glycan. Glucose (Glc) residues are also temporarily incorporated into the growing N-glycan during its synthesis in the ER, yet they are invariably removed as glucose residues have not been detected in a mature N-glycan isolated from cultured cells or tissues (Zuber et al. 2000). Occasionally, mature N-glycans can also be modified by the addition of sulfate or phosphate, generating determinants that modulate cell adhesion or glycoprotein localization in the cells. Interestingly, despite the deletion of the CMAH gene (needed for N-glycolylneuraminic acid (Neu5Gc) synthesis) 3 million years ago, this neuraminic acid variant is still regularly detected in trace amounts in human glycans (Angata and Varki 2002). This is due to dietary consumption of Neu5Gc-containing animal products (e.g., red meat and dairy products) and its incorporation into newly synthesized glycans (Banda et al. 2012). Perhaps unsurprisingly, the highest Neu5Gc levels are detected in epithelial and endothelial cells that line the intestine and blood (and lymph) vessels, respectively.

2.2 Precursor Synthesis and Its Attachment to Nascent Polypeptide Chains

The N-glycosylation of nascent polypeptides in the ER lumen relies on the prior assembly of a lipid-linked oligosaccharide (LLO) precursor (Glc3Man9-GlcNAc2) onto a membrane-embedded dolichyl phosphate (Dol-P) carrier (Fig. 7.1). This set of events is orchestrated by the Alg-family of ER-localized, membrane-associated glycosyltransferases (Kelleher et al. 2007). They stepwise assemble the LLO using nucleotide sugars (UDP-GlcNAc, GDP-Man, Dol-P-Man, and Dol-P-Glc) as donor substrates. The LLO assembly begins on the cytoplasmic face of the ER membrane by the formation of a GlcNAc2-PP-Dol intermediate from GlcNAc-1-phosphate and GlcNAc. These additions are catalyzed by Dpagt1 (Alg7) and Alg13p/Alg14p UDP-GlcNAc-transferases, respectively. The three enzymes exist as hexamers with a stoichiometry of 2:2:2 (Noffz et al. 2009). Alg14 appears to be the central unit, capable of recruiting other enzymes to the complex (Lu et al. 2012). Next, ER mannosyltransferases (Alg1, Alg2, and Alg11) that also form complexes with each other (Gao 2004) add five mannose residues from GDP-Man donors to form a Man5GlcNAc2-PP-Dol intermediate. Thus, LLO precursor synthesis on the cytoplasmic face of the ER membrane involves three main enzyme complexes, one formed by Dpagt1/Alg13/Alg14 and the other two either by Alg1/Alg2 and Alg1/Alg11. This arrangement likely ensures that each mannose residue will be linked correctly to the precursor despite the coexistence of several competing enzymes on the same membrane.

Fig. 7.1
figure 1

A schematic representation of the N-glycan biosynthetic pathway in the ER and the Golgi apparatus. The figure shows the gradual maturation of an N-glycan and the various steps involved. For more details, please see the text

Full size image

The next step involves translocation of the Man5GlcNAc2-PP-Dol intermediate into the ER lumen, a process that is thought to be mediated by a protein termed as the Rft1, but it is still uncertain whether it acts as a bonafide flippase protein (Helenius et al. 2002). In the ER lumen, mannosyltransferases (Alg3/Alg9/Alg12) and glucosyltransferases (Alg6/Alg8/Alg10) further elongate the LLO precursor by attaching four additional mannose residues and three glucose residues, respectively. This completes the precursor synthesis and yields the Glc3Man9GlcNAc2-PP-Dol structure, which will be used later as the donor substrate for en bloc transfer of an N-glycan to a suitable polypeptide chain. It is noteworthy that unlike the initial catalytic steps on the cytosolic face of ER, the completion of the LLO precursor synthesis in the ER lumen does not use nucleotide sugars as donors. Rather, membrane-embedded Dol-P-Man and Dol-P-Glc are used as sugar donors in this case. Their synthesis takes place also on the cytoplasmic side of the ER membrane (from GDP-Man and UDP-Glc, respectively) before they are translocated (flipped) to the luminal side (Helenius et al. 2002).

The most preferred acceptor asparagine residues for N-glycosylation are the ones within the Asn-X-Ser/Thr motif (where X ≠ proline) (Zielinska et al. 2010). Of these two, the Asn-X-Thr sequon is preferred over Asn-X-Ser, mainly because the interaction between the side chain methyl group of threonine and the asparagine-lysine (NK) motif in the binding pocket of the oligosaccharyltransferase (OST) increases the stability of the complex (Kasturi et al. 1995, 1997; Medus et al. 2017). The identity of the amino acid X and flanking amino acids also contribute to the glycosylation of a given sequon. In addition, the position of the sequon within the polypeptide, the secondary and tertiary structure of the protein, and its final destination in a cell can impair or enhance the likelihood of whether that site becomes glycosylated or not (Rao and Wollenweber 2010). Thus, the presence of sequons alone cannot be used as an adequate predictor of N-glycosylation. Indeed, roughly one-third of the identified sequons in secreted glycoproteins remain non-glycosylated (Schulz 2012).

The transfer of the completed precursor oligosaccharide is catalyzed by ER membrane-localized OST complex. It is an octamer consisting of a single catalytic subunit and seven accessory subunits, each important for optimal glycosylation efficiency. Most multicellular animals (sponges are an exception) possess two such complexes due to an ancient duplication of the gene encoding the catalytic subunit. The STT3A and STT3B (for OST-A and OST-B complexes, respectively) have different kinetic properties, acceptor substrate preferences, and partially non-overlapping roles in glycosylation (Shrimal et al. 2013a). The accessory subunit compositions between the two complexes also differ. OST-A complex associates with Sec61 core components of the ER translocon complex and co-translationally glycosylates the nascent polypeptide in accessible sequons during polypeptide chain translocation into the ER lumen. Sequons within the last ~50–55 residues of the C-terminus are, however, inside the translocon and hence inaccessible for STT3A. Instead, STT3B transferase in OST-B complexes can posttranslationally add N-glycans to such sequons. It also can use internal sequons that are skipped by the STT3A as acceptors (Lu et al. 2018; Shrimal et al. 2013b). Often, these include closely spaced sequons adjacent to signal cleavage site or sequons with cysteine residues nearby or inside the motif (i.e., the N-C-T/S motif (Shrimal et al. 2013a).

2.3 N-Glycan Processing in the ER and Quality Control

The newly attached Glc3Man9GlcNAc2 N-glycan structure is further modified once the polypeptide is translocated to the ER lumen and begins to fold. The first step involves the removal of the terminal glucose residue by a transmembrane enzyme α-glucosidase I. The second glucose is then rapidly removed by the soluble α-glucosidase II (Grinna and Robbins 1979; Janssen et al. 2010) The resulting mono-glucosylated glycan is a preferred ligand for the carbohydrate-recognizing molecular chaperones calnexin and calreticulin. These chaperones readily associate also with the protein ERp57 (Ruddock and Molinari 2006), a disulfide isomerase that catalyzes the formation of inter- and intramolecular disulfide bonds, thereby helping proper folding of the nascent glycoprotein. Calnexin binding appears to happen irrespective of the folding state of the glycoprotein (Zapun et al. 1997), suggesting that it most likely interacts with the nascent polypeptide as soon as it arrives in the ER lumen. During the folding process (Fig. 7.2), the last glucose residue is removed by the α-glucosidase II. Unfolded or misfolded proteins display exposed hydrophobic patches that are recognized by UDP-glucose-glycoprotein glucosyl-transferase (UGGT) (Caramelo et al. 2003), an enzyme that can glucosylate the same mannose residue again in that N-glycan. By doing so, it recreates the Glc1Man9GlcNAc2 structure that is again acted upon by the chaperone-disulfide isomerase complex. This removal and re-addition of glucose residues can continue for several cycles until the protein is properly folded. Once this is achieved, the glycan is finally trimmed by ER mannosidase I (ERMan1) that removes the terminal mannose residue from the middle branch of the N-linked oligosaccharide. The resulting Man8GlcNAc2 structure then can be recognized by the ERGIC-53 (LMAN1), a mannose-specific lectin of the LMAN1/MCFD2 cargo receptor complex (Zheng et al. 2010), thereby facilitating packaging and transport of the native N-glycosylated glycoprotein into COPII (Coat Protein complex II)-coated vesicular carriers that ferry cargo from the ER to the Golgi via the ER-Golgi intermediate compartment (ERGIC) (Hanna et al. 2018; Peotter et al. 2019).

Fig. 7.2
figure 2

A glycan-based quality control system in the ER that distinguishes correctly folded glycoproteins from unfolded or misfolded ones. CNX calnexin, CRT calreticulin, G-I-II α-glucosidases I and II

Full size image

An ER stress caused by various factors (e.g., altered calcium homeostasis, redox state and glucose deprivation or mutations) is characterized by accumulation of misfolded or unassembled proteins in the ER and can be detrimental to cell viability. Metazoan cells can, however, cope with this stress by launching an ER stress response that suppresses the rate of translation and increases the expression of molecular chaperones to ease protein folding in the ER. In acase these maneuvers still fail, terminally misfolded glycoproteins will be directed to degradation via an ER-associated degradation (ERAD) pathway (Benyair et al. 2015). It starts when ERMan1 mannosidase and EDEM (ER degradation enhancing mannosidase-like) proteins (EDEM1/2/3 in mammals) are recruited to cleave off two mannose residues (instead of one) from an N-glycan. As a result, the glycan becomes unrecognizable by ERGIC-53, thereby preventing glycoprotein transport to later secretory compartments. The exposed α(1,6)-linked mannose in the Man7GlcNAc2 structure is now recognized by the OS-9/XTP3-B lectin complex that directs the bound glycoprotein to the transient ERAD (ER-associated degradation) protein complex at the ER membrane with ubiquitin ligase activity. ERAD complex then ensures that the glycoprotein is returned back to the cytoplasmic side of the ER membrane by tagging it for proteasomal degradation through ubiquitination (Benyair et al. 2015).

The other route for degrading of misfolded glycoproteins relies on malectin, a membrane-associated, ER stress-induced lectin first identified in 2008 (Schallus et al. 2008). It is highly conserved in metazoans (Yang et al. 2018) and it shows high specificity toward di-glucosylated N-glycans (Glc2Man9GlcNAc2) (Schallus et al. 2008, 2010). Malectin forms a stable complex with ribophorin I, a subunit of the OST complex with proposed chaperone activity based on its ability to recognize misfolded protein backbones (Qin et al. 2012; Galli et al. 2011) This complex seems to act as an early intervention mechanism for detecting and capturing nascent non-native N-glycoproteins before it delivers them to proteasomal degradation if initial attempts to fold will fail (Stanley 2016). Whether this malectin-ribophorin I-mediated removal mechanism involves a unique retro-translocation machinery different from the ERAD machinery is currently unclear.

3 N-Glycan Processing in the Golgi Apparatus

Correctly folded glycoproteins entering cis-Golgi compartment carry typically an N-glycan with eight mannose residues left (Man8GlcNAc2). While some N-glycans may exit the Golgi without being modified, their proportion is normally low in humans (Lee et al. 2014). Partly, this is due to a presence of a quality control mechanism that is present in the Golgi. Golgi membranes harbor a mannose-binding lectin VIP36, that can recycle high mannose type N-glycans back to the ER (Lee et al. 2014). In support of this, VIP36 also interacts with the ER-localized BiP chaperone (Nawa et al. 2007). By doing so, VIP36 can halt the secretion of improperly glycosylated glycoproteins to post-Golgi compartments. Another mechanism to prevent high mannose type N-glycoproteins from passing through the Golgi takes over when a mono-glycosylated N-glycan (Glc1Man9GlcNAc2) carrying glycoprotein arrives in the Golgi. The glycan part is cleaved internally by the Golgi endo-α-mannosidase between the two Man residues of the Glcα1–3Manα1–2Manα1–2 moiety, thereby yielding a Man8GlcNAc2 isomer that is different from that produced by ERMan1 in the ER (Thompson et al. 2012). Interestingly, experimental evidence also suggests that the calreticulin-based glycoprotein quality control may be functional also in the Golgi compartment, as calreticulin was found to co-localize with endo-α-mannosidase in the ERGIC and cis/medial-Golgi compartment at least in cultured rat liver cells (Zuber et al. 2000).

Normally, the vast majority of N-glycans are processed in the Golgi to complex and/or hybrid type N-glycans by a distinct set of glycosidases and N-acetylglucosaminyltransferases, also termed as MGAT1–5 (Kellokumpu et al. 2016; Khoder-Agha et al. 2019a). The processing involves complex mutual interplay between the MGAT homomers and heteromers, mannosidase II (ManII) acting as a central hub (Khoder-Agha et al. 2019a). Thus, upon arriving in the Golgi, ER-derived MGAT homomers form heteromeric complexes not only with other MGATs but also with relevant UDP-N-acetylglucosamine transporters. Thereby, they organize into multienzyme/multi-transporter assemblies in the Golgi membranes. Their interplay likely involves either distinct or dynamic complexes (Khoder-Agha et al. 2019a) to facilitate efficient processing and branching of N-glycans in the cis- and medial-Golgi.

The processing begins in the cis-Golgi by the removal of three mannose residues to yield the Man5GlcNAc2 structure. Golgi mannosidases IA-C are responsible for the cleavages. Then, the first GlcNAc is added by MGAT1, using nucleotide-activated N-acetylglucosamine as a donor substrate. Once this GlcNAc is added, two additional mannose residues are removed by the Golgi α-mannosidase II. This creates a scaffold for MGAT2 to add a second GlcNAc to the exposed mannose residue, yielding a precursor for all complex-type N-glycans. MGAT4 and MGAT5 can then initiate the synthesis of third and fourth GlcNAc branches, respectively. Alternatively, MGAT3 can add a bisecting GlcNAc at the tri-mannosyl core structure (Fig. 7.3, middle). If this bisecting GlcNAc is added before MGAT4 and MGAT5 have added theirs, the synthesis of the third and fourth GlcNAc branches by MGAT4 and MGAT5 is halted (Kizuka and Taniguchi 2018). Bisecting GlcNAc also cannot be further elongated with any other sugar residue. Its addition also significantly alters the conformation of an N-glycan and suppresses the addition of terminal sugar residues such as sialic acid and fucose. The human natural killer-1 epitope (HSO3-3GlcAβ1-3Galβ1-4GlcNAc), a sulfated trisaccharide structure that is extensively expressed in the nervous system, is another terminal epitope suppressed by the bisecting GlcNAc (Nakano et al. 2019). Bisecting GlcNAc is also known to inhibit α-mannosidase II, suggesting that this addition may be one reason for the synthesis of hybrid type N-glycans. Yet, not all hybrid-type N-glycans contain a bisecting GlcNAc. It is, therefore, possible that rapid elongation of the first GlcNAc branch by galactose can also inhibit the necessary removal of two terminal mannoses by α-mannosidase II and thus, the build-up of the other GlcNAc branches.

Fig. 7.3
figure 3

Three examples depicting the main N-glycan types present on the cell surface in higher eukaryotes. High-mannose type N-glycan is characterized by having not undergone any processing in the Golgi compartment. Hybrid-type N-glycan typically has only one branch that has been processed in the Golgi. Bulky complex-type N-glycans in turn have two to five branches that are made and terminated in the medial and trans-Golgi cisternae

Full size image

Normally, the GlcNAc branches are further elongated by adding galactose, N-acetylglucosamine, and sialic acid. Galactose is added to the GlcNAC nearly always with the β(1,4)-linkage. This structure, termed N-acetyllactosamine (LacNAc), can be repeated several times in one branch, forming poly-LacNAc structures. Poly-LacNAc motifs in turn can act as substrates for making additional branches to the antennae. This is done by a set of special enzymes called GCNT2s A-C (Dimitroff 2019) by adding extra GlcNAc residues with the β(1,6)- linkage to internal galactose residues. These GlcNAc residues can also be subsequently elongated by β(1,4)-galactosyltransferases to form additional LacNAc structures. This kind of branched N-glycan is termed an I-branched glycan. They are most frequently found in adult erythrocytes, mucosal epithelia, and cells of the eye and olfactory bulb (Dimitroff 2019). GalNAc residues are also occasionally found in N-glycans of mammals forming LacdiNAc (GalNAcβ(1,4)GlcNAc) type structures.

The antennae are often capped with sialic acid by various sialyltransferases. This blocks further elongation of the branches except in the case of polysialylation. Polysialylated N-glycans are commonly detected in neural cell adhesion molecules (NCAMs) of the nervous system (Kiss and Rougon 1997). Fucose is another residue that cannot be elongated further. It can be added by specific Golgi fucosyltransferases either to the asparagine-linked GlcNAc to produce the “core fucosylated” N-glycan, or to GlcNAc residues of the antennae.

In specific cases, sugar residues of the antennae can undergo further modifications such as sulphation, phosphorylation, and O-acetylation (Klein and Roussel 1998; Wang et al. 2017). For example, lysosomal acid hydrolases carry N-glycans with a phosphate that directs the enzymes to lysosomes. Lysosomal enzymes share common conformational lysine-containing motifs that are recognized by the cis-Golgi-localized GlcNAc-1-phosphotransferase enzyme. In the first catalytic step, GlcNAc-1-phosphotransferase transfers GlcNAc-1-P from UDP-GlcNAc to the C6 hydroxyl group of selected mannose residue present in the high mannose-type N-glycan (Oh 2015; Qian et al. 2010). In the second step, N-acetylglucosamine-1-phosphodiester α-N-acetyl-glucosaminidase (NAGPA) cleaves the GlcNAc, leaving only the phosphate group linked to the mannose. Man-6-phosphate (M6P) tag is the ligand for transmembrane Man-6-P receptors (MPRs) residing in the trans-Golgi network (TGN). Once recognized by the MPR, the receptor escorts the lysosomal hydrolase with its ligand to endosomes and eventually to lysosomes in clathrin-coated vesicles. In lysosomes, the enzyme is released at low pH and the receptor is recycled back to the trans-Golgi. Mutations that impair tagging of mannose with phosphate lead to lysosomal storage diseases, a group of over 70 rare diseases characterized by accumulation of macromolecules in lysosomes (Xu et al. 2016).

3.1 N-Glycosylation of Immunoglobulins

N-glycosylation is also an important modification of all immunoglobulin isotypes and contributes affecting their binding characteristics and effector functions. Although their synthesis might not be different in any way from other N-glycans, there are some special issues that are worth discussing. N-glycans attached to immunoglobulin G (IgG) are best characterized owing to IgG abundance in the serum and successful production of many IgG-based therapeutic antibodies by the biopharma industry. IgG N-glycans are typically found in the Fc region but a minor proportion (15–25%) of serum IgG can contain N-glycans within their variable domains. These so-called “Fab glycans” differ from the Fc region N-glycans by having a higher proportion of terminally galactosylated and sialylated N-glycans with a bisecting GlcNAc, while having a lower abundance of core-fucosylated N-glycans (van de Bovenkamp et al. 2016). Yet, it is not clear why the number of antennae in IgG N-glycans seems to be limited to only two antennae (or three if the bisecting GlcNAc is considered also as an own branch). One possibility that may explain this is that antibody-producing plasma cells do not express the MGAT4 or MGAT5 enzymes needed for further branching. Another explanation could be that the addition of bisecting GlcNAc (or some other regulatory system) will prevent further branching of IgG N-glycans. The existence of such a system would be logical, given that an increase in N-glycan “bulkiness” brought about by additional branching might interfere with the folding and pairing of the Fc regions in the ER, and thereby alter its conformation known to be important for its binding to Fc receptors and antibody effector functions. Similarly, it is unclear why the Fab N-glycans display a higher proportion of more mature (more completely processed) N-glycans than those of Fc N-glycans. Whether this difference stems from better accessibility of the Fab glycans over Fc glycans, or something else such as increased extracellular glycosylation or decreased degradation of glycosidases, remains to be explored.

4 Golgi Microenvironment Is Important for Normal Processing and Maturation of N-Glycans

Despite the rather homogenous nature of high mannose type N-glycans arriving in the Golgi, N-glycans leaving the Golgi are much less so. For example, a single glycoprotein can carry complex, hybrid, and high-mannose N-glycans on the same polypeptide. Hybrid and complex type N-glycans can also display a variable number of antennae in their structure that may, or may not, carry sialic acid and/or fucose. While we do not have a clear picture at the molecular level of what determines the outcome in each case, this heterogeneity reflects both protein- and cell-specific processing of N-glycans brought about for example by epigenetic changes that determine what enzymes are expressed by the cell. Other factors that also modulate N-glycan biosynthesis are discussed below.

4.1 Golgi pH Homeostasis

The environmental cues outside or inside the cells can also contribute to N-glycan diversity. Unlike the ER, the other secretory pathway compartments, including the ER-Golgi intermediate compartment (ERGIC), the Golgi apparatus itself, and secretory vesicles, have uniquely acidic lumens, their pH decreasing along the pathway toward the plasma membrane (Paroutis et al. 2004). This pH gradient is crucial for their efficient functioning in membrane trafficking, glycosylation, proteolysis, protein sorting, or cargo transport (Kellokumpu 2019). Altered glycosylation due to abnormal Golgi pH is also responsible for several human disorders identified recently (Khosrowabadi and Kellokumpu 2020). Proper pH in the Golgi lumen appears to be important especially for the activity and assembly of the glycosyltransferase complexes in the Golgi. Previously, we have shown that the trans-Golgi β(1,4)GalT1 galactosyltransferase not only forms homomers in the ER but also heteromers with either ST3Gal3 or ST6Gal1 sialyltransferases upon its arrival in the Golgi (Hassinen et al. 2011; Hassinen and Kellokumpu 2014). Interestingly, these two heteromeric complexes assemble only in the acidic pH of the Golgi lumen (pH < 6.5). In the former case, complex formation could be prevented by increasing Golgi pH only by 0.2 pH units (Hassinen et al. 2011; Hassinen and Kellokumpu 2014). This increase was also sufficient to redirect the ST3Gal3 enzyme from the Golgi to post-Golgi compartments, consistent with their oligomerization-driven retention in the organelle (Rivinoja et al. 2009). The loss of the enzyme heteromers and enzyme mislocalization also coincided with reduced α(2,3)-sialylation and increased α(2,6)-sialylation of carcinoembryonic antigen (CEA) (Rivinoja et al. 2009). A similar decrease and increase in α(2,3)- and α(2,6)-sialylation, respectively, in CEA N-glycans has also been observed in cancer tissues in vivo (Kobata et al. 1995). Since Golgi resting pH is often elevated in cancer cells (Rivinoja et al. 2006), these findings suggest that Golgi resting pH may be used to regulate what linkage type will be used to link sialic acids to an N-glycan. This kind of switch from one linkage type to another one can have dramatic effects on cell behavior. For example, increased expression of α(2,6)-linked sialic acid in N-glycans can inhibit tumor cell apoptosis and activate growth factor pathways (Francisco et al. 2020; references therein).

Interestingly, the formation of the β(1,4)GalT1/ST6Gal heteromer was shown to increase markedly the catalytic activity of the β(1,4)GalT1 perhaps via substrate channeling (Hassinen et al. 2011). Alternatively, heteromer formation may also increase the accessibility of the donor or acceptor substrates to the active site of the β(1,4)GalT1, even though it is not directly involved in homodimer formation (Harrus et al. 2018). Yet, the active site is more exposed in the β(1,4)GalT1/ST6Gal heterodimers than it is in homodimers (Khoder-Agha et al. 2019b). In addition to 3D structures, this view is supported by the observation that a single mutation in the active site (H243) was able to abolish homodimer formation but not heterodimer formation.

Acidic Golgi resting pH is also needed the keep certain glycosyltransferases active. Accordingly, Golgi acidity (pH < 6.5) is essential for the full catalytic activity of ST6Gal1 sialyltransferase but not for β(1,4)GalT1, nor the MGATs (Hassinen et al. 2011). Partly, this can be explained by the pH-sensitive interactions between the β(1,4)GalT1 and the two sialyltransferases acting on N-glycans (Hassinen and Kellokumpu 2014). Yet, it is likely that pH-dependent conformational changes in the tertiary structure of ST6Gal1 also contribute to the activity loss if the Golgi resting pH is close to neutral. Collectively, these data suggest that the main role of the decreasing pH gradient from the cis-to-trans side of the Golgi compartments (pH 6.7–pH 6.3) is to orchestrate mutual interactions between glycosyltransferases, to promote their active conformation, and to get them correctly localized, in accord with their suggested oligomerization-mediated retention in the Golgi (Nilsson et al. 2009).

4.2 Golgi Ion Homeostasis

Golgi lumen contains high amounts of calcium, magnesium, and manganese ions (Van Baelen et al. 2004; Pizzo et al. 2010; Vangheluwe et al. 2009). The presence of these divalent cations is important for cargo concentration and sorting (Chanat and Huttner 1991) as well as for glycosylation (Vanoevelen et al. 2007). The cations are transported into the Golgi lumen by the SERCA2 and SPCA1/2 type Ca2+/Mn2+ pumps. Of these two, SERCA2 is enriched in the cis-Golgi, while SPCA1 is mainly present in the trans-Golgi (Vangheluwe et al. 2009). Unlike SERCAs, SPCAs are also engaged in Mn2+ transport (Vangheluwe et al. 2009; Wong et al. 2013). In addition to SPCAs, recent evidence suggests that TMEM165 mutations in patients cause a type II congenital disorder of glycosylation in humans by interfering with Mn2+ and Ca2+/H+ transport (Dulary et al. 2017; Thines et al. 2018). Manganese is an essential trace metal and important co-factor needed for the catalytic activity of many inverting Golgi glycosyltransferases such as β(1,4)GalT1. The DXD motif typically present in these enzymes plays a key role in Mn2+-mediated donor substrate (UDP-Gal) binding (Breton et al. 2005). Based upon the recent structure of the β(1,4)GalT1 homodimer (Harrus et al. 2018), Mn2+ appears to regulate transitions of the lid and the “Trp loop” that define the open (inactive) and closed (active) states of the enzyme. Accordingly, the Met340H mutant form of the enzyme that binds Mn2+ 25 times more avidly, blocks the β(1,4)GalT1 in the closed state, inactivates the enzyme, and prevents its ability to form homodimers.

4.3 Golgi Redox State

Reactive oxygen species and low oxygen tension (hypoxia) also contribute to Golgi glycosylation potential. Most often, their effects are mediated by hypoxia-inducible factors (HIF1–3) that regulate the expression of a number of N-glycosylation-associated genes, including MGATs (MGAT2, MGAT-3, and MGAT5a and 5b), fucosyltransferases (FuT1, 2 and 7), sialyltransferases (ST3Gal1 and ST6Gal1) as well as nucleotide sugar transporters for UDP-galactose, CMP-sialic acid and UDP-N-acetylglycosamine (Koike et al. 2004; Shirato et al. 2010; Belo et al. 2015). Based upon these observations, Taniguchi et al. (Taniguchi et al. 2016) introduced the term “Glyco-redox” to link altered glycosylation with oxidative stress generated by hypoxia or reactive oxygen species (ROS). Their close association may also contribute to neurodegenerative disorders such as Parkinson’s disease, Alzheimer’s disease, and amyotrophic lateral sclerosis (ALS). Hypoxia (or HIFs) may also induce cleavage of cell surface N-linked glycans and thereby affect cell–extracellular matrix interactions (Taniguchi et al. 2016; Eguchi et al. 2002, 2005). Oxidative stress and altered glycosylation have also been linked to high-fat diet, obesity, and the onset of type II diabetes mellitus (Ohtsubo and Marth 2006; Ohtsubo 2010; Ohtsubo et al. 2011). Marth and co-workers showed in their studies (Ohtsubo 2010; Ohtsubo et al. 2011) that high levels of free fatty acids inhibit the expression of MGAT4a, a glycosyltransferase needed for β(1,4)-GlcNAc branching of N-glycans as well as GLUT-2 glucose transporter in pancreatic β-cells. The β(1,4)-GlcNAc branch is normally required for cell surface localization of the glucose transporter, and thus for glucose transport into cells. Without the β(1,4)-GlcNAc branch, the GLUT-2 remains intracellular, leading to decreased glucose import, insulin export, and accumulation of glucose in the blood.

Recent evidence indicates that hypoxia can modulate N-glycosylation also in a HIF-independent manner via affecting the oxidative potential of the Golgi lumen (Hassinen et al. 2019). Surprisingly, in normoxic conditions, it is higher than that of the ER (the main site of disulfide bond formation in the cells). In hypoxic cells, however, Golgi oxidative potential equals that of the ER in normoxic cells. The cells also displayed less sialic acid in their cell surface N-glycans. Interestingly, this was shown to be associated with reduced formation of surface-exposed disulfide bonds in ST6Gal1 (and likely also in some other sialyltransferases including ST3Gal3), loss of its catalytic activity, and inability to interact with β(1,4)GalT1 (Hassinen et al. 2019). Therefore, the high oxidative potential in the Golgi lumen appears to be necessary for the catalytic activity of certain sialyltransferases. This “redox switch” guarantees that the ST6Gal1 remains inactive until it reaches the Golgi compartment where it is expected to function. Likewise, the β(1,4)GalT1 enzyme acquires full activity also in the acidic Golgi compartment after interacting with the ST6Gal1 sialyltransferase.

5 Concluding Remarks

N-glycosylation is a frequent and complex modification of proteins, and essential for both uni- and multicellular life. It regulates a plethora of cellular functions that range from protein folding, trafficking, sorting, localization, half-life, and signaling to proliferation, migration, and adhesion with its surroundings. Therefore, it is also not surprising that we currently know a vast number of human disorders that are caused by, or are associated with, altered N-glycosylation. While previous work has provided us a clear overall picture of the basic principles in N-glycan biosynthesis, there is a big gap in our understanding of the factors that underlie cell-, tissue-, or organism-specific glycosylation patterns and their dynamic variability that starts during embryonic development continuing thereafter throughout our lives. We currently know that factors such as pH, redox potential, and changes in ion fluxes mainly in the Golgi compartment fundamentally affect and regulate the functioning and activity of glycosyltransferases expressed in a cell. Yet, there are many questions that remain unanswered. For example, how and why cells have evolved in such a complex way to make their N-glycans, needing removal of some sugar residues and replacing them with others instead of adding the right sugar in the beginning? Perhaps there is an evolutionary reason, as yeasts (an early eukaryote) produce mainly high mannose type N-glycans which needed to be modified to different ones in order to provoke immune responses only against them and other pathogens, and thereby survive. And what is the reason (or cause) of producing N-glycans which differ between two identical protein molecules? Does it result from a “sloppy” machinery that is prone to mistakes, or is there some purpose or benefit behind? Increasing biodiversity perhaps? Or is it a mark of ongoing evolution and trials to find the best fit for changing conditions? Is it regulated, or random? List is endless.

Nevertheless, these examples emphasize the need to understand in much more detail how glycans are made, how their synthesis is regulated and to what extent. An important issue also to keep in mind when one aims to produce optimally glycosylated antibodies for therapeutic use is to realize that yeasts, other lower eukaryotes or bacteria might not be the best choices to be used as hosts, as glycosylation is not just a simple outcome of enzymes present. It requires also conditions that support their full activity and complex mutual interplay necessary for their efficient functioning. Finally, we infer that there is an urgent need for developing more effective glycoengineering tools to edit glycans at will and thereby improve physicochemical and pharmacological properties of glycoprotein-based therapeutic compounds.