Generation and degradation of free asparagine-linked glycans

Abstract

Asparagine (N)-linked protein glycosylation, which takes place in the eukaryotic endoplasmic reticulum (ER), is important for protein folding, quality control and the intracellular trafficking of secretory and membrane proteins. It is known that, during N-glycosylation, considerable amounts of lipid-linked oligosaccharides (LLOs), the glycan donor substrates for N-glycosylation, are hydrolyzed to form free N-glycans (FNGs) by unidentified mechanisms. FNGs are also generated in the cytosol by the enzymatic deglycosylation of misfolded glycoproteins during ER-associated degradation. FNGs derived from LLOs and misfolded glycoproteins are eventually merged into one pool in the cytosol and the various glycan structures are processed to a near homogenous glycoform. This article summarizes the current state of our knowledge concerning the formation and catabolism of FNGs.

Introduction

In the ER lumen, nascent polypeptides emerging from the protein-conducting channel are frequently modified with glycans at selected asparagine residues within the consensus sequences (N-X-S/T, X ≠ P) (Fig. 1) [1, 2]. This process, which is referred to as N-glycosylation, is one of the most frequent co- and post-translational modifications in eukaryotes [3]. In the past three decades, extensive genetic and biochemical studies on mutant cells derived from Saccharomyces cerevisiae harboring N-glycosylation defects have led to the identification of most of the genes that are involved in the biosynthesis of the core glycan. In sharp contrast, our knowledge of aspects of degradation of N-glycans is very incomplete. It is generally believed that lysosomes are the predominant organelle for the degradation of various glycoconjugates [4]. However, an accumulating body of evidence suggests that the non-lysosomal degradation of N-glycans is ubiquitous [5, 6]. Above all, one of the outstanding questions that remain to be clarified is the molecular mechanisms underlying the generation and degradation of “free”, unconjugated forms of glycans, which are structurally related to N-glycans (free N-glycans; FNGs). Since our last comprehensive review on this subject [7], significant progress has been made and the findings reveal the existence of more diverse, nevertheless sophisticated, pathways than had ever been thought. In this review, we update our knowledge regarding the generation and degradation of FNGs with a focus on mammalian cells and S. cerevisiae.

Fig. 1.

Biosynthesis of N-glycans in the ER of mammalian cells. Reactions utilizing UDP-Glc, UDP-GlcNAc and GDP-Man are boxed with cyan, magenta and green, respectively (dashed lines). Reactions utilizing Dol-P-Man and Dol-P-Glc are indicated green and cyan arrows, respectively. Red arrow indicates an unidentified pathway

Biosynthesis of N-glycans in the ER

The biosynthesis of N-glycans in the ER consists of two steps in all eukaryotic cells: (1) the preassembly of LLOs as glycan donor substrates and (2) their transfer to polypeptide chains (Fig. 1). LLO biosynthesis is a biphasic process initiated by the synthesis of Man₅GlcNAc₂-PP-Dol at the cytosolic side of the ER membrane by the sequential action of DPAGT1 (Alg7 in S. cerevisiae) [8, 9], the Alg13–Alg14 heterodimer complex [10–14], Alg1 [15–17], Alg2 [18, 19] and Alg11 [20, 21]. The donor substrates for all of these glycosyltransferases are two nucleotide sugars (UDP-GlcNAc and GDP-Man) (Fig. 1, red and green dashed boxes). In this connection, it is particularly noteworthy that the ALG13, ALG14, ALG1, and ALG2 genes from S. cerevisiae have been successfully expressed in Escherichia coli, and as a result, Man_3-4GlcNAc₂-PP-undecaprenol can be produced in the engineered E. coli [22]. However, the issue of specifically how the fourth mannose residue is added remains to be clarified.

Man₅GlcNAc₂-PP-Dol, the final product in the cytosolic process of LLO biosynthesis, is flipped into the luminal side of the ER membrane. Genetic studies using S. cerevisiae revealed that Rft1 is required for the transbilayer movement of LLOs [23], while biochemical reconstitution studies suggest the occurrence of an RFT1-independent flipping process [24–26]. Irrespective of the mechanisms involved, Man₅GlcNAc₂-PP-Dol oriented toward the ER lumen undergoes four additional mannose transfer reactions catalyzed by enzymes utilizing Dol-P-Man, Alg3 [19, 27, 28], Alg9 [29, 30] and Alg12 [31–34] (Fig. 1, green arrows). The synthesized LLO intermediate, Man₉GlcNAc₂-PP-Dol, is further glucosylated by enzymes utilizing Dol-P-Glc Alg6 [19, 35, 36], Alg8 [37, 38] and Alg10 [39] (Fig. 1, cyan arrows), generating the fully assembled Glc₃Man₉GlcNAc₂-PP-Dol. The core glycan is transferred en bloc onto growing polypeptide chains in the ER lumen by the oligosaccharyltransferase (OST), leaving PP-Dol as a byproduct. The PP-Dol is then dephosphorylated to P-Dol by DOLPP1 (Cwh8 in S. cerevisiae) [40–42] and recycled as the acceptor substrate in LLO biosynthesis [43]. This cycle is generally referred to as the “dolichol cycle”.

In mammals and S. cerevisiae, OST associates with the translocon complex [44–49], and often catalyzes co-translational N-glycosylation [47, 50]. OST also catalyzes the post-translational N-glycosylation of the acceptor sites that are located in a flexible region of a protein [47, 51]. In co-translational glycosylation, the acceptor sites should be located > 15 amino acid residues away from the luminal surface of the ER membrane [52]. The cryo-electron microscopic structure of S. cerevisiae OST indicates that the active site may be located ~3 nm (corresponding to ~15 amino acids) above the membrane [53], while the crystal structures of bacterial OST suggest that the active site is formed at the juxtamembrane region [54]. The minimum acceptor sequence for fulfilling the requirement for N-glycosylation is a tripeptide (N-X-S/T, X ≠ P) [55], and OST can modify many proteins carrying the N-glycosylation consensus sequences. By contrast, OST shows a strict specificity for the donor substrate and preferentially utilizes the fully assembled LLO in vitro [56, 57]. Although OST also can utilize truncated LLOs as short as GlcNAc₂-PP-Dol with a reduced efficiency in vitro [58, 59], an ordered assembly of LLOs (Fig. 1) ensures the preferential transfer of the fully assembled oligosaccharide onto a protein by OST under normal conditions [34, 60]. However, a subset of glycoproteins carrying N,N′-diacetylchitobiose has been detected in several types of glucose-starved cancer cells [61–63], implying that such conditions somehow allow OST to utilize GlcNAc₂-PP-Dol, which is synthesized at the cytosolic side of the ER membrane, for N-glycosylation reactions in the ER lumen. The N-glycosylation reaction in vitro is usually activated by the addition of manganese ion [55], while the endogenous co-factor(s) for OST has not yet been determined.

In animals and S. cerevisiae, OST is the multi-membrane protein complex consisted of the conserved catalytic subunit Stt3 and several accessary subunits (Table 1) [1]. In contrast, some protists, such as Trypanosoma brucei and Leishmania major, possess only the Stt3 subunit as the OST enzyme, but instead have multiple numbers of STT3 genes (Table 1) [64–68]. The exogenous expression of T. brucei Stt3 and L. major Stt3 can suppress lethality as well as N-glycosylation defects in stt3∆ yeast mutant cells [66–68]. The multiplication of the STT3 gene is also found in animals and plants, where they have two STT3 genes (STT3A and STT3B) [57]. However, S. cerevisiae and other fungi have a single copy of the STT3 gene, which is closely related to metazoan STT3B. In mammalian cells, Stt3A and Stt3B exclusively form two distinct OST complexes. The Stt3A complex is believed to exclusively catalyze co-translational N-glycosylation by associating with the translocon [47], while the Stt3B complex can mediate the post-translational N-glycosylation of the sites that are skipped by the Stt3A complex [47, 69–72]. In S. cerevisiae, Stt3 forms two OST complexes containing either Ost3 or its paralog, Ost6 [45, 73], both of which have thioredoxin activity and enhances the efficiency of the N-glycosylation of sites near cysteine residues, presumably by hydrophobic interactions and the formation of mixed-disulfide bonds with substrate polypeptides [74, 75]. The mammalian Stt3B complex also associates with a thioredoxin-like protein MagT1 [57, 76] that forms mixed-disulfide bonds with substrate proteins [72]. The issue of why eukaryotic cells use multiple OST complexes is not fully understood, but it has been proposed that they complement substrate specificity and maximize N-glycosylation efficiency [77].

Table 1.

Genes encoding OST subunits in S. cerevisiae, Homo sapiens, L. major, T. brucei and C. jejuni

Genes encoding OST subunits
S. cerevisiae	H. sapiens	L. major	T. brucei	C. jejuni
STT3	STT3A STT3B	LMJF_35_1130 (STT3A) LMJF_35_1140 (STT3B) LMJF_35_1150 (STT3C) LMJF_35_1160 (STT3D)	Tb927.5.890 (STT3A) Tb927.5.900 (STT3B) Tb927.5.910 (STT3C)	pglB
WBP1	DDOST/OST48	–	–	–
SWP1	RPN2	–	–	–
OST1	RPN1	–	–	–
OST2	DAD1	–	–	–
OST3/OST6	TUSC3/MAGT1	–	–	–
OST4	OST4	–	–	–
OST5	N.D.	–	–	–

N.D. not determined

The pathogenic bacterium Campylobacter jejuni expresses N-glycosylation machinery that is encoded by the pgl (protein glycosylation) gene cluster [78]. In C. jejuni, OST is a single subunit-type enzyme encoded by the pglB gene (Table 1), which is homologous to the eukaryotic STT3 [79]. LLOs synthesized at the inner membrane of C. jejuni are quite distinct from those found in eukaryotes in terms of sugar compositions (GlcNAc, Man and Glc in eukaryotes and bacillosamine, GalNAc and Glc in C. jejuni) [80], as well as lipid structure (dolichol in eukaryotes and undecaprenol in C. jejuni) [81]. Unlike STT3 in mammals and S. cerevisiae, pglB is not required for the cell growth of C. jejuni under normal culture conditions, while the deletion of pglB causes severe growth defects under conditions of osmotic stress [82].

Glycoprotein quality control in the ER

In the ER lumen of mammalian cells and S. cerevisiae, the folding processes of nascent polypeptides are strictly monitored by a system called ER quality control (ERQC) [83, 84]. In this system, concerted actions of ER-localized chaperones, modifying enzymes and lectins play critical roles in distinguishing native glycoproteins from misfolded or unfolded ones and in selectively transporting the former to their final destinations. On the other hand, the terminally misfolded glycoproteins that are formed in the ER lumen are eliminated by ER-associated degradation (ERAD), which involves the retrotranslocation of misfolded glycoproteins back to the cytosol followed by their proteasomal degradation. The detailed molecular mechanisms of ERQC and ERAD have been summarized in excellent review articles [85–87].

In both ERQC and ERAD, there are N-glycan structures that dictate the folding status of the carrier polypeptides (Fig. 2) [88, 89]. After N-glycosylation in the ER lumen, two glucose residues of the Glc₃Man₉GlcNAc₂ (Fig. 2, residues 13 and 14) are rapidly removed by ER glucosidases (GI and GIIα/β in mammals; Gls1 and Gls2/Gtb1 in S. cerevisiae). Calnexin (CNX) and calreticulin (CRT) recognize the glucose residue of the G1M9 isomer (Fig. 2, residue 12) and may bind to the nascent glycoproteins, which facilitates productive folding. In contrast, Cne1, the yeast ortholog of CNX, is dispensable for the folding of nascent glycoproteins, although this lectin binds to the G1M9 isomer [90, 91]. The removal of the last glucose residue, also catalyzed by ER glucosidase II (Fig. 2, residue 12), occurs relatively slowly compared to the first reaction [92]. If the nascent glycoproteins bearing non-glucosylated N-glycans are recognized as being in unfolded states by the UDP-glucose:glycoprotein glucosyltransferase (UGGT), they are reglucosylated by this enzyme, thus allowing CNX/CRT to interact again with unfolded glycoproteins. It appears that the genome of S. cerevisiae does not have a gene ortholog of UGGT. During the CNX/CRT cycle [93, 94], the external α1,2-linked mannose residue at the B-arm (Fig. 2, residue 9) may be trimmed, presumably by ERManI (Mns1 in S. cerevisiae) [95] or the ER degradation-enhancing mannosidase-like protein 2 (EDEM2) [96]. When glycoproteins fold properly, they exit the ER and are transported to their final destinations. If glycoproteins fail to acquire correct folding or subunit assembly states, the external α1,2-linked mannose residue at the C-arm (Fig. 2, residue 11) is removed by EDEM1/3 (Htm1 in S. cerevisiae), which results in the α1,6-linked mannose residue being exposed (Fig. 2, residue 10) [89, 97, 98]. The degradation lectins (OS9 and/or XTP3-B/Erlectin in mammals; Yos9 in S. cerevisiae) recognize the exposed α1,6-linked mannose residue at the C-arm (Fig. 2, residue 10) and direct misfolded glycoproteins to ERAD by association with SEL1L (Hrd3 in S. cerevisiae) [99, 100], an ER membrane protein that forms a complex with the “retrotranslocon” [84, 101, 102]. Especially for misfolded (glyco)proteins generated in the ER lumen, they need to cross the ER membrane before their proteasomal degradation. The molecular mechanism underlying the retrotranslocation process is not currently understood. However, the central player of the retrotranslocation channel is the subject of debate, and so far HRD1 (Hrd1 in S. cerevisiae), Derlin-1 (Der1 in S. cerevisiae) and the Sec61 complex (Sec61α–Sec61β–Sec61γ in mammals and Sec61–Sbh1–Sss1 in S. cerevisiae) have been proposed to be involved in this process [103]. HRD1 is an E3 ubiquitin ligase that adds and elongates ubiquitin chains on misfolded (glyco)proteins during their retrotranslocation, and the polyubiquitination recruits the AAA-ATPase p97/VCP (Cdc48 in yeast) and 26S proteasomes to the ERAD machinery, promoting the extraction and degradation of misfolded (glyco)proteins [104]. During substrate extraction, N-glycans on misfolded glycoproteins are removed by the cytosolic peptide:N-glycanase (PNGase) [105–107], releasing FNGs in the cytosol. In mammalian cells, PNGase forms a complex with p97, HR23, and 26S proteasomes at the periphery of the ER membrane [108–110]. It has been postulated that the removal of bulky side chains, such as N-glycans, may facilitate the degradation of proteins by cylinder-shaped 26S proteasomes, as the proteins must enter the inside the cylinder, where the proteolytic activities reside. Yeast Png1 also forms a complex with Rad23, an ortholog of HR23, which may couple deglycosylation and proteasomal degradation [111, 112]. It should be noted that the ERAD process is implicated, not only in the degradation of misfolded proteins, but also in the generation of FNGs in the cytosol [6, 113, 114].

Fig. 2.

ERQC and ERAD for glycoproteins in mammalian cells. In the ER, glycan trimming is a temporally regulated process that generates glycan epitopes, which play crucial roles in protein folding, sorting and ERAD. Nascent glycoproteins are subjected to a productive folding process aided by lectin chaperones (CNX and CRT), GII and UGGT. When glycoproteins are terminally misfolded, they are segregated from the correctly folded molecules by the actions of EDEMs and degradation lectins (OS9 and XTPB-3), which initiate the retrotranslocation of misfolded glycoproteins. During the retrotranslocation of misfolded glycoproteins, the bulky N-glycans are removed by PNGase, releasing neutral FNGs in the cytosol. The deglycosylated polypeptides are degraded by proteasomes. The number for each sugar indicates the order of their assembly in LLO biosynthesis. A, B and C: A-, B- and C-arms. Red arrow indicates an unidentified pathway

Generation of neutral FNGs in the ER lumen of mammalian cells

In addition to the N-glycosylation function of OST, it has been proposed that OST may have hydrolytic activity for LLOs to generate neutral FNGs in the ER lumen (Fig. 3). This assumption is based on experimental findings showing that the LLO-degrading enzyme preferentially hydrolyzes Glc₃Man₉GlcNAc₂-PP-Dol over biosynthetic intermediates, requires divalent cations, and is inhibited by N-glycosylatable peptides [115], all of which are in good agreement with the enzymatic properties of OST. Major questions that remain to be clarified concern (1) the hydrolytic activity of OST, (2) the STT3 isoform that is responsible for the hydrolysis, and (3) the biological function of the LLO hydrolysis.

Fig. 3.

Formation and processing of neutral and phosphorylated FNGs derived from LLOs and misfolded glycoproteins in mammalian cells. In the ER lumen, a putative enzyme, proposed to be OST, hydrolyzes the fully assembled LLO and releases Glc₃Man₉GlcNAc₂. The luminal FNG is rapidly trimmed by GI/II and may be ERManI or EDEM2. The deglucosylated FNGs are transported to the cytosol through a putative transporter. In the cytosol, neutral FNGs are generated from misfolded glycoproteins by PNGase. Neutral FNGs derived from LLOs and misfolded glycoproteins are processed by ENGase and Man2C1, generating the M5B′ isomer (in solid box) in the cytosol. LLOs are also degraded by a putative LLO-PP’ase. The topology of the LLO-PP’ase reaction is tentatively depicted on the cytosolic side of the ER membrane. The fate of phosphorylated FNGs is not known, but ENGase and a putative phosphatase are potential candidate enzymes for the catabolism of phosphorylated FNGs. An example for Gn1 and Gn2 types of the M9 isoform (M9′ and M9, respectively) is shown in the yellow box. Mannose residues removed by Man2C1 are boxed with dashed line. The number for each sugar indicates the order of their assembly in LLO biosynthesis. A, B and C: A-, B- and C-arms. Red arrows indicate unidentified pathways

The only way to unambiguously demonstrate the hydrolytic activity of OST would be to purify the enzymatically active OST and test whether the OST produces FNGs from LLOs. In the case of S. cerevisiae OST, FLAG tagging of the C-terminus of OST4 permits the immuno-purification of both of the OST3 and OST6 complexes to be achieved in a single step [116]. Although no such facile method for purifying mammalian OST has been developed, mammalian STT3A and STT3B complexes have been separately purified from canine pancreatic microsomes by conventional chromatography [57]. It is also extremely important to identify the STT3 isoform that catalyzes LLO hydrolysis in vivo, which would definitely provide deeper mechanistic insights into glycoprotein biosynthesis. We envision that simple knockdown/knockout studies for STT3s would not provide conclusive evidence, because the cytosolic deglycosylating enzymes, i.e., PNGase [117] and ENGase [118], are also involved in the generation of FNGs in living cells (see below). Therefore, both of the deglycosylating enzymes would need to be ablated prior to gene manipulation for STT3s and the evaluation of the effects on the production of FNGs.

In mammalian cells, the efficiency of LLO hydrolysis is estimated to be ~35 % of that of N-glycosylation [119], indicating that a substantial fraction of the mature LLO is hydrolyzed and not used for N-glycosylation. It appears that this energy-waste does not interfere with N-glycosylation reactions, since excess amounts of the mature LLO are normally synthesized [120, 121]. The question then arises as to what is the biological function of LLO hydrolysis? Previous studies have proposed that, in the absence of a sufficient supply of acceptor substrates for OST, the hydrolysis of Glc₃Man₉GlcNAc₂-PP-Dol occurs, thus balancing the amounts of donor and acceptor substrates [119, 122]. However, this hypothesis is most unlikely, because incubation with translation inhibitors completely inhibits the generation of FNGs [123]. Moreover, the translation attenuation evoked by interferon-inducible RNA-dependent protein kinase (PKR)-like ER kinase (PERK) signaling does not reduce the extent of LLO hydrolysis [124]. Apart from these observations, Gao et al. [121, 124, 125] proposed that the frequency of LLO hydrolysis is regulated by the intracellular levels of Man 6-P. Their initial observations showed that, in streptolysin O-permeabilized cells, the addition of Man 6-P (10–100 μM) specifically induced the reduction of mature LLO and an increase in the corresponding FNG [121]. A remarkable elevation of Man 6-P has been detected in herpes simplex virus 1 (HSV-1)-infected mouse embryonic fibroblasts (MEFs) [124] and mannose 6-phosphate isomerase (MPI)-deficient mouse embryonic fibroblasts (MEFs) that had been cultured in medium supplemented with extremely high concentrations (500 μM) of mannose [126]. The MPI-deficient MEFs produce abnormally low levels of mature LLO by unknown mechanisms, while the loss of LLO has no effect on global N-glycosylation [126]. On the other hand, an HSV-1-infection apparently induces the hypoglycosylation of a viral glycoprotein in wild-type MEFs [124]. Although the N-glycosylation status of proteins produced in host cells needs to be carefully analyzed, it is tempting to speculate that, in virus-infected cells, Man 6-P-mediated LLO degradation may prevent the production of viral glycoproteins [124, 125].

Transport of neutral FNGs from the ER lumen to the cytosol in mammalian cells

When free Glc₃Man₉GlcNAc₂ is generated in the ER lumen, the FNG is rapidly processed by ER mannosidases and/or glucosidases and transported out to the cytosol (Fig. 3). Since FNGs are bulky hydrophilic molecules, they should be able to cross the ER membrane through a transporter. This translocation process is thought to be mediated by a putative ATP- and Ca²⁺-dependent FNG transporter [127]. In addition to ATP, unidentified cytosolic factors are also required in FNG transport assay using reconstituted microsomes [128]. Although the biological function, as well as the molecular mechanism, of FNG transport remains to be clarified, one can imagine that the persistent retention of neutral FNGs in the ER may impair the glycoprotein quality control machinery regulated by glycan-recognizing proteins.

In vitro experiments indicate that the FNG transporter recognizes mannose, but not GlcNAc residues of FNGs [129], and non-glucosylated FNGs were found to be the preferred substrates over glucosylated ones [127, 128]. What then would be the fate of glucosylated FNGs that are formed in the ER lumen upon the inhibition of ER glucosidases? It has been reported that, under such conditions, glucosylated FNGs somehow reach the Golgi apparatus and the Glc_1-3Man unit at the A-arm (Fig. 2, residues 7 and 12–14) of the FNGs may be cleaved off by a Golgi endo-α-mannosidase [130, 131]. The cleavage allows the structural remodeling of high-mannose-type FNGs to sialylated, complex-type FNGs (sialyl FNGs) by Golgi glycosidases and glycosyltransferases and the sialyl FNGs are then secreted to and detected in the culture medium [132]. While the fate of the extracellular sialyl FNG is currently unknown, it is tempting to speculate that sialyl FNGs in sera (see the section “Occurrence of sialyl FNGs inside and outside of mammalian cells” below) may be formed by a similar mechanism.

Generation of neutral FNGs in the cytosol of mammalian cells

It has been postulated that the deglycosylation reaction catalyzed by PNGase facilitates the degradation of the polypeptide backbone by proteasomes [105, 133, 134]. PNGase-mediated deglycosylation has been observed for a few model glycoproteins [118, 135–138]. Of these, a plant-derived, mutant form of the ricin toxin A chain (RTA) is the only model misfolded glycoprotein determined so far that is degraded in a PNGase-dependent manner [118]. Interestingly, in the absence of PNGase, ENGase deglycosylates the mutant form of RTA, producing the aggregation-prone, N-GlcNAc-modified protein [118]. This finding clearly demonstrates that ENGase can also act directly on glycoproteins in the cytosol of mammalian cells to generate FNGs (Fig. 3). However, the physiological substrates for ENGase, if any, need to be identified.

The knockdown of NGLY1 (the gene encoding human PNGase) in human cells results in a partial reduction of very specific neutral FNGs (Man₉GlcNAc₂ and Glc₁Man₉GlcNAc₂), whereas a large majority of other FNGs remain unaffected by the NGLY1 knockdown [117]. This observation implies that misfolded glycoproteins that had been trapped and released during the calnexin/calreticulin cycle (Fig. 2) may be targeted to ERAD and undergo PNGase-mediated deglycosylation. Although complete knockout studies on NGLY1 will be required to draw any conclusions, the experimental findings indicate that, in mammalian cells, bulk FNGs are produced in a PNGase-independent manner, and indicate that PNGase-mediated deglycosylation may occur in a glycan structure-specific fashion. In any event, the biological importance of PNGase in mammalian systems is now well appreciated because patients harboring NGLY1 mutations have now been identified [107, 139–143]. The patients showed multiple symptoms including global developmental delay, multifocal epilepsy, involuntary movement, abnormal liver function, and the absence of tears [139, 141, 143]. However, the pathological mechanism responsible for the NGLY1 disorder is unclear. The establishment of model organisms with NGLY1 mutations and the identification of endogenous substrates for NGLY1 would be expected to provide mechanistic insights into the pathology of the NGLY1 disorder.

Processing of neutral FNGs in the cytosol of mammalian cells

It has long been believed that lysosomes are the predominant organelle for the degradation of biological macromolecules, including glycoconjugates. In this context, the occurrence of cytosolic glycosidases had attracted less attention in glycan catabolism, especially after the demonstration that lysosomes function as degradation factories [5, 144].

In the cytosol, FNGs derived from LLOs and misfolded glycoproteins are eventually merged into one pool and are then sequentially processed to a near homogenous glycoform (Fig. 3, the M5B′ isoform; Man₅GlcNAc₁) by the cytosolic endo-β-N-acetylglucosaminidase (ENGase) and Man2C1 (Fig. 3). Extensive structural studies of neutral FNGs have shown that there are two types of FNGs [115, 119, 127, 145–148]: One contains an N,N′-diacetylchitobiose core at the reducing end of FNGs (Gn2 type) and does the other a single GlcNAc (Gn1 type) predominated in the cytosol (Fig. 3, structural models shown in yellow box). The Gn2-type structure can be generated by either the hydrolysis of LLOs or the deglycosylation of misfolded glycoproteins by PNGase, but such FNGs rarely accumulate in the cytosol. On the other hand, high levels of the Gn1-type structure are found in the cytosol and are believed to be primarily the product of ENGase [149, 150], which converts Gn2-type FNGs to the Gn1 type by removing the innermost GlcNAc residue (Fig. 3, ENGase). The cytosolic ENGase has been biochemically purified and characterized [151], and the gene encoding this enzyme (Engase) has been identified [149, 150]. Engase knockdown in mammalian cells results in a dramatic accumulation of the Gn2 type of cytosolic neutral FNGs, which is accompanied by an increased number of mannose residues (Fig. 3, M9/(M8); Man_8–9GlcNAc₂) [117].

The “cytosolic” neutral α-mannosidase, referred to as Man2C1, occurs in various animal sources [152–165]. It has been shown that the knockout of Man2C1 in mice results in a dramatic accumulation of Gn1 type of FNGs with Man_8–9 (Man_8–9GlcNAc₁) in the cytosol of all tissues tested [166], strongly indicating that the action of Man2C1 is responsible for generating M5B′ in the cytosol (Fig. 3). Biochemical studies of Man2C1 indicate that this enzyme preferentially trims Gn1-type FNGs over the Gn2 type [157, 158, 161, 167]. This substrate preference may explain the reason why Engase knockdown resulted in accumulation of M8B and M9 (Fig. 3) in the cytosol [117]. Man2C1 preferentially trims mannose residues on the B- and C-arms of FNGs with the reversed order of the mannose assembly in LLO biosynthesis (Fig. 3, Gn1 type in yellow box) [157].

In addition to the accumulation of larger Gn1-type FNGs, Man2C1 knockout mice exhibit multiple tissue damage, such as the degeneration of neurons, the glia, liver and kidney [166]. Although the molecular mechanism underlying these organ defects in Man2C1-null mice is currently unclear, it has been reported that the suppression of Man2C1 expression in cultured cells results in cell growth retardation and apoptosis [168–171]. In contrast to the cases for gene suppression, the overexpression of Man2C1 has been implicated in tumorigenesis [171–173]. In prostate cancer, Man2C1 is prone to be overexpressed, directly binds to PTEN, a tumor suppressor catalyzing the dephosphorylation of phosphatidylinositol 3,4,5-triphosphate, and suppresses its phosphatase activity, thereby promoting tumor growth [173]. In this study, it was suggested that multiple domains of Man2C1 including its catalytic site are involved in this PTEN interaction, whereas the issue of whether the catalytic activity of Man2C1 is required for tumor progression is not known with certainty. Interestingly, Man2C1 has an enzyme-independent function in suppressing mitochondria-dependent apoptosis [169], which may aid in Man2C1-overexpressing cancer cells to escape from programmed cell death. Recent studies have identified inhibitors for Man2C1 [174, 175]. These compounds show an inhibitory effect in intact cells, serving as new tools for evaluating the biological functions of Man2C1.

It has been repeatedly reported that Man2C1 overexpression alters glycosylation patterns: A reduction in concanavalin A staining in splenocytes from human Man2C1-transgenic mice [172] and the underglycosylation of a null Hong Kong variant of α1-antitrypsin in Man2C1-overexpressing HeLa cells [176]. Although the precise mechanism of how the overexpression of Man2C1 can affect the N-glycosylation process occurring in the luminal side of the membrane remains to be clarified, it has been implied that excess amounts of Man2C1 may inhibit the glucosylation steps of LLO biosynthesis [176]. It should be noted that Man2C1 is present, not only as a soluble form in the cytosol, but also as a membrane-associated form [168, 177]. Although it has been proposed that cytosolic Man2C1 is the likely precursor of an ER mannosidase [177, 178], no definitive evidence for the luminal occurrence of Man2C1 has been reported.

Transport of neutral FNGs from the cytosol to the lysosomes in mammalian cells

It has been shown that the final product of ENGase and Man2C1 in the cytosol, which is mainly composed of M5B′ (Fig. 3), is eventually degraded via a mechanism involving a cytosol–lysosome transport of FNGs mediated by an ATP-driven lysosomal transporter [179, 180]. However, the molecular nature of this transporter remains unknown. Biochemical studies showed that the transport process accepts both Gn2 and Gn1 types of FNGs, shows a strong preference for FNGs with shorter mannose residues, does not tolerate the phosphorylation of the reducing end of the innermost GlcNAc residue, and is inhibited by GlcNAc, but not mannose [180]. The lysosomal degradation of the transported FNGs results in the production of monosaccharides. Although their precise fates are currently unknown, it is reasonable to assume that the monosaccharides are recycled to synthesize nucleotide sugars for glycosylation reactions.

Occurrence of sialyl FNGs inside and outside of mammalian cells

The occurrence of sialyl FNGs, which contain galactose and sialic acid residues, has been reported in the urine from patients of lysosomal storage diseases, such as sialidosis and galactosialidosis [181–185]. It is generally thought that these sialyl FNGs leak from lysosomes upon the lysis of sick cells [186]. On the other hand, the intracellular accumulation of sialyl FNGs was also observed in mouse liver [147], various cultured cells [187] autophagy-defective cells [188], and various cancer tissues, but not in normal epithelial tissues [189, 190]. It is tempting to speculate that the sialyl FNGs found in cancer tissues may serve as useful, specific tumor markers. It has also been shown that the knockdown of Slc17a5, the gene encoding a lysosomal sialic acid transporter (known as sialin), causes a significant delay in the accumulation of sialyl FNGs upon induction of autophagy defect [188]. More recently, it was shown that the protein level of sialin is increased by shutting down of Atg5 expression, which causes a loss of autophagosome formation [191]. Sialin upregulation is not caused by general defects of lysosomes, as the protein level of LAMP1, another lysosomal membrane protein, remains unchanged under the same conditions [191]. How sialin levels are regulated by the autophagy process, as well as how it is correlated with the cytosolic accumulation of sialyl FNGs, is an intriguing question. All of these findings clearly show the occurrence of FNGs composed of, not only high-mannose-type glycans, but also complex-type glycans with galactose and sialic acid residues in the cytosol of mammalian cells. However, it is not known whether the catabolic pathway for the sialyl FNGs resides in the cytosol, except that the overexpression of cytosolic sialidase, Neu2, can reduce the levels of the FNGs that have accumulated in the cytosol of human stomach cancer-derived cultured cells MKN45 [187]. Further studies will clearly be needed to develop a better understanding of the molecular mechanism underlying the catabolic pathway of complex-type FNGs.

More recently, “extracellular” complex-type FNGs, which are structurally related to complex-type N-glycans, were identified in human sera [192]. While the most intracellular, complex-type FNGs are of the Gn1 type [147, 181–185, 187–190], the extracellular complex-type FNGs were exclusively the Gn2 type [192]. This finding implies that the intra- and extracellular complex-type FNGs may have different origins. How these complex-type FNGs are formed, as well as the nature of their biological roles, will be interesting issues that remain to be explored.

FNGs generated in the cytosol of S. cerevisiae

In S. cerevisiae, a large majority (~95 %) of the cytosolic FNGs, which are exclusively of the Gn2 type due to the lack of ENGase, are produced in a PNGase-dependent manner (Fig. 4a) [193–195]. The importance of PNGase-mediated deglycosylation in the glycoprotein ERAD was first demonstrated in S. cerevisiae: The deletion of PNG1 (the yeast ortholog of PNGase) appears to decelerate ERAD of a plant-derived ricin toxin A chain [112]. Although a few Png1-dependent ERAD substrates are now available [112, 196–199], no physiological substrates for Png1 have been identified. Therefore, the current model of glycoprotein ERAD in S. cerevisiae is thoroughly dependent on studies in which a limited number of non-physiological substrates have been used. It has been proposed that the sequential mannose trimming by Mns1 and/or Htm1 plays crucial roles in targeting misfolded glycoproteins to ERAD by recruiting a degradation lectin Yos9 [88, 198, 200]. On the other hand, when we analyzed the detailed structures of FNGs as a read-out of a global glycoprotein ERAD, it was found that the most abundant FNG was Man₈GlcNAc₂ lacking one external α1,2-linked mannose at the B-arm (M8B, Fig. 4b), but not Man₇GlcNAc₂ (M7C, Fig. 4b), the product of Mns1 and Htm1 [195]. This experimental finding clearly indicates that not all N-glycans that are present on misfolded glycoproteins are necessarily trimmed by Htm1. Interestingly, ~10 % of FNGs were found to contain an internal α1,6-linked mannose residue on their A-arm, which is added by the Golgi-resident mannosyltransferase Och1 (Fig. 4b) [195, 201]. This observation indicates that some misfolded glycoproteins travel to the Golgi prior to retrotranslocation. Although it is unclear what factors control mannose trimming and addition, Och1-modified M8B (Fig. 4b) is increased upon the induction of the unfolded protein response (UPR), while the amounts of Och1-modified M7C (Fig. 4b) were negligibly affected by UPR [195].

Fig. 4.

Formation and processing of neutral FNGs derived from LLOs and misfolded glycoproteins in S. cerevisiae. a OST hydrolyzes Glc₃Man_8–9GlcNAc₂-PP-Dol. The released Glc₃Man_8–9GlcNAc₂ is trimmed by ER glycosidases (Gls1, Gls2 and Mns1). The processed FNGs are transported to the cytosol through an uncharacterized transporter and catabolized to Man₁GlcNAc₂ by Ams1. Png1 also generates neutral FNGs by deglycosylating misfolded glycoproteins in the cytosol, which are also processed by Ams1. b Glycan isoforms for Man_7–8GlcNAc₂ and their Och1-modified forms produced in S. cerevisiae. Mannose residues added by Och1 were indicated by down arrowheads. The number for each sugar indicates the order of their assembly in LLO biosynthesis. A, B and C: A-, B- and C-arms. Red arrow indicates an unidentified pathway

As discussed above, FNGs serve as a powerful tool for the dissection of glycoprotein ERAD in S. cerevisiae. However, it is clear that this concept is not applicable to mammalian cells, as bulk FNGs appear to be produced independent of the action of PNGase [117]. However, confusingly, there are several literature reports claiming that the FNGs found in mammalian cells represent the flow of glycoproteins to ERAD [123, 202–204]. In mammalian cells, very specific FNGs (G_0–1Man₉GlcNAc₂) are produced in a PNGase-dependent manner [117]. Therefore, extensive structural and quantitative analyses on these FNG structures, together with the knockout of NGLY1 (PNGase ortholog in mammals), should be empirically carried out to assess the contribution of glycoprotein ERAD to the generation of FNGs in mammalian cells.

An overall picture of FNG catabolism, as well as the degradation of glycoproteins, is not fully understood in S. cerevisiae, since none of the glycosidases involved in these processes have been identified, except for the cytosol/vacuolar α-mannosidase (Ams1) [205]. Ams1 removes all α-linked mannose residues from FNGs and generates Man₁GlcNAc₂ (Fig. 4a) [193–195, 206]. Ams1 is first synthesized in the cytosol and is then transported to vacuoles via a cytosol-to-vacuole targeting (CVT) pathway mediated by Atg19 [207–211]. When ATG19 is deleted to block CVT pathway, Man_1–3GlcNAc₂ largely accumulates in the mutant cells [195]. This circumstantial evidence suggests that FNG catabolism can occur in the cytosol. However, it should be noted that S. cerevisiae produces Atg34 [212, 213], a paralog of Atg19, which functions as a cargo of Ams1 only under starvation conditions [213]. Further studies will be needed to clarify the involvement of Atg34 in the catabolism of FNG. The conversion of FNGs to Man₁GlcNAc₂ has been observed in post-diauxic and stationary phases [206] when the expression of Ams1 is upregulated [195]. Although the biological function of Ams1 is currently unknown, it is tempting to speculate that the Ams1-mediated catabolism of FNGs may generate mannose as a carbon source to compensate for the depletion of glucose in the stationary phase.

FNGs generated in the ER of S. cerevisiae

In S. cerevisiae, ~5 % of the FNGs are produced in an OST-dependent manner (Fig. 5) [194]. The S. cerevisiae OST complex is composed of 8 subunits including 5 essential gene (STT3, WBP1, SWP1, OST1 and OST2) and 3 non-essential gene (OST4, OST5 and either OST3 or its paralog OST6) products (Table 1) [1, 2, 77]. It was found that the maximal N-glycosylation activity of OST is required for the efficient hydrolysis of LLOs [194]. In yeast, Ost3 and Ost6 are exclusively incorporated into two distinct OST complexes, resulting in the formation of Ost3- and Ost6-containing OST complexes [45, 73, 214]. Although the deletion of either OST3 or OST6 causes the hypoglycosylation of a subset of glycoproteins [74, 215], only the deletion of OST3 drastically reduces FNG levels [194]. The underlying mechanism responsible for this is unknown, but it has been reported that the Ost3-containing OST complex is much more abundant (~80 % of the total OST) than the Ost6-containing complex [73]. In addition to OST subunits, LLO structures also modulate the hydrolytic activity of OST. It has been shown that the deletion of ALG6, which results in the production of Man₉GlcNAc₂-PP-Dol, causes hypoglycosylation [36] and a severe reduction in FNG levels [194]. However, the overproduction of the fully assembled LLO does not result in a proportional increase in FNGs and N-glycans [194], indicating that neither the hydrolysis nor the N-glycosylation is regulated by LLO levels.

Fig. 5.

Model for the metabolic regulation of the degradation of LLOs by LLO-PP’ase. Under normal-glucose conditions (left), GDP-Man is efficiently synthesized from glucose (Glc) and utilized by ALG2 in LLO biosynthesis. When glucose availability becomes low under glucose starvation (right), GDP-Man biosynthesis is downregulated, causing the arrest of the second ALG2 reaction by unknown mechanisms

The regulation of the hydrolytic activity of yeast OST appears to be quite complex. However, protozoan OST may serve as a powerful tool for dissecting the regulatory mechanism, because these organisms have a much simpler OST composition than yeast OST [1, 64–68, 77] and the OST has a relaxed specificity with respect to glycan donor substrates [66–68, 216, 217]. As discussed above, the genome of L. major encodes four STT3 genes (STT3A-D) as the sole OST subunit [65, 67, 68]. It has been reported that L. major Stt3D can be functionally expressed in S. cerevisiae [67, 68] and that non-glucosylated LLOs are utilized as donor substrates for this N-glycosylation [67]. Interestingly, it was found that L. major STT3D produces FNGs from Man₉GlcNAc₂-PP-Dol in alg6Δ cells where the endogenous OST barely generates FNGs [194]. The mutagenesis of L. major STT3D revealed that an Ala substitution at Glu102, which consists of a putative Asp–Xaa–Asp-like motif (¹⁰²Glu-Phe-Asp¹⁰⁴), causes the nearly complete loss of N-glycosylation in vivo, whereas the hydrolytic activity was reduced by only ~50 % [194]. To ensure normal N-glycosylation, the OST-mediated hydrolysis of LLOs should be tightly regulated. Further studies will be needed to clarify precisely how such regulation is achieved.

Processing and catabolism of FNGs produced in the ER lumen of S. cerevisiae

In S. cerevisiae, the Glc₃Man₉GlcNAc₂ released from the fully assembled LLO by OST is rapidly trimmed by ER glycosidases (Fig. 4a) [194]. Such glycan trimming can be considered to be a part of the catabolic pathway of FNGs, because no apparent catabolic α-glucosidases are produced by S. cerevisiae. Based on the glycan structures of FNGs identified in png1∆ ams1∆ cells (M7C, M8B and their Och1-modified forms, see Fig. 4b for their glycan structures), it appears that ER glucosidases (Gls1 and Gls2) and ER mannosidase (Mns1) are involved in the glycan trimming [194]. For the formation of the M7C isoform, the involvement of Htm1 is unlikely, because this mannosidase has a very low specificity for FNGs [200]. Although it is unclear how M7C is generated in an Htm1-independent manner, an LLO analysis indicated that the lipid-linked triglucosylated M8C (Fig. 4a) may be the source of this isoform. It should be noted that the luminal FNGs acquire the Och1 modification [194]. The fact that no Och1-modified LLOs were detected [194] suggests that Och1 directly modifies FNGs. This assumption is consistent with the observation that fluorescently labeled oligosaccharides can serve as acceptor substrates for Och1 in vitro [218].

What is the fate of luminal FNGs? It has been reported that luminal FNGs are intracellularly catabolized to Man₁GlcNAc₂ by the cytosolic form of Ams1 [194]. In this context, the ER-to-cytosol transport of FNGs should be required to allow the cytosolic Ams1 to gain direct access to substrates. Although a similar transport system has been identified in mammalian cells [127, 128], it is unclear whether mammals and yeast share a common molecular mechanism, as the gene encoding a FNG transporter has not yet been identified in either organism. In addition to the ER-to-cytosol transport, FNGs are subject to Och1 modification [194], suggesting that they may be transported to the Golgi via a secretory pathway. However, it is currently unknown whether Och1-modified FNGs are retrieved back from the Golgi to the ER or whether they are further transported out of the cells.

Generation of phosphorylated FNGs

The occurrence of phosphorylated FNGs was first reported by Hsu et al. [219]. It was found that, during the incubation of [¹⁴C]-mannose with microsomes from a mouse myeloma, a considerable amount of [¹⁴C]-mannose is converted into a soluble oligosaccharide, estimated to be Man₅GlcNAc₂-P. It was suggested that a phosphate group is attached to the reducing end of the innermost GlcNAc, since glucosaminitol was generated after sodium borohydride treatment only when the phosphate group was removed by treatment with alkaline phosphatase. Cacan et al. [220] subsequently showed that similar phosphorylated FNGs are produced in intact rat-spleen lymphocytes. It was also shown that an endo-H treatment of the phosphorylated FNGs generates GlcNAc-P, supporting the earlier idea that the phosphate group is attached to the reducing end of the GlcNAc residue. However, no direct experimental evidence to show the number of phosphate groups and the attachment site has yet appeared in the literature. Based on the structure of phosphorylated FNGs, it was proposed that LLOs are degraded by an unidentified LLO-pyrophosphatase (LLO-PP’ase) [220]. Indeed, such enzymatic activity has been detected in microsomes from yeast [221] and the human liver [222].

Although the occurrence of LLO-PP’ase has been known for more than three decades, the LLO-PP’ase gene has not yet been identified. This is largely due to the fact that the biological relevance of the degradation reaction is still unclear. However, recent studies have shown that the generation of phosphorylated FNGs occurs when LLO biosynthesis is impaired. Peric et al. [223] and Vleugels et al. [222] reported that phosphorylated FNGs are actively produced in cells derived from congenital disorders of glycosylation type I (CDG-I), but not in healthy donor cells. CDG-I is caused by mutations in genes that are involved in the biosynthesis of LLOs and their transfer to proteins [140]. It has been reported that the glycan structures of phosphorylated FNGs (Man_1–7GlcNAc₂-P) are highly correlated with those of LLO intermediates that are abnormally produced in CDG-I cells [222, 223]. Interestingly, it was found that phosphorylated FNGs are almost exclusively detected in the cytosol [222, 223], although the substrate LLOs are assumed to be distributed on the both sides of the ER membrane. Identification of the LLO-PP’ase gene will be required to answer this question.

It has long been known that, in mammalian cells, glucose starvation causes the arrest of the maturation of LLOs, thereby resulting in hypoglycosylation [224–229]. Quite recently, we reported that, in mouse embryonic fibroblasts (MEFs), glucose starvation results in a dramatic loss of LLOs and a substantial accumulation of Man₂GlcNAc₂-P in the cytosol [120]. Because only low levels of the phosphorylated form of FNG are generated under normal-glucose conditions, this indicates that the degradation of LLOs is tightly regulated by the availability of glucose. It was also found that glucose starvation results in a reduction of intracellular GDP-Man levels, and that the selective shutdown of the GDP-Man biosynthetic pathway is sufficient to induce the degradation of Man₂GlcNAc₂-PP-Dol, even when glucose levels are normal [120]. This observation clearly indicates that glucose starvation downregulates GDP-Man biosynthesis, which causes the arrest of the maturation of LLOs with the synthesis of Man₂GlcNAc₂-PP-Dol and facilitates its degradation by LLO-PP’ase (Fig. 5). Man₂GlcNAc₂-PP-Dol was found to be a biosynthetic intermediate synthesized by ALG2 (Fig. 1). This finding implies that whether or not the second ALG2 reaction proceeds may be a fate determinant for Man₂GlcNAc₂-PP-Dol to be degraded or subjected to biosynthesis. Although the mechanism underlying the fate determination is unknown, one plausible explanation is that the availability of intracellular GDP-Man is involved in this process (Fig. 5). Kinetic studies of each ALG2 reaction will be required to validate this attractive hypothesis.

In eukaryotes, reports on the occurrence of phosphorylated FNGs have been limited to mammalian cells thus far [120, 219, 220, 222, 223]. While LLO-PP’ase activity has been detected in microsomes prepared from S. cerevisiae [221], no experimental evidence has been provided to show that S. cerevisiae produces phosphorylated FNGs. The issue of how widespread is the distribution of LLO-PP’ase in eukaryotes remains unclear at present.

FNGs in other organisms

It should be noted here that there are also abundant reports of the occurrence of FNGs in organisms other than mammals and budding yeast.

Non-mammalian vertebrates

The detailed structures of cytosolic FNGs in the hen oviduct have been determined [148]. Major structures observed were high-mannose-type, Gn1-type FNGs, while Gn2-type high-mannose-type FNGs were also detected [148]. The major structure of FNGs was M5B′ (Fig. 6), and the structures of other FNGs were also found to be very similar to those found in mammalian cells [148, 174, 230]. The cytoplasmic PNGase [231], ENGase [149, 151, 232] and Man2C1 [164] in the hen oviduct have also been biochemically characterized.

Fig. 6.

Two glycan isoforms of Man₅GlcNAc₁ found in D. melanogaster, hen oviduct and mammalian cells. In D. melanogaster and C. elegans, M5A′ (left) is abundantly found in the soluble oligosaccharide fraction. In contrast, M5B′ (right) is dominant in the cytosol of hen oviduct and various mammalian cells. The number for each sugar indicates the order of their assembly in LLO biosynthesis

Sialylated, complex-type FNGs are found in hen’s egg yolk [233]. The structures have been determined to be α2-6-linked disialyl-biantennary glycans with either Gn1 or Gn2 types. Based on structural similarity, these sialylated complex-type FNGs are most likely derived from sialylglycopeptide (SGP), an abundant glycopeptides found in egg yolk. However, the responsible enzymes involved in the formation of sialyl FNGs have not been identified [233].

In fish, Gn2-type sialylated complex-type FNGs have also been identified in unfertilized eggs or embryos of various fish species [234–239]. PNGase activity is potentially involved in the formation of Gn2-type complex-type FNGs, which were identified in early embryos of the medaka fish [240]. The PNGase from medaka embryos requires an acidic pH for optimal activity, and is most likely a lysosomal enzyme and, therefore, distinct from the cytosolic PNGase [241]. In some cases, Gn1-type sialylated complex-type FNGs are also observed [234, 239]. In this case, ENGase activity, which is potentially responsible for the formation of the Gn1-type FNGs, was observed in Zebrafish embryos [239]. However, the nature of this enzyme remains unclarified.

Plants

It has long been proposed that FNGs have biological activities with respect to growth, differentiation, fruit ripening and senescence in plants [242–249], and FNGs have been identified in seedlings, stems, developing fruits and culture cells of various plants [250–264]. FNGs in plants are classified into two subclasses according to their structures: Gn2-type, plant complex-type FNGs bearing a core α1,3-linked fucose or a β1,2-linked xylose and Gn1-type, high-mannose-type FNGs. The former is believed to be generated by the action of a plant-specific acidic PNGase [246, 247, 252, 254, 265–277]. Interestingly, multiple genes encoding acidic PNGases can be found in plants—2 genes Arabidopsis thaliana and 6 genes in Lycopersicum esculentum—and it has been proposed that there may be two types of acidic PNGases present in different subcellular locations in plants [263]. In contrast, Gn1-type, high-mannose-type FNGs are mainly observed in the cytosol [261], and the cytosolic ENGase is believed to be responsible for their formation. ENGase activity has been widely reported in plants [246, 247, 254, 261, 264, 270, 272, 276, 278–287]. Arabidopsis thaliana possesses two ENGase genes, both of which are found to be active [264, 286]. Arabidopsis thaliana bearing mutations in both ENGase genes was generated [264], revealing that the conversion of Gn2-type to Gn1-type FNGs is suppressed in this mutant plant. It was also found that there is no significant reduction in the levels of FNGs, implying that the action of ENGase to glycoproteins or possibly LLOs may not be a dominant source of FNGs in this plant.

More recently, the third type of FNGs has been identified; Gn1-type, complex-type FNGs were found in a rice cell culture medium [263]. In their paper, the authors proposed two possible mechanisms for the occurrence of this unusual glycan structure in the extracellular space. One possibility concerns the existence of a putative plant-specific chitobiase, which catalyzes the hydrolysis of the chitobiosyl linkage of GlcNAcβ1,4(Fucα1,3)GlcNAc of plant-type, complex-type FNGs [263]. Another possibility is that the extracellular Gn1-type, complex-type FNGs may be originated from the cytosol: The cytosolic Gn1-type, high-mannose-type FNGs produced by ENGase are translocated to the ER lumen by unknown mechanisms, processed by Golgi-resident glycosidases and glycosyltransferases, and are then secreted to the extracellular space [263]. It will be extremely interesting to know how these glycans are formed and distributed in plants, especially because the cytosolic ENGase that has been characterized thus far cannot act on plant complex-type glycans [249].

Plants also have the gene encoding the cytosolic PNGase. Arabidopsis thaliana PNGase exhibits a PNGase activity when expressed in yeast [288, 289]. How much this enzyme contributes to the generation of FNGs in plants remains to be determined.

Insects

It has been shown that high-mannose-type FNGs with both Gn1 and Gn2 types are present in the soluble fraction of larva of Drosophila melanogaster [290]. The most abundant structure observed was Man₅GlcNAc₁ (M5A′, Fig. 6), which is structurally different from the one abundantly found in the cytosol of mammalian cells (M5B′, Fig. 6). When a mutant of pngl, a gene encoding an ortholog of the cytosolic PNGase in D. melanogaster, is analyzed, a similar amount of FNGs are observed to wild-type flies, strongly indicating that most of the FNGs in flies are formed in a PNGase-independent manner.

Worms

A detailed structural analysis of FNGs has been carried out in Caenorhabditis elegans [291, 292]. In the wild-type worms, all of the FNGs detected were found to be the high-mannose-type, 94 % of which were the Gn1 type [291]. The most abundant structure observed, as was the case with D. melanogaster, was M5A′ (Fig. 6) [291]. When a mutant of the gene encoding the cytosolic ENGase [150] was generated, major FNGs became the Gn2 type, together with some reduction (<50 %) in the total amount of FNGs, clearly indicating that the cytosolic ENGase is involved in their catabolism [291]. It is interesting to note that ~30 % of FNGs detected still remained as the Gn1 type. Enzyme(s) responsible for the formation of these Gn1-type FNGs remains unclarified. In C. elegans, the cytosolic PNGase has been shown to be dual functional as PNGase and thioredoxin [293, 294]. While mutants for the gene encoding the PNGase has been isolated as the ones exhibiting the abnormal axon branching of the specific neuron [295], FNGs in this mutant have not yet been analyzed.

Bacteria

N-Glycosylation reactions are not restricted to eukaryotes, but occur over all domains of life including bacteria and archaea [77]. The key catalytic enzymes (STT3 in eukaryotes; PglB in bacteria; AglB in archaea) are orthologous to each other. As discussed above, the first bacterial N-glycosylation system encoded by the pgl gene cluster was described in C. jejuni [78]. In C. jejuni, FNGs comprise as much as 2.5 % of the dry cell weight [296], and the ratio of C. jejuni FNGs to N-glycans was determined to be 10:1 under standard laboratory growth conditions [297], indicating that the hydrolytic activity of PglB proteins is relatively efficient, compared to yeast OST [194]. It was also shown that Campylobacter species contain an N-glycosylation system and also release structurally identical FNGs, indicating that the formation of FNGs is a general phenomenon for these bacteria [296, 298]. Since the FNGs produced by C. jejuni are not detected in the culture medium [82], it is possible that they are not secreted into the extracellular space. FNGs cannot be detected in other pgl mutants producing smaller N-glycan structures, while the N-glycosylation of proteins has been observed [299].

Interestingly, in addition to neutral FNGs, mono- and di-phosphorylated FNGs are also detected by MS analysis in Campylobacter lari [296, 298], while the exact site for the phosphorylation has not yet been fully determined. Whether the similar phosphorylated FNG-releasing system from lipid-linked oligosaccharides found in mammals exists in bacteria remains to be seen.

Perspectives

Extensive efforts in the last decade have resulted in a better understanding of the mechanisms responsible for the formation and processing of FNGs in eukaryotic and prokaryotic cells. In particular, the hypothetical hydrolytic activity of OST has now been experimentally demonstrated both in C. jejuni and S. cerevisiae. Major questions in this context include whether the OST-mediated hydrolysis of LLOs is a universal phenomenon, and how OST balances the hydrolysis and N-glycosylation. To address these issues, an empirical validation of the hydrolytic activity of OST in each organism and the identification of key amino acid residues or co-factors defining the balance, as well as a crystallographic analysis of the OST-LLO-peptide complex, will be needed. Another important issue concerns the unidentified components involved in the formation and transport of FNGs, such as LLO-PP’ase and ER/lysosome transporters for FNGs. The action of LLO-PP’ase appears to be limited under normal conditions, presumably to avoid the undesired degradation of LLOs, and may have important roles under patho/physiological conditions, such as CDG-I and glucose starvation. FNG transporters are closely involved in the catabolism of FNGs. The identification of their genes will open a new avenue for researches to investigate the biological importance of FNG production and catabolism.

The issue of whether FNGs or their catabolic pathways have any biological functions remains an unanswered question. We believe that a careful phenotypic analysis of organisms that have impairments in FNG catabolism should be carried out to elucidate the function of FNGs and their catabolism. For example, the mutation of ENGase in C. elegans dramatically changes the FNG pool from Gn1 to Gn2 structures and this structural change slightly shortened the life span [291]. Interestingly, the supplementation of GlcNAc has been shown to prolong the life of C. elegans by improving ER protein quality control, increasing ERAD activity, and inducing autophagy [300]. Therefore, it is tempting to speculate that the impairment of the Gn2-to-Gn1 conversion of FNGs in mutant worms may reduce the GlcNAc pool, thereby somehow affecting their life span. This attractive hypothesis, however, requires further experimental validations. In the case of C. jejuni, the deletion of pglB, which causes the loss of LLO-hydrolytic and N-glycosylation activity, renders the cells susceptible to osmotic stress [82]. It, therefore, would be extremely interesting to determine which enzymatic activity (i.e., LLO hydrolysis vs N-glycosylation) is important for osmotic regulation. Regarding non-lysosomal glycan degradation, the knockout of Man2C1 in mice apparently causes damage to multiple tissues, which are accompanied by the accumulation of considerable levels of Gn1-type of FNGs with untrimmed mannose residues [166]. As Man2C1 has both enzyme-dependent [168] and -independent [169] functions, knock-in studies will be necessary to test whether an enzyme-dead form of Man2C1 can complement the phenotypes in Man2C1-null mice.

As discussed above, PNGase (NGLY1) biology has attracted widespread interest of researchers in diverse fields since patients harboring mutations in this gene have been identified. The development of clinical approaches for the treatment of these patients is an urgent issue, while the knowledge of physiological functions of PNGase has been quite limited in mammals. Further basic researches directed to the functions of PNGase are required to accelerate clinical research. In any sense, it is imperative to note here that before the discovery of NGLY1 genetic deficiency patients, there were very few who believed that the cytoplasmic PNGase is a biologically important molecule, especially because png1 mutant yeast cells did not exhibit any significant phenotypes [197]. The unexpected findings of the importance of NGLY1 serve as good testimony that we will never be able to predict the biological significance of molecules involved in novel biological pathways, and we, therefore, should be prepared to make tenacious efforts to identify new genes that are involved in the generation and degradation of FNGs. We feel fortunate that there are so many interesting questions remaining to be explored regarding this subject.

It should also be noted that, among the distinct FNG generation pathways, a dominant one can be completely different between mammalian cells and S. cerevisiae, i.e., the PNGase-independent pathway appears to be the predominant path for the former, while most FNGs are generated by PNGase in S. cerevisiae. It is, therefore, important to clarify the mechanism underlying the generation and degradation of FNGs in different organisms, which promises to provide deeper insights into the evolutional conservation, as well as the biological importance, of these processes.