Oligosaccharide-based Information in Endoplasmic Reticulum Quality Control and Other Biological Systems*

Oligosaccharides as Information Carriers Glycosidically linked sugar polymers are well known as nutritional and structural molecules, and it is now clear that they also have essential roles as carriers of biological information. The constituent sugar residues contain multiple hydroxyl groups capable of forming complex arrangements of hydrogen bonds. Sugars are often modified with amino, N-acetyl, carboxyl, phosphate, and sulfate groups, permitting more varied interactions than those achieved with hydroxyls. Many different sugars occur in nature, and these can be coupled in numerous ways through a or b glycosidic linkages of their hydroxyls to form oligosaccharides (with relatively few sugars) and polysaccharides (with many sugars) (1). For example, there are eight different ways to couple the anomeric carbon (no. 1) of one residue of glucose to the nonanomeric carbons (no. 2, 3, 4, or 6) of another. Sugar polymers are also distinguished from other biological polymers by facile formation of both linear and branched structures. For example, the b1,4-linked mannose residue in asparagine (N)-linked oligosaccharides is always linked to at least three other sugars as indicated in Fig. 1. Oligosaccharides that carry information are usually coupled to chemically distinct units termed aglycones that themselves are not carbohydrates, but typically are proteins or lipids, and whose biological properties can be dramatically changed by the oligosaccharide. The purpose of this minireview is to explore the roles of oligosaccharides as carriers of intraand intercellular information with emphasis on the relationships between oligosaccharide metabolism, quality control, and stress responses of the endoplasmic reticulum (ER).

low). Many glycoproteins require Glc 1 Man 9 GlcNAc 2 -dependent interactions with CNX or CRT for efficient folding, assembly, and export from the ER (3,4,7). CNX (8) and CRT (9) are clearly lectins, and Glc 1 Man 9 GlcNAc 2 can bind directly whether it is free or linked to protein. Association constants for CNX are in the range of 4 -5 ϫ 10 5 M Ϫ1 (10). Glc 1 Man (5-9) GlcNAc 2 , but not Glc 1 Man 4 GlcNAc 2 , bind to CNX (11) and CRT (9), and the three mannosyl residues that form the "arm" to which glucose is attached also contribute to binding (11). Because only oligosaccharides with a single ␣-1,3-linked glucosyl residue bind, oligosaccharide ligands for these lectin-chaperones are referred to collectively as monoglucosylated oligosaccharides.
In addition to their lectin activities, both CNX (12) and CRT (13) have efficient oligosaccharide-independent chaperone activities in vitro. There is abundant evidence that many glycoprotein folding intermediates first interact with CNX and CRT in a lectin-dependent manner, followed by formation of oligosaccharide-independent complexes (3,8). It is probable that the chaperone activities of CNX and CRT contribute to glycoprotein folding in such complexes. Oligosaccharide-independent interactions with CNX and CRT would likely involve hydrophobic surfaces on glycoprotein folding intermediates. However, such hydrophobic interactions cannot be easily distinguished experimentally from irrelevant hydrophobic interactions that would also be anticipated with partially unfolded glycoproteins. Hence, the two-state mechanism has been difficult to prove (3). The situation has been complicated further by reports of some "lectin-only" and "lectin-independent" complexes of glycoprotein folding intermediates with CNX (3). It is possible that the mechanism used depends upon the specific glycoprotein in question. Resolution of this controversy may require the determination of three-dimensional structures of lectin-chaperone⅐glycoprotein complexes.
Glc 1 Man 9 GlcNAc 2 is an excellent substrate for the Golgi endomannosidase, which releases a Glc␣1,3Man disaccharide to yield * This minireview will be reprinted in the 2001 Minireview Compendium, which will be available in December, 2001 1 The abbreviations used are: ER, endoplasmic reticulum; AT, antitrypsin; CNX, calnexin; CRT, calreticulin; CSN, castanospermine; eIF, eukaryotic initiation factor; GPD, glucose-P-dolichol; GPT, GlcNAc-1-P transferase; GRP, glucose-regulated protein; LLO, lipid-linked oligosaccharide; MPD, mannose-P-dolichol; PERK, PKR-like ER kinase; TN, tunicamycin; UPR, unfolded protein response. 2 The term "lectin-like chaperone" was originally introduced into the literature when the binding of CNX and CRT to glycoproteins was known to require specific oligosaccharide-dependent interactions, but the lectin activities for CNX and CRT remained to be demonstrated. Because the lectin activities of CNX and CRT have now been proven, the term "lectin-chaperone" will be used. the A-isomer of Man 8 GlcNAc 2 ( Fig. 1) (4). Because of the Golgi apparatus location of the endomannosidase (14), glycoproteins with Glc 1 Man 9 GlcNAc 2 that have escaped further glycosidic processing, but have been properly folded and exported from the ER, can be digested to enable Golgi-type processing (4).
Man 9 GlcNAc 2 -The presence of this oligosaccharide, formed by glucosidase II digestion of Glc 1 Man 9 GlcNAc 2 , indicates that the glycoprotein in question should be inspected for exposure of hydrophobic surfaces not present in the native glycoprotein. If such surfaces are found, the oligosaccharide is enzymatically reglucosylated to regenerate Glc 1 Man 9 GlcNAc 2 for an additional round of binding to CNX/CRT. Reglucosylation is carried out by a single remarkable ER resident enzyme, UDP-glucose:unfolded glycoprotein glucosyltransferase (3). The existence of this enzyme was originally suggested by reports of direct transfer of glucose from UDP-glucose to glycoproteins (15,16). The product, Glc 1 Man 9 GlcNAc 2 , is the same structural isomer as that achieved by glucosidase II processing of Glc 2 Man 9 GlcNAc 2 (17). Although Man 9 GlcNAc 2 oligosaccharides on improperly folded glycoproteins are good substrates, Man 9 GlcNAc 2 on properly folded glycoproteins and free Man 9 GlcNAc 2 oligosaccharides are poor substrates for the glucosyltransferase. Thus, the enzyme has a catalytic site for glucose transfer and a separate site that interacts with non-native surfaces on glycoprotein folding intermediates. For some misfolded glycoproteins, several rounds of reglucosylation-deglucosylation can occur (3).
Man 9 GlcNAc 2 is also the preferred oligosaccharide on coagulation factors V and VIII (18) and a cathepsin Z-related protein (19) needed for interactions with ERGIC-53, a lectin that is a resident of the ER-Golgi intermediate compartment and is involved in trafficking of these glycoproteins. In both cases interactions with ER-GIC-53 were hindered by treatments with CSN but not with the ER mannosidase I inhibitor deoxymannojirimycin.
Man 8 GlcNAc 2 -The presence of Man 8 GlcNAc 2 on a misfolded glycoprotein indicates that it should be degraded. The B-isomer of Man 8 GlcNAc 2 is generated by digestion of Man 9 GlcNAc 2 by a kifunensine-and deoxymannojirimycin-sensitive ER mannosidase I (Fig. 1). For most glycoproteins digestion by ER mannosidase I is the last step in glycan processing before export to the Golgi apparatus and occurs at a time when folding and assembly should be complete. Abundant biochemical (Refs. 20 and 21, and references therein) and genetic (22) evidence implicates Man 8 GlcNAc 2 in the degradation of many types of misfolded glycoproteins by cytoplasmic proteasomes, whereas properly folded glycoproteins bearing Man 8 GlcNAc 2 escape degradation. Two models have been proposed, distinguished operationally by whether or not CNX binding stabilizes the glycoprotein. As first shown with myosin class I heavy chains (23) and subsequently with other systems (Refs. 20 and 21, and references therein), CNX binding stabilizes many misfolded glycoproteins. It has been suggested that degradation might involve interaction with a Man 8 GlcNAc 2 -specific lectin, which remains to be identified. This model is supported by the finding that misfolded carboxypeptidase Y bearing Man 9 GlcNAc 2 , Man 7 GlcNAc 2 , or Man 6 GlcNAc 2 was more stable than with Man 8 GlcNAc 2 (22). Degradation of misfolded variants of ␣ 1 -antitrypsin (AT) in hepatoma cells also requires formation of Man 8 GlcNAc 2 (24). However, ␣ 1 -AT variants are destabilized by interaction with CNX. It has been proposed (24) that formation of Man 8 GlcNAc 2 -␣ 1 -AT promotes CNX binding and accelerates degradation of ␣ 1 -AT because, after reglucosylation, Glc 1 Man 8 GlcNAc 2 -␣ 1 -AT is predicted to be more resistant to glucosidase II digestion than Glc 1 Man 9 GlcNAc 2 -␣ 1 -AT. This model is based upon the observation that the relative activity of glucosidase II toward free oligosaccharides is Glc 1 Man 9 GlcNAc Ͼ Glc 1 Man 8 GlcNAc Ͼ Glc 1 Man 7 GlcNAc (25). In the same study glucosidase II activity was found to be strongly influenced by the attachment of substrate oligosaccharides to protein. Thus, it would be interesting to extend these results with purified glucosidase II and purified glycoproteins bearing known Glc 1 Man x GlcNAc 2 structural isomers. Recent studies have implicated Man 7 GlcNAc 2 , the ER mannosidase II product of Man 8 GlcNAc 2 , in a distinct mechanism involving nonproteasomal degradation of the ␣ 1 -AT variant PI Z (26).

Endoplasmic Reticulum Oligosaccharide Quality Control
The preceding section reviewed the role of N-linked Glc 3 Man 9 GlcNAc 2 in quality control of nascent ER glycoproteins. Conversely, cells have several strategies for maintaining the quantity of Glc 3 Man 9 GlcNAc 2 -P-P-dolichol, the direct precursor of Nlinked Glc 3 Man 9 GlcNAc 2 (Fig. 2). Glc 3 Man 9 GlcNAc 2 -P-P-dolichol synthesis occurs in a stepwise manner from dolichol-P, requiring four donor substrates in this order: UDP-GlcNAc (2 eq), GDPmannose (5 eq), mannose-P-dolichol (MPD; 4 eq), and glucose-Pdolichol (GPD; 3 eq). MPD and GPD are formed by transfer of mannose or glucose from GDP-mannose or UDP-glucose, respectively, to dolichol-P. Each sugar residue in Glc 3 Man 9 GlcNAc 2 -P-Pdolichol is added by a separate transferase (5). This section will review regulation of priming of dolichol-P by its conversion to GlcNAc-P-P-dolichol, extension of LLO intermediates, and transfer of Glc 3 Man 9 GlcNAc 2 from its dolichol-P-P carrier to appropriate asparaginyl residues in proteins.
Priming and Extension-LLO synthesis is primed by UDP-Glc-NAc:dolichol-P GlcNAc-1-P transferase (GPT), which transfers GlcNAc-1-P from UDP-GlcNAc to dolichol-P to yield GlcNAc- P-P-dolichol (27). Tunicamycin (TN) is a selective inhibitor of GPT (27), although high concentrations also inhibit protein palmitoylation (28). Several forms of regulation of GPT have been reported (27). Considerable information is available for the stimulation of GPT in vitro by exogenously added MPD (29). Stimulation can be achieved with the physiological ␤-isomer of MPD but not with the ␣-isomer (30) or with GPD (31). GlcNAc-P-P-dolichol was recently shown to stimulate MPD synthase in vitro (29). The reciprocal stimulations of GPT and MPD synthase by their products may help maintain a balance of LLO substrates; GlcNAc-P-P-dolichol is needed for priming, and MPD is needed for extension. The intriguing possibility remains that related regulatory relationships exist for GPD synthesis.
Priming is dependent upon the dolichol-P concentration because LLO synthesis in cultured cells is increased by supplementation with dolichol-P (27). The concentration of dolichol-P, also required for the MPD and GPD synthases and hence extension of LLO intermediates, is regulated in response to various biological stimuli associated with increased synthesis of N-linked glycoproteins by cis-isoprenyltransferase, the enzyme system catalyzing the chain elongation stage in de novo dolichol-P biosynthesis (32)(33)(34). cis-Isoprenyltransferase regulation is most likely at the level of mRNA synthesis, but direct transcriptional studies remain to be performed. LLO extension can also be controlled by the unfolded protein response (below) (35).
Oligosaccharide Transfer-Oligosaccharide transfer from Glc 3 Man 9 GlcNAc 2 -P-P-dolichol to appropriate asparaginyl residues of nascent proteins is catalyzed by the enzyme oligosaccharyltransferase (5). Even under conditions where Glc 3 Man 9 GlcNAc 2 -P-P-dolichol represents a minor fraction of the steady-state LLO pool, the selectivity of oligosaccharyltransferase for Glc 3 Man 9 GlcNAc 2 -P-P-dolichol ensures that most nascent proteins will be modified with Glc 3 Man 9 GlcNAc 2 . As an example of the importance of oligosaccharyltransferase selectivity in quality control, studies from the author's laboratory with primary human dermal fibroblasts showed that if Glc 3 Man 9 GlcNAc 2 -P-P-dolichol was only 10% of the LLO pool, it still corresponded to essentially all of the transferred oligosaccharide (35). However, lower percentages resulted in transfer of incomplete oligosaccharide intermediates that caused an unfolded protein response (see below). Mechanisms also exist to eliminate excess Glc 3 Man 9 GlcNAc 2 -P-P-dolichol (36,37).

Defective Synthesis or Processing of ER Information-carrying Oligosaccharides Triggers the Unfolded Protein Response (UPR)
Interference with protein folding, assembly, or misfolded protein degradation in the ER triggers multiple aspects of the UPR (38 -40). The first UPR aspect to be established involves transcriptional activation of genes encoding ER lumen chaperones and enzymes involved in protein folding. Reports of transcriptional control now exist for many genes, including those needed for ER maintenance and ER-associated degradation (41,42). UPR regulation of transcription occurs in all eukaryotes examined so far. A novel ER-tonucleus signaling pathway in Saccharomyces cerevisiae has been elucidated, involving the ER transmembrane kinase Ire1p/Ern1p. Corresponding pathways in mammalian cells are more complex because at least two forms of Ire1p, ␣ (43) and ␤ (44), and the transcription factor ATF6 (45) are involved. All three are synthesized as ER transmembrane proteins. There is no evidence for proteolytic processing of Ire1p in S. cerevisiae. By comparison, activation of the UPR in mammalian cells causes release of the cytoplasmic domains of the mammalian Ire1 proteins (46) and ATF6 (45) by "regulated intramembrane proteolysis" (47). The released cytoplasmic domains then enter the nucleus to facilitate transcription of mammalian UPR genes. The cytoplasmic domain of ATF6, which clearly becomes a transcription factor, is released by the site 1 and site 2 SREBP proteases (48), raising the interesting possibility of a relationship between sterol metabolism and the UPR. Mammalian cells with various presenilin-1 mutations have been reported to have impaired Ire1p function (46,49), suggesting a potential role for presenilin-1 in Ire1p cleavage, although contradictory results have recently appeared (50). The cytoplasmic domains of the Ire1 proteins are bifunctional kinases-endonucleases that undergo autophosphorylation upon UPR activation. In S. cer-evisiae the Ire1p endonuclease activity propagates the UPR by participating in splicing of the mRNA encoding the transcription factor Hac1p (51). A similar chain of events appears to occur in mammalian cells (46), although the endonuclease substrate remains to be identified.
Biochemical evidence exists for other UPR aspects in mammalian cells: (i) inhibition of translation because of phosphorylation of eIF2␣ by the transmembrane PKR-like ER-associated kinase "PERK" (52), (ii) apoptosis involving CHOP (53) and ER-associated caspase-12 (54), and (iii) enhanced extension of dolichol-linked oligosaccharide intermediates (35). Translational inhibition aids in resistance to ER stress (55) by lowering the load of misfolded protein and by allowing the selective expression of genes involved in amino acid metabolism (56). PERK is also responsible for cell cycle arrest due to the UPR (57). The mechanisms of activation for PERK and Ire1p by misfolded proteins are similar (58). Under normal conditions the lumenal domains of PERK and Ire1p interact dynamically with BiP. Misfolded proteins that accumulate in the ER lumen form complexes with BiP, lowering the amount of BiP available for interaction with Ire1p and PERK. In the absence of bound BiP, the lumenal domains dimerize. Consequently, the cytoplasmic kinase domains are brought into proximity, allowing autophosphorylation and activation of downstream events.
All aspects of the UPR can be caused by interference with the synthesis of N-linked oligosaccharides. (i) Inadvertent depletion of glucose from culture medium, associated with deficient protein glycosylation, led to the discovery of stress induction of GRP78 (BiP) and GRP94 as well as their designations as glucose-regulated proteins (59,60). (ii) TN, which blocks protein N-glycosylation by inhibiting GPT, is a highly potent UPR inducer causing transcription of BiP, phosphorylation of eIF2␣, inhibition of translation, and cell death. The effect of the UPR on the dolichol pathway cannot be studied with TN due to direct inhibition. Commercial TN preparations typically contain mixtures of homologues that block N-glycosylation but paradoxically vary in their abilities to inhibit translation (61). It remains to be determined whether these homologues also vary in other aspects of UPR activation. (iii) CSN, an inhibitor of ER glucosidases I and II, eliminates aspects of ER quality control requiring processing of asparagine-linked Glc 3 Man 9 GlcNAc 2 . CSN and other glucosidase inhibitors are less potent UPR inducers than TN, because their use does not cause inhibition of translation (62) or cell death (63). This suggests that the N-linked Glc 3 Man 9 GlcNAc 2 oligosaccharide itself contributes to glycoprotein folding and assembly, perhaps by enhancing the hydrophilicity of folding intermediates. However, CSN treatment does activate BiP transcription and the dolichol pathway (35), confirming the importance of N-linked Glc 0 -2 Man 9 GlcNAc 2 in ER protein quality control. (iv) A selection for S. cerevisiae mutants with activated UPRs yielded cells with defects in genes required for N-glycosylation (64).

N-Linked Glycosylation as an Aspect of UPR
Screens for S. cerevisiae genes activated by the UPR not only revealed genes encoding various ER chaperones and enzymes involved in protein folding and assembly but also genes involved in other functions associated with ER maintenance including the synthesis and processing of Glc 3 Man 9 GlcNAc 2 -protein ( Fig. 2) (41). This evidence is supported by biochemical studies from the author's laboratory (35). Activation of the UPR in primary cultures of adult dermal fibroblasts stimulated extension of LLO intermediates to Glc 3 Man 9 GlcNAc 2 -P-P-dolichol and increased the fraction of nascent ER proteins bearing Glc 3 Man 9 GlcNAc 2 . The metabolic step(s) in LLO synthesis activated by the UPR has yet to be identified.

Insights from Other Biological Systems: Stable and
Transient Forms of Oligosaccharide Information A survey of selected biological functions of oligosaccharides reveals that oligosaccharides can carry information by either of two modes, stable or transient. Oligosaccharides carrying stable information are not altered in the course of serving a biological role and are usually directly involved in the function of the molecule or cell to which they are attached or from which they originate. Examples include sialyl Lewis X -like glycans bound by selectins (65), sialylconjugates recognized by siglecs (66), and sulfated GalNAc on glycoprotein hormones that interact with specific clearance receptors (67). Chitin-like Nod factors secreted by Rhizobia are recognized by specific receptors on the root hairs of legumes to promote nodulation (68,69). Recently, Notch was shown to be modulated by its GlcNAc␤ 1,3Fuc O-glycan, formed by the glycosyltransferase activity of Fringe (70,71).
In contrast, oligosaccharides that carry transient information undergo structural changes during the course of biological activity and may not necessarily be involved in the ultimate biological functions of the molecules or cells that bear them. Therefore, the critical modification might not be detected by biochemical analysis of the mature glycoconjugate. The glucose residues on N-linked oligosaccharides, which as noted by Spiro (4) are transient, are definitive examples. Another example is the lysosomal sorting signal mannose-6-P found on high mannose oligosaccharides of lysosomal enzymes, which is dephosphorylated once the lysosome has been reached (72). Monosaccharides can also carry information transiently. O-GlcNAc on cytoplasmic and nuclear proteins can be removed and replaced many times during the life of a protein (73).

Summary and Perspectives
This article attempted to demonstrate how oligosaccharides can carry biological information. The large variety of oligosaccharide structures that can be created by just a few glycosidically linked monosaccharides, and the potential for ionic as well as hydrogen bonds, makes them highly suitable for this purpose. In particular, novel oligosaccharides and oligosaccharide modifications are very likely to carry essential information and should never be ignored. Because they can perform so many functions, there is no way to predict how oligosaccharides may be involved in any particular biological system under study. For this reason, a fruitful area for future research will be the development of new reagents and techniques that can identify information-carrying oligosaccharides in action.