Proc. Nati. Acad. Sci. USA Vol. 91, pp. 913-917, February 1994 Cell Biology Role of N-linked oligosaccharide recognition, glucose trimming, and calnexin in glycoprotein folding and quality control (Ifluenza hemagglutinio/vesicular stomafitis virus glycoproten/glucosidasc lnhlbitors/chaperones/endoplasmic reticulum) CRAIG HAMMOND, INEKE BRAAKMAN*, AND AmI HELENIUS Department of Cell Biology, Yale School of Medicine, 333 Cedar Street, New Haven, CT 06510 Communicated by Richard D. Klausner, September 24, 1993

ABSTRACT Using a pulse-chase approach combined with brane-bound ER chaperone called calnexin (or p88 or 1P90). immunoprecipitation, we showed that newly synthesized influ- Calnexin, a 64-kDa type I membrane protein, has been shown enza virus hemagglutinin (HA) and vesicular stomatitis virus G to interact transiently with a variety ofmembrane and soluble protein ssociate transiently during their folding with calnexin, glycoproteins in the ER, and it is thought to have a chaper- a membranebound endoplasmic reticulum (ER) chaperone. one-like function in the retention and oligomeric assembly of Inhibitors ofN-linked glycosylation (tunicamycin) and glucosi- several important membrane glycoproteins (10-15). The re- dases I and II (castanospermine and 1-deoxynojirimycin) prew sults reported here lead us to propose a hypothesis which vented the association, whereas inhibitors of ER a-mannosi- includes calnexin as one of several key players in a folding dases did not. Our results indicated that binding of these viral and quality control system. The model helps explain why glycoproteins to calnexin correlated closely with the composi- proteins need carbohydrates for maturation and why tion oftheir N-linked oligosaccharide side chains. Proteins with N-linked oligosaccharides undergo trimming in the ER. monoglucosylated oligosaccharides were the most likely bind- ing species. On the basis of our data and existing information concerning the role of monoglucosylated oligosaccharides on MATERIALS AND METHODS glycoproteins, we propose that the ER contains a unique folding Cell Lines, Viruses, and Reagents. The X31 strain of influ- and quality control machinery in which calnein acts as a enza virus, wild-type VSV, and the temperature-sensitive chaperone that binds proteins with partially glucose-timmed ts045 mutant of VSV were all propagated and used as carbohydrate side chains. In this model glucosidases I and H previously described (16, 17). Chinese hamster ovary (CHO) serve as signal modifiers and UDP-glucose:glycoprotein gluc- 15B cells are defective in N-acetylglucosaminyltransferase I osyltransferase, as a folding sensor. and thus allow trimming of N-linked carbohydrates to Man5GlcNAc2 without further modification (17, 18). All The lumen of the endoplasmic reticulum (ER) provides a reagents were purchased from Sigma with the exceptions of highly specialized compartment for the folding and oligo- 3-[(3-cholamidopropyl)dimethylammonio]-l-propanesulfo- meric assembly of secretory proteins, plasma membrane nate (CHAPS) from Pierce and 1-deoxynojirimycin and proteins, and proteins destined for the various organelles of 1-deoxymannojirimycin from Boehringer Mannheim. A mix- the vacuolar system. Their conformational maturation is a ture of [35S]methionine and [35S]cysteine used for metabolic complex process determined not only by the amino acid labeling was purchased from ICN. The rabbit anti-calnexin sequence but also by post- and cotranslational modifications, antibody, raised against a C-terminal peptide, was kindly by the intralumenal milieu, and by a variety of chaperones provided by John J. M. Bergeron (McGill University). and folding enzymes present (for recent reviews, see refs. 1 Metabolic Labeling and Immunoprecipitation. Infection of and 2). The ER possesses an efficient but still poorly under- CHO and CHO 15B cells with X31 influenza virus or VSV stood "quality control" system to ensure that transport is and subsequent metabolic labeling have previously been limited to properly folded and assembled proteins (3). described in detail (16, 17). With the exception of the exper- Of the covalent modifications that occur in the ER, the iment shown in Fig. 5, cells were metabolically labeled for 1 cotranslational addition of N-linked oligosaccharides in the hr prior to infection to allow detection of calnexin, which is form of a 14-saccharide core unit (Glc3Man9GlcNAc2) is one the short times used for viral ofthe most common (4). Glycosylation inhibitors, mutant cell not efficiently labeled with pulse lines, and site-specific mutagenesis of consensus glycosyla- proteins. Cells were lysed and precipitated with anti-calnexin tion sequences have shown that it is crucial for the folding of antibodies as described (15). Viral proteins were precipitated many, but not all, glycoproteins (5, 6). Without added sugars, as described (16). SDS/PAGE, fluorography, and quantita- proteins misfold, aggregate, and get degraded without trans- tion was performed as described (16). port to the Golgi complex. For optimal folding and secretion, Glycosidase Digestion. After immunoprecipitation of sam- some glycoproteins must undergo a series of early trimming ples, the wash buffer was removed by washing the staphy- steps involving the removal of the three terminal glucose lococcal protein A-Sepharose beads once with deionized residues (7-9). Trimming, catalyzed by glucosidases I and II, water. Then 20 Al of0.05 M sodium citrate buffer, pH 4.4, and begins on the nascent chain and is followed by a series of ER 0.28 unit of jack bean a-mannosidase were added to each a-mannosidase cleavages (4). sample. Samples were incubated at 370C for 24 hr before As part of our studies on the folding of influenza hemag- analysis by SDS/PAGE. glutinin (HA) and vesicular stomatitis virus (VSV) G protein in living cells, we have analyzed the interaction between Abbreviations: ER, endoplasmic reticulum; dNM, 1-deoxynojirimy- these well-characterized viral glycoproteins and a mem- cin; HA, influenza virus hemagglutinin; VSV, vesicular stomatitis virus; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-pro- panesulfonate. The publication costs of this article were defrayed in part by page charge *Present address: University of Amsterdam, Department of Bio- payment. This article must therefore be hereby marked "advertisement" chemistry, Academic Medical Center, Meibergdreef 15, 1105 AZ in accordance with 18 U.S.C. §1734 solely to indicate this fact. Amsterdam, The Netherlands. 913 Downloaded by guest on September 23, 2021 914 Cell Biology: Hammond et al. Proc. Natl. Acad Sci. USA 91 (1994) RESULTS with anti-calnexin. Precipitation ofNT was only partial (25% at 0-min chase). As already shown for other glycoproteins To analyze the interaction between the viral glycoproteins (10, 15, 21, 22), it was apparent that HA associated tran- and ER chaperones, we used a pulse-chase approach already siently with calnexin. Quantitation by densitometry (Fig. ic) extensively utilized in our laboratory to study the folding and indicated that the til2 for calnexin binding of HA was 5 min, oligomerization of HA and G protein (16, 17). Both are type with minimal binding observed after 15 min. For G I membrane proteins with the majority of their mass in the protein, lumen ofthe ER. HA has seven N-linked carbohydrates and the binding peaked about 5 min after the pulse and dropped six intrachain disulfides (19), whereas G protein has two rapidly thereafter. These results showed that the viral pro- N-linked carbohydrates and, most likely, seven intrachain teins do not undergo initial folding as free monomers in disulfides (20). CHO 15B cells were first labeled with P5S]me- solution, but they do so in association with calnexin. thionine for 60 min to incorporate radioactivity into resident Recent studies by Bergeron and co-workers (15) have lumenal proteins. The cells were then infected with virus and, shown that in HepG2 cells only glycoproteins associate with 5 to 6 hr after infection, pulsed with [35Simethionine for 2-3 calnexin. This conclusion was supported by our observation min and chased in medium containing 5 mM unlabeled me- that tunicamycin, a glycosylation inhibitor, blocked the as- thiomne and0.5 mM cycloheximide. At different chase times, sociation of HA and G protein with calnexin (Fig. 2). Since the cells were treated with a membrane-penetrating alkylat- both proteins are unable to fold properly when synthesized in ing agent, N-ethylmaleimide, to block any free sulfhydryl the presence of tunicamycin (23, 24), it was apparent that groups, and lysed with a zwitterionic detergent, CHAPS. The incomplete folding alone is not sufficient for their attachment postnuclear supernatant was immunoprecipitated by using to calnexin; the proteins must also have N-linked carbohy- antibodies to the viral glycoproteins or calnexin. drates. When analyzed under nonreducing conditions, the anti-HA This raised the possibility that calnexin binds to the precipitates gave a series of heterogeneous bands on SDS/ N-linked oligosaccharides of newly synthesized proteins. To PAGE faster than the fully reduced proteins (Fig. la). explore it further, we tested the effects of castanospermine According to the previously established nomenclature (16), and 1-deoxynojirimycin (dNM), which inhibit glucosidases I the incompletely oxidized folding intermediates were labeled and II and thus prevent the trimming of the three glucoses IT1 and IT2, while the band corresponding to the fully from the core oligosaccharide (7). The mobility differences oxidized HA was labeled NT for native. The glycoproteins displayed by the HA bands in inhibitor-free controls (Fig. 3a, that had reached the medial Golgi apparatus or beyond were lanes 1 and 2) and castanospermine- and dNM-treated cells distinguished as a separate faster-mirating band labeled G. (lanes 3 and 4, 5 and 6, respectively), showed that both As previously described, the shift from fT1 and 1T2 to NT, compounds inhibited the trimming of N-linked sugars. In which occurs within the first 10 min of chase, is caused by both reduced and nonreduced samples, the mobility of HA disulfide bond formation and folding of the glycopolypeptide was slower than in the controls. Castanospermine had a chains (16). The move from NT to G is caused by trimming greater effect than dNM, suggesting that dNM, being less of N-linked oligosaccharides in the Golg apparatus. efficient (7), allowed one or two glucoses to be removed, as When the lysates were precipitated with anti-calnexin, a previously described (8). labeled calnexin band at 90 kDa could be observed (Fig. lb). The inhibitors did not seem to have a noticeable effect on Below it were HA bands corresponding to fT1, IT2, and NT. HA folding; the ratios between fT1, 1T2, and NT were More than 801% of IT1 and IT2 could be coprecipitated with comparable to controls. Cleavage of the labeled HA into its calnexin, while the Golgi form of HA did not coprecipitate two subunits HA1 and HA2 when the cells were exposed to min chase at 37C trypsin (25), indicated, in agreement with previous studies (26, 27), that the HA molecules were transported normally to i-&i 21 5-l17 0 I 1 1201 the cell surface in the presence of castanospermine (not a ITl shown). "M.14 -.11:0, .:. 11 2- .. mmhk.. ""' 4 NT- ...:",:" .: 1. 011,; 0, G w;w-W.. 11*11 calnexin b b IT IL= w t _ __ IT2- NT *- sh

tunicamycin C +- + - - + c 100 C 1 2 3 4 5 6 7 8 ~80 calnexin _ 60 HA... ~40 - tun HA - _ VSV G 20 - tun G NP -*

0 5 10 115 20 chase time at 370 (min) FIG. 2. Tunicamycin inhibits binding of calnexin to HA and G. CHO 15B cells infected with X31 influenza virus (lanes 1-4) or VSV FIG. 1. Kinetics of HA and G association with calnexin. CHO (lanes 5-8) were pretreated with tunicamycin at 5 pg/ml for 45 min 15B cells were metabolically labeled for 1 hr, then infected with X31 and then pulse labeled in the presence of the inhibitor for 5 min at influenza virus or VSV. At 5.5 hr after infection, cells were pulse- 3WC. Lysates were inmunoprecipitated with antibodies against X31 labeled with [5Sjmethionine for 2 min and chased at 370C for 0 to 20 HA, VSV G, or calnexin and analyzed by reducing SDS/7.5% min. Lysates were precipitated with antibodies to HA (a) or calnexin PAGE. The positions of tunicamycin-treated HA and G are marked (b) and analyzed by nonreducing SDS/7.5% PAGE. (c) Quantitation tun HA and tun G, respectively. The band below G protein in lane by n densitometry offluorographs from anti-calnexin precip- 5 (*) is a soluble form of G, Gs. Influenza virus nucleoprotein is itations of HA (e) and VSV G (o). Amount of viral protein precip- marked NP. Because of the limited availability of anti-calnexin itated is expressed as a percentage of the maximal amount bound. antibodies, only a fraction of the total calnexin was precipitated. Downloaded by guest on September 23, 2021 Cell Biology: Hammond et al. Proc. Natl. Acad. Sci. USA 91 (1994) 915

Z minchase 1 L|1 |15| S U U cZn - + + + + mannosidase -j.1jL- + I inin chase 0 120 20 0 20 Ijj a untrimmed HAHA 40 l 40.W.....*, ...... NT -1^ 1 ..MM0

b calnexin Fi-11-1- - -_-_ r'r 2- I ..z l 2 3 4 5 6 NT- 4J.' anti- anti-HA calnexin 1 2 3 4 5 6 FIG. 5. N-linked oligosaccharides on calnexin-bound HA contain FIG. 3. Glucosidase inhibitors block the interaction between HA terminal glucose residues. CHO 15B cells were pulse labeled for 2 and calnexin. CHO 15B cells were pretreated for 45 min with either min and chased for 0 or 15 min. Lysates were immunoprecipitated 1 mM castanospermine (CST) or 1 mM dNM, then pulse labeled for with antibodies to either calnexin (lanes 1 and 2) or HA (lanes 3-6). 3 min and chased for either 0 or 20 min in the presence of the The immunoprecipitated HA was digested with jack bean a-man- inhibitors. Lysates were immunoprecipitated with antibodies to HA nosidase and analyzed by reducing SDS/7.5% PAGE. Calnexin is not (a) or calnexin (b) and analyzed by nonreducing SDS/7.5% PAGE. detected in lanes 1 and 2 because cells were not prelabeled in this experiment. However, the most interesting observation was that the glucosidase inhibitors blocked the binding of HA to calnexin not take place (Fig. 4, lanes 6 and 8), confirming that (Fig. 3b, lanes 3-5). In contrast, 1-deoxymannojirimycin, an incompletely folded proteins do not bind to calnexin unless inhibitor of ER a-mannosidases, had no effect on the binding they have partially trimmed N-linked sugars. of HA to calnexin (data not shown). In the second approach, we treated HA that had been This result indicated that calnexin binds proteins that have coprecipitated with anti-calnexin antibodies with jack bean partially orfully deglucosylated oligosaccharide chains. Ifthe a-mannosidase. It is an exoglycosidase that removes all dNM effect on trimming was only partial, as suggested by the terminal a-mannoses from N-linked carbohydrates but does intermediate mobility shift between control and castanosper- not digest the portions of the triantennary core oligosaccha- mine-treated samples, we reasoned that trimming by both rides that retain terminal glucose residues (30, 31). The glucosidase I and II was needed to make a glycopolypeptide results in Fig. 5 show that the calnexin-associated HA capable ofbinding to calnexin. In other words, calnexin was molecules had a small mobility shift (lane 2) after mannosi- most likely to bind HA or G molecules that had either dase treatment. This shift was, most likely, caused by diges- monoglucosylated or fully deglucosylated oligosaccharides. tion of terminal mannose residues from the two glucose-free To distinguish between these possibilities we used two antennae of the core oligosaccharides. Following a 15-min approaches. One took advantage of the fact that misfolded chase, by which time nearly all labeled HA is dissociated proteins that remain in the ER tend to have predominantly from calnexin (Fig. 1), treatment with a-mannosidase re- Glc1MangGlcNAc2 oligosaccharide side chains (28, 29). Suh sulted in a larger mobility shift (lane 6), indicating that et al. (28) have shown that the sugar side chains of G protein terminal glucose residues had been removed from the ma- of tsO45 VSV are predominantly in the monoglucosylated jority of the oligosaccharides, leaving the mannose residues form when synthesized at the nonpermissive temperature. in all three antennae ofthe side chain susceptible to digestion. This mutant G protein has a defect in the ectodomain which The intermediate mobility shift seen after mannosidase treat- makes it temperature conditional for folding. When precipi- ment indicated that calnexin-bound HA had at least one tating with calnexin antibodies, we found this misfolded G glucose left on many of its oligosaccharides. We concluded protein associated with calnexin (Fig. 4). Unlike the wild- that calnexin possesses binding activity specific for proteins type G protein, whose binding to calnexin was transient, the with partially deglucosylated N-linked side chains, with the ts045 G protein remained calnexin associated for at least 60 monoglucosylated form being the most likely ligand. A direct min (Fig. 4, lane 7). When glucosidase inhibitors were present chemical analysis ofthe oligosaccharides present in calnexin- during the pulse, binding of tsO45 G protein to calnexin did bound proteins was precluded by the small amounts of anti-calnexin antibodies available to us. E the co, co, A final piece of information about binding specificity U U U u was obtained from cells chased for 20 min in the presence of + + + + minchase dNM. In the presence of this relatively inefficient inhibitor, at 39.50 O | O | 60 | 60 0jodo0j60 6ol some of the labeled HA that was not initially associated with calnexin (Fig. 3a, lane 5) was seen to associate (Fig. 3a, lane calnexin- 6). The bound HA had the mobility of oxidized and glucose- trimmed NT. Apparently, trimming was merely delayed by dNM, and molecules that were already extensively folded but ts045 G still present in the ER started to bind to calnexin belatedly. Gs- This observation showed that proteins do not necessarily 1 2 3 4 5 6 7 need to bind to calnexin directly after their translocation; | anti -VSV G | anti - calnexin | they can engage the chaperone long after they have been synthesized. Moreover, it was evident that a protein that was FIG. 4. Misfolded G protein binds to calnexin. Binding is inhib- already folded could bind to calnexin if it had the appropri- ited by castanospermine. CHO cells were infected with the ts045 ately trimmed N-linked sugars. mutant of VSV, pulse labeled at the nonpermissive temperature (39.50C) for 10 min, and chased either 0 or 60 min at 39.5°. Where indicated the cells were pretreated with 1 mM castanospermine DISCUSSION in the inhibitor during the pulse and chase. (CST) and maintained to calnexin Lysates were precipitated with antibodies to VSV G (lanes 1-4) or Our results showed that HA and G protein bind calnexin (lanes 5-8) and analyzed by reducing SDS/7.5% PAGE. during the first 5-10 min after synthesis. This is a period Downloaded by guest on September 23, 2021 916 Cell Biology: Hammond et al. Proc. Nad. Acad. Sci. USA 91 (1994) during which most of their post-translational folding takes These trimming events occur almost immediately after chain place (16, 32). Intrachain disulfide bonds are formed and addition, in many cases on the nascent chain (see ref. 4). conformational epitopes appear. When they dissociate from In the likely case that the protein is still unfolded when all calnexin, they are extensively folded and oxidized and are three glucoses are removed, the oligosaccharide serves as a ready to leave the ER. Ourresults are similar to those already substrate for the UDP-glucose:glycoprotein glucosyltrans- published for class I major histocompatibility complex anti- ferase, which adds a glucose residue to the side chain. The gens and other proteins that associate with calnexin (12, 14, added glucose is subsequently removed by glucosidase II and 21). For both HA and G, the timing of release is close to the the protein enters the re- and deglucosylation cycle observed half-time of trimerization [til2 = 7-15 min for HA, ti/2 = 6-8 by Suh et al. (28). Several misfolded transport-incompetent min for G (16, 33)], leaving open the possibility that oligo- proteins in the ER have been shown to retain mainly merization occurs while the glycoproteins are calnexin- Glc1Man9GlcNAc2 oligosaccharides (28, 29, 37), and it is bound. known that even for normal proteins the innermost glucose is That binding to calnexin includes both soluble and mem- removed considerably more slowly than the two terminal brane-bound proteins, and that it is restricted to glycopro- glucoses (38). The glycoprotein is released from the re- and teins, was recently demonstrated by Bergeron and co- deglucosylation cycle when it has folded and is no longer a workers (15) for secretory proteins in Hep G2 cells. Our substrate for the glucosyltransferase. Alternatively, the oli- results with viral glycoproteins confirmed the latter obser- gosaccharide may be modified by mannosidases, as mannose vation and revealed an additional level of specificity; the removal is known to make oligosaccharides less effective or N-linked side chains must be partially trimmed by glucosi- entirely incompatible as glucose acceptors forthe transferase dases. We concluded that preferential binding most likely (35). involves proteins with monoglucosylated oligosaccharides, The re- and deglucosylation cycle would be futile were it irrespective of their folding status. Incomplete folding alone not linked to the binding of proteins with monoglucosylated does not result in binding. oligosaccharides to calnexin. Calnexin's role, according to Suh et al. (28), who studied the oligosaccharides of mis- our model, and as already suggested by others (10, 15, 22), is folded ts045 G protein at nonpermissive temperature, ob- to serve as a chaperone for the newly synthesized glycopro- served that not only are the majority of the oligosaccharides teins during their folding and/or oligomeric assembly so that of this protein monoglycosylated but the single glucose they do not aggregate, or leave the ER prematurely. Left to residues in these oligosaccharides are rapidly turning over. their own devices, nonglycosylated proteins are known to where are either aggregate irreversibly or leave the ER without assem- Apparently, they are subject to a cycle they bling into oligomers (see refs. 3, 23, and 24). continuously removed by glucosidase II and replaced by a The model implies that glucose trimming provides a way glucosyltransferase. An abundant and ubiquitously ex- for resident ER factors to keep track of the folding and pressed glucosyltransferase in the ER lumen, called UDP- maturation status of newly synthesized glycoproteins. The glucose:glycoprotein glucosyltransferase, has, in fact, been detailed configuration of the N-linked oligosaccharides may isolated and characterized by Parodi and co-workers (34, 35) reflect the degree offolding, the state ofoligomerization, the as a soluble homodimer with 150-kDa subunits. This enzyme time spent in the ER, etc. It is interesting to note in this has a remarkable property: it uses as its substrate Man9g context that glucosidase inhibitors not only prevent glyco- 7GIcNAc2 oligosaccharide side chains, but only when these proteins from binding to calnexin, they also increase the are linked to misfolded or incompletely folded glycoproteins degradation rate of some proteins (39). a-Mannosidase in- (36). We think this enzyme may be intimately associated with hibitors have the opposite effect: they enhance the rate ofER calnexin binding of newly synthesized and misfolded pro- degradation of some transport-incompetent glycoproteins teins. (40). This suggests that ER degradation may also be con- One model for glycoprotein maturation and quality control nected to trimming and/or calnexin association. consistent with the available data is outlined below and It is important to point out that, since calnexin has been depicted schematically in Fig. 6. Here, calnexin plays a shown to exist in a complex with at least three other resident central role as a chaperone, and the N-linked oligosaccha- ER membrane proteins (11), it is possible that it is not itself rides serve as signals of a glycopolypeptide's folding status. the component of the complex responsible for binding to When a nascent polypeptide enters the lumen ofthe ER, the partially folded proteins. However, cross-linking studies oligosaccharide transferase attaches the core saccharide show only calnexin coupled to newly synthesized major units to asparagines in the consensus sequences. The sugar histocompatibility complex class I heavy chain (10), suggest- moieties are immediately trimmed by glucosidase I, which ing close contact. Whether calnexin binds directly to mono- removes the al,2fterminal glucose residue, thus making the glucosylated oligosaccharides remains an interesting and, as chain a substrate for glucosidase II, which proceeds to yet, open question. Its sequence has no homology with remove one or both of the remaining al,3 glucose residues. known lectins. Our data are most consistent with a lectin-like UIDP1- ;lc:glccoprotein Glucosidale I ,lucosidase II G(lucotidase II glucas*ltransferase FIG. 6. Proposed pathway for cal- nexin-dependent folding and quality control as depicted for a transmem- brane protein with one N-linked oli- gosaccharide. There are four main W;(;(;~~~I};tW---il stages to the (i) stepwise trim- __------_...... ------.-- --- model: ----- I F E I I _ nung ofglucoses, (i) calnexin binding (alInexiun ofproteins withmonglucosidated side compflehxI chains, (iii) reglucosylation of glu- .-...... Incompletely folded Folded cose-free side chains on incompletely ty1 copoIypeptides g}I N. lcop oINlppiC t i (C3 folded glycopeptide chains, and (iv) release of folded chains from the re- gl and deglucosylation cycle. Downloaded by guest on September 23, 2021 .....-...... Cell Biology: Hammond et al. Proc. Natl. Acad. Sci. USA 91 (1994) 917 activity for calnexin, but we cannot exclude that calnexin 8. Gross, V., Andus, T., Tran-Hi, T.-A., Schwartz, R. T., recognizes structural characteristics in polypeptides that Decker, K. & Heinrich, P. C. (1983) J. Biol. Chem. 258, appear and disappear coincidently with the processing of 12203-12209. oligosaccharides. A direct demonstration oflectin activity for 9. Lodish, H. F. & Kong, N. (1984) J. Cell Biol. 98, 1720-1729. 10. Degen, E. & Williams, D. B. (1991) J. Cell Biol. 112, 1099- calnexin will require further investigation. 1115. In mature folded glycoproteins, the carbohydrate side 11. Wada, I., Rindress, D., Cameron, P. H., Ou, W.-J., Doherty II, chains are generally located on the surface. They are often J. J., Louvard, D., Bell, A. W., Dignard, D., Thomas, D. Y. & structurally dispensable. If calnexin does attach to proteins Bergeron, J. J. M. (1991) J. Biol. Chem. 266, 19599-19610. via these appendices, it may give the newly synthesized 12. Degen, E., Cohen-Doyle, M. F. & Williams, D. B. (1992) J. chains extensive conformational and steric freedom to fold Exp. Med. 175, 1653-1661. and interact with other folding enzymes and chaperones such 13. Hostenbach, F., David, V., Watkins, S. & Brenner, M. B. as protein disulfide isomerase and immunoglobulin heavy (1992) Proc. Natl. Acad. Sci. USA 89, 4734-4738. chain binding protein/glucose-regulated protein 14. Galvin, K., Krishna, S., Ponchel, F., Frohlich, M., Cummings, (BiP/ D. E., Carlson, R., Wands, J. R., Isselbacher, K. J., Pillai, S. Grp78). In other cases it is known that calnexin-associated & Ozturk, M. (1992) Proc. Natl. Acad. Sci. USA 89, 8452-8456. proteins can assemble with other subunits to form quaternary 15. Ou, W.-J., Cameron, P. H., Thomas, D. Y. & Bergeron, complexes (12). We have shown previously that VSV G J. J. M. (1993) Nature (London) 364, 771-776. protein associates transiently during folding with BiP/Grp78 16. Braakman, I., Hoover-Litty, H., Wagner, K. R. & Helenius, (32), which raises the possibility that both chaperones coop- A. (1991) J. Cell Biol. 114, 401-411. erate in assisting the folding ofthis protein. Present in the ER 17. De Silva, A., Braakman, I. & Helenius, A. (1993) J. Cell Biol. membrane as monomers or small oligomers (15), calnexins 120, 647-655. may provide dynamic, multivalent, and functionally flexible 18. Balch, W. E., Elliott, M. M. & Keller, D. S. (1986) J. CellBiol. scaffolds designed to assist the folding and oligomeric as- 261, 14681-14689. 19. Wiley, D. C. & Skehel, J. J. (1987) Annu. Rev. Biochem. 56, sembly of glycoproteins. 365-395. Finally, it is important to emphasize that, although all 20. Rose, J. K. & Gallione, C. J. (1981) J. Virol. 39, 519-528. newly synthesized glycoproteins so far tested bind tran- 21. Hochstenbach, F., David, S., Watkins, S. & Brenner, M. (1992) siently to calnexin, only a subgroup show folding and secre- Proc. Natl. Acad. Sci. USA 89, 4734-4738. tion defects in the presence of glucosidase inhibitors which 22. David, V., Hochstenbach, F., Rajagopalan, S. & Brenner, inhibit this interaction (6). Glucosidase inhibitors have only M. B. (1993) J. Biol. Chem. 268, 9585-9592. mild effects on cell viability and secretion in many cases (6). 23. Gibson, R., Schlesinger, S. & Kornfeld, S. (1979) J. Biol. Cell lines with mutations in glucosidases and N-linked gly- Chem. 254, 3600-3607. fX cosylation are usually relatively normal (41). Evidently, 24. Hurtley, S. M., Bole, D. G., Hoover-Litty, H., Helenius, A. & Copeland, C. S. (1989) J. Cell Biol. 108, 2117-2126. when prevented from binding to calnexin many glycoproteins 25. Matlin, K. S. & Simons, K. (1983) Cell 34, 233-243. find alternative ways of folding and assembling in the ER 26. Burke, B., Matlin, K., Bause, E., Legler, G., Peyrieras, N. & lumen. The high concentrations of classical chaperones and Ploegh, H. (1984) EMBO J. 3, 551-556. redox enzymes present in the ER may allow them to fold 27. Pan, Y. T., Hori, H., Saul, R., Sanford, B. A., Molyneux, R. J. without the calnexin pathway (1, 2). The apparent redun- & Elbein, A. D. (1983) Proc. Natl. Acad. Sci. USA 22, 3975- dancy in the ER folding machinery does, however, raise a 3984. question as to the real function ofthe calnexin system. While 28. Suh, P., Bergmann, J. E. & Gabel, C. A. (1989) J. Cell Biol. more work is needed to provide a full answer, it seems most 108, 811-819. to us while involved in and 29. Rizzolo, L. J. & Kornfeld, R. (1988) J. Biol. Chem. 263, likely that, intimately folding 9520-9525. oligomeric assembly of glycoproteins, the calnexin system 30. Li, Y.-T. (1967) J. Biol. Chem. 242, 5474-5480. may have evolved mainly to ensure quality control, an 31. Kornfeld, K., Reitman, M. L. & Kornfeld, R. (1981) J. Biol. important but yet elusive function of the ER. Chem. 256, 6633-6640. 32. Machamer, C. E., Doms, R. W., Bole, D. G., Helenius, A. & We are thankful to Dr. J. Bergeron for the anti-calnexin antibodies Rose, J. K. (1990) J. Biol. Chem. 265, 6879-6883. and for sharing data with us prior to publication. We thank Dr. David 33. Doms, R. W., Keller, D. S., Helenius, A. & Balch, W. E. Bole for antibodies to BiP, Jonne Helenius for technical assistance, (1987) J. Cell Biol. 105, 1957-1969. and Henry Tan for help with photography. This study was supported 34. Parodi, A. J., Mendelzon, D. H., Lederkremer, G. Z. & Mar- by grants from the National Institutes of Health (5 RO1 GM38346-13 tin-Barrientos, J. (1984) J. Biol. Chem. 259, 6351-6357. and P01 CA46128). 35. Trombetta, S. E. & Parodi, A. J. (1992) J. Biol. Chem. 267, 9236-9240. 1. Gething, M.-J. & Sambrook, J. (1992) Nature (London) 355, 36. Sousa, M. C., Ferrero-Garcia, M. A. & Parodi, A. J. (1992) 33-45. Biochemistry 31, 97-105. 2. Helenius, A., Marquardt, T. & Braakman, I. (1992) Trends Cell 37. Tufaro, F., Snider, M. D. & McKnight, S. L. (1987) J. Cell Biol. 2, 227-231. Biol. 105, 647-657. 3. Hurtley, S. M. & Helenius, A. (1989) Annu. Rev. Cell Biol. 5, 38. Hubbard, S. C. & Robbins, P. W. (1979) J. Biol. Chem. 254, 277-307. 4568-4576. 4. Kornfeld, R. & Kornfeld, S. (1985) Annu. Rev. Biochem. 54, 39. Moore, S. E. & Spiro, R. G. (1993) J. Biol. Chem. 268, 3809- 631-664. 3812. 5. Paulson, J. C. (1989) Trends Biochem. Sci. 14, 272-276. 40. Su, K., Stoller, T., Rocco, J., Zemsky, J. & Green, R. (1993) 6. Elbein, A. D. (1991) FASEB J. 5, 3055-3063. J. Biol. Chem. 268, 14301-14309. 7. Elbein, A. D. (1991) Semin. Cell Biol. 2, 309-317. 41. Stanley, P. (1984) Annu. Rev. Genet. 18, 525-552. Downloaded by guest on September 23, 2021