Calnexin fails to associate with substrate proteins in glucosidase-deficient cell lines.

Increasing evidence shows that calnexin, a membrane-bound chaperone in the endoplasmic reticulum, is a lectin that binds to newly synthesized glycoproteins that have partially trimmed N-linked oligosaccharides. It specifically attaches to core glycans from which two glucoses have been removed by glucosidases I and II. Several recent reports suggest, however, that it can also bind to proteins devoid of N-linked glycans. To investigate the extent of glycan-independent binding, we have analyzed two mutant cell lines (Lec 23 and PhaR 2.7) that are unable to process the core glycans because they lack glucosidase I or glucosidase II, respectively. In contrast to parental cell lines, calnexin binding of substrate proteins was found to be virtually nonexistent in these cells. Neither cellular nor viral proteins associated with the chaperone. It was concluded that glycans are crucial for calnexin association and that the vast majority of substrate proteins are therefore glycoproteins.

Increasing evidence shows that calnexin, a membrane-bound chaperone in the endoplasmic reticulum, is a lectin that binds to newly synthesized glycoproteins that have partially trimmed N-linked oligosaccharides. It specifically attaches to core glycans from which two glucoses have been removed by glucosidases I and II. Several recent reports suggest, however, that it can also bind to proteins devoid of N-linked glycans. To investigate the extent of glycan-independent binding, we have analyzed two mutant cell lines (Lec 23 and Pha R 2.7) that are unable to process the core glycans because they lack glucosidase I or glucosidase II, respectively. In contrast to parental cell lines, calnexin binding of substrate proteins was found to be virtually nonexistent in these cells. Neither cellular nor viral proteins associated with the chaperone. It was concluded that glycans are crucial for calnexin association and that the vast majority of substrate proteins are therefore glycoproteins.
The lumen of the endoplasmic reticulum (ER) 1 contains a large number of chaperones and folding enzymes (see Refs. 1 and 2). They assist the folding, oligomeric assembly, and quality control of newly synthesized proteins, the majority of which are glycoproteins. Some of these folding factors, such as BiP/ GRP78 and GRP94, are members of classical chaperone families while others are unique to the ER. The latter include calnexin, a membrane-bound protein that interacts transiently with a variety of soluble and membrane-bound substrate molecules (see Ref. 3).
Our studies have indicated that calnexin differs from previously characterized chaperones in being a lectin; it binds to glycoproteins with partially trimmed N-linked oligosaccharides (4,5). The specificity for monoglucosylated chains (Glc 1 Man 9 -7GlcNAc 2 ) was recently confirmed by Ware et al. (32) using purified N-linked oligosaccharides. These results are significant because they link the folding of glycoproteins and their quality control in the ER directly to the process of oligosaccha-ride trimming. Based on these observations, we have proposed that the de-and reglucosylation cycle, known to operate in the lumen of the ER (6), is connected to calnexin binding and that the cycle plays a central role in the maturation of newly synthesized glycoproteins (4,7).
Recent studies have confirmed the requirement for glycans in substrate binding to calnexin. These rely on glycosylation inhibitors (tunicamycin), ␣-glucosidase inhibitors castanospermine (CST) and 1-deoxynojirimycin, and mutagenesis to remove consensus glycosylation sequences (8 -10). It has also been reported, however, that polypeptide chains that have no N-linked oligosaccharides can associate with calnexin. The latter observations include the redistribution of the unglycosylated CD3⑀ subunit of the T cell receptor in cells expressing a mutant version of calnexin, and the co-immunoprecipitation of unglycosylated thyroglobulin, class I and II major histo compatibility complex subunits and P-protein with calnexin (11)(12)(13)(14)(15). It has therefore been suggested that calnexin binding may occur in more than one way and may not require the presence of sugars.
To address the mode of substrate association with calnexin, we have analyzed mutant cells with glucosidase defects. Since they fail to remove glucoses from the core oligosaccharide, the glycoproteins do not reach the monoglucosylated form and are therefore unable to bind to calnexin via the lectin binding sites. Hence, these cells provide an opportunity to determine whether glycan-independent binding does take place under physiological conditions. Tests can be made without resorting to inhibitors and overexpression systems.

MATERIALS AND METHODS
Cell Lines, Viruses, and Reagents-Chinese hamster ovary (CHO) cells were grown in ␣-MEM with 8% fetal calf serum (JRH Biosciences), Lec 23 cells in ␣-MEM with nucleosides, and 8% fetal calf serum and BW 5147 and Pha R 2.7 cells in Dulbecco's modified Eagle's medium with 10% horse serum (Gemini Biocorporation Inc., Calabasas, CA). All media had also 100 units/ml penicillin and 100 g/ml streptomycin (Life Technologies, Inc.). The X31/A/Aichi/1968 strain of influenza A virus and the temperature-sensitive tsO45 mutant of vesicular stomatitis virus (VSV) were used as described previously (16,17). Polyclonal antibodies were raised in rabbits against a peptide consisting of the C-terminal 19 amino acids of canine calnexin (18). The polyclonal VSV G protein and the anti-influenza antisera used have been described previously (19,20).
Metabolic Labeling and Immunoprecipitation-Lec 23 and CHO cells were infected with influenza virus and VSV as described (16,17). After 5 h of infection, cells were transferred for 1 h to cysteine-and methionine-free medium with or without 1 mM CST. They were then pulsed for 10 min with 0.1 Ci 35 S-labeled ProMix (Amersham Corp.)/60-mm dish in a total volume of 0.5 ml. Cells were lysed with 2% CHAPS (Pierce) in HBS buffer (50 mM Hepes, 0.1 M NaCl, pH 7.6) containing a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 10 g/ml each of chymostatin, leupeptin, antipain, and pepstatin) and precleared by rotating for 30 min at 4°C with Zysorbin (Zymed Laboratories, Inc., San Francisco, CA). One-third of the lysates were used for immunoprecipitation with 10 l of anti-calnexin antisera incubating overnight at 4°C as described (21). After incubation, precipitates were washed three times with 0.5% CHAPS in HBS buffer. Influenza hemagglutinin (HA) and G protein were immunoprecipitated, and SDS-polyacrylamide gel electrophoresis was performed as described (16). Pha R 2.7 and BW 5147 cells were infected with VSV tsO45. Cells were pulsed for 1 h with 0.15 Ci of [ 35 S] ProMix. Otherwise the procedures were the same as for Lec 23 and CHO cells.
Immunofluorescence-CHO and Lec 23 cells were plated on alcian blue-treated glass coverslips as described previously (22). BW 5147 and Phar R 2.7 were washed once with serum-free medium. 1 ϫ 10 5 cells in

RESULTS AND DISCUSSION
The two mutant cell lines used in this study were Lec 23 and Pha R 2.7. Lec 23 are mutant CHO cells defective in glucosidase I (23), and Pha R 2.7 cells are mutants of the BW 5147 mouse lymphoma cell line defective in glucosidase II (24). Previous work has established that glycoproteins are synthesized in both mutant cell lines, but glucose trimming is defective. Both mutant cell lines grow well in culture and synthesize normal or slightly increased amounts of calnexin compared with the parental lines (see below). Indirect immunofluorescence microscopy using anti-calnexin antibodies showed that the calnexin is distributed in a typical reticular ER pattern with nuclear rim staining (Fig. 1). The distribution of calnexin was identical to that seen in the parental cells.
To analyze the interaction of newly synthesized cellular proteins with calnexin in Lec 23 cells and in the parental CHO cells, monolayers were pulse-labeled for 10 min with 35 S-labeled ProMix. CST was included in some of the cultures to prevent the action of glucosidases I and II. A CHAPS-solubilized lysate was prepared and immunoprecipitated using anticalnexin antibodies, and the proteins were analyzed by SDSpolyacrylamide gel electrophoresis and fluorography.
As reported previously (8), several labeled protein bands were precipitated by the anti-calnexin antibodies from labeled wild type CHO cells (Fig. 2, lane 1). The band migrating at 90 kDa corresponded to calnexin itself, whereas the others were coprecipitating substrate proteins. When labeling was performed in the presence of CST (lane 2), only the calnexin band was seen, confirming the requirement for glucose trimming and the specificity of the antibodies.
In Lec 23 cells, the coprecipitation of proteins was dramatically reduced (lane 3). Few if any substrate proteins associated with calnexin. The background was, however, further suppressed by the addition of CST (lane 4), suggesting that it was specific. The presence of small background staining was consistent with the report that Lec 23 cells are somewhat leaky, i.e. the glucosidase I activity is not completely inhibited (23).
In infected wild type CHO cells, we have previously reported efficient binding of two viral glycoproteins, influenza HA and VSV G protein, to calnexin (4,8). This is also seen in lanes 5 and 9 (Fig. 2). Binding of these viral membrane proteins was blocked if they were synthesized in the presence of CST (lanes 6 and 10).
When Lec 23 cells were infected with the same viruses, calnexin binding of the two glycoproteins was virtually nonexistent (lanes 7 and 8 and 11 and 12). That the cells were, indeed, infected is shown in Fig. 3. The amounts of labeled HA and G protein after immunoprecipitation with anti-HA and G-protein antibodies show that the amounts were comparable with or without CST (compare lanes 1 and 3 and 5 and 6 and  lanes 2 and 4 and 7 and 8). As expected, the apparent molecular weights of HA and G protein were slightly higher in Lec 23 cells due to the lack of glucose trimming.
With the glucosidase II-deficient Pha R 2.7 cells, the results were similar to those seen for Lec 23. Whereas numerous newly synthesized proteins coprecipitated with anti-calnexin from the parental BW 5147 cells (Fig. 4, lane 3), virtually no labeled bands (except for calnexin itself) were seen in the Pha R 2.7 cells (lane 4). After infection of the wild type cells with VSV, the G protein was coimmunoprecipitated with anti-calnexin as expected (lane 1), whereas coprecipitation from the mutant cells was virtually undetectable (lane 2). That the cells were infected, and that they expressed VSV G protein, was shown by immunoprecipitation using anti-G protein antibodies (Fig. 1,  lanes 9 and 10). Both the membrane-bound, intact G protein (G m ) and the cleaved (G s ) forms were seen. Similar experiments were attempted with influenza virus, but being lymphoid, these cells could not be efficiently infected by this virus.
Taken together, the results showed that the majority, if not all, of the substrate proteins that normally associate with cal-  1 and 2) and Lec 23 cells (lanes 3 and 4) were immunoprecipitated with anti-calnexin antibody. CHO and Lec 23 cells were infected by influenza virus and VSV and immunoprecipitated with anti-calnexin antibodies (lanes 4 -12). Immunoprecipitates from castanospermine-treated cells are marked with a ϩ.
nexin in CHO and mouse lymphocytes must first be processed by glucosidases I and II. This confirms that substrate proteins were glycoproteins. Given that the results were obtained under normal tissue culture conditions without added inhibitors and without the use of overexpression, virus vectors, or mutant proteins, they probably reflect the normal binding specificity of calnexin.
The results are consistent with data obtained with CST and other glucosidase inhibitors (4,10). As shown in Fig. 2 such inhibitors effectively block the binding of virtually all newly synthesized substrate proteins to calnexin. Pha R 2.7 cells have recently been used by Kearse et al. (33) to probe the specificity of T cell receptor ␣ and ␤ chain binding to calnexin. Consistent with our results they found that coimmunoprecipitation of these two glycoproteins was decreased compared to the parental cell.
The reasons why association of calnexin with nonglycosylated proteins has been observed in several systems (11)(12)(13)(14) remain unclear. Some of the studies in question have been performed using tunicamycin as a glycosylation inhibitor or dithiothreitol as a folding inhibitors. Transient overexpression of glycoproteins has also been used. These experimental conditions are less physiological than those employed here. Tunicamycin causes global inhibition of N-linked glycosylation and massive aggregation of misfolded proteins in the ER lumen as well as a blockade in their transport to the Golgi complex (25)(26)(27). Similar aggregation has been reported in the presence of dithiothreitol (28). Mutant proteins and protein subunits that fail to oligomerize also often end up in large aggregates. It has, moreover, been our experience that transient expression of wild type proteins using the T7 vaccinia virus expression system (29) frequently results in aggregated products (27). When the substrate proteins are aggregated, it is difficult to know whether the coimmunoprecipitation of a specific nonglycosylated substrate protein with calnexin occurs because of a direct contact between the two proteins or due to indirect association mediated by other proteins. Further studies using isolated proteins in vitro are needed to determine the various modes of calnexin binding to its substrate.
That Lec 23 and Pha R 2.7 cells are viable is interesting in view of their glucosidase defects and the lack of calnexinmediated folding. These cells have been selected for growth and may therefore contain higher amounts of other folding factors. It is known that many glycoproteins can fold correctly, albeit usually less rapidly and less efficiently, in cells treated with glucosidase inhibitors (see Refs. 30 and 31). Evidently, they can fold and reach the mature conformation using other chaperones and folding enzymes.  Fig. 2. Immunoprecipitations were performed using anti-influenza antibodies (lanes 1-4) and anti-VSV G antibody (lanes [5][6][7][8]. Immunoprecipitates with anti-VSV G protein antibody from BW 5147 and Pha R 2.7 cell lysates are shown in lanes 9 and 1, respectively.  1 and 2). Anti-calnexin immunoprecipitates from noninfected BW 5147 and Pha R 2.7 cells are shown in lanes 3 and 4, respectively.