Chaperone Selection During Glycoprotein Translocation into the Endoplasmic Reticulum

See allHide authors and affiliations

Science  14 Apr 2000:
Vol. 288, Issue 5464, pp. 331-333
DOI: 10.1126/science.288.5464.331


A variety of molecular chaperones and folding enzymes assist the folding of newly synthesized proteins in the endoplasmic reticulum. Here we investigated why some glycoproteins interact with the molecular chaperone BiP, and others with the calnexin/calreticulin pathway. The folding of Semliki forest virus glycoproteins and influenza hemagglutinin was studied in living cells. The initial choice of chaperone depended on the location of N-linked glycans in the growing nascent chain. Direct interaction with calnexin and calreticulin without prior interaction with BiP occurred if glycans were present within about 50 residues of the protein's NH2-terminus.

Proteins synthesized in the endoplasmic reticulum (ER) rely during their folding and assembly extensively on the assistance of molecular chaperones and other factors present in the ER lumen. Interactions with such ER-resident folding factors begin, typically, as growing nascent chains enter the lumen through the translocon complex and continue until the newly synthesized proteins are folded, assembled, and ready to leave the ER. Proteins interact differently with the molecular chaperones. Some interact first with BiP and then with calnexin (Cnx) (1, 2), some bind Cnx and calreticulin (Crt) but do not associate with BiP (3–5), some are sequentially assisted by BiP and GRP94 (6). Formation of intramolecular disulfide bonds can be catalyzed by protein disulfide isomerase (PDI), ERp57, or by both oxidoreductases acting together (7). The rules that govern such differential chaperone selection are only poorly understood.

To learn more about selection of folding factors, the maturation of two glycoproteins, p62 and E1 of the Semliki forest virus (SFV), was analyzed in infected Chinese hamster ovary (CHO) cells. SFV suppresses host protein translation without affecting protein folding and maturation. In contrast to noninfected cells, infected cells only express viral proteins (Fig. 1A). Among these, E1 and p62 are translocated into the ER. They are type I membrane glycoproteins with four and one N-linked glycans, respectively, and they have numerous intrachain disulfide bonds in their ectodomain.

Figure 1

Folding and interaction with ER-resident chaperones of SFV glycoproteins. (A) Lane 1, CHO cells were pulsed for 2 min with 500 μCi of [35S]methionine and cysteine and chased for 30 min without radioactivity. Cell extract was prepared as described in (7) and analyzed by nonreducing gel electrophoresis. Lane 2, CHO cells were infected with 30 plaque-forming units per cell of SFV 4 hours before pulse-chase. Infected cells mainly express viral structural proteins (indicated). Cellular or nonstructural viral proteins are indicated with asterisks 1, 2, and 3. (B) Whole-cell extract (lanes 1 to 5), oxidative folding of SFV glycoproteins E1 and p62 was analyzed by nonreducing SDS-PAGE. Viral proteins in different oxidation states are labeled. E2 derives from p62 processing in the Golgi compartment (lanes 4 and 5). Cell extracts were precipitated with antibody to BiP (lanes 6 to 8), Cnx (9 to 11) and Crt (12 to 14). The E1 form coprecipitating with the lectins after 5- and 9-min chase has faster electrophoretic mobility than the E1 forms coprecipitating with BiP.

Infected cells were pulse-labeled for 1 min with [35S]methionine and cysteine and chased for 1 to 72 min (Fig. 1B). Cellular extracts were prepared and analyzed by nonreducing SDS–polyacrylamide gel electrophoresis (SDS-PAGE) as described (7). After a 1-min chase, most of the label was in nascent chains. Because of their heterogeneous size, the nascent chains run as diffuse smears. The labeled, full-length E1 and p62 differed in the extent of intrachain disulfide bonding and were thus distributed in multiple bands. The folding of E1 involved three intermediates (E1red, E1ox1, and E1ox2). E1red had the same electrophoretic mobility as the E1 protein run under reducing conditions. E1red was rapidly converted to E1ox1, a partially oxidized intermediate that persisted for 6 to 10 min before being converted into E1ox2, which had the same mobility as mature E1. In contrast to E1, the oxidation of p62 took place during the first 20 min without distinct intermediates. That the spike glycoproteins followed a productive maturation pathway was shown by the cleavage of p62 into E2, by the formation of noncovalent E1:p62 heterodimers (t 1/2 = 15 min), by the transition of E1 and p62 from a dithiothreitol (DTT)-sensitive to a DTT-resistant conformation (t 1/2 = 15 to 20 min), and by the release of mature viral particles in the culture medium.

We established by coimmunoprecipitation that BiP associated with early intermediates of E1 (nascent chains, E1red, and E1ox1, but not E1ox2; Fig. 1B). In contrast, Cnx and Crt associated with nascent chains, folding intermediates of p62, and some late forms of E1 (E1ox2), indicating that Cnx and Crt associated with E1 only after it had been released from BiP. Thus E1 and p62 expressed different chaperone preferences, and the differences were already manifested when they were growing nascent chains.

To provide detailed information about cotranslational events, we investigated the interactions occurring between chaperones and growing nascent chains, using a two-dimensional (2-D) SDS-PAGE technique in which the first dimension is nonreducing and the second reducing (7, 8). After a 1-min pulse and a 1-min chase (Fig. 2A), the full-length viral proteins appeared as spots, whereas the nascent chains formed streaks starting on the diagonal in the bottom left corner. The presence of radioactive streaks below the diagonal indicated that intramolecular disulfide bonds were forming cotranslationally. Oxidation started when nascent chains were about 35 kD in size (see arrowhead 1, Fig. 2A).

Figure 2

Cotranslational folding and interaction with ER-chaperones of SFV glycoproteins analyzed by 2-D SDS-PAGE. Whole-cell extract (A), αBiP (B), and αCnx (C) precipitates were subjected to 2-D SDS-PAGE in which the first dimension was nonreducing and the second, reducing (7). The full-length labeled viral glycoproteins, as well as E1 (arrow a) and p62 growing nascent chains (arrows b and c) are indicated.

When the extracts were precipitated with antibody to BiP, the 2-D gel pattern showed coprecipitation of E1 nascent chains with molecular masses of 30 kD and larger (Fig. 2B, arrow a). Since these nascent chains migrated on or close to the diagonal, we concluded that BiP associated with E1 nascent chains that had not been extensively oxidized. Although it has been speculated that BiP binds to growing nascent chains, these data suggest that such binding can occur in cells.

The chaperone molecule BiP also coprecipitated with early folding intermediates of E1 (E1red and E1ox1 inFigs. 1B and 2B) and interchain disulfide-bonded complexes visible above the diagonal (Fig. 2, A and B). The complexes corresponded to mixed disulfides between E1 and PDI (7). Evidently they contained BiP as a third component that was not covalently bound. The fully oxidized form of E1 (E1ox2) did not coprecipitate with BiP (Figs. 1B and 2B), consistent with preferential BiP-binding to less folded forms of E1.

Nascent chains of p62 coprecipitated with Cnx (in Fig. 2C, arrow b indicates p62 nascent chains without intrachain disulfides and arrow c, p62 nascent chains that are partially oxidized). Unlike E1 oxidation, p62 oxidation started cotranslationally when 30 to 50% of the polypeptide chain had entered the ER lumen. Previous work has shown that oxidation of p62 is catalyzed by the oxidoreductase ERp57 (7).

When access to the Cnx/Crt chaperone system was blocked by adding the reducing agent DTT to the medium (9), both p62 and E1 were seen to associate with BiP (Fig. 3A). When the DTT was washed out, the glycoproteins were rapidly transferred to the Cnx/Crt chaperone system, and proper folding was reestablished. The same was observed when access to Cnx and Crt was blocked by addition of castanospermine, an inhibitor of glucose trimming (Fig. 3B): both E1 and p62 showed prolonged association with BiP. Thus, the reason BiP failed to bind to p62 normally was not because p62 lacked BiP binding sites but rather that the bound Cnx and Crt prevented the interaction.

Figure 3

SFV glycoproteins expression in the presence of DTT or castanospermine. (A) The binding of E1 and p62 to BiP, Cnx, and Crt was tested in control cells (left) and in cells treated with 5 mM DTT during pulse and 5-min chase (right). (B) BiP rapidly released E1 folding intermediates before complete oxidation (t 1/2 = 5 min, refer also to Fig. 1B). The viral-derived radioactivity associated with Cnx and Crt increased during the first 15 min of chase as the E1 released by BiP completed its folding, in association with p62, upon binding to the ER lectins. In the presence of castanospermine (right), association of p62 with Cnx and Crt was blocked. As observed in the presence of DTT, p62 bound to BiP.

In trying to explain these results, we noted that one difference between E1 and p62 was that p62 possessed glycosylation sites close to the NH2-terminus that might provide a site for early lectin binding to nascent chains. To test the notion that the position of N-linked glycans in the sequence of a glycoprotein correlates with differential chaperone binding, we made use of mutants of influenza hemagglutinin (HA) in which consensus sites for N-linked glycosylation were altered either close to the NH2-terminus or more toward the COOH-terminus [legend of Fig. 4 and (10)]. The results with wild-type HA confirmed that BiP does not bind despite the presence of multiple, potential BiP binding sequences [Fig. 4 and (3,11–13)]. When maturation of the mutants was analyzed, it was found that removal of glycosylation sites in positions 8 and 22, or—even better—in positions 8, 22, and 38, resulted in binding of BiP. As already observed for SFV E1 (Figs. 1B and 2B), and for vesicular stomatitis virus (VSV) glycoprotein (1), it was the less oxidized forms of M1 and M2, that is, those migrating slower in the nonreducing gel (labeled IT1 and IT2 in Fig. 4), that preferentially associated with BiP. The fully oxidized HA form (NT) was mostly associated with Cnx (10).

Figure 4

Deletion of N-linked glycans and glycoprotein association with BiP. (A) The position of N-linked glycans in the four constructs used is shown. Proteins were transiently expressed in CHO cells using a recombinant vaccinia virus with T7 polymerase and cowpox hr gene as described in (10). (B) The different mobility in the reducing gel (left panel) corresponds to the different number of N-linked glycans each of them contributing approximately 2 kD to the protein molecular mass. (C) In unchased samples, folding of the HA molecules was still incomplete, and in the nonreducing gel (central panel) the different oxidation states (IT1, IT2, and NT) of wild-type and mutant HAs are indicated. Note that oxidative folding was somewhat faster for the mutants than for the wild-type HA (10). (D) Association with BiP was determined by double immunoprecipitation (αBiP, then αHA). Only M1 and M2, in which the glycans located at the NH2-terminus were deleted, associated with BiP during the early step of oxidative folding.

Taken together, the results indicated that, during translocation of a glycoprotein in the ER, a choice is made between chaperone systems. One of the systems comprises BiP and possibly PDI, and the other consists of Cnx, Crt, and ERp57. The chaperones compete for binding sites on the incoming nascent chain. Whichever binds first apparently dominates during early stages of co- and postranslational folding. Its association delays or prevents binding of the other chaperones. Later in the folding process a transfer from one pathway to the other can, however, take place as observed for E1 here and VSV glycoprotein in a previous study (1).

The results indicate, moreover, that glycoproteins such as p62 and HA that possess glycans close to the NH2-terminus enter the Cnx/Crt pathway directly without prior binding to BiP. Glycoproteins like E1 and VSV glycoprotein that have their glycans more COOH-terminally in the sequence associate first with BiP. Inhibitors of the Cnx/Crt pathway, such as castanospermine, direct proteins such as p62 and HA that normally do not bind BiP to the BiP pathway. This shows that the presence of N-linked glycans, as such, does not prevent BiP binding but rather promotes the association of the alternative chaperones. A literature survey of glycoproteins for which chaperone requirements are known is consistent with the notion that proteins with N-linked glycans within the first 50 residues do not undergo transient interaction with BiP (14).

Thus, in summary, our findings demonstrated that to understand the molecular basis for chaperone selection in the ER it is important to analyze growing nascent chains and to consider the precise timing of cotranslational modifications such as N-linked glycosylation and oligosaccharide trimming. Attention must be paid particularly to the location of glycosylation sites in the linear sequence.

  • * To whom correspondence should be addressed. E-mail: ari.helenius{at}


View Abstract

Navigate This Article