Novel covalent chaperone complexes associated with human chorionic gonadotropin beta subunit folding intermediates.

Molecular chaperones facilitate the folding of proteins in the endoplasmic reticulum (ER) of mammalian cells. The glycoprotein hormone chorionic gonadotropin β subunit is a secretory protein whose folding in the ER has been demonstrated (Huth, J. R., Mountjoy, K., Perini, F., and Ruddon, R. W. (1992) J. Biol. Chem. 267, 8870-8879). Because folding of wild type hCG-β subunit occurs in the ER with a t1/2 = 4-5 min, stable association of ER chaperones with hCG-β have been difficult to detect probably because they have a short half-life. However, β-chaperone complexes containing the ER chaperones BiP, ERp72, and ERp94 have been detected in slow folding mutants of hCG-β subunit that lack both of the N-linked oligosaccharides (Feng, W., Matzuk, M. M., Mountjoy, K., Bedows, E., Ruddon, R. W., and Boime, I. (1995) J. Biol. Chem. 270, 11851-11859). The questions addressed here are 1) whether the detection of chaperone-containing complexes is related to the absence of carbohydrate or to the rate of hCG-β subunit folding, 2) whether such complexes are dead-end or whether they lead to formation of a secreted, mature hCG-β form, and 3) what the nature of the hCG-β-chaperone binding is. The data obtained indicate that the amount of detectable hCG-β-chaperone complexes correlates with the rate or extent of folding, that the complexes of hCG-β with ER chaperones lead to the formation of secretable β, and that the complexes of hCG-β with chaperones involve the formation of intermolecular disulfide bonds.

In the endoplasmic reticulum (ER) 1 two classes of proteins assist polypeptide folding. These proteins include foldases such as peptidylprolyl cis/trans-isomerase and protein disulfide isomerase, which catalyze rate-limiting isomerization steps in protein folding, and binding proteins called molecular chaperones that stabilize unfolded or partially folded structures and prevent the formation of inappropriate folding interactions (1)(2)(3).
Molecular chaperones are proteins that are expressed in cellular compartments where protein folding occurs, and those of the heat-shock protein families (e.g. hsp-70 and hsp-90) are frequently induced to higher levels during response to stress. They function in protein folding and translocation by mechanisms that involve binding to hydrophobic surfaces of unfolded proteins. This prevents aggregation and fosters conformational changes that lead to native structure (1)(2)(3). Chaperone-mediated protein folding appears to proceed by multiple rounds of binding and release of nonnative forms (4) and by sequential binding of various chaperones that may be involved in different steps of the folding pathway (5). However, binding of partially folded intermediates to chaperones can facilitate protein folding while the folding intermediate is still chaperone-bound (6).
The molecular chaperones of the hsp-70 and hsp-90 families in eukaroytic cells function in the stabilization, translocation, and degradation of partially folded intermediates during polypeptide folding and assembly. BiP is a member of the hsp-70 chaperone family in the ER of eukaryotic cells. It has been proposed that BiP associates transiently with unfolded or misfolded proteins to modulate protein folding (5,7). ERp94, a member of hsp-90 chaperones, appears to function with BiP to assist protein folding in the ER lumen (8). ERp72 has been identified as an ER protein containing protein disulfide isomerase homology units (9). In addition, calnexin (also called p88, IP90), an ER transmembrane protein, plays a chaperone-like role by binding with monoglucose-containing N-linked glycans of newly synthesized glycoproteins (5,10). Calreticulin, another ER Ca 2ϩ -binding protein, has also been implicated as a molecular chaperone (9,11).
The folding of the ␤ subunit of hCG has proven to be an excellent model for studying the events that lead to the production of a biologically active hormone (12,13). It has been reported to be the only mammalian protein for which a folding pathway in intact cells has been established (14). Because it is a secretory glycoprotein with six intramolecular disulfide bonds and must assemble with the glycoprotein hormone ␣ subunit to become biologically active, several lessons may be learned about the role of disulfide bonds, oligosaccharides, and chaperones in the in vivo folding, assembly, and secretion of such proteins. Using the formation of disulfide bonds as an index of conformational changes during protein folding, the intracellular kinetic folding pathway of the ␤ subunit of hCG has been determined (12,13,15), as shown below with the disulfide bonds formed between Cys residues (e.g. Cys 34 -Cys 88 ) at each step indicated above the arrows: Wild type hCG-␤ subunits recovered from the ER possess two high mannose-containing asparagine (N)-linked oligosaccharide chains at residues Asn 13 and Asn 30 (16). We have previously examined the kinetics of hCG-␤ folding in Chinese hamster ovary (CHO) cells transfected with wild type or mutant hCG-␤ genes lacking either or both N-linked glycosylation sites (17) as well as CHO cells containing hCG-␤ genes mutated at each of the cysteines involved in the formation of the six intramolecular disulfide bonds (18,19). Folding of the ␤ subunit is inhibited or slowed with many of these mutants. In the case of the glycosylation mutants, we found that the t1 ⁄2 for folding of the first detectable ␤ folding intermediate, p␤1, into the second major hCG-␤ folding intermediate, p␤2 (the rate-determining step in hCG-␤ folding), was 5-7 min for wild type ␤ but 33 min for ␤ lacking both N-linked glycans. However, the t1 ⁄2 of conversion of p␤1 3 p␤2 was 7-8 min in CHO cells expressing hCG-␤ subunits missing only the Asn 13 -linked glycan and 10 min for hCG-␤ subunits missing only the Asn 30 -linked glycan (17). Moreover, we reported that the ER chaperones BiP, ERp74, and ERp94, but not calnexin, are co-immunoprecipitated with unglycosylated hCG-␤ folding intermediates (17).
In this report, we examine whether the formation of stable complexes of hCG-␤ with ER chaperones occurs in several CHO cell lines transfected with the WT ␤ gene or a variety of mutated hCG-␤ genes: 1) slow folding mutants in which both Cys residues of intramolecular disulfide bonds were replaced by Ala, 2) glycosylation mutants in which Asn residues of N-linked glycosylation sites were converted to Gln, or 3) a previously undescribed CHO cell line transfected with a Pro 3 Gly mutant of hCG-␤ (P73G) that slows hCG-␤ folding. We have determined that the amount of hCG-␤-chaperone complex correlates with the rate or extent of hCG-␤ subunit folding. The slower folding mutants formed a greater detectable amount of chaperone complex, but the nonfolding mutants did not form detectable amounts of chaperone complex. Moreover, the hCG-␤-chaperone complexes appear to be productive in that hCG-␤ contained in them is ultimately folded and secreted from CHO cells. Finally, the hCG-␤-chaperone complexes involve the formation of intermolecular disulfide bonds.

EXPERIMENTAL PROCEDURES
Cell Culture-The terminology used here is: CHO ␤ WT, ␤ Asn1, ␤ Asn2, and ␤ Asn(1ϩ2) for CHO cells transfected with wild type hCG-␤ genes or hCG-␤ genes containing mutations Asn 3 Gln, at Asn 13 , Asn 30 , or both glycosylation sites, respectively (20). CHO ␤C9A/C90A, ␤C38A/ C57A, and ␤C23A/C72A refer to cells in which Cys 9 and Cys 90 , Cys 38 and Cys 57 , and Cys 23 and Cys 72 residues were replaced by Ala residues (18). These stably transfected CHO cells were grown and maintained in Ham's F-12 medium supplemented with the antibiotic G418 as described previously (17,18). CHO ␤ P73G denotes the conversion of the hCG-␤ Pro 73 residue to a Gly residue. CHO P73G cells were transfected with the pGS plasmid (a gift of Dr. Tyler White, Scios Nova, Mountainview, CA) and were grown and maintained in UltraCulture medium (BioWhittaker, Inc) supplemented with 50 M methionine sulfoximine.
Biosynthetic Labeling-CHO cells were metabolically labeled as described previously (17). Briefly, 100-mm Petri dishes of 90% confluent CHO cells were starved in cysteine-free and serum-free Dulbecco's modified Eagle's medium without G418 for 30 min. These cells were then pulse-labeled for 5 min with 200 -400 Ci/ml L-[ 35 S]cysteine (1100 Ci/mmol; DuPont NEN) in serum-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) without G418 and lacking cysteine and chased for the times indicated in the text. Then cells were rinsed with cold phosphate-buffered saline and lysed in 5 ml of phosphate-buffered saline (pH 8.0) containing detergents (1% Triton X-100, 0.5% sodium deoxycholate, and 0.1% SDS), protease inhibitors (20 mM EDTA and 2 mM phenylmethylsulfonyl fluoride), and 50 mM sodium iodoacetate (pH 8.3) to alkylate free sulfhydryl groups of folding intermediates and to prevent further disulfide bond formation or rearrangement.
Immunoprecipitation of Cell Lysates and Separation of ␤ Folding Intermediates by SDS-PAGE-The cell lysates or chase media were immunoprecipitated with an anti-hCG-␤ polyclonal antibody (1:1000), which recognizes all forms of the hCG-␤ subunit, for 16 h at 4°C, and the immune complexes were precipitated with protein A-Sepharose (Sigma), as described previously (17).
Purification of hCG-␤ Folding Intermediates-The protein A-Sepharose beads containing hCG-␤ immune complexes were eluted with 6 M guanidine hydrochloride (pH 3) (Sequanal grade; Pierce) for 16 h with rotation at room temperature. Eluates were purified by reversed-phase high performance liquid chromatography (HPLC) using a Vydac 300-Å C 4 column with elution by an acetonitrile gradient as described previously (12). The fractions containing hCG-p␤1 or -p␤2 subunits and chaperone-hCG-␤ complexes were collected and concentrated by Speed-Vac® concentrator as described previously (17).
SDS-PAGE and Western Blot Analysis-To detect the presence of proteins that co-immunoprecipitated with ␤ folding intermediates, CHO ␤ WT, ␤ Asn1, ␤ Asn2, or ␤ Asn(1ϩ2) cell lysates were immunoprecipitated with polyclonal hCG-␤ antisera, and the hCG-␤ subunits were eluted from protein A-Sepharose beads with 6 M guanidine (pH 3.0) for 16 h with rotation at room temperature and purified by C 4 reversed-phase HPLC (12,17). The 80 -95 min and 100 -107 min fractions (termed C1 and C2, respectively) containing ␤-chaperone complexes were collected and separated by SDS-PAGE under reducing conditions. The resolved proteins were transferred to Immobilon polyvinylidene difluoride transfer membranes (Millipore) in a Trans-Blot® Cell (Bio-Rad) at 480 mA for 1-2 h at 4°C. The membranes were immunoblotted with either rat anti-BiP (1:500), rabbit anti-ERp72 (1: 100), rabbit anti-ERp94 (1:100), or rabbit anti-calnexin (1:2000) overnight at 4°C with gentle shaking. Rat anti-BiP polyclonal antibody was kindly provided by Dr. David Bole (University of Michigan, Ann Arbor, MI). Rabbit anti-ERp72 (against the 16 C-terminal amino acids of murine ERp72) and rabbit anti-ERp94 (against the 16 C-terminal amino acids of murine ERp94) were provided by Dr. Michael Green (St. Louis University Medical Center, St. Louis, MO). The rabbit anticalnexin (against the C-terminal 19 amino acids of canine calnexin) was provided by Dr. Ari Helenius (Yale University, New Haven, CT). Membranes were washed several times with buffer containing 20 mM Tris, 150 mM NaC1, 1% nonfat milk, and 0.2% Tween (pH 7.4) and incubated with anti-rat or anti-rabbit IgG peroxidase conjugate (1:1000, Sigma) for 30 min at 4°C. Enhanced chemiluminescence (ECL Western blotting kit, Amersham Life Science) was used to identify the blotted proteins. Aliquots of the HPLC purified hCG-␤ folding intermediates and chaperone complexes were also analyzed by reducing and nonreducing SDS-PAGE followed by autoradiography as described previously (17).
Determination of Amount of Association of Chaperones with ␤ Folding Intermediates-[ 35 S]Cysteine-labeled hCG-␤ folding intermediates (p␤1 and p␤2) and C1 and C2 chaperone complexes were quantitated by summation of the radioactivity detected in each of the respective peaks eluting in the C 4 -HPLC chromatogram. The values were normalized by expressing the percentage of total radioactivity of each form depicted in the chromatogram.

Molecular Chaperones Are Involved in Unglycosylated hCG-␤ Subunit
Folding-To investigate the kinetics of association of molecular chaperones with ␤ folding intermediates, CHO cells, transfected with the wild type (␤ WT) or a mutated hCG-␤ gene [␤ Asn(1ϩ2)] containing mutations N13Q and N30Q at the two hCG-␤ N-linked glycosylation consensus sequences (20), were pulse-labeled for 5 min with [ 35 S]cysteine and chased for periods of 0 -120 min. The cell lysates were then immunoprecipitated with polyclonal hCG-␤ antisera followed by protein A-Sepharose precipitation. Bound immunocomplexes were eluted with 6 M guanidine and purified by C 4 reversed-phase HPLC. As previously seen (17), wild type p␤1 (Fig. 1A) converted to p␤2 more efficiently than unglycosylated p␤1 (p␤1 (0) ) ( Fig. 1,  B-D). Two additional peaks, C1 and C2, were more prevalent in the HPLC profile of the ␤ Asn(1ϩ2) mutant (Fig. 1, B-D) than in that of WT ␤ (Fig. 1A). The new point that we make here is that the C1 and C2 complexes are present as early as 5 min into the chase before much p␤2 is formed and that the C1 and C2 complexes disappear with the same kinetics as folding of p␤1 into p␤2. The C1 peak contained hCG-␤ and co-immunoprecipitated proteins, three of which were identified as the ER chaperones BiP, ERp72, and ERp94 (Fig. 2). The C2 fraction also contained these chaperones as well as hCG-␤ (17). The fact that the C1 and C2 peaks diminished with increasing chase time as the earliest folding intermediate, p␤1, was converted to the next folding intermediate, p␤2, (Fig. 1, B-D) suggested that the association of unglycosylated ␤ with certain ER chaperones is an intermediate step in the hCG-␤ folding process. It should be noted that calnexin was not detected in the C1 and C2 complexes of unglycosylated ␤ (data not shown).
To determine whether the absence of both N-linked oligosaccharides was necessary to form the C1 and C2 hCG-␤-chaperone complexes, a similar experiment to that shown in Fig. 1 was carried out with CHO cells containing the ␤ subunit gene mutated at only one N-linked glycosylation site (either the Asn 13 or Asn 30 codon). CHO cells containing the wild type ␤ gene or these mutants were pulsed for 5 min with [ 35 S]Cys and chased for 5 min, and the HPLC profiles of the anti-␤ immunoprecipitates analyzed as noted above. As can be seen in Fig.  3, there was only a small amount of immunoprecipitated radioactivity in the C1 and C2 loci for the wild type samples (Fig.  3A), intermediate amounts in the C1 and C2 complexes for the monoglycosylated mutants ␤ Asn1 and ␤ Asn2, (Fig. 3, B and  C), and the highest amount in the C1 and C2 complexes for the ␤ Asn(1ϩ2) mutant (Fig. 3D). It should be noted that the rate of p␤1 to p␤2 conversion is as follows: ␤ WT Ն ␤ Asn1 Ͼ ␤ Asn2 Ͼ Ͼ ␤ Asn(1ϩ2) (17), and this is reflected by the data in Fig. 3 in that the formation of p␤2 was observed for wild type, ␤ Asn1, and ␤ Asn2 but was barely detectable for ␤ Asn(1ϩ2) after a 5-min chase.
Quantitation of the Relative Amounts of hCG-␤ Present as Folding Intermediates and as Chaperone Complexes in ␤ WTand Mutant ␤-containing CHO Cells-To quantitate the amount of hCG-␤ forms present as p␤1 and p␤2 folding intermediates or contained in C1 and C2 chaperone-containing complexes, experiments similar to those described in Fig. 3 were carried out with additional slow folding or nonfolding hCG-␤ mutants. CHO cells were pulsed for 5 min with [ 35 S]Cys and , or 120 min. The cell lysates were immunoprecipitated with anti-hCG-␤ antibody (1:1000) followed by protein A-Sepharose precipitation. The immunocomplexes were eluted from protein A-Sepharose beads with 6 M guanidine and purified on reversed-phase HPLC. The C1 complex from the HPLC (fraction 80 -95) was analyzed by reducing SDS-PAGE and transferred to polyvinylidene difluoride membranes for Western blotting. The membranes were cut horizontally by using a prestained molecular weight marker (Bio-Rad) as a guide. The top portions of the membranes containing higher molecular weight proteins (M r ϭ Ͼ35,000) were probed using antibodies against three ER molecular chaperones: BiP, M r ϭ 78,000 (panel A, I), ERp72, M r ϭ 72,000 (panel B, I), and ERp94, M r ϭ 94,000 (panel C, I). The bottom portion of the membrane was subjected to Western blot analysis using hCG-␤ antibody (deglycosylated hCG-␤ M r ϭ 18,000 (panels A-C, II)). Nonimmune sera were used as controls and demonstrated the absence of bands at the loci detected by the specific chaperone or ␤ antibodies (data not shown). chased for 5 min. Then the immunoprecipitated, HPLC purified p␤1, p␤2, C1, and C2 peaks were quantitated by determining the amount of radioactivity in each peak, and the relative amounts of radioactivity in each peak (compared with the total hCG-␤ immunoprecipitated CPM in the entire HPLC profile) were calculated for each peak and plotted as a percentage of total CPM (Fig. 4). The relative amount of immunoprecipitated CPM in the C1 ϩ C2 fractions known to contain ␤ and chaperones ( Fig. 2 and Ref. 17) correlated with the rate or extent of hCG-␤ folding. The P73G mutant (Fig. 4, lane 5) has a rate of p␤1 3 p␤2 conversion of t1 ⁄2 ϭ 13 min and is similar to the ␤ Asn1 and ␤ Asn2 mutants (Fig. 4, lanes 2 and 3) in the amount of C1 and C2 formed. Interestingly, the nonfolding Cys mutants C38A/C57A (Fig. 4, lane 7) and C9A/C90A (Fig. 4, lane 8) did not produce detectable amounts of C1 and C2 complexes. The C23A/C72A mutant p␤1 only partially folds into p␤2 (Յ25% during a 30-min chase) (18) and produced small amounts of C1 and C2 complexes (Fig. 4, lane 6).
Characterization of hCG-␤-Chaperone Complexes-To determine whether BiP was present in the HPLC samples eluted in the C1 peaks of the anti-␤ immunoprecipitates from the wild type and glycosylation mutants, the samples eluting at 80 -95 min (C1) shown in Fig. 3 (A-D) were analyzed by nonreducing or reducing SDS-PAGE (Fig. 5). The top part of the gel was Western blotted with anti-BiP and developed by ECL (Fig. 5A). The bottom half of the gel was dried and developed by autoradiography to show the radiolabeled ␤ subunits present (data not shown). Aliquots from identically prepared samples were developed by autoradiography to detect the amount of [ 35 S]Cyslabeled ␤ (Fig. 5, B-E). BiP was present in the C1 fractions from ␤ WT and from each of the glycosylation mutants (Fig.  5A).
The radioactively labeled hCG-␤ present in the C1 and C2 fractions was present primarily as high molecular weight complexes when analyzed by nonreducing SDS-PAGE (Fig. 5, B  and D), but under reducing conditions most of the hCG-␤ present in the C1 and C2 complexes collapsed to monomeric hCG-␤ forms (Fig. 5, C and E). The separation of hCG-␤ folding intermediates from chaperones under a combination of denaturing and reducing conditions but not under denaturing conditions alone indicates that disulfide bond formation is an essential characteristic of the molecular chaperone-␤ interaction.
Unglycosylated ␤ Is Released from Chaperones and Secreted-When the 2-or 24-h chase media were collected from [ 35 S]Cys-labeled, ␤ Asn(1ϩ2)-containing CHO cells, immunopurified, and separated by HPLC, only forms that elute at the position of mature ␤ were observed (Fig. 6). Neither the C1 and C2 complexes nor degradation products of ␤ were observed in the chase media. Previous studies have shown that most of the C1 and C2 complexes disappear from intracellular lysates by 5 h, the time at which most of the hCG-␤ forms present in these complexes have been converted into p␤2 and before there is any   FIG. 3. Detection of hCG-␤-chaperone complexes. CHO cells (␤ WT, ␤ Asn1, ␤ Asn2, ␤ Asn(1ϩ2)) were radiolabeled with [ 35 S]cysteine for 5 min and chased for 5 min. The polyclonal anti-hCG-␤ antibody was used to immunoprecipitate cell lysates. Immunocomplexes were eluted from protein A-Sepharose beads and purified on C 4 reversed-phase HPLC. Panel A, ␤ WT; panel B, ␤ Asnl; panel C, ␤ Asn2; panel D, ␤ Asn(1ϩ2). p␤1 (1) , p␤1 lacking the Asn 13 -linked glycan; p␤2 (1) , p␤2 lacking the Asn 13 -linked glycan; p␤1 (2) , p␤1 lacking the Asn 30linked glycan; p␤2 (2) , p␤2 lacking the Asn 30 -linked glycan; p␤1 (0) , p␤1 lacking both N-linked glycans; p␤2 (0) , p␤2 lacking both N-linked glycans; C1 and C2, protein complexes containing hCG-␤ and chaperone-like proteins. are as follows: n ϭ 3 for WT, ␤ Asn1, ␤ Asn2, and ␤ Asn(1ϩ2); n ϭ 2 for P73G and C23A-C72A; n ϭ 1 for C38A/C57A and C9A/C90A. Only a single experiment is shown for the latter two mutants because no conversion of p␤1 to p␤2 has been observed in several previous experiments (18,19) and no C1 or C2 peaks were noted in duplicate experiments (not shown). The S.D. bars for the p␤2 values of the ␤ Asn2 (lane 3) and C23A-C72A (lane 6) mutants were too small (Յ 1.5) to show a deviation from the mean. The P values (by Student's t test) for the WT and oligosaccharide mutants are as follows: WT versus Asn1, p Ͼ 0.2 Ͻ 0.5; WT versus Asn2, p Ͼ 0.1 Ͻ 0.2; WT versus Asn(1ϩ2) p Ͻ 0.01; Asn1 versus Asn(1ϩ2), p Ͼ 0.05 Ͻ 0.10; Asn2 versus Asn(1ϩ2), p Ͻ 0.01. evidence of intracellular degradation of hCG-␤ (17). Thus, the kinetic precursor-product relationship between the disappearance of the C1 and C2 complexes and the appearance of the p␤2 folding intermediate as well as the inability to detect these complexes in the culture medium even after 24 h indicates that the C1 and C2 complexes were dissociated and suggests that association of hCG-␤ with chaperones is a transient step in the folding pathway of unglycosylated hCG-␤ subunits. The two secreted ␤ peaks seen in Fig. 6 represent p␤2-like, folded forms that appear to differ in the amount of alkylation by iodoacetate. 2

DISCUSSION
The data presented here extend a previous observation that hCG-␤ subunits lacking both N-linked oligosaccharides form tight complexes with ER chaperones (17) and further demonstrate that WT ␤ and ␤ subunits lacking only one of the two N-linked chains form similar complexes that include the ER chaperone BiP. The amount of the C1 and C2 complexes containing monoglycosylated ␤ and chaperones were intermediate between wild type ␤ and unglycosylated ␤ and appeared to be related to the rate at which these mutant ␤ forms fold in transfected CHO cells. The mutant lacking both N-linked glycans has the slowest t1 ⁄2 of folding (33 min) and the highest amount of C1 and C2 hCG-␤-containing complexes (Figs. 3 and 4).
These data raise the question whether the amount of ␤-chaperone complexes is related to the rate of folding or to the lack of N-linked glycans, which may simply allow for the availability of more hydrophobic sites to bind the chaperones. To test this, other slow folding or nonfolding mutants of ␤ were examined for the amount of ␤-chaperone complexes formed in ␤ genetransfected CHO cells (Fig. 4). The data indicate that the cysteine mutants ␤ C38A/C57A and ␤ C9A/C90A, which do not fold to form p␤2, do not form detectable, stable ␤-chaperone complexes (Fig. 4). Table I summarizes the relationship of the rate of hCG-␤ folding (p␤13p␤2), the relative amount of hCG-␤ contained in the C1 and C2 complexes, and the relative amount of BiP present in the C1 and C2 complexes. These results 2  indicate that the rate or extent of ␤ folding determines the amount of stable binding of hCG-␤ to chaperones. Although the chaperone primarily probed for in the complexes formed by the various ␤ forms was BiP (Fig. 5), it is likely that other ER chaperones including ERp72 and ERp94 are also present in the C1 and C2 ␤-chaperone complexes formed with WT ␤ and the P73G mutant because these complexes separate by HPLC with elution times identical to those containing these chaperones and unglycosylated ␤ (data not shown).
The C1 and C2 ␤-chaperone complexes are not just dead-end complexes leading to ␤ degradation because these disappear as p␤1 is converted into p␤2 (Fig. 1) at a time when total recovery of ␤ does not change and no ␤ degradation products are detected (17). Furthermore, unglycosylated ␤ does fold efficiently, though more slowly than wild type, into a mature, secretable form ( Figs. 1 and 6).
Interestingly, the C1 and C2 unglycosylated ␤-chaperone complexes survive treatment with 0.1% SDS (during the CHO cell lysis procedure) and 6 M guanidine (during the elution step of immunoconjugates from protein A-Sepharose beads) and boiling in 1% SDS prior to SDS-PAGE (Fig. 5). These results indicate that the C1 and C2 ␤-chaperone complexes are covalently linked forms. This was demonstrated by the fact that these complexes were only dissociated by reduction (Fig. 5), suggesting that intermolecular disulfide bonds are formed between unfolded ␤ and ER chaperones or that lattices of ␤ oligomers are formed by intermolecular disulfide bonds, thus trapping chaperones in these lattices. There is precedent for the formation of such intermolecular disulfide bond-linked hCG-␤ oligomers being formed by unfolded or incompletely folded forms of ␤. For example, we have previously shown that cysteine-mutant forms of hCG-␤ C34A/C88A, C38A/C57A, and C9A/C90A, which do not fold into p␤2, form such oligomers (19). However, the fact that these mutants do not form detectable C1 and C2 complexes does not support the lattice hypothesis but rather supports the idea that intermolecular disulfide bonds are formed between hCG-␤ and chaperones.
Wild type hCG-␤ also forms a BiP-containing complex (Fig.  5A), suggesting that these tight complexes are involved in the folding pathway of WT ␤ as well. It is not clear why the C23A/C72A mutant forms only a small amount of stable C1 and C2 complexes because this mutant also folds, albeit incompletely, to form p␤2. It may be that the cysteines at positions 23 and 72 are involved in the intermolecular disulfide bonds formed between ␤ and the chaperones or that this mutant has a conformation that does not favor stable interaction with chaperones. The mutants that do not fold from p␤1 to p␤2 (C38A/C57A and C9A/C90A) most likely do not form stable C1 and C2 complexes because these mutant ␤ forms do not achieve a conformation that allows stable binding to chaperones, which may relate to the fact that they do not progress down the folding pathway.
When protein folding occurs under suboptimal conditions, for example, when cellular ATP levels are depleted (21) or when BiP or protein disulfide isomerase are present in substoichiometric amounts in relation to substrate (22), disulfide-linked aggregates of substrates can occur. In some instances, these disulfide-cross-linked aggregates can be rescued by restoring more favorable folding conditions, e.g. by adding ATP (21). What we are proposing here is unique in that our data suggest that disulfide-linked substrate-chaperone complexes are normal intermediates in the folding of wild type proteins to their native structures.
Finally, the possible differences between the C1 and C2 hCG-␤-chaperone complexes should be noted. It is not clear why these two complexes migrate differently on reversed-phase HPLC because they both contain hCG-␤ and the three chaperones BiP, ERp72, and ERp94 (17). There are some differences, however, between the C1 and C2 complexes. For example, the high molecular weight bands seen by SDS-PAGE are somewhat different between the two fractions (Fig. 5, B-E), suggesting that the array of chaperones or other binding proteins present in the two complexes may differ. Thus, it will be interesting to determine if there is a precursor-product relationship between the hCG-␤ forms contained in the two complexes and whether the different chaperones contained in the C1 and C2 complexes act at different steps in the folding pathway.
a BiP was detected on a Western blot similar to that shown in Fig. 5. b Conversion of p␤1 3 p␤2 does not occur in these hCG-␤ mutants (18,19).