Multiple molecular chaperones complex with misfolded large oligomeric glycoproteins in the endoplasmic reticulum.

Thyroglobulin (Tg), the major protein secreted by thyroid epithelial cells and precursor of thyroid hormones, is a large dimeric glycoprotein with multiple disulfide bonds. The folding and assembly of this complex molecule begins in the endoplasmic reticulum (ER) and is likely to involve a variety of reactions catalyzed by molecular chaperones (Kuznetsov, G., Chen, L. B., and Nigam, S. K. (1994) J. Biol. Chem. 269, 22990-22995). By coimmunoprecipitation in rat thyroid cells, we were able to demonstrate that BiP, grp94, ERp72, and grp170, four proteins believed to function as specific molecular chaperones, complex with Tg during its maturation. The same complex of the four putative chaperones with Tg was observed in cells treated with tunicamycin, indicating that these four ER chaperones stably associate with Tg when it is misfolded/misassembled due to inhibition of its glycosylation. BiP, grp94, and ERp72 were also found to associate with Tg in cells in which misfolding was induced by perturbing ER calcium stores. To determine if the assembly of a complex between the four chaperones and Tg under conditions of misglycosylation was unique to the maturation of this particular secretory protein or a more general phenomenon, adenovirus-transformed rat thyroid cells that do not synthesize Tg were analyzed. In these transformed cells, the only protein these same four chaperones were found to complex with was a protein of approximately 200 kDa. This protein was subsequently identified as thrombospondin, which, like Tg, is a large oligomeric secreted glycoprotein with multiple disulfide bonds. We therefore propose that these ER chaperones complex together with a variety of large oligomeric secretory glycoproteins as they fold and assemble in the ER.

The ER 1 contains a number of lumenal proteins, many of which have been proposed to function as molecular chaperones in the folding and/or assembly of membrane and secretory proteins (1)(2)(3)(4)(5)(6). Molecular chaperones are postulated to bind transiently to newly synthesized proteins, interact with the polypeptides as they fold through sequential cycles of binding and release, and finally release the proteins upon their folding and/or assembly (7). Some of these ER chaperones are thought to bind to exposed hydrophobic regions of proteins (grp78/BiP) (8,9), whereas others are thought to participate in disulfide isomerization (protein disulfide isomerase, ERp72) (10,11) and peptidyl-prolyl isomerization (FKBP13) (12,13). Many of the ER chaperones have been shown to associate with secretory proteins transiting the ER (examples include BiP, grp94, ERp72, and calnexin) (6, 14 -16), but it has been unclear whether they work together as part of a large complex or separately. One difficulty has been that it has often been necessary to resort to cross-linking approaches to demonstrate associations of multiple ER chaperones with secretory proteins (6,16,17). Whether stable associations of multiple chaperones with secretory proteins occur during their folding and assembly in the ER is a question of considerable interest. Furthermore, it is unknown whether particular sets of chaperones work together on certain classes of proteins or whether the case for each protein is different.
In a previous study, we reported that several unrelated unfolded protein substrates bind to the same set of resident ER proteins on affinity columns (5). These ER proteins, which all exhibited characteristics of molecular chaperones (such as binding to denatured substrates and elution from them in an ATP-dependent manner), included BiP, grp94, ERp72, protein disulfide isomerase, and calreticulin. One of the substrates used in the study was thyroglobulin (Tg), a protein precursor of thyroid hormones produced by thyroid epithelial cells. In support of these in vitro findings, we subsequently demonstrated by cross-linking and immunoprecipitation that BiP, grp94, and ERp72 also associate with Tg in intact thyroid cells (6). Similar associations between Tg and these chaperones were observed in a cell culture model for epithelial ischemia when ATP levels declined to less than 10% of control (18). Nevertheless, it remained unclear whether these associations exist in the absence of cross-linking or severe ATP depletion.
Coordinate binding to substrates and coelution of a set of ER molecular chaperones (5), as well as their coordinate induction under conditions that perturb the internal environment of the ER (13), prompted us to hypothesize the existence of large macromolecular chaperone-containing complexes that are involved in maturation of certain classes of secretory and membrane proteins. The association of multiple resident ER proteins with Tg by cross-linking is consistent with this hypothesis. In the present study we investigated the involvement of multiple ER molecular chaperones in the maturation of large oligomeric secreted glycoproteins under conditions of misfolding in the ER. Thyroglobulin (19) typifies this class of proteins, made by many specialized secretory cells. Like immunoglobulin chains (20,21) and thrombospondin (22), it is a large oligomeric glycoprotein with multiple disulfide bonds. The size (Tg is a ϳ660-kDa homodimer consisting of two subunits of ϳ330 kDa) and apparent complexity (extensive glycosylation, disulfide bonding, and other posttranslational modifications) of Tg suggest that its folding and assembly may involve many reactions catalyzed by multiple ER chaperones. Indeed, it appears as though Tg does, at least transiently, associate with BiP, grp94, ERp72, and calnexin during its maturation (6,23); however, with the exception of BiP (6,23,24), stable associations with Tg have not been observed. Since such associations seemed more likely to occur when the ER transit time of Tg was prolonged, we examined conditions that perturb the internal environment of the ER and lead to accumulation of immature proteins in the ER lumen. Here, we report that following treatment of cells with tunicamycin, an agent that inhibits N-linked glycosylation in the ER, the same set of four ER molecular chaperones (BiP, grp94, ERp72, and grp170) stably complexes with two secreted proteins that exhibit strikingly similar structural characteristics. We therefore propose that this set of four ER chaperones is involved in a dynamic "folding complex" for many large oligomeric glycoproteins.

EXPERIMENTAL PROCEDURES
Materials-Rabbit polyclonal antiserum against BiP used for immunoprecipitation was purchased from Affinity Bioreagents (Neshanic Station, NJ). The mouse monoclonal antibody against BiP used for Western immunoblot analysis and rat monoclonal antibody against grp94 were obtained from Stress-Gen (Victoria, British Columbia). Rabbit polyclonal antiserum against grp170 was a gift from Dr. J. Subjeck (Roswell Park Cancer Institute). Rabbit polyclonal antiserum against ERp72 was kindly provided by Dr. M. Green (St. Louis University). Rabbit polyclonal anti-thrombospondin antiserum used for immunoprecipitation was a gift from Dr. J. Lawler (25) (Brigham and Women's Hospital, Harvard Medical School), and the rabbit polyclonal antithrombospondin antiserum used for Western immunoblot analysis was kindly provided by Dr. V. M. Dixit (University of Michigan Medical Center). Rabbit polyclonal antiserum against thyroglobulin was purchased from DAKO Corp. (Carpinteria, CA). Secondary antisera were obtained from Amersham Corp. The following materials were purchased from Sigma: thyroid-stimulating hormone, insulin, hydrocortisone, transferrin, somatostatin, glycyl-L-histidyl-L-lysine acetate, leupeptin, pepstatin A, antipain, and tunicamycin. DSP was obtained from Pierce. A23187 and thapsigargin were from Calbiochem.
Cross-linking and Immunoprecipitation-Cell monolayers were washed twice with the phosphate-buffered saline (PBS) and suspended in 1 ml of PBS, pH 7.4, containing 5 mM EDTA and 5 mM EGTA (6). The cells were incubated with the cross-linking agent, DSP (100 g/ml), or vehicle (dimethyl sulfoxide, final concentration of 1%) for 30 min at room temperature. The cross-linking reaction was stopped by incubation with 100 mM Tris/HCl, pH 7.5, for 15 min. Cells were then lysed on ice by addition of 1% Triton X-100 in the presence of protease inhibitors (1 mM phenylmethylsulfonyl fluoride and 5 g/ml each leupeptin, pepstatin A, and antipain). Prior to immunoprecipitation, samples were incubated with protein A-Sepharose CL-4B (Pharmacia Biotech Inc.) to remove any material that binds nonspecifically to protein A. Samples were incubated with various antisera for 16 h at 4°C followed by incubation with protein A-Sepharose for 1 h at 4°C. The pellets were washed in PBS, suspended in sample buffer (29) containing 50 mM dithiothreitol, heated to 100°C for 10 min and analyzed by SDS-PAGE (29). Protein-associated radioactivity was analyzed on a PhosphorImager using the Image-Quant software package (Molecular Dynamics). Images were processed using the Adobe Photoshop TM software and printed with the Fujix Pictography 3000 color printer.
Western Immunoblot Analysis-Electrophoresed proteins were transferred to polyvinylidene fluoride membranes (Immobilon, Millipore Corp.) by electroblotting at 30 V for 18 h at room temperature. Membranes were incubated with various primary antisera followed by incubation with secondary horseradish peroxidase-labeled anti-immunoglobulin antisera as described (5). Immunoblots were developed using the ECL kit (Amersham Corp). The proteins were detected using Hyperfilm ECL (Amersham Corp). The exposed film was scanned using the HP Scan Jet IIcx, and the images were processed using the Adobe Photoshop software and printed with the Fujix Pictography 3000 color printer.
Sucrose Density Gradient Analysis-Cell lysates were layered on top of linear 5-20% sucrose gradients prepared in PBS, pH 7.4, containing 5 mM EDTA and 5 mM EGTA (6). The gradients were centrifuged at 32,000 rpm for 24 h at 4°C using a Beckman SW40TI ultracentrifuge rotor. 15 fractions (0.8 ml) were collected using a Buhler Instruments Auto Densi-Flow IIC gradient fractionator. Trichloroacetic acid was added to fractions at a final concentration of 10% to precipitate proteins. Trichloroacetic acid precipitates were washed once with ice-cold acetone, dissolved in sample buffer containing 50 mM dithiothreitol (DTT), heated to 100°C for 10 min, and analyzed by SDS-PAGE followed by Western Immunoblotting with anti-Tg of anti-BiP antisera.
Immunofluorescence-FRTL-5 cells grown on coverslips were fixed in methanol (Ϫ20°C, 5 min), washed twice with PBS, pH 7.4, and incubated with rabbit polyclonal anti-Tg antiserum for 1 h at 37°C in a humidified chamber. Cells were then washed in PBS three times and incubated with rhodamine-conjugated anti-rabbit immunoglobulin antiserum. Coverslips were mounted onto a glass slide in glycerol-gelatin medium (Sigma). Staining was examined using a confocal laser scanning microscope LSM 410 (Zeiss, Germany). The images were processed using the Adobe Photoshop software and printed with the Fujix Pictography 3000 color printer.

RESULTS AND DISCUSSION
Under conditions which perturb the folding and secretion of Tg, existing interactions of this large dimeric secretory protein with ER molecular chaperones would be expected to be prolonged and more stable, and therefore easier to detect without the use of cross-linking reagents. Inhibitors of glycosylation and agents that perturb ER Ca 2ϩ stores create such conditions, and we have employed both types of agents to study the involvement of ER chaperones in Tg maturation. The nucleoside antibiotic tunicamycin inhibits N-linked glycosylation in the ER by blocking the formation of the lipid-linked oligosaccharide donor through specific inhibition of the transfer of GlcNAc-1phosphate from UDP-GlcNAc to dolichyl phosphate (30). After tunicamycin treatment, misglycosylated proteins accumulate in the ER (31), which is accompanied by the increased synthesis of ER molecular chaperones (13,36), including BiP, grp94, and ERp72. Thus, since mature Tg is extensively glycosylated, tunicamycin treatment would be expected to severely affect its conformation and stability and to facilitate detection of interactions of Tg with molecular chaperones of the ER without the use of cross-linking agents. Two rat thyroid epithelial call lines that synthesize and secrete Tg were used in our study: FRTL-5 and PCCl3. Our findings in both cell lines were virtually identical; therefore, below we present results obtained with only one type of cells.
A relatively low dose of tunicamycin (1 g/ml) was chosen. A recently published study (24) reported that as little as 0.5 g/ml tunicamycin results in a complete inhibition of N-linked glycosylation of Tg. At the same time, a number of studies in recent years have shown that concentrations of tunicamycin similar to the one we have chosen result in virtually complete inhibition of N-linked glycosylation without a significant effect on overall protein synthesis (32,33). When FRTL-5 cells were treated with 1 g/ml tunicamycin, Tg was found intracellularly in a lower molecular weight (unglycosylated) form that could be clearly distinguished by SDS-PAGE from its glycosylated counterpart (Fig. 1A, lanes 10 and 12). Inhibition of N-linked glycosylation of Tg persisted for up to 5 h after the removal of tunicamycin from the culture medium (Fig. 1A, lane 11). Secretion of unglycosylated Tg was almost completely (Ͼ95%) blocked in tunicamycin-treated cells (Fig. 1A, lanes 3, 6, and 9) as well as in cells pretreated with tunicamycin for 16 h and then transferred to tunicamycin-free medium for up to 5 h (Fig.  1A, lanes 2, 5, and 8). Thus, under the conditions used in our study, tunicamycin induces complete inhibition of N-linked glycosylation of Tg and completely blocks its secretion, and these two effects are not reversed during the time periods employed in the experiments described below. By immunofluorescence, the unglycosylated Tg localized largely to the ER (Fig. 1B). Furthermore, both Tg and BiP appeared to associate into larger macromolecular complexes when analyzed by sucrose density gradient centrifugation (Fig. 1C). In control cells, Tg sedimented largely in fractions 8 -12 with a small amount of the protein present in fractions 13-15 and in fraction P. Fraction P represents material pelleted at the bottom of the gradient as a result of transient association of Tg into large complexes in the course of its normal maturation (34). The majority of intracellular BiP in untreated cells was found in fractions 3-8, with a small amount present in heavier fractions (fractions 9 -12 and the pellet fraction, P). Following treatment of cells with tunicamycin, both Tg and BiP have shifted to heavier fractions of the gradient, and there was a marked increase in the amounts of both proteins found in the pellet fraction of the gradient. Together, these data suggested that tunicamycin treatment led to formation of large macromolecular complexes containing unglycosylated Tg and possibly ER molecular chaperones that were retained in the ER.
To ensure that the formation of Tg-containing macromolecular complexes in tunicamycin-treated cells does not represent nonspecific protein aggregation resulting from cell injury, we measured cell viability and overall cellular secretory activity in both treated and untreated cells. To assess cell viability, a trypan blue exclusion assay was performed following treatment of FRTL-5 cells with 1 g/ml tunicamycin for 16 h. In both the untreated and tunicamycin-treated cell preparations, less than 1% of cells were non-viable (have incorporated the dye). The assay was performed in triplicate (data not shown).
To estimate the effect of tunicamycin on the overall cellular secretory activity, FRTL-5 cells were pretreated without or with tunicamycin (1 g/ml) for 16 h, pulse-labeled with [ 35 S]methionine/cysteine, and chased in the presence or absence of tunicamycin for 4 h (Fig. 2). Aliquots of media were collected and subjected to SDS-PAGE followed by autoradiography ( Fig. 2A). Amounts of radioactivity associated with Tg secreted into the culture medium as well as with various other secretory products of FRTL-5 cells were quantified (Fig. 2, B  and C). As shown in Fig. 2A, Tg represents the major, although not the sole secretory protein product of untreated thyroid cells FIG. 1. Inhibition of N-linked glycosylation and secretion of thyroglobulin by tunicamycin. A, PCCl3 cells were pretreated with (lanes 2, 3, 5, 6, 8, 9, 11, and 12) or without (lanes 1, 4, 7, and 10) 1 g/ml tunicamycin for 16 h. At the end of the pretreatment period, cells were washed twice with fresh culture medium and incubated for additional 5 h in the absence (lanes 1, 2, 4, 5, 7, 8, 10, and 11) or presence (lanes 3, 6, 9, and 12) of 1 g/ml tunicamycin. Cell lysates (lanes 10 -12) and aliquots of culture medium (lanes 1-9) collected at various times were analyzed on SDS-PAGE followed by Western immunoblotting with anti-Tg antiserum. B, PCCl3 cells grown on coverslips were treated with 1 g/ml tunicamycin for 16 h. At the end of the treatment period, the cells were fixed in methanol and analyzed for Tg immunofluorescence as described under "Experimental Procedures." Immunofluorescence was examined using a confocal laser scanning microscope. The untreated cells incubated with anti-Tg antiserum showed mostly a typical ER staining (data not shown). C, PCCl3 cells were incubated with or without 1 g/ml tunicamycin for 16 h. At the end of the treatment period, tunicamycin-treated and untreated cell lysates were layered on top of linear 5-20% sucrose gradients and centrifuged to equilibrium as described under "Experimental Procedures." 15 fractions were collected; fraction P represents material pelleted at the bottom of each gradient. Fractions were analyzed on 5% SDS-PAGE followed by Western immunoblotting with anti-Tg or anti-BiP antisera. (lane 1). Additional polypeptide bands were present on the SDS gel, although by far the Tg band (ϳ300 kDa) was the most prominent. In control cells, the 300-kDa band represented approximately 75% of total labeled material secreted (Fig. 2B). Treatment of cells with tunicamycin reduced secretion of all polypeptides, but to various degrees ( Fig. 2A, lanes 2 and 3). Secretion of three other secretory products of FRTL-5 cells (denoted by arrows designated p1, p2, and p3) was compared to that of Tg. In tunicamycin-treated cells, the amount of Tg present in culture medium was reduced by almost 200-fold (Fig. 2C). In contrast to Tg, secretion of polypeptides p1 (ϳ112 kDa), p2 (ϳ100 kDa), and p3 (ϳ80 kDa) was reduced by only 1.7-2.5 fold (Fig. 2C). As shown in Fig. 2B, the secretory profile of tunicamycin-treated cells is significantly different from that of control preparations. Following treatment with tunicamycin, Tg becomes a minor secretory product, representing only ϳ5% of total secreted protein, while other polypeptides (like bands p1, p2, and p3) represent a substantial portion of total secretory material. Thus, different secretory proteins of thyroid cells exhibit vastly different sensitivities to tunicamycin treatment. These findings indicate that the inhibition of N-linked glycosylation by tunicamycin results in selective inhibition of secretion rather than in a nonspecific alteration of cellular function. Together with the lack of toxicity of tunicamycin (as estimated by the trypan blue exclusion assay), these findings suggest that the action of tunicamycin on thyroid cells is specific and also support the validity of using tunicamycin as a tool to induce stable associations of Tg with the ER molecular chaperones.
The proteins associated with the unglycosylated Tg intermediate retained in the ER of tunicamycin-treated cells were then analyzed by coimmunoprecipitation with anti-Tg antibodies in the absence and presence of a thiol-cleavable chemical crosslinking agent (DSP). For initial identification of likely proteins associated with Tg in control and tunicamycin-treated cells, coimmunoprecipitation of Tg from metabolically labeled cells was performed. For subsequent positive identification of the coimmunoprecipitated polypeptides, Tg was immunoprecipitated from unlabeled cells, and the polypeptide bands were analyzed by Western immunoblotting. Fig. 3 and 4, respectively, show coimmunoprecipitations from metabolically la-beled FRTL-5 cells and Western immunoblot analysis of the immunoprecipitates. In the absence of tunicamycin, the association of BiP, grp94, and ERp72 with Tg was seen after chemical cross-linking (Fig. 3, lanes 2, 5, and 9; Fig. 4, lanes 7, 11,  and 15), as previously reported (6). It is important to note that, apart from these bands, no others were consistently immunoprecipitated with anti-Tg antiserum, suggesting that the associations we observed are highly specific and authentic. Furthermore, to eliminate the possibility that coprecipitation of BiP with anti-Tg results from binding of BiP (otherwise known as immunoglobulin binding protein) to immunoglobulins added during immunoprecipitation, we have conducted immunoprecipitation of Tg in the thyroid cells that lack the ability to synthesize and secrete Tg (Fig. 6, lanes 1-4) but contain normal amounts of BiP. In these cells, neither Tg nor any other molecular chaperones could be immunoprecipitated with anti-Tg antibodies. In Tg-containing cells, following tunicamycin treatment, BiP, grp94, and ERp72 could be seen even without cross-linkers (Fig. 3, lanes 3, 8, and 12; Fig. 4, lanes 8, 12, and  16), although the associations were further enhanced by crosslinking (Fig. 3, lanes 4, 7, and 11; Fig. 4, lanes 9, 13, and 17). In addition, a band of approximately 150 kDa was coimmunoprecipitated with Tg in tunicamycin-treated cells in the absence and presence of cross-linkers (Fig. 3, lanes 3 and 4). This band was identified as the putative ER chaperone grp170 (Fig. 4B), a member of the glucose-regulated stress protein family that is thought to participate in folding and/or assembly of immunoglobulin chains (35). Thus, the large aggregates of unglycosylated Tg in the ER appeared to stably complex with four ER chaperones. No other polypeptides were present in Tg immunoprecipitates from 35 S-labeled cell lysates (Fig. 3). The same complex containing Tg, BiP, grp94, ERp72, and grp170 could also be coimmunoprecipitated with anti-BiP antibodies in tunicamycin-treated cells in the absence of cross-linkers (Fig. 3,   FIG. 3. Associations of Tg with the molecular chaperones of the ER in stressed and unstressed thyroid cells. Uncross-linked (lanes 1, 3, 6, 8, 10, and 12) and cross-linked (lanes 2, 4, 5, 7, 9, and 11) lysates from [ 35 S]methionine-labeled, untreated (lanes 1, 2, 5, 6, 9, and 10) or tunicamycin-treated (lanes 3, 4, 7, 8, 11, and 12) FRTL-5 cells were subjected to immunoprecipitation with anti- Tg (lanes 1-4), anti-BiP (lanes 5-8), or anti-ERp72 (lanes 9 -12) antisera. Note that lanes 5-12 containing cross-linked samples precede lanes containing uncrosslinked samples. Arrows denote positions of Tg, p200, grp170, grp94, BiP, and ERp72; the asterisk denotes an unidentified polypeptide band of approximately 180 kDa that was immunoprecipitated with anti-ERp72 antiserum but not with the anti-Tg or anti-BiP antisera.   Fig. 4, lane 12), though once again the associations were enhanced by cross-linking (Fig. 3, lane 7; Fig. 4, lane 13). Nevertheless, a stable complex of these four ER proteins with Tg appears to develop when normal Tg folding and/or assembly is perturbed by inhibition of its N-linked glycosylation. In addition, in BiP and ERp72 immunoprecipitates, a 200-kDa band was consistently observed. This band was subsequently identified (see below). The identity of an approximately 180-kDa polypeptide coimmunoprecipitated with ERp72 only has not yet been established (Fig. 3, lanes 9 -12). It is possible that this 180-kDa polypeptide represents a secretory protein that specifically associates with ERp72 in the ER.
Some interactions between Tg and molecular chaperones, already increased after tunicamycin treatment, were further enhanced by cross-linking more than others. For example, treatment of tunicamycin-treated cells with cross-linker resulted in a marked further increase in amount of grp94 coimmunoprecipitated with Tg and BiP (Fig. 4C, lanes 8 , 9, 12, 13) or in the amount of BiP coprecipitated with ERp72 (Fig. 4D,  lanes 16, 17). The amount of grp170 co-precipitated with either Tg or BiP was unaffected by cross-linking (Fig. 4B, lanes 8 , 9, 12, 13). Thus, the presumed misfolding of the Tg caused by misglycosylation affects the stability of interactions of the protein with molecular chaperones to different degrees.
grp170 is a glycoprotein (35), and after tunicamycin treatment it was indeed detected intracellularly as a lower molecular weight (unglycosylated) form (Fig. 4B, lanes 4, 5 versus  lanes 2, 3). We could not detect an association of grp170 with either Tg or BiP in unstressed cells in the absence of crosslinking (Fig. 4B, lanes 6 and 10). However, treatment of unstressed cells with the cross-linking agent caused a small amount of the glycosylated form of grp170 to coimmunoprecipitate with both Tg and BiP (Fig. 4B, lanes 7 and 11). After treatment with tunicamycin, a substantial amount of grp170 coimmunoprecipitated with Tg and BiP (Fig. 4B, lanes 8 and  12), and this was not further increased by cross-linking (Fig.  4B, lanes 9 and 13). However, by Western immunoblot analysis, no grp170 was detected in any of the ERp72 immunoprecipitates (Fig. 4B, lanes 14 -17), although a trace amount of polypeptide of identical molecular weight to grp170 was present in ERp72 immunoprecipitates from 35 S-labeled, tunicamycin-treated cells (Fig. 3, lanes 11 and 12). Since grp170 and ERp72 were present in both anti-Tg and anti-BiP immunoprecipitates from tunicamycin-treated cells, failure to confirm by Western immunoblotting an association between ERp72 and grp170 suggested by autoradiographic analysis was puzzling. It is possible, however, that the amounts of these two ER proteins associating with each other is extremely low, and Western blotting is less sensitive than autoradiography in detecting these associations.
A long-term treatment with tunicamycin (such as used in the above described experiments) has been shown to induce the mRNA levels for ER molecular chaperones (including BiP, grp94, and ERp72) in various experimental systems (13,36) as well as in rat thyroid epithelial cells (18). It is currently believed that the accumulation of misfolded and/or misassembled secretory proteins in the lumen of the ER serves as the signal and/or the trigger for the induction (36 -38). Stable long-term associations of the ER molecular chaperones with their misfolded/misassembled protein substrates may also be a determining factor in signaling the induction of mRNA levels for these chaperones. We have therefore analyzed the associations of Tg with ER chaperones following treatment with other agents that induce their mRNA levels (presumably through causing accumulation of misfolded secretory proteins in the ER lumen): a Ca 2ϩ ionophore A23187 and an inhibitor of the ER Ca 2ϩ -ATPase thapsigargin (39,40). Both agents result in depletion of Ca 2ϩ from the ER (41,42) and are not known to induce aggregation in the ER lumen; in fact, it has been argued that calcium depletion may have the opposite effect (43). Following treatment of FRTL-5 cells with A23187, thapsigargin, or tunicamycin for 16 h, Tg was immunoprecipitated from uncross-linked cell lysates, and the immunoprecipitates were analyzed by Western immunoblotting with antisera against Tg (Fig. 5A) or ER molecular chaperones (Fig. 5, B-D). Similar to tunicamycin, both Ca 2ϩ -depleting agents resulted in coimmunoprecipitation of BiP, grp94, and ERp72 with Tg (Fig. 5). grp170 was not coimmunoprecipitated with Tg in cells treated with either A23187 or thapsigargin (data not shown). Stable association of grp170 and Tg thus appears to be a specific result of tunicamycin treatment, although both treatments (inhibition of N-linked glycosylation and Ca 2ϩ depletion) have been shown to induce grp170 (35).
To determine if the assembly of a complex involving the four chaperones was specific to Tg or represented a more general phenomenon, a group of adenovirus-transformed rat thyroid cells, PCE1As (27), were studied. These cells have lost their ability to synthesize and secrete Tg and no longer depend on thyroid-stimulating hormone for growth following transfection with the virus (27). As expected for these cells, no Tg was detected by immunoprecipitation (Fig. 6, lanes 1-4) or Western immunoblotting (data not shown). However, in [ 35 S]methionine-labeled immunoprecipitates, the same complex of BiP, grp94, ERp72, and grp170 was observed in association with other bands (Fig. 6, lanes 8 and 12). An unidentified 180-kDa band was only seen in ERp72 immunoprecipitates (Fig. 6, lanes  9 -12), but the most consistently observed band seen in both BiP and ERp72 immunoprecipitates migrated at approximately 200 kDa (Fig. 6, lanes 5-12). This band had also been visualized in BiP and ERp72 immunoprecipitates of [ 35 S]methionine-labeled FRTL-5 cells that secrete Tg (Fig. 3, lanes  [5][6][7][8][9][10][11][12]. The p200 might have represented a resident ER protein that functions as a chaperone like BiP or ERp72. Alternatively, p200 could be another secretory product of thyroid cells that utilizes the same ER molecular chaperones in its maturation. The latter seemed a more plausible hypothesis since we could not detect an association of p200 with Tg (data not shown). Like Tg, the 200-kDa protein only entered standard polyacrylamide resolving gels under reducing conditions, suggesting that it too was a highly disulfide-bonded protein that possibly oligomerized under non-reducing conditions (Fig. 7, lanes 1 versus 2). One such protein made by thyroid and other cells is a member of the thrombospondin protein family, thrombospondin-1 (TSP-1), a trimeric, multiply disulfide-bonded glycoprotein (22,44). Native TSP-1 is a homotrimeric protein of 450 kDa, and its monomeric size of approximately 190 -200 kDa corresponds to the p200 polypeptide band present on reducing gels. By coimmunoprecipitation and Western blotting, the 200-kDa band was subsequently indeed identified as TSP-1 (Fig. 7, panels B and C).
To summarize, the existence in thyroid cells of complexes containing these four ER chaperones was not dependent on Tg. In searching for other proteins that might associate with the same four chaperones, it was very striking to us that the only major protein detected turned to have many characteristics similar to Tg, for TSP-1 is also a large oligomeric secreted glycoprotein. Thus, TSP-1 would be expected to require reactions very similar to Tg for proper folding and/or assembly in the ER, consistent with our finding that the same set of ER chaperones complexes with both secretory proteins. Based on these results, we propose that BiP, grp94, ERp72, and possibly grp170 participate in complex(es) involved in the maturation of these two secretory glycoproteins. On the basis of similarities between Tg and TSP we hypothesize that, in general, large oligomeric glycoproteins may follow the same pattern of interaction with these four molecular chaperones as they pass through the ER. Two types of interactions are possible: concurrent and sequential. In the case of thyroglobulin, one would expect a heterogeneous population present in the ER lumen at all times, each subpopulation consisting of a certain intermediate in post-translational processing of the protein. Each subpopulation would then associate with a chaperone or a subset of the four chaperones. In support of this hypothesis is the fact that coimmunoprecipitations with different antibodies yielded different sets of proteins. For example, while all four chaper-ones coprecipitated with Tg, grp170 did not coprecipitate with ERp72. ERp72 and grp170 might therefore bind different intermediates in Tg maturation that are present at different time points along the maturation pathway, while binding of Tg to BiP and to grp94 occupies a broader time period and overlaps with binding to both ERp72 and grp170. Sequential action of chaperones like BiP and grp94 has been suggested in maturation of immunoglobulin light chains (17), and sequential action of calnexin and BiP has been suggested in maturation of thyroglobulin (23). Our findings, and the notion that chaperones recognize distinct features within polypeptides, are also consistent with a model of interaction in which various chaperones or sets of chaperones successively complex with various intermediates in maturation of oligomeric glycoproteins.