Reduction of BiP levels decreases heterologous protein secretion in Saccharomyces cerevisiae.

Increased levels of the endoplasmic reticulum-resident protein folding chaperone BiP would be expected to either increase protein secretory capacity by improved solubilization of folding precursors or decrease secretory capacity by binding and retaining misfolded proteins. To address this question, the relationship between BiP levels and heterologous secretion in yeast was determined. A yeast strain was constructed in which BiP expression is tunable from 5 to 250% of wild-type levels, and this strain was used to explore the effect of varying BiP level on overall secretion of three heterologous proteins: human granulocyte colony-stimulating factor, Schizosaccharomyces pombe acid phosphatase, and bovine pancreatic trypsin inhibitor. For all three proteins examined, reduction in BiP expression below wild-type level diminished overall secretion, whereas 5-fold BiP overexpression from a constitutive glycolytic promoter did not substantially increase or decrease secretion titers. These results are consistent with a positive role for BiP in promoting membrane translocation and solubilization of folding precursors but are inconsistent with a negative role in proofreading and improper retention of heterologous secreted proteins.

Increased levels of the endoplasmic reticulum-resident protein folding chaperone BiP would be expected to either increase protein secretory capacity by improved solubilization of folding precursors or decrease secretory capacity by binding and retaining misfolded proteins. To address this question, the relationship between BiP levels and heterologous secretion in yeast was determined. A yeast strain was constructed in which BiP expression is tunable from 5 to 250% of wildtype levels, and this strain was used to explore the effect of varying BiP level on overall secretion of three heterologous proteins: human granulocyte colony-stimulating factor, Schizosaccharomyces pombe acid phosphatase, and bovine pancreatic trypsin inhibitor. For all three proteins examined, reduction in BiP expression below wild-type level diminished overall secretion, whereas 5-fold BiP overexpression from a constitutive glycolytic promoter did not substantially increase or decrease secretion titers. These results are consistent with a positive role for BiP in promoting membrane translocation and solubilization of folding precursors but are inconsistent with a negative role in proofreading and improper retention of heterologous secreted proteins.
The native route of synthesis of most proteins of therapeutic interest is via secretion; growth factors, hormones, thrombolytic and clotting factors, and antibodies are all secreted proteins. The eucaryotic secretory pathway mediates the folding, assembly, glycosylation, proteolytic processing, and conformational proofreading of secreted polypeptides. Thus, eucaryotic secretion expression systems represent an efficient means of producing proteins with high fidelity to the native form. Unfortunately, the specific productivity of mammalian, insect, and yeast secretion systems is generally 1-2 orders of magnitude lower than that of bacterial systems (Hodgson, 1993). The rate-limiting step in eucaryotic protein secretion is generally transport from the endoplasmic reticulum to the Golgi apparatus (Lodish et al., 1983;Shuster, 1991). Unfolded proteins are prevented from leaving the ER 1 by a proofreading mechanism (reviewed by Helenius (1994)), so it is generally the rate and efficiency of protein folding and assembly in the ER that determines the overall rate of transport through the secretory pathway. The lumen of the ER is an environment adapted to the assistance of folding by the presence of a high concentration of protein folding chaperones (BiP, GRP94, and calnexin) and foldases (protein disulfide isomerase, FKBP, ERp72, and cyclophilin) (Helenius et al., 1992;Gething and Sambrook, 1992). Previous attempts to enhance secretory capacity have focused on manipulating levels of lumenal chaperones and foldases (Dorner et al., 1988(Dorner et al., , 1992Robinson et al., 1994;Schultz et al., 1994;Hsu et al., 1994).
Perhaps the most thoroughly studied ER-resident chaperone is BiP (or heavy chain binding protein). Initially discovered in a complex with unassembled IgG heavy chains (Haas and Wabl, 1984), BiP was soon identified as an Hsp70 chaperone that is present in the endoplasmic reticulum of all eucaryotes. The KAR2 gene for the S. cerevisiae BiP homolog was identified separately by homology to mouse BiP (Normington et al., 1989) and by functional and sequence similarity of BiP to a karyogamy gene (Rose et al., 1989). BiP is induced by stressors that cause the accumulation of unfolded proteins in the lumen of the ER and is often present in stable complexes with misfolded or unassembled proteins while interacting only transiently with efficiently folded proteins (Blount and Merlie, 1991;Bole et al., 1986;Dorner et al., 1987;Gething et al., 1986;Hendershot, 1990;Ng et al., 1992). BiP seems to be most involved in the early stages of protein conformational maturation in the ER (Hammond and Helenius, 1994;Melnick et al., 1994;Kim and Arvan, 1995;Simons, et al., 1995) and has been demonstrated to play a key role in membrane translocation (Vogel et al., 1990;Sanders et al., 1992;Brodsky and Schekman, 1993;Panzner et al., 1995).
It might be expected that increasing BiP levels would increase the protein folding capacity of the ER, because chaperones act essentially as detergents to solubilize folding intermediates and circumvent kinetic trapping of aggregates (reviewed by Hartl et al. (1994)). BiP expression is induced by increased protein flux into the ER, consistent with this hypothesis (Dorner et al., 1989;Tokunaga et al., 1992;Robinson and Wittrup, 1995;Watowich et al., 1991). Furthermore, stressors that cause the accumulation of unfolded protein in the ER also induce BiP expression (Kozutsumi et al., 1988;Rose et al., 1989;Normington et al., 1989;Kohno et al., 1993). However, the capability to bind to unfolded polypeptides might instead be interpreted as serving to retain unfolded proteins in the ER, such that BiP acts as an element of the proofreading apparatus (Dorner and Kaufman, 1994).
Dorner, Kaufman, and colleagues have provided evidence that the chaperone BiP can serve a proofreading role in heterologous secretion. Reduction of BiP levels by antisense transcript expression increases secretion of a mutant tissue plasminogen activator lacking glycosylation sites (Dorner et al., 1988), and overexpression of BiP decreases secretion of factor VIII and von Willebrand factor (Dorner et al., 1992). These effects are protein-specific; a negative function for BiP in secretion appears to correlate with the detection of stable BiPprotein complexes (Dorner et al., 1992) and a high ATP requirement for secretion (Dorner et al., 1990). Dorner and Kaufman conclude that BiP does not serve a positive role in assisting folding and secretion but instead serves to bind and retain misfolded proteins (Dorner and Kaufman, 1994).
We were interested in determining the role of BiP in heterologous protein secretion in Saccharomyces cerevisiae. Towards this end, we constructed a yeast strain wherein the levels of BiP protein are tunable by a copper-inducible promoter. We examined the secretion of three proteins in this strain: human granulocyte colony-stimulating factor (GCSF), Schizosaccharomyces pombe acid phosphatase (PHO), and bovine pancreatic trypsin inhibitor (BPTI). We find that reduction of BiP below normal levels progressively inhibits secretion of all three proteins, whereas 5-fold overexpression of BiP from a strong glycolytic promoter essentially does not affect secretion. These data are consistent with a model wherein BiP functions to promote folding and secretion of proteins by competing with unfavorable aggregation rather than act predominantly in a negative proofreading capacity (Robinson and Wittrup, 1993). We discuss our results in light of a growing body of evidence that BiP functions primarily at the earliest stages of protein folding and secretion and attempt to reconcile the apparent contradiction between Dorner and Kaufman's data and our own in terms of structural differences among the proteins studied.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-The strain BJ5464 (␣ ura3-52 trp1 leu2⌬1 his3⌬200 pep4::HIS3 prb1⌬1.6R can1 GAL) was obtained from the Yeast Genetic Stock Center (Berkeley, CA) and used as the basis of new strain construction. The integrating vector pMR2281, containing a BiP gene with all coding sequence deleted, was partially digested by the restriction endonuclease AflII and transformed into BJ5464, yielding a strain with one wild-type and one nonfunctional BiP gene in tandem with an intervening URA3 marker. This strain was transformed with pGalKar2-LEU, a galactose-inducible BiP expression plasmid , and then grown on galactose and 5-fluoroorotic acid to select for cells in which homologous recombination has "looped out" the URA3 marker (Rothstein, 1991). Colonies were selected that exhibited galactose-dependent growth on ura ϩ medium, and this strain is called JBY001. BiP levels fall to 1% of wild-type within four generations when JBY001 is transferred from galactose to glucose (data not shown).
The plasmid pMR1341 (gift of M. Rose) was used to construct the plasmid pCUPKAR2, in which Kar2 is under the control of the inducible promoter CUP1, derived from the yeast metallothionein gene (Butt and Ecker, 1987). The CUP1 promoter was amplified from the plasmid pYR-CUP (gift of J. Sambrook) by polymerase chain reaction. Unique restriction sites were incorporated to facilitate insertion into the pMR1341 vector. The forward primer 5Ј-CTCGACGTCGGATCCCAT-TACCG-3Ј introduced an AatII site to the 5Ј end of the CUP1 promoter, and the reverse primer 5Ј-GGGTCGACGTACAGTTTGTTTTTC-3Ј introduced a SalI site to the 3Ј end of the CUP1 promoter. The cloned PCR product was subcloned in place of the SalI-AatII GAL1,10 promoter of pMR1341, to create pCUPKAR2.
The plasmid pMR1341 was also used in the construction of pGAPDH-KAR2, in which Kar2 is under control of the strong constitutive promoter glyceraldehyde phosphate dehydrogenase (GAPDH). To facilitate cloning, an internal SalI site was first removed from the GAPDH promoter in the vector pUC119␣G6 (GCSF) (gift of S. Elliot, Amgen) by a SalI digest followed by Klenow treatment. The product was purified using DNA Clean-Up (Promega), and blunt ends were religated with high concentration DNA Ligase (Boehringer Mannheim). The ligation mixture was transformed into HB101 (BRL), and product plasmid was screened by an EcoRI-BamHI digest. The promoter was then amplified by "hot start" polymerase chain reaction (Chou et al., 1992), and unique restriction sites were introduced to permit insertion into the pMR1341 vector. The forward primer 5Ј-GGCGGACGTCAAGGTCGAGTTTAT-CAT-3Ј introduced an AatII site to the 5Ј end of the GAPDH promoter, and the reverse primer 5Ј-CTCGTCGACCGTCGAAACTAAGT-3Ј introduced a SalI site to the 3Ј end. The initial three cycles were run at an annealing temperature of 37°C because of the AT-rich primers, followed by thirty cycles at an annealing temperature of 55°C to improve specificity for the full-length product. The product was purified using DNA Clean-Up (Promega), digested with AatII and SalI and subcloned in place of the SalI-AatII GAL1,10 promoter of pMR1341. The plasmid pGAPDH-KAR2 was screened on LB-AMP (50 g/ml) and identified by restriction digests with PstI and EcoRI.
A low copy expression plasmid for human granulocyte colony-stimulating factor (pCEN-GCSF) was constructed by subcloning the BamHI ϩ EcoRI fragment of the plasmid pUC119␣G6 (gift of S. Elliott, Amgen) containing the GAPDH promoter, ␣-factor signal sequence and leader, and human GCSF coding sequence, and ␣-factor transcriptional terminator into the polylinker site in the pRS314 centromere shuttle vector (Sikorski and Hieter, 1989) bearing the TRP1 selectable marker.
A low copy expression plasmid for S. pombe acid phosphatase (pCEN-PHO) was constructed by inserting a partial BamHI fragment from the plasmid pYE␣FPHO (gift of S. Elliot, Amgen) containing the ␣-factor promoter and leader and S. pombe acid phosphatase coding sequence into the polylinker BamHI site of the vector pNN342 (Elledge and Davis 1988) bearing the TRP1 selectable marker. The BPTI expression plasmid was constructed as described previously .
Single transformant colonies of JBY100 and a control strain JBY200 (see Table I) were grown to saturation in 5 ml of synthetic galactose medium supplemented with 2 ϫ SCAA. 250-ml baffled Erlenmeyer flasks containing 25 ml of synthetic glucose ϩ 2 ϫ SCAA medium at indicated concentrations of copper sulfate were inoculated with JBY100 or JBY200 culture to an A 600 of 0.1. In order to maintain the cultures in exponential growth phase, all cultures were serially diluted to A 600 ϭ 0.1 when the A 600 reached approximately 1.0. Samples were taken after 22 h (approximately 6 -7 doublings), at which point intracellular levels of BiP and the levels of secreted protein were measured.
BiP Immunoassay-Protein extract samples were prepared from whole cell cultures using a previously described adaptation (Robinson et al., 1994) of a trichloroacetic acid lysis procedure. Aliquots were stored at Ϫ70°C and were subjected to no more than one freeze-thaw cycle. For Western blots, equivalent (A 600 ) (ml) of extract were separated by SDS-polyacrylamide gel electrophoresis. Protein extract samples were boiled for 10 min in buffer containing 1.67% SDS, 1.67 mM dithiothreitol, and 1 mg/ml bovine serum albumin to prevent loss of cellular protein samples by adsorption. Following SDS-polyacrylamide gel electrophoresis, proteins were electrophoretically transferred to 0.2 M nitrocellulose membrane. Primary anti-Kar2 IgG (gift of M. Rose) was used at 1:10,000 dilution, followed by goat anti-rabbit secondary antibody conjugated to horseradish peroxidase at 1:2000 dilution (Sigma). Detection of the antigen-antibody complex was performed with enhanced chemiluminescence (ECL, Amersham Corp.), and images were recorded on film (Hybond ECL, Amersham Corp.). For quantitation, multiple exposures of each blot were scanned (Ofoto 1.0, Apple 8 bit B&W Scanner), and values in the linear range of the film were analyzed (Image 1.44, NIH). In order to assess the variability of the quantitative immunoassay, duplicate cultures of BJ5464 were grown, samples of the cell suspension were lysed in triplicate, and SDS-polyacrylamide gel electrophoresis followed by Western detection probing for yeast BiP was performed. The variation among the samples was less than 8%. The linearity of this assay has been demonstrated previously .
Acid Phosphatase Assay-At the desired time point, the culture A 600 was recorded, and 150 l of cell culture was assayed in duplicate. 600 l of 2 mg/ml para-nitrophenyl phosphate (Sigma) in 50 mM sodium acetate buffer, pH 4, was preincubated at 30°C for 10 min prior to the addition of cell culture. The reaction was incubated at 30°C for 10 min and then stopped by transferring the assay tube to wet ice (0°C) and by the addition of 150 l of 25% trichloroacetic acid. Saturated sodium carbonate (700 l) was added to bring the solution to alkaline pH (Robinson et al., 1994). The reaction mixtures were transferred to cuvettes containing 1.4 ml of saturated sodium carbonate and the absorbance of the yellow-colored product measured at 435 nm. Volumetric acid phosphatase activity was calculated as A 435 /(A 600 of original culture). Control BJ5464 cells not expressing S. pombe acid phosphatase give negligible activity by this assay.
GCSF Immunoassay-Supernatant samples were obtained by centrifugation of 3 ml of cell culture for 5 min at 5000 ϫ g followed by the addition of dithiothreitol to 40 M and SDS to 0.001% and storage in 1-ml aliquots at Ϫ70°C. The amount of GCSF present was measured for dilutions in the linear range by enzyme-linked immunosorbent assay (Quantikine GCSF kit, R&D Systems) following the supplied protocol.
BPTI Activity Assay-To assay for BPTI activity, the culture was centrifuged for 5 min at 10,000 ϫ g, and 0.5 ml of supernatant was added to 2.5 ml of buffer (15 mM CaCl 2 , 0.2 M triethanolamine, pH 7.8) at 30°C in a thermostatted cuvette holder. 0 -20 g of bovine trypsin (L-(tosylamido 2-phenyl) ethyl chloromethyl ketone-treated, Worthington Biochemical) was then added, and the mixture was incubated for 30 min to allow trypsin-BPTI binding. Reaction was initiated by the addition of 150 l of 32 g/l of N-␣-benzoyl-arginine-p-nitroanilide (Sigma), a synthetic trypsin substrate, and the increase in absorbance at 405 nm was recorded for 4 min. Because BPTI is a competitive inhibitor of trypsin with essentially irreversible binding BPTI activity is equal to the difference between the added and detected trypsin activities.
KAR2 mRNA Quantification-To quantify KAR2 mRNA levels, ribonuclease protection assay (Ambion) was used. A 300-base pair EcoRI-HindIII fragment of the full KAR2 gene was transcribed by T7 polymerase as a probe for KAR2 mRNA. As a control for varying RNA extraction efficiency, a probe for the DPM1 dolichol phosphate mannose synthase was constructed as a 550-base pair XhoI-HindIII fragment. MAXIscript kit (Ambion) and [ 32 P]UTP were utilized to synthesize the probes. Total RNA was isolated from each flask of cells grown in a CupKar2 dilution experiment. As per ribonuclease protection assay II kit (Ambion), radioactive probes were allowed to hybridize with Kar2 and DPM1 mRNA before RNase T1 was added to digest unhybridized probe. Samples were separated on 5% acrylamide, 8 M urea gels and exposed to a PhosphorImager screen. Several exposures were scanned to ensure unsaturated signals, and ImageQuant (Molecular Dynamics) was used to quantify both Kar2 and DPM1 mRNA in each sample.

Construction of a Yeast Strain with Variable BiP Levels-A
strain of S. cerevisiae was constructed in which BiP expression is controlled via a CUP1-inducible promoter across a broad range in order to examine the relationship between heterologous protein secretion and BiP levels. A gratuitous inducer such as copper is preferable to minimize changes in concentration due to metabolism of the inducer, so inducible promoters controlled by carbon source (GAL and ADH2) were deemed unsuitable for these studies. The CUP1 promoter is induced by the addition of Cu 2 SO 4 to the growth medium, and expression monotonically increases with Cu 2 SO 4 concentration across a 25-50 fold dynamic range with maximal transcription equivalent to a strong constitutive glycolytic promoter (Butt and Ecker, 1987).
Because BiP expression is induced by secretion of heterologous proteins (Dorner et al., 1989;Tokunaga et al., 1992;Robinson and Wittrup, 1995), it is necessary to delete the chromosomal KAR2 gene coding for yeast BiP in order to remove native BiP regulatory effects and explore the consequences of low BiP levels in the context of heterologous protein expression. Because KAR2 is an essential gene in yeast, BiP expression in the deletion strain was provided from a galactose-inducible plasmid-borne KAR2 gene, which expresses 10 -12-fold excess of BiP protein when grown on galactose . A CUP1-KAR2 expression cassette on plasmid pCUPKAR2 was transformed into this strain to provide copper controlled BiP levels on glucose by varying the concentration of Cu 2 SO 4 in the growth medium. This strain is subsequently referred to as JBY100 (Table I).
A stationary phase culture of galactose-grown JBY100 ϩ pNN342 was split to inoculate nine independent cultures in glucose medium containing 0 -1.5 mM Cu 2 SO 4 (pNN342 is a control CEN TRP1 plasmid lacking a heterologous protein expression cassette). Cultures were serially transferred to maintain exponential growth. Growth rates were similar for 30 h following transfer to glucose at Cu 2 SO 4 concentrations of 30 M and above, whereas growth was slightly slowed at 0 and 15 M Cu 2 SO 4 (data not shown.) In cultures of JBY001 (lacking the pCUPKAR2 plasmid), BiP levels drop to less than 1% of normal 20 h following transfer to glucose medium (data not shown). Thus, BiP levels sampled 22 h following transfer to glucose medium are entirely due to expression from the copper-inducible KAR2 gene. Western blots with ␣-Kar2p antisera of cell extracts taken 22 h following transfer to glucose show a smooth increase in BiP levels with copper concentration (Fig. 1A). Densitometric quantitation of several exposures of this blot show that BiP levels are controlled across a 33-fold range by varying copper concentration (Fig. 1B). BiP protein levels are linearly related to KAR2 mRNA levels, as measured by ribonuclease protection assay (Fig. 1C). To control for artifactual physiological effects of copper on BiP expression, control cultures of JBY200 (BJ5464 ϩ pLac33)ϩ pNN342 were grown at 0, 0.5, 1.0, and 1.5 mM Cu 2 SO 4 . No effect on native KAR2 expression was observed at any of the Cu 2 SO 4 concentrations used, although cultures at 1.5 mM exhibited a slightly reduced growth rate (data not shown).
Effect of Varying BiP Levels on Heterologous Protein Secretion-Three heterologous proteins representing a range of structural characteristics were selected for these studies. Human GCSF is a 19-kDa monomer lacking N-linked glycosylation sites but possessing O-linked glycosylation sites. The PHO polypeptide is 49 kDa with ten N-linked glycosylation sites and forms a tetramer. BPTI is 6.5 kDa and not glycosylated. In addition to their differing structural features, these proteins also differ in their processing characteristics in the yeast secretory pathway. We have shown previously that optimal secretion is a protein-specific function of synthesis level . GCSF secretion is increased by multicopy expression relative to single copy; BPTI secretion titers are similar by multi or single copy expression; and PHO secretion is actually reduced by multicopy expression relative to single copy . Although some of these effects may be related to instability of the 2-m multicopy vector used, it is clear that the processing capacity of the yeast secretory pathway differs for these three proteins. Another distinction among the proteins is evident from their interaction with the foldase PDI. Namely, PHO secretion is increased 5-fold by protein disulfide isomerase overexpression, whereas GCSF secretion is unaffected (Robinson et al., 1994). Thus, the structural and processing characteristics of these three model proteins differ qualitatively. These proteins are also representative of the major classes of proteins suitable for expression in yeast: nonglycosylated cytokines, hormones, inhibitors, and fungal enzymes for industrial applications (Hodgson, 1993). Secretion of GCSF was examined at varying BiP levels. Cultures of JBY100 transformed with a low copy GCSF expression plasmid (pCEN-GCSF) were grown and sampled as described for the experiment represented in Fig. 1. GCSF secreted to the growth medium was measured by immunoassay, and BiP levels were quantitated by densitometry of Western blots as described under "Experimental Procedures." GCSF secretion was normalized to cell density to eliminate any effects due to differential cell growth. The addition of 0, 0.5, 1.0, or 1.5 mM Cu 2 SO 4 had no detectable effect on growth, BiP expression, or GCSF secretion in control JBY200 ϩ pCEN-GCSF cells. Maximal GCSF secretion is attained at BiP levels equivalent to those found in the JBY200 control strain (Fig. 2). Reduction of BiP below this level steadily reduces GCSF secretion to 10% of the maximal secretion at the lowest BiP level attained. A similar experiment was performed with S. pombe acid phosphatase with similar results: decreasing levels of BiP decreases specific secretory productivity (Fig. 3). BPTI secretion depends on BiP levels in a similar fashion (Fig. 4).
It should be noted that in cells overexpressing any of the three heterologous proteins the maximum level of CUP-driven BiP expression is 2.5-fold that of cells possessing the native chromosomal KAR2 gene, although in cells not secreting foreign proteins the expression level is similar from the two promoters. This is likely a result of a previously observed tendency for levels of ER chaperones and foldases to decrease with prolonged constitutive secretion of heterologous proteins (Robinson et al., 1994;Robinson and Wittrup, 1995).
BiP Overexpression Does Not Significantly Affect Heterologous Secretion-Because copper toxicity effects preclude BiP expression from the CUP1 promoter greater than 2-3-fold above wild-type levels (Figs. 1-4), we constructed an expression plasmid with the KAR2 gene transcribed by the strong constitutive GAPDH glycolytic promoter (Bitter et al., 1987). This construct drives BiP expression to levels 3-5-fold higher than wild type as measured by quantitative Western blots (data not shown). Secretion of the three model proteins was measured in cells overexpressing BiP, and the results are shown in Table II. Essentially no effect of BiP overexpression is seen for secretion of any of the proteins, either for single or multicopy protein secretion. DISCUSSION We have constructed a strain of S. cerevisiae in which levels of the ER chaperone BiP can be continuously varied over a 33-fold range by use of the CUP1 copper-inducible promoter. Using this strain, we have examined the relationship between BiP levels and secretion of three heterologous proteins: GCSF, PHO, and BPTI. For all three proteins, secretion increases with increasing BiP levels up to a saturating point beyond which further increases in BiP provide no significant benefit. Furthermore, maximal secretion is obtained at BiP levels equivalent to FIG. 1. A, BiP protein levels as a function of added Cu 2 SO 4 concentration in JBY100, an S. cerevisiae strain constructed for copper-regulated BiP expression (Table I). Cell extracts were performed on samples taken after 30 h of growth on glucose and analyzed by Western blot with ␣-BiP antibody (as described under "Experimental Procedures"). B, densitometric quantitation of a series of exposures of the Western blot shown in A (as described under "Experimental Procedures"). Error in quantitation is typically less than 10%. C, BiP mRNA transcribed from the CUP1 promoter is linearly related to BiP protein present in JBY100 (open circles). JBY100 (open circles) was grown as described in A. BiP/Kar2p protein was determined by Western blot analysis as described for A and B. Total mRNA was purified, and KAR2 and DPM1 mRNA levels were quantified by ribonuclease protection assay (Ambion) as described under "Experimental Procedures." The closed circle represents JBY200 possessing the wild-type chromosomal KAR2 gene.
FIG. 2. Human GCSF secretion decreases when BiP levels in JBY100 fall below those in control JBY200 cells. Cultures of JBY100 (open circles) and JBY200 (closed circle) were transformed with a low copy GCSF expression vector and then grown and sampled as described in the legend to Fig. 1. Secreted GCSF was measured in cellular supernatants by enzyme-linked immunosorbent assay assay (Quantikine GCSF kit, R&D Systems) as described under "Experimental Procedures" and normalized by cellular growth. BiP protein levels were analyzed by quantitative Western blot as described in the legend to Fig. 1 and under "Experimental Procedures." BiP and GCSF were measured in four independent control JBY200ϩpCEN-GCSF cultures (closed circle), and error bars encompass clonal variability as well as measurement precision. A duplicate experiment showed similar trends in GCSF secretion. wild-type levels from an unmodified strain.
Our results indicate a positive role for BiP in heterologous protein secretion, which differs from the findings of Dorner and co-workers (Dorner and Kaufman, 1994;Dorner et al., 1988Dorner et al., , 1992. It is unlikely that this discrepancy is due to fundamental differences between the role of BiP in yeast and in mammalian cells, because expression of mammalian BiP has been shown to complement the kar2-1 mutation in yeast (Normington et al., 1989). Further evidence for the similarity of the yeast and mammalian lumenal environment is that expression of mammalian protein disulfide isomerase complements depletion of yeast protein disulfide isomerase, another essential gene product involved in protein folding in the ER (Gunther et al., 1993).
A more plausible explanation for the discrepancy between our results and those of Dorner and co-workers lies in the nature of the proteins studied. In fact, BiP may have multiple functions in the ER: promoting translocation, solubilizing folding precursors, stabilizing unassembled subunits, direct proofreading, retention, degradation, and perhaps other as yet un-explored functions. The particular facet of BiP function observed could depend on the structural characteristics of the substrate polypeptide studied. The proteins for which BiP has been shown to impede secretion in Chinese hamster ovary cells are either mutant (tissue plasminogen activator with three N-linked glycosylation sites deleted) or very large proteins (factor VIII is 250 kDa; von Willebrand Factor is 300 kDa) (Dorner et al., 1988(Dorner et al., , 1992. Furthermore, the processing and assembly of fVIII and vWf are complex; in mammalian cells, fVIII and vWf form a stabilizing complex together, and vWf is packaged in secretory granules for induced secretion (Voorberg et al., 1993). By contrast, the proteins we have examined are considerably smaller (GCSF, 19 kDa; PHO, 49 kDa; BPTI, 6.5 kDa) and perhaps more typical of a majority of proteins of pharmaceutical interest. It should also be noted that BiP overexpression in Chinese hamster ovary cells increases secretion of macrophage colony-stimulating factor (46 kDa) an average of 25-fold, although this effect was attributed to unexpected increases in macrophage colony-stimulating factor mRNA (Dorner et al., 1992).
Convergent lines of evidence indicate that BiP acts early in the secretory process, as the nascent polypeptide chain is translocated across the ER membrane into the lumen. Depletion of BiP from yeast or the shift to nonpermissive temperatures of certain temperature sensitive alleles of KAR2 result in a translocation block (Vogel et al., 1990), and BiP can be cross-linked to trapped translocation intermediates in contact with the Sec61p translocon component (Sanders et al., 1992). Genetic interactions between SEC63 and KAR2 imply an interaction similar to that found between the cytoplasmic bacterial chaperones DnaJ and DnaK (Scidmore et al., 1993), and the SEC63 gene product, possessing homology to DnaJ in a lumenal domain, is required for translocation in yeast (Feldheim et al., 1992). Furthermore, a Sec63p-BiP complex is a necessary component in reconstituted translocation reactions (Brodsky and Schekman, 1993;Panzner et al., 1995). It appears that the mitochondrial hsp70 plays a similar role in mitochondrial membrane translocation (Ungermann et al., 1994). Secreted proteins have been shown to undergo sequential interaction first with BiP and then with other lumenal chaperones. IgG polypeptides form transient complexes first with BiP and then with GRP94 (Melnick et al., 1994), whereas VSV G protein has been shown to bind first to BiP and then calnexin during folding (Hammond and Helenius, 1994). Thyroglobulin also interacts sequentially with BiP and calnexin while it folds in the ER (Kim and Arvan, 1995).
The function of BiP in translocation might involve direct binding of the nascent chain or alternatively cycling or assembly of translocon components. A proposed mechanism for translocation termed the Brownian ratchet has been proposed, wherein the driving force for membrane translocation is random thermal motion biased toward the lumenal direction by events on the lumenal side that sterically prevent backwards motion out of the ER (Simon et al., 1992). Binding by BiP could serve in this role. Given that BiP binds to relatively short  polypeptide stretches (7 amino acids) (Flynn et al., 1991) in an extended conformation (Landry et al., 1992), BiP could bind nascent polypeptide chains as they emerge in the lumen. An alternative model is that BiP serves as a "translocation motor" that generates force to pull polypeptides into the ER (Glick, 1995). If the primary role for BiP is in membrane translocation, the lumenal requirement for BiP would be stoichiometrically related to the number of translocation sites in the ER membrane rather than the nature or quantity of proteins secreted. Our data are consistent with this interpretation, because the relationship between secretion and BiP levels is similar for all three proteins examined, despite substantial structural differences among these proteins.
The role of BiP in protein folding and secretion has previously been examined via a mathematical model that incorporates in vitro binding constants reported in the literature (Robinson and Wittrup, 1993) The predictions of this model are qualitatively similar to the results shown in Figs. 2-4. The model predicts that unless binding by BiP blocks folding to the extent that a polypeptide's tendency to aggregate is unchanged after release from BiP, secretion increases with increasing BiP to a plateau level. Thus, if a protein is less likely to aggregate after BiP release because some degree of folding took place while in complex with BiP, then decreasing BiP levels are predicted to decrease secretion efficiency, as our data indicate. The model also predicts the saturation effect observed, that increasing BiP levels beyond a certain point provides no incremental improvements in secretion.
Experimental evidence for a positive role for BiP in protein folding has also been obtained. Carboxypeptidase Y is bound by yeast BiP ATPase mutants at the restrictive temperature, essentially preventing the folding of carboxypeptidase Y and enhancing the tendency of the reduced form to aggregate (Simons et al., 1995). This result suggests that BiP plays an active role in carboxypeptidase Y folding and maturation.
From a practical perspective, it is unfortunate that manipulation of BiP levels does not improve heterologous protein secretion relative to unmodified strains. By contrast, overexpression of protein disulfide isomerase has been shown to result in substantial improvements in secretion of heavily disulfide bonded proteins (Schultz et al., 1994;Robinson et al., 1994). It is possible that limiting levels of Sec63p or an as yet unidentified ER GrpE homolog constrain the beneficial effects of BiP overexpression. However, it is also possible that chaperones such as GRP94 and calnexin that act later in the conformational maturation of proteins will play a more significant role in determining the capacity of the ER lumen to solubilize and assist in the folding of secreted proteins.