Protein Folding Stability Can Determine the Efficiency of Escape from Endoplasmic Reticulum Quality Control*

A fraction of each secreted protein is retained and degraded by the endoplasmic reticulum (ER) quality control apparatus that restricts export to correctly folded proteins. The intrinsic biophysical attributes that determine efficiency of escape from this proofreading process have been examined by expressing mutants of bovine pancreatic trypsin inhibitor (BPTI) in yeast. Secretion efficiency is strongly correlated with thermodynamic stability for a series of six point mutations of BPTI. No correlation of secretion efficiency with either oxidative folding or refolding rates in vitro is found; both the rapidly folded Y35L BPTI mutant and the slowly unfolded G36D BPTI mutant exhibit low secretion efficiency. Elimination of cysteines 14 and 38 by mutagenesis does not increase secretion efficiency, indicating that intramolecular thiol/disulfide rearrangements are not primarily responsible for retention and degradation of destabilized BPTI variants. Mutant yeast strains with diminished ER-associated degradation do not secrete BPTI more efficiently, indicating that retention and degradation are separable processes. These data support a model for ER quality control, wherein protein folding is functionally reversible and the relative rates of folding, unfolding, vesicular export, and retention determine secretion efficiency.

A fraction of each secreted protein is retained and degraded by the endoplasmic reticulum (ER) quality control apparatus that restricts export to correctly folded proteins. The intrinsic biophysical attributes that determine efficiency of escape from this proofreading process have been examined by expressing mutants of bovine pancreatic trypsin inhibitor (BPTI) in yeast. Secretion efficiency is strongly correlated with thermodynamic stability for a series of six point mutations of BPTI. No correlation of secretion efficiency with either oxidative folding or refolding rates in vitro is found; both the rapidly folded Y35L BPTI mutant and the slowly unfolded G36D BPTI mutant exhibit low secretion efficiency. Elimination of cysteines 14 and 38 by mutagenesis does not increase secretion efficiency, indicating that intramolecular thiol/disulfide rearrangements are not primarily responsible for retention and degradation of destabilized BPTI variants. Mutant yeast strains with diminished ER-associated degradation do not secrete BPTI more efficiently, indicating that retention and degradation are separable processes. These data support a model for ER quality control, wherein protein folding is functionally reversible and the relative rates of folding, unfolding, vesicular export, and retention determine secretion efficiency.
The quality control system of the endoplasmic reticulum allows export of only correctly folded and assembled proteins (1.) The components of the quality control system include the calnexin/glucosylation cycle (2)(3)(4), persistent binding to endoplasmic reticulum (ER) 1 -retained chaperones such as BiP or GRP94 (5)(6)(7), and exclusion of large aggregates from transport vesicles budding off from the ER (8 -9). In addition to complete retention of grossly misfolded proteins, some fraction of every secreted protein is retained and degraded by the quality control system. Secretion efficiency varies markedly among proteins and can be particularly low for overexpressed heterologous, secreted proteins (10 -15).
It is not obvious what intrinsic biophysical properties determine a given protein's secretion efficiency. Folding rate could predominate, if folded protein rapidly escapes from the proofreading apparatus by exiting the ER promptly upon achieving a folded conformation. Alternatively, thermodynamic folding stability could determine secretion efficiency by defining a rapidly equilibrated partition between conformations subject to two irreversible processes: degradation of the unfolded form versus export of the folded form. Yet another possibility is that the varying strength of currently ill-defined ER export signals affects ER residence time (16), hence exposing proteins to potential retention and degradation for varying lengths of time. In all likelihood, each of these properties influences secretion efficiency to some extent, with particular attributes accentuated for particular proteins.
It was recently shown that secretion efficiency in yeast for bovine pancreatic trypsin inhibitor mutants lacking disulfide bonds is correlated with thermodynamic stability but not with reported in vitro folding rates (17). Because removal of cysteines by site-directed mutagenesis qualitatively alters the oxidative folding pathway, the linkage between stability and secretion efficiency was tested further in the present study by examining six point mutants of BPTI that destabilize the surface 14 -38 disulfide (18). For these mutants, secretion efficiency is strongly correlated with the destabilization free energy of the 14 -38 disulfide but is not predicted by in vitro folding or unfolding rates. These results support a model for BPTI quality control in which a variable fraction of the protein is degraded while the folded form awaits export from the ER, implying that folding of even relatively stable proteins is functionally reversible in the ER.
Plasmids and Mutant Construction-The plasmid used to express EA-BPTI in Saccharomyces cerevisiae was constructed by subcloning the expression cassette from the vector pUC-G-BPTI (23) into the multicopy yeast shuttle vector YEplac112 (24) as an EcoRI-BamHI fragment. The resulting plasmid is referred to as YE112-GPD-BPTI. BPTI is expressed using the constitutive yeast glyceraldehyde-3-phosphate dehydrogenase promoter. A synthetic preproleader is used to target BPTI to the ER (23). The proregion contains a dibasic Lys-Arg site at its C terminus to allow cleavage by the Kex2p protease. This results in a * This work was funded by the National Institutes of Health Grant GM50673 and the National Science Foundation Grant BES 95-31407. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Glu-Ala extension at the N terminus of the secreted, mature BPTI. The construction of the C14/38A BPTI plasmid has been described previously (17). This plasmid is referred to as YE112-GPD-BPTI-C14/38A.
BPTI Trypsin Inhibition Assays-The level of BPTI secreted in the supernatant was measured as described (23). Briefly, the transformants were grown in selective growth medium (SD-CAA ϩ uracil) in 5-ml test tube cultures for 96 h to an A 600 of approximately 10.0. Saturated A 600 did not vary with expression of different mutants, and quintuplicate replicates varied less than 5% in A 600 . Increase in intensity of L-BAPA (Sigma), a synthetic trypsin substrate, at 405 nm was used to measure trypsin activity. A standard curve was obtained for a given level of trypsin (36 mg) and varying amounts of standard native BPTI (0 -10 mg) obtained from Worthington Biochemicals. Trypsin inhibition obtained with the supernatant was compared with the standard curve to estimate the level of BPTI present in the culture. Since BPTI-trypsin binding is effectively irreversible, BPTI activity is equal to the difference between the added and detected trypsin activities.
Pulse-Chase Radiolabeling Kinetic Experiments-For metabolic labeling of cells, an exponentially growing culture of BJ5464 in selective minimal media was used. The cells were transformed with either the expression plasmid plasmids for WT, P13S, or Y35L BPTI. The cell cultures were switched to minimal media lacking methionine 2 h before labeling. Cells were pulsed with 50 mCi of L-[ 35 S]methionine (Redivue, Amersham Pharmacia Biotech) per 1 A 600 units of cells for 1 min. For the chase, methionine was added to a final concentration of 10 mM. Intracellular samples were taken immediately following the addition of unlabeled methionine to determine initial BPTI synthesis. Supernatant samples were taken for times between 0 min and 2 h to record the appearance of secreted BPTI. During labeling and chase periods, the cells were incubated at 30°C with shaking at 14,000 rpm using an Eppendorf Thermomixer.
For the intracellular samples, one A 600 unit was pelleted by centrifugation at 16,000 ϫ g for 15 s. The supernatant was removed and 100 l of ice-cold stop buffer was rapidly mixed with the pellet (20 mM Tris-HCl (pH 8.0), 50 mM ammonium acetate, 500 mM iodoacetamide, 20 mM sodium azide, 0.45 mM cycloheximide, 5 mM EDTA). The sample was then frozen in a dry ice ethanol bath.
Cell extracts were prepared as described previously (26). Triplicate samples were thawed by addition of ice-cold cell extract buffer and vortexing (final concentrations: 10 mM Tris-HCl (pH 8.0), 25 mM ammonium acetate, 500 mM iodoacetamide, 20 mM sodium azide, 0.45 mM cycloheximide, 20 mM EDTA, 10% trichloroacetic acid, 4 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 1 mM phenylmethylsulfonyl fluoride, 1 mg/ml pepstatin A, 3 mg/ml leupeptin, 1 mg/ml antipain). This suspension was transferred to a tube containing 600 l of zirconium oxide beads (BioSpec Products, Bartlesville, OK). The cells were mechanically lysed by two 50-s cycles in a Bead-Beater (BioSpec Products) with cooling on wet ice between cycles. The supernatant was transferred to a new tube, and the beads were washed twice with 500 ml of wash buffer (10 mM Tris-HCl (pH 8.0), 25 mM ammonium acetate, 2 mM EDTA, 10% trichloroacetic acid). The washes were combined with the original supernatant, and the mixture was centrifuged for 5 min at 16,000 ϫ g. The pellet was resuspended in 40 l of trichloroacetic acid resuspension buffer (3% SDS, 100 mM Tris (pH 11), 10 mM dithiothreitol) and boiled for 5 min. After boiling, iodoacetamide was added to a final concentration of 50 mM, then 160 l of dilution buffer was added (60 mM Tris (pH 7.4), 190 mM NaCl, 6 mM EDTA, 1.25% Triton X-100). Ten l of Omnisorb (CalBiochem) was added, and the mixture was incubated with rotation for 10 min. This mixture was centrifuged at 16,000 ϫ g for 5 min, and the supernatant was prepared for electrophoresis. In control immunoprecipitations, the bands corresponding to intracellular and secreted BPTI were identified. Immunoprecipitation was not necessary for the quantitation presented here because no abundant cellular proteins were present in the molecular weight range of BPTI and pro-BPTI.
Supernatant samples were concentrated 10-fold in a Microcon-3 concentrator (Amicon) and prepared for electrophoresis. The sample was mixed with an equal volume of 2ϫ Tricine sample buffer (Novex, San Diego, CA) with dithiothreitol added to a final concentration of 65 mM and boiled for 5 min. This mixture was then loaded onto a 10 -20% gradient Tricine gel (Novex, San Diego, CA) and was electrophoresed at 125 V for 90 min using Tricine running buffer (10 mM Tris, 10 mM Tricine, 0.05% SDS, pH 8.3).
The gel was placed on 3 MM Whatman paper and covered with plastic wrap. The gel was then dried on a gel dryer (Hoeffer) for 2 h at 80°C and allowed to cool for 30 min under vacuum only. The dried gel was placed in a PhosphorImager exposure cassette (Molecular Dynamics) for 24 -72 h. The cassette was scanned and analyzed using a Molecular Dynamics PhosphorImager with ImageQuant software.

Yeast Secretion Efficiency of Six BPTI Point Mutants-A
series of point mutations that destabilize the 14 -38 disulfide of BPTI by 0.6 -3.7 kcal/mol (18) were selected for examination of secretion efficiency from yeast. These mutations were introduced into a yeast expression and secretion vector for BPTI (23), and BPTI secretion levels were determined (Table I.) The observed variation in secreted BPTI was not because of variable stability in the culture medium because secreted BPTI mutant activities were stable for at least one week in the growth culture under these experimental conditions, in the presence or absence of 1 mg/ml bovine serum albumin as a carrier protein.
To confirm that the observed variation in secreted BPTI levels is attributable to variable secretion efficiency, pulsechase radiolabeling experiments were performed for WT and two mutant forms of BPTI. Yeast cultures expressing wild-type BPTI, P13S BPTI, or Y35L BPTI were labeled with [ 35 S]me- This result indicates an absence of substantial variation in pre-ER processes such as plasmid stability, transcription, translation, or ER membrane translocation. These percentages are in quantitative agreement with earlier measurements for C14/38A, C30/51A, and C5/55A BPTI in the same expression system under the same experimental conditions (17). The rate of appearance of radiolabeled BPTI in the culture supernatant is shown in Fig. 1. Two h following the radiolabeling pulse, levels of secreted radiolabeled P13S and Y35L BPTI relative to wild-type BPTI parallel total secretion levels in Table I. Wild-type BPTI continues to accumulate in the medium 30 min following the radiolabeling pulse, whereas the less stable mutant P13S BPTI reaches its maximum secretion level within 30 min. Similar secretion kinetics were observed previously for wild-type BPTI relative to the destabilized mutants C14/38A and C30/51A BPTI (17), with the less stable mutants ceasing accumulation in the medium by 1 h of chase time, while WT BPTI continues to accumulate.
Secretion levels are strongly correlated with the reported extent of destabilization of the 14 -38 disulfide (Fig. 2). Secretion efficiency is not predicted by the in vitro reduction rates because G36D BPTI actually displays a slower in vitro reduction rate for the 14 -38 disulfide (18) yet is secreted less efficiently than WT BPTI, while each of the other five mutants have faster reduction rates than WT BPTI. Secretion efficiency is also not predicted by the in vitro oxidative folding rate because Y35L BPTI has been shown to fold an order of magnitude faster than wild-type BPTI in vitro (27). The correlation between thermodynamic stability and secretion shown in Fig.   2, together with the absence of such a correlation with in vitro oxidative folding or reduction rate, supports the hypothesis that a pseudo-steady-state equilibration between the folded and unfolded states determines the efficiency of escape from the ER quality control system for this series of BPTI mutants.
Secretion Efficiency Does Not Depend on Intramolecular Thiol/Disulfide Exchange with Cysteines 14 and 38 -Removal of cysteines 14 and 38 by mutagenesis indicates that folding instability rather than destabilization of the 14 -38 disulfide per se is responsible for reduced secretion efficiency. Following reduction of the 14 -38 disulfide, BPTI unfolding can proceed both by intramolecular attack on the buried 30 -51 or 5-55 disulfides by the newly liberated cysteine 14 or 38 thiols, or by intermolecular attack of the buried 30 -51 and 5-55 disulfides by a reducing agent (28). To determine whether an increased presence of free cysteine 14 and 38 thiols is responsible for the observed reduction in secretion efficiency of P13S and Y35L BPTI, we introduced the C14/38A mutations into these expression constructs. If the intramolecular rearrangement pathway were primarily responsible for misfolding, retention, and degradation of P13S and Y35L BPTI, then removal of cysteines 14 and 38 should produce a marked stabilization of these forms, leading to increased secretion efficiency. This is not the case, as indicated in Table I and Fig. 3. Loss of the 14 -38 disulfide reduces secretion of wild-type, P13S, and Y35L BPTI by similar proportions. These results indicate that destabilization of the folded conformation rather than the increased presence of Cys-14 and Cys-38 thiols is primarily responsible for reduced secretion. As discussed in further detail below, thermodynamic linkage between disulfide redox equilibria and folding stability have been established previously for other mutants of BPTI, DsbA, and thioredoxin.
Increased PDI Levels Decrease Secretion of All Mutants Proportionally-Overexpression of PDI in yeast can increase secretion of some proteins by as much as an order of magnitude (20,29), yet overexpression of PDI reduces secretion of BPTI (Ref. 17, Table I, Fig. 4). If either oxidative folding or reductive unfolding rates alone determined the efficiency of quality con-  Table I, and plotted against the previously reported 14 -38 disulfide destabilization free energy determined previously by equilibration with glutathione and/or dithiothreitol (Ref. 18). trol escape for these BPTI mutants, then increased PDI levels might be expected to exert differential effects among the mutants. However, PDI overexpression reduces secretion of wildtype, Y35F, P13S, A16V, G37A, Y35L, P13S ϩ C14/38A, Y35L ϩ C14/38A BPTI proportionally (Fig. 4.) The correlation of secretion efficiency with thermodynamic stability therefore persists under substantially altered conditions for redox folding catalysis in the ER. Of course, a redox folding catalyst such as PDI accelerates the oxidative folding and unfolding kinetics of a protein without altering its underlying thermodynamic stability. The mechanism for reduced BPTI secretion upon PDI overexpression is not clear, particularly given the beneficial effects of PDI overexpression for platelet-derived growth factor, Schizosaccharomyces pombe acid phosphatase, and antistasin (20,29). 2 One possible explanation is the "antichaperone" activity of PDI (30), which has also been invoked to explain negative effects of mutant PDI expression on antibody secretion in insect cells (31).
ER-associated Degradation Mutations Do Not Increase BPTI Secretion Efficiency-Because the extent of retention and proteolysis of unfolded protein determines secretion efficiency (Fig. 1), one might expect that loss of key components of the degradation pathway might increase secretion efficiency. This was not found to be the case however (Fig. 5.) Secretion of wild-type, P13S, and Y35L BPTI was unaffected by loss of function of the DER1 and UBC7 genes, which are required for efficient ER-associated degradation (32)(33). Thus, BPTI secretion efficiency is not increased in the absence of a fully functional ER-associated degradation apparatus. It is interesting that greater clonal variation of WT BPTI secretion occurred in the ER degradation-deficient mutant strains; this is because of the presence of several cultures in which secretion dropped precipitously to Ϸ 5 mg/liter. Such variation was not observed for P13S or Y35L BPTI secretion in these strains.

DISCUSSION
The data presented here supports a model for ER quality control whereby thermodynamic stability can determine the efficiency of escape from retention and degradation. Such a correlation was also observed for mutants of BPTI with cysteine/alanine replacements (17). A kinetic model for this pathway is represented schematically in Fig. 6. If folded protein were promptly exported from the ER by vesicular transport, one would expect secretion efficiency to be determined chiefly by the rate of folding. However, BPTI mutants that fold more rapidly than wild-type in vitro (Y35L and C30/51A BPTI) are secreted less efficiently than wild-type (Table I, Fig. 2; Ref. 17). The ER export rate for folded protein is determined by an ill-characterized process of protein cargo concentration into vesicles budding from the ER membrane (16). The nature of the required export signals and possible cargo receptors for soluble proteins are still uncertain, although Emp24p is a candidate yeast cargo receptor for some proteins (34 -35). If a protein folds rapidly but is delayed in export from the ER because of slow or inefficient cargo loading into transport vesicles, then a greater fraction of protein may be unfolded and degraded while awaiting export. Consistent with this scenario, the half-time for appearance of WT BPTI in the growth medium is approximately 30 min (Fig. 1) whereas the half time for folding, as assayed by binding to trypsin-Sepharose, is 2-5 min. 2 BPTI therefore persists intracellularly with a half time of approximately 30 min following completion of folding, during which time it may be partially unfolded, retained, and degraded, with the extent of such losses correlated with stability.
Retention and degradation are represented as separable processes in Fig. 6 because loss of components of the degradation pathway (Der1p and Ubc7p) does not shift the balance toward secretion (Fig. 5). ER retention could result either from physical binding to an ER-resident component of the quality control apparatus or exclusion from export vesicles. A constant fraction of WT, P13S, and Y35L BPTI are retained by the quality control apparatus even in the absence of a degradation sink for the retained material (Fig. 5), indicating that any retention receptors are not saturable at this level of multicopy expression. Calnexin has been shown to serve as a retention receptor for glycosylated proteins; however, BPTI is not glycosylated. Furthermore, deletion of the yeast calnexin homolog CNE1 does not increase the secretion efficiency of wild-type, P13S, or Y35L BPTI (data not shown). The identity of a quality control receptor responsible for BPTI retention has not been determined.
It should be emphasized that although secretion is correlated with thermodynamic stability, true equilibrium between the folded and unfolded forms cannot exist in the ER because of the presence of irreversible rate processes: degradation and ER export. Rather, a pseudo-steady-state partition between folded and unfolded proteins is attained. Of course, if the rates of degradation and export are much slower than the rates of folding and unfolding, then the pseudo-steady-state ratio of folded to unfolded protein could closely approximate the equilibrium value. Given wild-type BPTI's extreme stability (T m ϭ 105°C, Ref. 36), even the 3.7 kcal/mol of 14 -38 disulfide destabilization produced by the Y35L mutation should yield a protein that is mostly folded at equilibrium at 30°C. The precipitous drop in secretion efficiency for Y35L BPTI would be difficult to explain strictly on the basis of the small fraction of unfolded protein at equilibrium in vitro. However, the presence of high concentrations of protein folding chaperones such as BiP and PDI in the ER are likely to shift the folding equilibrium toward the unfolded state by Le Châ telier's principle, stabilizing partially unfolded conformers through chaperone binding free energy. Chaperones as structurally and functionally diverse as GroEL and SecB have been shown to catalyze deuterium exchange with normally protected amide protons in folded proteins (37)(38). Such "unfoldase" activity would be strongly present in the ER lumen given the high concentration of protein folding chaperones in this organelle. Therefore, the unfolded protein fraction vulnerable to retention and degradation would be expected to be larger than the equilibrium unfolded fraction in vitro.
There are several precedents for thermodynamic stability determining the half-life of proteins in vivo. Mutants of the bacteriophage cro protein with varying stability exhibit a correlation between increasing stability and decreasing degradation in the cytoplasm of Escherichia coli (39.) Degradation and resulting expression yield in the E. coli periplasm also correlates with thermodynamic stability of mutants of barnase (40). The metabolic stability in E. coli of T4 lysozyme mutants correlates roughly with their thermal stability with several clear exceptions (41); the same general correlation between thermal and metabolic stability with several distinct exceptions is also observed for stability of radiolabeled T4 lysozyme mutants injected into HeLa cells (42).
The available data indicate that secretion efficiency depends on the stability of the BPTI fold itself, rather than directly on the stability of the 14 -38 disulfide. Because removal of cysteines 14 and 38 by mutagenesis does not itself substantially reduce secretion efficiency of otherwise wild-type BPTI (Table I and Fig. 3, Ref. 17), destabilization of this disulfide per se is unlikely to directly reduce secretion efficiency. Furthermore, misfolding because of intramolecular thiol/disulfide rearrangements with free cysteines 14 and 38 cannot be responsible for reduced secretion of P13S or Y35L BPTI because the C14/38A mutation does not increase secretion of these mutants (Table I, Fig. 3). Destabilization of the 14 -38 disulfide therefore must correspond to destabilization of the folded form of BPTI in the absence of the 14 -38 disulfide. Such thermodynamic linkage between disulfide stability and folding stability has been demonstrated for several BPTI mutants previously (43.) This linkage has also been demonstrated for mutants of DsbA (44) and thioredoxin (45).
The kinetic model presented in Fig. 6 leads to several alternative limiting regimes depending on a given protein's rates of folding and unfolding relative to the rates of export and retention. For BPTI and the mutants examined to date, it appears that the folding and unfolding rates in vivo are substantially more rapid than export and retention rates, leading to a rapid equilibration relative to the irreversible processes leading to export or degradation. For proteins that fold more slowly or are exported more rapidly following folding, such equilibration would not occur and the rate of folding would be expected to determine secretion efficiency. The intrinsic biophysical attributes of folded BPTI that lead to its slow export kinetics from the ER are currently under investigation.
FIG. 6. Kinetic model for ER quality control. For a rapidly folding but slowly exported protein such as BPTI, thermodynamic stability determines secretion efficiency. For a slowly folded but rapidly exported protein, folding rate might instead determine secretion efficiency.