The BiP Molecular Chaperone Plays Multiple Roles during the Biogenesis of TorsinA, an AAA+ ATPase Associated with the Neurological Disease Early-onset Torsion Dystonia*

Background: The ΔE mutation in the AAA+ ATPase torsinA is associated with the neurological disease torsion dystonia. Results: BiP and its co-factors maintain torsinA and torsinAΔE stability, glycosylation, and solubility. Conclusion: torsinA/ΔE, a chaperone-like protein, requires the assistance of other chaperones to fold. Significance: Therapeutics that modulate BiP may counteract torsinAΔE-associated physiological defects. Early-onset torsion dystonia (EOTD) is a neurological disorder characterized by involuntary and sustained muscle contractions that can lead to paralysis and abnormal posture. EOTD is associated with the deletion of a glutamate (ΔE) in torsinA, an endoplasmic reticulum (ER) resident AAA+ ATPase. To date, the effect of ΔE on torsinA and the reason that this mutation results in EOTD are unclear. Moreover, there are no specific therapeutic options to treat EOTD. To define the underlying biochemical defects associated with torsinAΔE and to uncover factors that might be targeted to offset defects associated with torsinAΔE, we developed a yeast torsinA expression system and tested the roles of ER chaperones in mediating the folding and stability of torsinA and torsinAΔE. We discovered that the ER lumenal Hsp70, BiP, an associated Hsp40, Scj1, and a nucleotide exchange factor, Lhs1, stabilize torsinA and torsinAΔE. BiP also maintained torsinA and torsinAΔE solubility. Mutations predicted to compromise specific torsinA functional motifs showed a synthetic interaction with the ΔE mutation and destabilized torsinAΔE, suggesting that the ΔE mutation predisposes torsinA to defects in the presence of secondary insults. In this case, BiP was required for torsinAΔE degradation, consistent with data that specific chaperones exhibit either pro-degradative or pro-folding activities. Finally, using two independent approaches, we established that BiP stabilizes torsinA and torsinAΔE in mammalian cells. Together, these data define BiP as the first identified torsinA chaperone, and treatments that modulate BiP might improve symptoms associated with EOTD.


Early-onset torsion dystonia (EOTD) is a neurological disorder characterized by involuntary and sustained muscle contrac-
tions that can lead to paralysis and abnormal posture. EOTD is associated with the deletion of a glutamate (⌬E) in torsinA, an endoplasmic reticulum (ER) resident AAA ؉ ATPase. To date, the effect of ⌬E on torsinA and the reason that this mutation results in EOTD are unclear. Moreover, there are no specific therapeutic options to treat EOTD. To define the underlying biochemical defects associated with torsinA⌬E and to uncover factors that might be targeted to offset defects associated with torsinA⌬E, we developed a yeast torsinA expression system and tested the roles of ER chaperones in mediating the folding and stability of torsinA and torsinA⌬E. We discovered that the ER lumenal Hsp70, BiP, an associated Hsp40, Scj1, and a nucleotide exchange factor, Lhs1, stabilize torsinA and torsinA⌬E. BiP also maintained torsinA and torsinA⌬E solubility. Mutations predicted to compromise specific torsinA functional motifs showed a synthetic interaction with the ⌬E mutation and destabilized torsinA⌬E, suggesting that the ⌬E mutation predisposes torsinA to defects in the presence of secondary insults. In this case, BiP was required for torsinA⌬E degradation, consistent with data that specific chaperones exhibit either pro-degradative or pro-folding activities. Finally, using two independent approaches, we established that BiP stabilizes torsinA and torsinA⌬E in mammalian cells. Together, these data define BiP as the first identified torsinA chaperone, and treatments that modulate BiP might improve symptoms associated with EOTD.
In order to function proteins need to acquire proper secondary and tertiary structures. Although the amino acid sequence of a protein is a major determinant of its final fold, most proteins acquire intermediate folding states during translation that require the help of molecular chaperones and folding enzymes (1)(2)(3)(4)(5). Chaperones not only maintain protein solubility and facilitate folding but they also recognize misfolded proteins and target them for degradation via protein quality control pathways (6 -11). Indeed, cell survival depends on the efficiency of these pathways, as defects in pathway function are lethal when cells are exposed to various stresses. Moreover, misfolded proteins that are not cleared from the cell can severely impact cellular physiology. These proteins can aggregate, they can gain a dominant negative function, and/or they can deplete chaperones, thereby exacerbating the accumulation of misfolded proteins and compromising protein homeostasis or "proteostasis" (12)(13)(14)(15). Among all the proteins in the eukaryotic cell, approximately one-third enter the secretory pathway through the endoplasmic reticulum (ER) 4 and either remain in the ER, are secreted from the cell, or are ultimately distributed among the plasma membrane, endosomes, vacuole/lysosomes, or Golgi apparatus. In the secretory pathway, misfolded proteins are cleared through diverse mechanisms, beginning with ER-associated degradation (ERAD), which represents a critical step in the protein quality control of secretory proteins (10,11,16).
A key chaperone that is required for protein folding in the ER is the lumenal Hsp70, BiP/GRP78 (also known as Kar2 in Saccharomyces cerevisiae) (17)(18)(19). Kar2/BiP is also a major contributor to general ER homeostasis, not only by participating in protein folding and protein translocation, assembly, and quality control, but also by regulating calcium homeostasis and ER * This work was supported, in whole or in part, by National Institutes of Health To identify chaperones that aid in torsinA folding and that might be targeted to correct disease phenotypes associated with the ⌬E allele, we co-opted the genetic tools available in the yeast S. cerevisiae and designed a new torsinA and torsinA⌬E expression system. Yeasts have been used to define the molecular basis underlying several human diseases, including amyotrophic lateral sclerosis, antitrypsin deficiency, cancer, and Huntington, Alzheimer, and Parkinson diseases, among many others (69 -72). Using this system, we found that Kar2/BiP and its Hsp40, Scj1, as well as the Kar2/BiP-associated NEF, Lhs1, contribute to torsinA and torsinA⌬E stabilization in the ER. We also found that Kar2/BiP plays a dual role in controlling torsinA stability, depending on the presence of the ⌬E mutation and/or of secondary mutations in unique functional motifs. Furthermore, Kar2/BiP affects torsinA and torsinA⌬E solubility and N-linked glycosylation, consistent with a role in mediating protein folding. Finally, by depleting BiP levels in a mammalian cell model, we confirmed that BiP facilitates torsinA and torsinA⌬E biogenesis. Together, these data represent the first demonstration of the cellular chaperones that orchestrate torsinA maturation.

EXPERIMENTAL PROCEDURES
Plasmid and Yeast Strain Construction-The pcDNA3.1-torsinA and pcDNA3.1-torsinA⌬E expression plasmids were kindly provided by Dr. Xandra Breakefield (65). The torsinA and torsinA⌬E open reading frames (ORF) were subcloned into the yeast expression vector pRS426GPD (73) by double restriction enzyme digestion with EcoRI/XhoI (Fermentas, Thermo Scientific) of pcDNA3.1-torsinA and pcDNA3.1-torsinA⌬E and ligation into EcoRI/XhoI-linearized pRS426GPD (Fig. 1B). To construct HA-tagged torsinA and torsinA⌬E vectors, a single HA tag was introduced at the C terminus of torsinA and torsinA⌬E by in vivo recombination in S. cerevisiae following a previously published protocol (74). Briefly, primers LZJB12 and -13, encoding the HA tag sequence (Table 1), were annealed and co-transformed into S. cerevisiae together with NotI-digested pRS426GPD-torsinA or torsinA⌬E. Recombined plasmids were extracted from S. cerevisiae and transformed into E. coli DH5␣ for amplification.
All S. cerevisiae strains used in this study are described in Table 3. To construct the strains containing a deletion of the PDR5 or PEP4 ORFs in the kar2-1 strain background, the pdr5⌬::KanMX and pep4⌬::KanMX deletion cassettes containing Ͼ200 bp of homology region upstream and downstream of each ORF were PCR-amplified (Pfu Turbo, Agilent) from DNA extracted from the corresponding deletion mutant in the haploid BY4742 S. cerevisiae deletion strain collection (Open Biosystems, Thermo Scientific) using primer pairs LZJB69 and -70 (for PDR5) and LZJB34 and -35 (for PEP4) (Table 1), and transformation of these cassettes into the kar2-1 strain ( Table 3). The correct genotype of the mutants was confirmed by PCR analysis of genomic DNA of strains resistant to the antibiotic geneticin sulfate G418 (Research Products International Corp.) using primer pairs LZJB71 and -72 (for PDR5) and LZJB44 and -45 (for PEP4), as well as the KanB primer (CTGCAGCGAG-GAGCCGTAAT), which anneals to the KanMX gene ( Table 1). The phenotypes of the constructed strains were verified as follows. Deletion of PEP4 was verified by a defect in the maturation of the vacuolar protease aminopeptidase 1 (Ape1) (76), and deletion of PDR5 was verified by a general increase in the level of ubiquitinated proteins in MG132 (Peptide Institute Inc., Japan)-treated cells (data not shown) (77,78). All yeast transformations were performed using lithium acetate/PEG3350, and DNA extractions were performed following standard protocols (79,80).
Media and Growth Conditions-Yeast strains were grown at 26 -28°C on YPD medium (1% yeast extract, 2% peptone, 2% dextrose) or on synthetic complete (SC) medium lacking specific amino acids required for auxotrophic selection, as described previously (81). For selection of KanMX-expressing strains, YPD plates were supplemented with 250 g/ml G418.
Biochemical Methods-To assess protein stability, cells were grown overnight in selective medium at 28°C, diluted to an initial A 600 of 0.2 in fresh medium, and incubated at the same temperature for ϳ6 h to an A 600 of ϳ1. Cycloheximide (CHX) (Sigma) was added to a final concentration of ϳ190 g/ml, and cells were incubated at 37°C for 90 min. A 1-ml aliquot was taken before adding CHX (time 0 min) and every 30 min after adding CHX. The cells were then pelleted by centrifugation at 18,000 ϫ g for 1 min at 4°C. The cell pellet was immediately frozen in liquid nitrogen. Samples were processed for Western blot analysis as described below. To inhibit the proteasome, cells were preincubated in the presence of MG132 for 30 min at 37°C before addition of CHX.
Protein extractions from mammalian cells were performed as follows. Cells were washed with cold PBS, resuspended in RIPA buffer with protease inhibitors (Roche Applied Science), rocked at 4°C for 40 min, and harvested. The cell suspension was then centrifuged at 18,000 ϫ g for 10 min at 4°C, and the supernatant was used for Western blotting.
To monitor the acquisition of N-linked glycosylation, whole cell protein extracts were digested with 50 milliunits of endoglycosidase H (Roche Applied Science) for 1 h at 37°C in the presence of 1 mM PMSF and 100 mM KOAc. Samples were resolved by SDS-PAGE and used for Western blotting, as described above.
To measure protein extraction from microsomal membranes, ER-derived microsomes were prepared from cells grown overnight in selective medium at 28°C, diluted to an A 600 of 0.2, and incubated at the same temperature for ϳ6 h to an A 600 of ϳ1.2. A total of 60 absorbance units of cells were harvested and frozen. The pellets were thawed on ice and resuspended in 600 l of lysis buffer (20 mM Hepes, pH 7.4, 50 mM KOAc, 2 mM EDTA, pH 8, 100 mM sorbitol, 1 mM dithiothreitol (DTT)). A total of ϳ500 l of glass beads were added, and samples were agitated on a Vortex mixer seven times for 60 s, with a 60-s incubation on ice between each treatment. The lysate was collected in an Eppendorf tube and centrifuged twice at 3000 rpm in a tabletop centrifuge at 4°C for 5 min to remove cell debris and unbroken cells. The supernatant was then centrifuged at 18,000 ϫ g at 4°C for 20 min to collect the microsomes, which were washed in 500 l of Buffer 88 (20 mM Hepes, pH 6.8, 150 mM KOAc, 250 mM sorbitol, 5 mM MgOAc) and centrifuged as above. Membranes were resuspended in Buffer 88, and the protein concentration was adjusted spectrophotometrically by measuring the A 280 in 2% SDS. Next, 20 l of the isolated microsomes were resuspended in 1 ml of the following buffers: Buffer 88, pH 6.8, 1% Triton X-100 (Sigma) in Buffer 88, pH 6.8, 0.1 M freshly prepared sodium carbonate in water, 25 mM CHES buffer, pH 9.5, 25 mM CAPS buffer, pH 10.5, or 6 M urea in 37.5 mM Tris, pH 8, 2 mM EDTA. After the samples were incubated for 30 min on ice, the samples were centrifuged at 100,000 ϫ g for 1 h at 4°C in a Sorvall RC M120EX ultracentrifuge. The pellets were washed with 500 l of Buffer 88, and re-centrifuged at 146,000 ϫ g for 15 min at 4°C. The supernatants were transferred to Eppendorf tubes, and total protein was precipitated by adding 110 l of 100% TCA and incubating on ice for 15 min, followed by a centrifugation at 18,000 ϫ g for 10 min at 4°C. Pellets from samples containing Triton X-100 were washed with acetone. All pellets were resuspended in 35 l of TCA sample buffer, and proteins were resolved by SDS-PAGE and examined by Western blotting. All buffers were supplemented with a proteinase inhibitor mixture. For protease protection assays, protease inhibitors were removed from the microsomes by washing them twice with 500 l of Buffer 88. The washed microsomes were incubated on ice for 30 min in the presence or absence of 1% Triton X-100, followed by a 1-h incubation on ice in the presence or absence of 50 g/ml proteinase K (Sigma). Proteins were precipitated with 500 l of ice-cold 10% TCA and centrifuged at 18,000 ϫ g for 10 min at 4°C. Pellets were resuspended in 30 l of TCA sample buffer. Samples were resolved by SDS-PAGE and analyzed by Western blotting.
For protein co-immunoprecipitation studies, yeast cells were grown overnight in selective medium at 28°C, diluted to an A 600 of 0.2, and grown at the same temperature for ϳ6 h to an A 600 of ϳ1.2. A total of ϳ45 absorbance units of cells were harvested and frozen. The pellets were thawed on ice and resuspended in 600 l of lysis buffer supplemented with 0.5% Triton X-100 and a protease inhibitor mixture and transferred to borosilicate tubes containing ϳ500 l of glass beads. Samples were agitated on a Vortex mixer 9 times for 60 s, with Ͼ60 s of incubation on ice between each treatment. The lysate was collected in an Eppendorf tube and centrifuged twice at 3000 rpm in a tabletop centrifuge at 4°C for 5 min to remove cell debris and unbroken cells. The supernatant was further cleared by centrifuging twice at 18,000 ϫ g at 4°C for 20 min. The protein concentration was adjusted spectrophotometrically by measuring the A 280 of an aliquot in 2% SDS. Extracts were precleared by incubation for 1.5 h at room temperature with 30 l of protein G-agarose (Roche Applied Science) and then incubated overnight with 30 l of protein G-agarose (Roche Applied Science) in the presence or absence of 1 l of anti-Kar2 antiserum. Beads were washed four times with 500 l of wash buffer (20 mM Hepes, pH 7.4 (buffered with KOH), 150 mM NaCl, and supplemented with a protease inhibitor mixture). 30 l of TCA sample buffer was added to the beads, and proteins were extracted by heating the beads for 5-7 min at 75°C. Samples were resolved by SDS-PAGE and analyzed by Western blot.
Indirect Immunofluorescence Microscopy-The preparation of yeast for indirect immunofluorescence was performed following a previously published protocol (86). Briefly, cells were grown overnight in selective medium at 28°C, diluted to an A 600 of 0.3 in fresh medium, and grown at 28°C for ϳ5 h to an A 600 of ϳ0.7. To fix the yeast, 37% formaldehyde was added (final concentration of 4%), and the cells were incubated for 10 min at 28°C with shaking. The cell suspension was centrifuged, and the pelleted yeast were resuspended in 5 ml of KM solution (40 mM KPO 4 , pH 6.5, 0.5 mM MgCl 2 ) with 4% formaldehyde and incubated for 1 h at 28°C with shaking. The cells were washed twice with KM solution and once with KM solution supplemented with 1.2 M sorbitol (KM ϩ sorbitol) and were resuspended in 500 l of KM ϩ sorbitol. Next, the cell walls were digested with 30 l of 10 mg/ml Zymolyase (20T, MP Biomedicals LLC) for no more than 25 min at 37°C, and the spheroplasts were washed with KM ϩ sorbitol buffer, resuspended in 500 l of the same buffer, and incubated overnight at 4°C. A 20-l aliquot of the cell suspension was spotted on polylysine-coated slides, and the cells were dehydrated with methanol/acetone, blocked with PBS, pH 7.4, supplemented with 0.5% BSA, 0.5% ovalbumin, and 0.1% Triton X-100, and incubated at 37°C for 1 h. Primary antibodies (mouse anti-HA 12CA5 (Roche Applied Science) (1:100) or rabbit anti-Kar2 (1:250)) diluted in blocking solution were applied, and the cells were incubated overnight at 4°C, before secondary antibodies (Alexa Fluor 488 anti-mouse and Alexa Fluor 569 anti-rabbit (1:250)) were added for 1 h at room temperature. The slides were mounted using Prolong Gold Antifade reagent with DAPI (Invitrogen). Pictures were taken using a Leica TCS SP5 confocal microscope, ϫ63 oil immersion objective, and analyzed with Adobe Photoshop (Version 7.0) software.

Analysis of TorsinA and TorsinA⌬E Stability and Biogenesis in HeLa Cells-HeLa
Tet-On cells (Clontech) were maintained in DMEM (Invitrogen) supplemented with antibiotics and 10% fetal bovine serum at 37°C in a 5% CO 2 humidified incubator. Vectors engineered for the transient expression of torsinA/⌬E or C-terminal HA-tagged versions of torsinA/⌬E (Table 2) were transfected using Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions (the medium was changed ϳ3.5-4 h post-transfection). siRNA transfections were performed using RNAiMax (Invitrogen) (the medium was changed ϳ7 h post-transfection). In these experiments, the torsinA expression vectors were transfected into siRNA-treated cells 24 h after the siRNAs had been introduced. The following Stealth siRNAs (Invitrogen) were used: HSS105077 and HSS179390 against BiP/GRP78, and Stealth RNAi siRNA Negative Control Medium GC (Invitrogen), at a concentration of 1.6 nM. Protein extracts were prepared from cells incubated for another 24 h (a total of 48 h after siRNA transfection). Although higher efficiency of BiP knockdown can be achieved, our protocol was adjusted to minimize cell toxicity (2). As an alternative means to deplete the levels of BiP, we employed the AB5 subtilase cytotoxin (87).
To monitor torsinA stability, pulse chase analyses were performed as previously published (63) and included a 30-min starvation period and a 15-min labeling period with 200 Ci/ well of [ 35 S]Met/Cys (ICN Biochemicals) in a 6-well plate. Protein levels at the indicated time points were analyzed after immunoprecipitation with mouse monoclonal anti-HA 16B12 (Covance) and protein G-agarose (Roche Applied Science) using a Typhoon FLA7000 and associated software (version 1.2) (General Electric). Where indicated, the active subtilase toxin or an inactive mutant (SubAA272B) was added to a final concentration of 0.5 g/ml during amino acid starvation and was present throughout the labeling and chase steps.
Statistical Analysis-Statistical analyses were performed using Student's t test (Microsoft Excel Software), assuming equal variances. A p value Ͻ 0.05 was considered significant. All statistical analyses of protein degradation assays were per-formed only with the values at the end of the chase, because they reflect maximal differences between data sets.

A Yeast Model to Study TorsinA and TorsinA⌬E Biogenesis-
To express torsinA and torsinA⌬E in S. cerevisiae, we subcloned the ORF of torsinA and torsinA⌬E into the multicopy yeast expression vector pRS426 (73), placing torsinA under the control of the constitutive glyceraldehyde-3-phosphate dehydrogenase promoter (Fig. 1B). In this system, torsinA and torsinA⌬E were expressed at similar levels and migrated by SDS-PAGE as an ϳ33-kDa species, which corresponds to the monomeric forms of the proteins (Fig. 1C). Endoglycosidase H treatment of torsinA and torsinA⌬E protein extracts produced a faster migrating band, indicating that both torsinA and torsinA⌬E were N-glycosylated in yeast, as shown previously in mammalian cells (Fig. 1C) (65). Also, in accordance with results in mammalian cells, torsinA-HA and torsinA⌬E-HA were localized in the ER/nuclear envelope compartments, as evidenced by co-localization with the ER chaperone, Kar2/BiP (Fig. 1D). Similar results were obtained using sucrose density gradients with untagged torsinA and torsinA⌬E and through live cell fluorescence imaging of GFP-tagged forms of torsinA and torsinA⌬E (data not shown).
TorsinA is a monotopic membrane protein (88) and behaves like a peripherally associated protein in mammalian cells (89). To determine whether this was also the case in yeast, we performed alkaline extractions using isolated ER-derived microsomal membranes from yeast expressing torsinA or torsinA⌬E, following a previously published method for torsinA (89). All proteins remained associated with the microsomal pellet fraction at pH 6.8, as expected ( Fig. 2A, lanes 1 and 2 and 9 and 10). At pH 9.5, both Pdi1, a soluble lumenal protein, and Kar2/BiP, a peripherally associated protein (90), were partially released into the supernatant. However, at pH 9.5, torsinA and torsinA⌬E remained associated with the pellet ( Fig. 2A, compare lanes 3 and 4 and 11 and 12). At pH 10.5, Pdi1 was almost completely released into the supernatant, whereas a considerable fraction of torsinA and torsinA⌬E remained associated with the pellet fraction, similar to Kar2/BiP ( Fig Fig. 8; K108 and E171, residues mutated in Fig. 8 that belong to the Walker motifs; D216, mutation that impacts ⌬E penetrance; E302 or E303, glutamate residue deleted in patients with EOTD. Walker-A and -B, and Sensor-I and -II motifs (for ATP binding and hydrolysis) are indicated by black lines. B, ORFs of torsinA and torsinA⌬E were cloned into the multicopy shuttle vector pRS426 and are under the control of the constitutive glyceraldehyde-3-phosphate dehydrogenase (GPD) promoter. C, protein samples were prepared from log phase wild-type yeast cells transformed with empty vector (V) or expression vectors for torsinA (T) or torsinA⌬E (⌬E) grown at 28°C and were treated with endoglycosidase H (EndoH) as indicated. The glycosylated species migrate at ϳ33 kDa and the unglycosylated species resolve at ϳ27 kDa. The * indicates a background band of the antibody. D, indirect immunofluorescence of log phase wild-type yeast cells transformed with expression vectors for torsinA-HA or torsinA⌬E-HA or an empty vector control grown at 28°C. The ER lumenal chaperone Kar2/BiP and the nuclear concentrated dye DAPI were used as controls.
remained associated with the pellet at each pH, as expected. Treatment with 6 M urea efficiently released Pdi1, Kar2/BiP, and both torsinA variants but not Sec61, as expected ( Fig. 2A, lanes [17][18][19][20]. Overall, torsinA and torsinA⌬E behaved more similarly to Kar2/BiP than to Pdi1, indicating that torsinA and torsinA⌬E are peripherally membrane-associated proteins in yeast. An N-terminal hydrophobic domain (HD; Fig. 1A) anchors torsinA to the lumenal side of the ER membrane (75,89,91). To test whether the HD was responsible for maintaining torsinA in the yeast ER, as anticipated, we examined whether torsinA, torsinA⌬E, and versions lacking the HD (torsinA-⌬24 -40 and torsinA⌬E-⌬24 -40) were secreted from cells by using a dot blot technique (82,92). Cells expressing each torsinA variant or harboring an empty vector control were spotted onto nitrocellulose membranes. After ϳ20 h of incubation on medium, we either lysed the cells in situ (to assess total protein) or removed the cells with vigorous washing (to examine secreted protein, which adhered to the membrane). We then performed an immunoblot for torsinA and the nonsecreted cytosolic protein G6PDH (Fig. 2B). As expected, we observed that torsinA-⌬24 -40 and torsinA⌬E-⌬24 -40 were secreted, while torsinA, torsinA⌬E, and G6PDH were not. Therefore, similar to mammalian cells, the N-terminal HD of torsinA prevents torsinA secretion to the extracellular medium, suggesting that this domain is properly folded and helps maintain torsinA and torsinA⌬E in the yeast ER (75,88).
Together, these data indicate that torsinA and torsinA⌬E are N-linked glycosylated ER resident proteins that are peripherally associated with the ER membrane in yeast, as in mammalian cells. Because these basic cellular properties of torsinA are conserved in yeast, we pursued subsequent studies to define how torsinA and torsinA⌬E impact cellular homeostasis and which cellular factors regulate their protein stability.
TorsinA can modulate the ER stress response, and torsinA overexpression leads to defects in ERAD and protein secretion (53,56,(93)(94)(95). To test whether torsinA or torsinA⌬E expression in yeast induces a stress response or leads to any overt phenotype, we compared the behavior of cells expressing torsinA, torsinA⌬E, or harboring an empty vector control. Cells expressing torsinA or torsinA⌬E showed no growth defect in the absence or presence of several ER stressors (including heat shock, tunicamycin, and DTT) compared with a strain transformed with an empty vector control (data not shown). Expression of torsinA or torsinA⌬E in yeast also did not alter the heat shock response of wild-type cells incubated at 39°C nor did it alter the unfolded protein response in wild-type cells incubated in the presence of tunicamycin (data not shown). These data are in-line with a recently published study in which the phenotypes of yeast expressing torsinA were analyzed (96). In addition, the degradation of the ERAD substrates CPY* and CFTR was unaltered by the co-expression of torsinA or torsinA⌬E. Finally, no morphological changes were evident in the nuclear envelope of strains expressing torsinA or torsinA⌬E, as determined by electron microscopy (data not shown) (53,97,98). These combined data indicate that torsinA is not toxic when expressed in yeast.
Stability of TorsinA and TorsinA⌬E Depends on the ER Chaperone Kar2/BiP-Most of the current knowledge on torsinA is concentrated on the function and biochemical properties of torsinA and on the effect the ⌬E mutation has on these properties (59,99). However, nothing is known about whether molecular chaperones facilitate the folding, stabilization, and/or degradation of torsinA and torsinA⌬E. Using yeast as a model system, we set out to identify which chaperones impact torsinA and/or torsinA⌬E biogenesis.
A critical ER resident chaperone that plays a role in both protein folding and degradation is Kar2/BiP (17-19, 100 -102). To test the role of Kar2/BiP on torsinA and torsinA⌬E stability, we performed CHX chase experiments in a wild-type (KAR2) strain and in a Kar2/BiP mutant (kar2-1) strain. The kar2-1 allele carries a temperature-sensitive mutation that compromises ERAD and protein folding due to a defect in the controlled release of bound peptide substrates (101,(103)(104)(105). We first noted that there was no significant difference in the degradation of torsinA and torsinA⌬E in wild-type cells (Fig. 3A). These results are in accordance with experiments in mammalian cells showing identical torsinA and torsinA⌬E turnover when measured during a short time course (63). However, in the kar2-1 strain both torsinA and torsinA⌬E were significantly destabilized (55% versus 81% for torsinA and 59% versus 84% for torsinA⌬E remained after 90 min in the mutant strain versus the wild-type strain, respectively) (Fig. 3A). The data suggest that Kar2/BiP helps maintain the stability of torsinA and torsinA⌬E in yeast.
We noticed that torsinA and torsinA⌬E expressed in the kar2-1 strain migrated as three distinct bands, with the highest molecular weight species migrating at the position of torsinA FIGURE 2. TorsinA and torsinA⌬E peripherally associate with the yeast ER membrane through an N-terminal hydrophobic domain. A, ER-derived microsomes were prepared from log phase wild-type yeast cells grown at 28°C that expressed torsinA or torsinA⌬E and were incubated in a pH 6.8 buffer, in 25 mM CHES, pH 9.5 buffer, 25 mM CAPS, pH 10.5 buffer, 0.1 M Na 2 CO 3 , pH ϳ11.5, or in 6 M urea for 30 min in ice. After centrifugation at 100,000 ϫ g, aliquots from the supernatant (S) or the pellet (P) fractions were analyzed by SDS-PAGE, and the presence of torsinA and torsinA⌬E was detected by Western blot. Pdi1 is a lumenal ER protein; Kar2 is a peripherally associated ER lumenal protein, and Sec61 is an integral membrane ER protein. B, nitrocellulose membranes spotted with overnight yeast cultures transformed with vector alone (V) or expressing torsinA (T) or torsinA⌬E (⌬E), were layered on top of solid selective medium and incubated for ϳ20 h at 30°C. To detect secreted proteins, cells were washed off the nitrocellulose membrane with water. To detect total protein (intracellular and secreted material), cells were lysed in situ and then washed. TorsinA and G6PDH were detected by Western ("dot") blot. Please note that secreted proteins yield a higher signal than their nonsecreted counterparts in the total protein blot (82). expressed in the wild-type strain (Fig. 3, B and C). Because torsinA can be N-glycosylated at two asparagines, Asn-143 and Asn-158 (Fig. 1A, and see below) (65, 106), we hypothesized that the faster migrating bands corresponded to mono-and unglycosylated torsinA. Indeed, the mobility of the middle band coincided with the mobility of the torsinA-N143Q mutant, which can only be N-monoglycosylated at the Asn-158 residue (ϳ30 kDa) (Fig. 3B). Similar results were obtained for torsinA⌬E (data not shown). This result indicates that, similar to mammalian cells, torsinA and torsinA⌬E are N-glycosylated at both asparagines. Furthermore, the migration of the lowest molecular weight species coincided with the mobility of endoglycosidase H-treated samples of torsinA and torsinA⌬E expressed in both the KAR2 and kar2-1 strains and with the mobility of a torsinA-N143Q,N158Q double mutant (ϳ27 kDa) (Fig. 3B). Therefore, a loss of Kar2/BiP function leads to inefficient N-linked glycosylation of torsinA and torsinA⌬E. Importantly, the effect of the kar2-1 mutation on torsinA is specific because the N-linked glycosylation of other substrates, such as CPY*, is unaffected by this mutation (data not shown) (107).
Although the kar2-1 allele does not appear to impact protein translocation into the ER (101), it was formally possible that the mono-and unglycosylated torsinA species in the kar2-1 strain arise from defective translocation. To exclude this possibility, we performed a protease protection assay in the KAR2 and kar2-1 strains expressing torsinA and torsinA⌬E. If indeed the mono-and unglycosylated torsinA species in the kar2-1 strain were caused by a translocation defect, they would be exposed on the surface of the microsomes and would be protease-accessible even in the absence of detergent. However, each of the torsinA and torsinA⌬E species was stable after protease treatment (Fig. 3C,  compare lanes 1 and 2 and 4 and 5, 7 and 8, and 10 and 11). In contrast, addition of 1% Triton X-100 initiated torsinA and torsinA⌬E degradation (Fig. 3C, compare lanes 2 and 3, 5 and 6, 8  and 9, and 11 and 12). Kar2/BiP was also protected from protease in the absence of detergent but was clipped and migrated faster in the presence of detergent. The ER transmembrane protein Bos1, which exposes an epitope to the cytosol (108), was proteolyzed in the presence and absence of Triton X-100, as anticipated. These results indicate that the differentially N-glycosylated torsinA and torsinA⌬E species are ER-encapsulated and support a role for Kar2/BiP in torsinA and torsinA⌬E N-linked glycosylation and stability. Because a loss of protein stability is associated with misfolding (109), our data strongly suggest that Kar2/BiP functions as a torsinA chaperone. replicates/experiment. The % of torsinA or torsinA⌬E remaining after CHX addition in wild-type KAR2 or kar2-1 yeast strains is graphed (*, p Ͻ 0.04 for KAR2 torsinA versus kar2-1 torsinA; p Ͻ 0.003 for KAR2 torsinA⌬E versus kar2-1 torsinA⌬E). Representative Western blots are shown. B, torsinA and torsinA⌬E N-linked glycosylation is Kar2-dependent. Protein samples were prepared from log phase KAR2 and kar2-1 yeast strains as before, except that they were treated with endoglycosidase H (EndoH) as indicated before Western blot analysis. Pdi1 is an N-glycosylated protein, whereas G6PDH is a cytosolic unglycosylated protein. * indicates a background band in the blot. C, torsinA and torsinA⌬E glycosylated species are ER-embedded. Microsomes from KAR2 and kar2-1 strains expressing torsinA and torsinA⌬E were incubated on ice for 30 min in the presence or absence of Triton X-100, followed by a 1-h incubation on ice in the presence or absence of proteinase K. Proteins were TCA-precipitated, resolved by SDS-PAGE, and analyzed by Western blot. Note that the ER protein Kar2 is clipped by proteinase K in the presence of Triton X-100. Bos1 is an ER transmembrane protein with an epitope exposed to the cytosol that is proteinase K-sensitive even in the absence of Triton X-100. UC, uncleaved Kar2/BiP; C, cleaved Kar2/BiP. D, Kar2/BiP co-immunoprecipitates with torsinA and torsinA⌬E. Whole cell extracts from KAR2 cells transformed with a vector control (V) or expression vectors for torsinA (T) or torsinA⌬E (⌬E) were immunoprecipitated (IP) under native conditions (0.5% Triton X-100) using anti-Kar2 antiserum. Immunoprecipitated material was resolved by SDS-PAGE and analyzed by Western blot. Controls from immunoprecipitations in which primary antiserum was absent (Ϫ) were also included.
If, as expected, Kar2/BiP is a torsinA chaperone, then both proteins should physically interact. To test this hypothesis, we obtained whole cell extracts from wild-type cells, immunoprecipitated Kar2/BiP under native conditions, and probed for torsinA and torsinA⌬E by Western blot. Both torsinA and torsinA⌬E co-precipitated with Kar2/BiP in the absence of cross-linking agents (Fig. 3D). Only a relatively small fraction of torsinA and torsinA⌬E co-precipitated with Kar2/BiP, suggesting a transient interaction. Prolonged Hsp70-substrate interactions normally lead to substrate degradation (110,111). However, transient interactions between chaperones and their substrates occur during substrate folding, as observed previously for Kar2/BiP (102,112,113). The combined data indicate that Kar2/BiP is a pro-folding chaperone for torsinA and torsinA⌬E.
The ER Lumenal Hsp40 Scj1 and the Nucleotide Exchange Factor Lhs1 Also Maintain TorsinA and TorsinA⌬E Stability-Kar2/BiP function in yeast is augmented by the action of cochaperones, including the Hsp40s Jem1 and Scj1 and the NEFs Lhs1 and Sil1 (21,114). Therefore, we compared torsinA and torsinA⌬E stability by CHX chase analysis in a jem1⌬ scj1⌬ strain to the wild-type strain (JEM1 SCJ1) (Fig. 4A). Similar to the results observed in the kar2-1 strain (Fig. 3A), torsinA and torsinA⌬E stability was significantly decreased in the jem1⌬ scj1⌬ strain compared with the wild-type strain (55% versus 84% torsinA, and 62% versus 80% torsinA⌬E remained after 90 min in the mutant and wild-type strains, respectively) (Fig. 4A).
To determine whether one of the two Hsp40s played a more important role in stabilizing torsinA and torsinA⌬E, we performed chase analyses in the jem1⌬ and scj1⌬ single mutant strains. In agreement with previous reports showing a more prominent role of Scj1 than Jem1 during protein maturation (27,29), torsinA and torsinA⌬E stability was exclusively SCJ1-dependent (89% versus 48% torsinA and 83% versus 46% torsinA⌬E remained after 90 min in the jem1⌬ and scj1⌬ strains, respectively) (Fig. 4B). Interestingly, we observed a glycosylation defect in the jem1⌬ scj1⌬ mutant but not in the jem1⌬ or scj1⌬ single mutants (Fig. 4, A and B). Therefore, although only Scj1 is required to stabilize torsinA and torsinA⌬E, both Hsp40s support efficient torsinA and torsinA⌬E glycosylation. These experiments suggest that torsinA instability is independent of the glycosylation defect caused by the absence of these chaperones.
above, that the defect in glycosylation is uncoupled from torsinA stabilization. Together, these data strongly suggest that Kar2/BiP functions with a cognate Hsp40, Scj1, as well as a NEF, Lhs1, to facilitate the folding of torsinA and torsinA⌬E in the ER.
Importantly, deletion of other ER chaperones had no impact on the stability or N-linked glycosylation of torsinA and torsinA⌬E, indicating that these phenomena are not the result of a general ER stress response ( Fig. 5 and data not shown). For example, deletion of the chaperone-like lectins Cne1 and Yos9 had no effect on these properties (Fig. 5). Preliminary data suggest that Mnl1 (the yeast EDEM1 homolog (116)) also had no effect on torsinA or torsinA⌬E stability or glycosylation (data not shown). Mutations in other ER resident enzymes, such as select protein-disulfide isomerases, also showed no effect on torsinA or torsinA⌬E (data not shown). Furthermore, similar to mammalian cells, the defects in N-linked glycosylation could also be induced by supplementing the growth medium with DTT (117), but this also occurred in an ire1⌬ yeast mutant in which the unfolded protein response cannot be induced (data not shown) (118). These results strongly support our conclusion that compromised torsinA and torsinA⌬E stability and folding are direct consequences of the lack of Kar2/ BiP, Scj1, and Lhs1 activity.
TorsinA and TorsinA⌬E Are Degraded by the Proteasome and Vacuole When Kar2/BiP Function Is Disabled-In mammalian cells, torsinA and torsinA⌬E are degraded in the lysosome via autophagy, but torsinA⌬E is also degraded by ERAD (58, 62).
To determine the role of the proteasome and the vacuole (the lysosome equivalent in yeast) during the turnover of torsinA and torsinA⌬E when Kar2/BiP function is compromised in yeast, we deleted PDR5 or PEP4 in the kar2-1 background. The PDR5 deletion allows for efficient MG132-mediated inhibition of proteasome function (119,120), and the PEP4 deletion inhibits nearly all vacuolar protease activity (121). We monitored torsinA and torsinA⌬E degradation by CHX chase in the kar2-1 pdr5⌬ strain treated with solvent (DMSO) or MG132 and found that both torsinA and torsinA⌬E were partially but significantly stabilized by MG132 (73% versus 58% torsinA and 80% versus 61% torsinA⌬E remained after 90 min in kar2-1 pdr5⌬ cells treated with MG132 versus DMSO, respectively) (Fig. 6A).
To determine whether vacuolar proteases contribute to torsinA and torsinA⌬E degradation, we performed a CHX chase in the kar2-1 pep4⌬ strain (Fig. 6B) and found that torsinA was stabilized but torsinA⌬E was only partially, yet still significantly stabilized (91% versus 55% torsinA and 76% versus 59% torsinA⌬E remained after 90 min in the kar2-1 pep4⌬ and the kar2-1 PEP4 strains, respectively) (Figs. 3A and 6B). Of note, torsinA degradation in the kar2-1 strain was significantly more dependent on PEP4 than torsinA⌬E (p Ͻ 0.04), similar to what has been observed in mammalian cells (58,62). No evidence of torsinA or torsinA⌬E aggregates was observed during chases in the absence of proteasome or vacuolar function (data not shown).
Defect in Kar2/BiP Function Alters TorsinA and TorsinA⌬E Membrane Extraction-The Kar2/BiP-dependent defects in torsinA and torsinA⌬E N-linked glycosylation and the observed destabilization of the proteins when Kar2/BiP function was curtailed suggest that this chaperone facilitates the folding of torsinA and torsinA⌬E. One method to monitor a change in the folding of peripheral, membrane-associated proteins is to determine the relative strength of membrane association. In other words, defects in folding might lead to greater insolubility, which would be evidenced by reduced membrane extraction in the presence of chaotropic agents (29,101,102,122). To this end, we compared torsinA and torsinA⌬E extraction from microsomes prepared from wild-type and kar2-1 strains after treatment with alkaline sodium carbonate buffer (pH ϳ11.5) (Fig. 7A). As expected, torsinA and torsinA⌬E associated with the pellet fraction after microsomes prepared from the kar2-1 strain were incubated at pH 6.8 (Fig. 7A, lanes 1, 2, 5, and 6). Although incubation with sodium carbonate released the majority of torsinA and torsinA⌬E from the membrane in the wild-type strain ( Figs. 2A and 7C), alkaline extraction was significantly less efficient in kar2-1 microsomes (64% versus 25% torsinA and 88% versus 31% torsinA⌬E remained in the membrane fraction in the kar2-1 and wild-type strains, respectively; p Ͻ 0.001) (Fig. 7, A, lanes 3, 4, 7, and 8, and C). Interestingly, torsinA⌬E associated more avidly with the membrane than torsinA in the kar2-1 strain (p Ͻ 0.02), suggesting that the presence of the ⌬E mutation causes a subtle folding defect, which is exacerbated when Kar2/BiP is disabled.
Kar2/BiP helps maintain substrate solubility, and in its absence substrates such as pro-␣-factor, CPY*, and CPY oligomerize or aggregate (29,101,102). To test if decreased Kar2/BiP function also triggers torsinA and torsinA⌬E aggregation, we extracted torsinA and torsinA⌬E from membranes collected from wild-type and kar2-1 cells using the nonionic detergent Triton X-100. In this experiment, torsinA and torsinA⌬E were readily extractable by 1% Triton X-100 from microsomes pre-pared from wild-type cells (Fig. 7B, lanes 1-4). However, in microsomes prepared from the kar2-1 strain, there was a significant increase in the fraction of torsinA and torsinA⌬E that remained associated with the pellet (51% versus 14% torsinA and 48% versus 18% torsinA⌬E in the kar2-1 versus the wildtype strains, p Ͻ 0.008) (Fig. 7, B, lanes 5-8, and C). All of the differentially glycosylated species of torsinA and torsinA⌬E in the kar2-1 strain were extracted to a similar degree both by detergent and alkali, indicating that the glycosylation state had no effect on torsinA and torsinA⌬E solubility (Fig. 7). Together, these results suggest that compromised Kar2/BiP function leads to folding alterations that decrease torsinA and torsinA⌬E solubility, which in turn initiates torsinA and torsinA⌬E aggregation. The data also provide further support for our hypothesis that Kar2/BiP is a torsinA chaperone in yeast.

Mutations in TorsinA Functional Motifs Exhibit a Synergistic Interaction with the ⌬E Mutation and Destabilize TorsinA-
Mutation of functional domains in torsinA has provided valuable information on torsinA function and on the role of the ⌬E allele on disease-associated phenotypes. For example, mutations in torsinA's C-terminal cysteines prevent a redox-dependent conformational change in torsinA that is critical for torsinA function (63). The ⌬E mutation yields the same phenotype in torsinA as the cysteine mutant, suggesting that it also impacts local conformation (63). Furthermore, some intragenic mutations may mimic the effects of cellular or ER stresses in torsinA⌬E-expressing cells, potentially replicating events that lead to the manifestation of the DYT1 mutation and EOTD development. Indeed, mutations in the N-linked glycosylation sites (Fig. 1A) can alter torsinA⌬E subcellular localization and inclusion formation in mammalian cells (106). Thus, an examination of secondary mutations in the context of the ⌬E allele in yeast may unveil previously undiscovered ⌬E phenotypes. In our yeast system and in some mammalian cell studies, no difference was observed in the stability of torsinA and torsinA⌬E (Figs. 3-5) (63). However, torsinA⌬E showed enhanced membrane association compared with torsinA in the kar2-1 strain (see above, Fig. 7). The data suggest that the lack of key chaperones can exacerbate subtle effects of the ⌬E allele, which appear otherwise indistinguishable from the wild-type allele by our methods. We reasoned that torsinA⌬E instability may also be magnified if the ⌬E allele is combined with additional intragenic mutations. Therefore, we set out to explore the effects on torsinA and torsinA⌬E stability when other mutations were engineered into distinct functional domains in torsinA.
TorsinA has two asparagines to which N-linked glycans are added (Asn-143 and Asn-158; Fig. 1A) (106,123). To examine how the presence of the ⌬E mutation affects torsinA stability in combination with mutations in the first (N143Q) or second (N158Q) N-linked glycosylation sites, we performed CHX chases in a wild-type strain (KAR2) transformed with the tor-sinA and torsinA⌬E alleles carrying either of these secondary mutations (Fig. 8A). The N143Q mutation had no significant effect on the stability of torsinA or torsinA⌬E (89% torsinA-N143Q versus 81% torsinA and 84% torsinA⌬E-N143Q versus 84% torsinA⌬E remained after 90 min) (Fig. 9B). The N158Q mutation also had no significant effect on torsinA stability (74% of torsinA-N158Q versus 81% torsinA remained after 90 min) (Figs. 8A and 9B). However, the N158Q mutation significantly decreased torsinA⌬E levels (42% torsinA⌬E-N158Q versus 84% torsinA⌬E remained after 90 min, p Ͻ 0.0004) (Figs. 8A and 9B). These results are in accordance with data suggesting a more prominent role of glycosylation at position Asn-158 in torsinA⌬E in mammalian cells (106) and indicate that glycosylation at position Asn-158 is critical to stabilize torsinA⌬E but not torsinA.
The integrity of the Walker motifs is essential for the function of AAA ϩ ATPases (48, 124 -126). Indeed, mutations in the Walker motifs that are predicted to disrupt ATP binding (K108A) or hydrolysis (E171Q) (Fig. 1A) (48,63,124,125,127) can alter torsinA and/or torsinA⌬E subcellular localization, inclusion formation, and/or binding to interacting partners in mammalian cells (56,60,61,63,(127)(128)(129). To test how the ⌬E mutation affects torsinA stability in the presence of mutations in the Walker motifs, we performed CHX chase experiments in a wild-type strain (KAR2) transformed with torsinA and torsinA⌬E alleles carrying either of these additional mutations (Fig. 8C). The K108A mutation led to a significant decrease in the stability of both torsinA and torsinA⌬E (40% torsinA-K108A versus 81% torsinA and 20% torsinA⌬E-K108A versus 84% torsinA⌬E remained after 90 min, p Ͻ 0.0007) (Figs. 8C FIGURE 7. Kar2/BiP maintains torsinA and torsinA⌬E membrane association and solubility. Microsomes prepared from log phase wild-type (KAR2) or kar2-1 yeast cells grown at 28°C and that express torsinA (T) or torsinA⌬E (⌬E) were incubated for 30 min in ice in pH 6.8 buffer or pH ϳ11.5 carbonate buffer (A) or in a pH 6.8 buffer containing 1% Triton X-100 (B). After centrifugation at 100,000 ϫ g, aliquots of proteins from the supernatant (S) or the pellet (P) fractions were separated by SDS-PAGE, and the presence of torsinA and torsinA⌬E was detected by Western blot analysis and quantified. Significant p values for the carbonate extraction are as follows: *, p Ͻ 0.001 for KAR2 torsinA versus kar2-1 torsinA; p Ͻ 0.001 for KAR2 torsinA⌬E versus kar2-1 torsinA⌬E; ¥ , p Ͻ 0.02 for kar2-1 torsinA versus kar2-1 torsinA⌬E; and for Triton extraction are as follows: *, p Ͻ 0.008 KAR2 torsinA versus kar2-1 torsinA; p Ͻ 0.002 for KAR2 torsinA⌬E versus kar2-1 torsinA⌬E. C, graph shows the means Ϯ S.E. of the fraction of torsinA or torsinA⌬E remaining in the pellet from at least three independent experiments. and 9B). On the contrary, although the E171Q mutation appeared to increase the stability of torsinA (94% torsinA-E171Q versus 81% torsinA remained after 90 min), it significantly destabilized torsinA⌬E (55% torsinA⌬E-E171Q versus 84% torsinA⌬E remained after 90 min, p Ͻ 0.0007) (Figs. 8C and 9B). Therefore, mutations in the Walker-B motif led to opposite effects on torsinA stability, depending on the presence of the ⌬E mutation. Importantly, both the K108A and E171Q mutations had a greater destabilizing effect on torsinA⌬E than on torsinA (p Ͻ 0.03) (Fig. 9B). K108A-mediated destabilization in torsinA and torsinA⌬E and the stabilizing effect of the E171Q mutation in torsinA are in agreement with results for other AAA ϩ ATPases (124,126,130). Overall, these experiments support our hypothesis that the ⌬E mutation harbors a subtle folding defect and that this defect is amplified in the presence of additional mutations that affect torsinA conformation.

JOURNAL OF BIOLOGICAL CHEMISTRY 12739
Kar2/BiP Plays a Dual Role during TorsinA and TorsinA⌬E Biogenesis-Loss of Kar2/BiP led to the formation of incompletely glycosylated forms of torsinA and torsinA⌬E and to a decrease in protein stability (Fig. 3). In our assays, all the differentially glycosylated species in the kar2-1 strain behaved similarly (Figs. 3 and 7). However, our results also indicate that robust glycosylation is required for torsinA⌬E stability in a wild-type strain (Fig. 8A). These observations can be reconciled by previous results in yeast that demonstrate that chaperone requirements can vary depending on a protein's glycosylation state (133,134). Thus, we hypothesized that although Kar2/BiP is required for stabilizing fully glycosylated torsinA and torsinA⌬E (Fig. 3A), it is also required to degrade the monoand unglycosylated torsinA⌬E species. To test this model, we performed CHX chases of torsinA and torsinA⌬E containing the N143Q or N158Q mutations in the kar2-1 strain (Figs. 8B and 9B). In agreement with our model, the lack of Kar2/BiP function significantly stabilized torsinA⌬E-N158Q (69% torsinA⌬E-N158Q versus 42% remained after 90 min in the kar2-1 or the wild-type strains, respectively, p Ͻ 0.002) (Figs. 8B and 9). We also found that torsinA-N143Q was significantly destabilized in the kar2-1 strain (75% versus 89% torsinA-N143Q remained after 90 min in the kar2-1 and the wild-type strains, respectively, p Ͻ 0.03) (Figs. 8B and 9B). This result suggests that a lack of N-linked glycans at N143 in torsinA does not impact the role of Kar2/BiP in maintaining torsinA stability. On the contrary, although there was a clear trend toward lower stability in the kar2-1 strain, neither torsinA-N158Q nor torsinA⌬E-N143Q stability was significantly altered in the kar2-1 strain (65% versus 74% torsinA-N158Q and 76% versus 85% torsinA⌬E-N143Q remained after 90 min in the kar2-1 or the wild-type strains, respectively). This result suggests that either lack of N-linked glycans at Asn-158 in torsinA and at Asn-143 in torsinA⌬E abrogates Kar2/BiP dependence for stability or that these species are protected from degradation in the absence of Kar2/BiP (Figs. 8 and 9). Thus, combined with the data provided above (Figs. 3 and 8, A and C), Kar2/BiP can play either a pro-degradative or a pro-folding role depending on the glycosylation state of torsinA and torsinA⌬E.
To determine whether Kar2/BiP's dual role was only associated with torsinA's glycosylation state or was more generally associated with protein folding, we measured the stability of torsinA and torsinA⌬E Walker-A and Walker-B mutants in the kar2-1 strain (Fig. 8D). Consistent with the acquisition of an altered conformation, the degradation of torsinA-K108A, torsinA⌬E-K108A, and torsinA⌬E-E171Q was significantly attenuated in the kar2-1 strain (57% versus 40% torsinA-K108A, 66% versus 20% torsinA⌬E-K108A, and 72% versus 55% torsinA⌬E-E171Q remained after 90 min in the kar2-1 and wild-type strains, respectively, p Ͻ 0.04) (Figs. 8D and 9B). On the contrary, torsinA-E171Q was significantly destabilized in kar2-1 yeast compared with the wild-type strain (59% versus 94% torsinA-E171Q remained after 90 min in the kar2-1 and wild-type strains, respectively, p Ͻ 0.01) (Figs. 8D and 9B). Taken together, these results indicate that Kar2/BiP plays a dual role during torsinA degradation, which is linked to the folding or conformational states of torsinA. BiP Also Promotes TorsinA and TorsinA⌬E Stability in Mammalian Cells-To test whether the mammalian BiP homolog also contributes to torsinA and torsinA⌬E biogenesis in mammalian cells, we first compared the levels of torsinA and torsinA⌬E in HeLa cells in which BiP expression was silenced using siRNA (Fig. 10A). We designed a protocol using the BiPspecific siRNA 077 to obtain a reduction of BiP levels by ϳ55% compared with a control siRNA, to prevent defects in cellular viability caused by greater BiP silencing (data not shown). These data are consistent with previous observations that BiP levels cannot be reduced to Ͻ40% of wild-type levels (2). Also consistent with previous data (135,136), the depletion of BiP was accompanied by the induction of Grp94, which is another major ER chaperone (Fig. 10A). Under these conditions, the steady state levels of both torsinA and torsinA⌬E were decreased by 70 -75% compared with the negative control (Fig.  10). Similar results were observed using another BiP-targeted siRNA (data not shown). These results suggest that BiP affects the stability of both torsinA and torsinA⌬E in mammalian cells.
To confirm these data, we implemented a second approach to deplete BiP. To this end, we performed a torsinA pulse-chase in HeLa cells transiently expressing torsinA or torsinA⌬E in which BiP levels were acutely reduced by means of the AB5 subtilase cytotoxin, also known as the SubAB toxin (Fig. 11). SubAB is a bacterial serine protease that cleaves and inactivates BiP rapidly and specifically (87). Indeed, only 45 min after the toxin was added to the culture medium (t ϭ "0" min of the chase), BiP levels dropped to ϳ35% of the levels of BiP in cells treated with the catalytically inactive toxin, SubAA272B (Fig.  11A). Moreover, BiP could not be detected by Western blot in toxin-treated cells 1 h into the chase, whereas the catalytically inactive toxin had no effect on BiP (Fig. 11A). As anticipated, the reduction of BiP upon SubAB treatment also significantly destabilized torsinA (26% versus 44% of torsinA remained after a 4-h chase in cells treated with SubAB or SubAA272B, respectively, p Ͻ 0.02) (Fig. 11, A and B). Similar results were observed if only the fully glycosylated bands were quantified (data not shown). Although the effect was less pronounced, there was also a difference in torsinA⌬E stability in the presence of active or inactive toxin (31% versus 39% torsinA⌬E remained after a 4-h chase in cells treated with SubAB or SubAA272B, respectively) (Fig. 11, A and B). We also noticed that the levels of unglycosylated torsinA and torsinA⌬E were higher at the beginning of the chase after toxin treatment compared with inactive toxin treatment or no treatment (72% versus 53% of unglycosylated torsinA and torsinA⌬E in SubAB versus SubAA272B-treated cells or -untreated cells, respectively) (Fig. 11, A and C, data not shown and (63)). Because the unglycosylated species appears to mature into the fully glycosylated form of torsinA, the more rapid disappearance of torsinA is likely not due to the selective turnover of the unglycosylated protein (Fig. 11A). In fact, the unglycosylated form degraded similarly irrespective of whether the sample was treated with SubAB or SubAA272B (data not shown). Furthermore, because of the acute nature of the subtilase treatment, the effect on torsinA maturation and stability is unlikely to be caused by the up-regulation of other chaperones. In contrast to the knockdown experiments, GRP94 levels did not vary during the Ͼ4-h treatment with the toxin (Fig. 10 and data not shown). Taken together, the results using toxin-mediated BiP depletion are in accordance with the results using RNAi-mediated depletion of BiP in mammalian cells and with the experiments in the yeast genetic system, and establish BiP as a torsinA chaperone.

DISCUSSION
EOTD, one of the most common forms of inherited primary dystonias, is associated with a dominant ⌬E302/303 mutation in the gene DYT1, which encodes the ER/nuclear envelope localized AAA ϩ ATPase torsinA (43,44). However, the variable phenotypic manifestation and low penetrance of the ⌬E muta- The graphed values were normalized as follows. The levels of BiP, torsinA, and torsinA⌬E for each replicate in each experiment were normalized first to the corresponding actin levels. Then the actin-normalized value from one siRNA control replicate was used to normalize the values of all other replicates. This procedure was performed for each torsinA and torsinA⌬E experiment independently. Actin was used as loading control. Grp94 levels were also monitored by Western blot. tion (45) indicate that additional environmental or genetic factors must play a role in EOTD onset. One hypothesis is that ⌬E impacts torsinA folding, and thus modifies torsinA's half-life, solubility, and/or interaction with chaperones, as evident in other genetic "conformational" diseases (34). Cellular chaperone levels are sensitive to genetic and environmental insults. Thus, altering chaperone levels could augment the phenotypes associated with the ⌬E mutation. By understanding the chaperone microenvironment required for the folding, stabilization, and degradation of wild-type torsinA and of the disease-associated torsinA⌬E variant, we hope to identify cellular factors that can be targeted with therapeutics. In this work, we report for the first time on ER chaperones required for torsinA and torsinA⌬E biogenesis.
We discovered that Kar2/BiP is required to maintain the N-glycosylation and solubility of torsinA and torsinA⌬E in a yeast model (Figs. 3 and 7) and that in the absence of Kar2/BiP, torsinA and torsinA⌬E are destabilized and degraded (Figs. 3  and 6). In addition, we obtained preliminary evidence that the proteins are ubiquitinated (data not shown), which is in agreement with a recent study showing that the E3 ubiquitin ligase, FBG1, is involved in the degradation of torsinA and torsinA⌬E (137). We also found that Kar2/BiP plays a pro-degradative role when the stability of torsinA and torsinA⌬E is compromised (Figs. 8 and 9). Finally, we provide evidence that BiP supports torsinA biogenesis in human cells (Figs. 10 and 11). Therefore, to our knowledge, BiP is the first ER molecular chaperone that has been reported to facilitate the maturation of torsinA and torsinA⌬E.
The ⌬E mutation affects torsinA's redox-sensitive conformational changes, potentially by altering the local structure of an ␣-helix in the C-terminal subdomain (63,94). Nevertheless, several studies have failed to uncover significant structural differences between torsinA and torsinA⌬E, suggesting that the ⌬E mutation confers a more subtle effect on torsinA's structure (91,94). This idea is supported by our observations that the ⌬E mutation increases torsinA membrane association in the absence of functional Kar2/BiP (Fig. 7) and that the ⌬E mutation and mutations in several functional motifs exhibit a synergistic interaction that alters torsinA stability (Figs. 8 and 9). These data also suggest that the effect of the ⌬E mutation could be amplified through secondary genetic or environmental insults, potentially mimicking the sequence of events that is required to manifest the dominant but poorly penetrant ⌬E mutation (45,91,94). Indeed, the secondary mutations we incorporated might mimic alterations in the availability or function of chaperones, enzymes, or interacting partners, which in turn could play a role in EOTD onset and/or progression.
N-Linked glycosylation is a post-translational modification that is intimately involved in ER protein quality control (4,138). Defects in N-glycosylation are associated with several human diseases (139 -141). Interestingly, we found that defects in Kar2/BiP or its associated Hsp40s (Scj1 and Jem1) altered torsinA and torsinA⌬E N-glycosylation, but this defect was uncoupled from destabilization ( Figs. 3 and 4). How do these chaperones impact N-linked glycosylation? The kar2-1 (BiP mutant) allele has not been shown to affect the glycosylation of other secreted substrates; therefore, a general effect on the glycosylation machinery can be ruled out. The efficiency of N-linked glycosylation is not only determined by the amino acid sequence to which the modification is appended (the NX(S/T) tripeptide motif) but is also affected by secondary and tertiary structure, hydrophobic collapse, and disulfide bond formation (142). Notably, torsinA has two asparagines that can be glycosylated, Asn-143 and Asn-158 ( Figs. 1 and 3) (65, 123). It is possible that the presence of the cysteine at position 162 (Fig. 1A) may affect glycosylation at the second acceptor site, N158VS, by altering the local conformation through the formation of an intramolecular disulfide bond (143) or by forming a covalent bond with the oligosaccharyltransferase during translocation (144). Any effect on glycosylation may be potentiated if the polypeptide is improperly folded due to compromised chaperone function, e.g. when torsinA is expressed in the kar2-1 yeast. This hypothesis is supported by the observations that treatment with the reducing agent, DTT, or deletion of the N-terminal hydrophobic domain in torsinA (Fig. 1A) leads to a similar glycosylation defect observed in the kar2-1 strain. 5 Furthermore, torsinA interacts with the chaperone and lectin calnexin in mammalian cells (60). S. cerevisiae lacks a true calnexin (145). Calnexin and calreticulin have been shown to compete with BiP for binding to specific glycoproteins (112), but calnexin also binds nonglycosylated substrates (146). This may explain why compromised Kar2/BiP function in yeast leads to a co-translational folding defect (evidenced by the appearance of different glycosylated species) that is not observed in mammalian cells (Figs. 3 and 9). Overall, our results indicate that torsinA and torsinA⌬E glycosylation is sensitive to alterations in folding caused not only by the absence of specific chaperones but also by environmental insults and by intragenic mutations.
Protein aggregation can lead to profound effects on cell viability (147). A critical role played by molecular chaperones, including Kar2/BiP, is to maintain the solubility of nascent polypeptides until their final conformations are achieved (29,148). The membrane extraction of torsinA and torsinA⌬E with carbonate or a nonionic detergent was severely impaired in the absence of Kar2/BiP (Fig. 7). In principle, this phenotype could be due to a translocation defect, but translocation of the proteins into the ER was unaffected in the kar2-1 strain (Fig. 3C). More likely, enhanced membrane association arose from conformational changes that expose hydrophobic regions, which would enhance membrane association or which could lead to aggregation (122). Alternatively, torsinA and torsinA⌬E might associate with other proteins that aggregate in the kar2-1 strain (29). This model would be consistent with the proposed chaperone function of torsinA (53)(54)(55)149), although no effect of torsinA or torsinA⌬E on the ERAD of select substrates was observed in yeast (data not shown), and only terminally misfolded ERAD substrates have been reported to aggregate in the kar2-1 yeast (29).
The introduction of mutations in torsinA and torsinA⌬E generally led to a more profound loss of stability in torsinA⌬E than in torsinA (Figs. 8 and 9). Lack of N-linked glycosylation at Asn-158 specifically affects torsinA⌬E, both in mammalian cells and in our yeast model (Fig. 8) (106). Defects in glycosylation can impact local folding, alter the interaction with lectins and chaperones, and reduce protein solubility (138). Thus, it is possible that the glycan at Asn-158 is required for the interaction of torsinA and torsinA⌬E with chaperones or lectins, which torsinA⌬E may require for folding. This effect, in turn, would lead to increased degradation of torsinA⌬E but not torsinA. Alternatively, altered folding caused by the lack of a glycan at Asn-158 may impact the ability of torsinA⌬E to form hexamers, which is the typical oligomeric state assumed by AAA ϩ ATPases, including torsinA (48,61,150). These unassembled torsinA⌬E monomers may be targeted for degradation via the ERAD pathway (10,151). This model may also explain why mutating the Walker-A motif triggers torsinA and torsinA⌬E degradation, whereas mutating the Walker-B motif stabilizes torsinA (Figs. 8 and 9). The K108A mutation in the Walker-A motif of AAA ϩ ATPases not only prevents ATP binding, it also impairs substrate interaction and hexamer formation (48,152). Conversely, the E171Q mutation in the Walker-B motif locks the protein in an ATP-and substratebound conformation and promotes hexamer stabilization (48,152). In the case of torsinA⌬E, the C-terminal conformational defect and the lack of binding to interacting partners required for ATPase activity (63, 127) may prevent efficient hexamerization, a phenotype that could be aggravated by the E171Q mutation. This phenomenon would explain why the E171Q mutation destabilizes torsinA⌬E (Fig. 9B).
Besides protecting nascent polypeptides and aiding in their folding, chaperones can also identify and target terminally misfolded intermediates for degradation (4,9). Thus, depending on the substrate, chaperones can have a pro-folding and/or a prodegradative role (109). For example, mammalian Hsp/c70 is a key pro-folding factor for CFTR that can also act as a pro-degradative chaperone, targeting CFTR for ERAD (153). The loss of stability of torsinA and torsinA⌬E containing additional mutations in functional motifs also uncovered a dual role of Kar2/BiP on the folding and stabilization of these proteins (Figs. 8 and 9). Kar2/BiP normally aids in folding torsinA and torsinA⌬E (Figs. 3 and 8 -11) but is instead a pro-degradative factor when the stability and folding of torsinA and torsinA⌬E are impaired by mutations in specific functional motifs (Figs. 8  and 9). This dual role of Kar2/BiP is not unprecedented. Kar2/ BiP functions as a pro-folding chaperone for the vacuolar protease CPY but as a pro-degradative factor of the ERAD substrate CPY* (100,102). Moreover, BiP aids in the folding and assembly of monomers into oligomers, while simultaneously contributing to the degradation of the unassembled, potentially misfolded monomeric subunits (154 -157). Thus, Kar2/BiP may help torsinA and torsinA⌬E to fold (Fig. 3) and hexamerize, while also facilitating the degradation of the torsinA and torsinA⌬E monomers that fail to assemble into hexamers due to, for instance, the presence of the additional mutations mentioned above (Figs. 8 and 9). Alternatively, Kar2/BiP may simply recognize misfolded features of the mutated torsinA and torsinA⌬E and aid in their degradation. These results illustrate the complex relationship between chaperones and their substrate clients, and underscore the delicate equilibrium between folding and degradation that ultimately dictates the fate of folding intermediates.
Because BiP is associated with the development of diseases as diverse as cancer and Parkinson disease (20, 31-33, 158, 159), there is growing interest in developing small molecule modulators of BiP. Chemical modulators of BiP activity have been developed and tested in vitro (87,160), and small molecules that enhance the activity of cytosolic Hsp70 exist (161), which may be adapted in the future to act on BiP. Alternatively, the expression levels of BiP can be modified, a procedure that has shown therapeutic potential both in vitro and in mouse models (162)(163)(164)(165)(166). Because our results demonstrate that BiP stabilizes torsinA and torsinA⌬E, and because patients afflicted with EOTD are heterozygous for the ⌬E mutation, the design of effective therapies may need to combine the modulation of BiP activity/expression levels with the modulation of secondary targets that selectively affect torsinA⌬E folding such as the glycosylation machinery (Fig. 8). A deeper understanding of the molecular and environmental factors that alter the biochemical properties of torsinA and of the factors that participate in torsinA folding and degradation will provide new clues to define the etiology of EOTD and may identify new therapeutic targets to prevent or treat the disease.