The Early Onset Dystonia Protein TorsinA Interacts with Kinesin Light Chain 1* □ S

Early onset dystonia is a movement disorder caused by loss of a glutamic acid residue (Glu 302/303 ) in the car-boxyl-terminal portion of the AAA (cid:1) protein, torsinA. We identified the light chain subunit (KLC1) of kinesin-I as an interacting partner for torsinA, with binding occur-ring between the tetratricopeptide repeat domain of KLC1 and the carboxyl-terminal region of torsinA. Co-immunoprecipitation analysis demonstrated that wild-type torsinA and kinesin-I form a complex in vivo . In cultured cortical neurons, both proteins co-localized along processes with enrichment at growth cones. Wild-type torsinA expressed in CAD cells co-localized with endogenous KLC1 at the distal end of processes, whereas mutant torsinA remained confined to the cell body. Subcellular fractionation of adult rat brain revealed torsinA and KLC associated with cofractionating membranes, and both proteins were co-immunoprecipi-tated after cross-linking cytoplasmically oriented proteins on isolated rat brain membranes. These studies suggest that wild-type torsinA undergoes anterograde transport along microtubules mediated by kinesin and may act as a molecular chaperone

The dystonias represent a varied group of movement disorders characterized by twisted and contracted movements and postures throughout the body (1), with early onset dystonia being the most common and severe form of the inherited dystonias. Onset of symptoms occurs in a limb between 5 and 20 years of age, and symptoms generalize by spreading to other limbs, physically incapacitating most patients within approximately 5 years (2,3). The disease follows an autosomal dominant mode of inheritance with reduced penetrance of 30 -40% (4). There are no signs of neurodegeneration in the brains of these patients (5,6), suggesting that neuronal dysfunction, rather than loss of neurons, underlies disease symptoms.
Most cases of early onset dystonia are caused by the deletion of 3 base pairs (⌬GAG) in the DYT1 gene that encode a single glutamic acid residue, Glu 302 or Glu 303 , near the C-terminal region of torsinA (7). This protein, which shares conserved domains with the AAA ϩ (HSP/Clp-ATPase-AAA) superfamily of chaperone-like proteins (8), is expressed in a broad range of tissues in mammals. Within the human brain, expression levels are highest in dopaminergic neurons of the substantia nigra (9). The DYT1 gene is a member of a gene family that includes three other related genes: TOR1B, adjacent to DYT1 on chromosome 9q34, encoding torsinB (70% identity); and TOR2A and TOR3A on chromosomes 9q34 and 1q24, respectively, encoding TORP1 and TORP2 (ADIR) (10), sharing 40% identity with torsinA (11). Following transient transfection in mammalian cells, wild-type torsinA localizes predominantly to the endoplasmic reticulum (ER), 1 as does endogenous torsinA (12), whereas mutant (⌬GAG) torsinA accumulates in the perinuclear region (13,14) and in large membranous structures in the cytoplasm (13)(14)(15)(16).
Functions of the torsin gene family have yet to be elucidated. In general, AAA ϩ proteins are critical to the assembly, disassembly, and operation of protein complexes and often form homo-oligomeric ring structures with substrate proteins or nucleic acids "threaded" through the hole in the ring (17). They participate in a wide variety of cellular processes, including membrane fusion, organelle biogenesis, movement along microtubule tracks, proteolysis and DNA replication. The C-terminal helical subdomain of AAA ϩ proteins in particular has been shown to be important for substrate interaction, and the contacts that this region forms in crystal structures of these proteins suggest a role in oligomerization (18). For some AAA ϩ proteases, mutations in this domain disrupt interaction with partner proteins in a dominant negative manner (19,20). Both dystonia-associated mutations identified in the DYT1 gene are in-frame deletions within the C-terminal region, including the common GAG deletion and an 18-bp deletion (Phe 323 -Tyr 328 del) in one patient with early onset dystonia with myoclonic and tic-like features (21). These deletions could have dominant negative effects on the function of a torsin AAAϩ oligomer (22).
To identify proteins that interact with torsinA, we undertook a yeast two-hybrid screen using the predicted C-terminal ␣-helical subdomain of human wild-type torsinA as a bait. Two isoforms of kinesin light chain (KLC) 1, KLC1-B and KLC1-C, were identified as specifically interacting proteins, and their interaction was confirmed by affinity precipitation assays, co-immunoprecipitation, and cross-linking experiments. KLC1 is part of the heterotetrameric motor protein kinesin-I (23) and is thought to be involved in cargo binding and/or regulation of kinesin-I activity (24,25). We show that KLC1 is a physiologically relevant binding partner for torsinA, with both proteins co-localizing in cultured primary cortical neurons and both concentrated in neuronal growth cones. Our results provide the first report of a direct interaction between torsinA and another protein.
GST in Vitro Pull-down Assays-HEK 293T/17 cells, transiently transfected with mammalian expression constructs encoding fulllength wild-type or ⌬GAG torsinA, were lysed in 0.1% Nonidet P-40 buffer (0.1% Nonidet P-40, 150 mM NaCl, 50 mM Tris, pH 8.0) or RIPA buffer (1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 150 mM NaCl, 50 mM Tris, pH 8.0) including Complete™ protease inhibitor mixture (Roche Applied Science), as indicated. Insoluble material was removed by centrifugation at 16,000 ϫ g for 30 min. Cell extracts were then incubated overnight with 600 pmol of either the indicated GST-KLC1 fusion protein or GST bound to glutathione-Sepharose 4B beads (Amersham Biosciences). Beads were washed five times with PBS. After an additional wash with 0.1% Nonidet P-40 buffer, bound proteins were eluted from the beads by resuspension in SDS-PAGE loading buffer and resolved by SDS-PAGE followed by immunoblotting using standard methods. For pull-downs of endogenous torsinA, PC12 cells were lysed in RIPA buffer.
Transfection and Immunofluorescence of CAD Cells-CAD and HEK293T/17 cells were transiently transfected as described (15). 12 h after transfection, cells were replated onto glass coverslips and switched to differentiation medium (serum-free Dulbecco's modified Eagle's medium/F-12). 48 -72 h later, cells were fixed in 4% paraformaldehyde at room temperature for 20 min. Immunofluorescence of fixed CAD cells was carried out as described (15), except that Alexa 594-conjugated goat anti-mouse and Alexa 488-conjugated goat antirabbit were used as secondary antibodies. Confocal images were captured using LSM 5 Pascal software coupled to a Zeiss LSM Pascal Vario 2 RGB confocal system. For quantification of co-localization of endogenous KLC to torsinA immunoreactive inclusions in CAD cells transiently transfected with ⌬GAG torsinA, 192 transfected cells from two independent experiments showing aggregates were scored visually for co-localization of both proteins by overlapping fluorescence.
Immunoprecipitations-For immunoprecipitations (IPs), PC12 cells, SH-SY5Y cells, or primary cortical neurons were lysed by resuspension in ice-cold RIPA buffer, as above. Lysates were subjected to IP overnight at 4°C. Subsequently, antibody-antigen complexes were precipitated by the addition of Protein A-agarose (for polyclonal antibodies) or Protein G-agarose (for monoclonal antibodies; Roche Applied Science) for 1 h. Beads were washed four times with PBS and resuspended in SDS sample buffer, and samples were resolved by SDS-PAGE followed by immunoblotting. Control human brain (hippocampus) tissue (anonymous donor, no diagnostic abnormality reported) was obtained with IRB approval from Dr. Jean-Paul Vonsattel (Massachusetts General Hospital). Human and embryonic (E16) or adult rat brain tissues were lysed in RIPA buffer and sonicated twice for 45 s at a pulse intensity of 6 using a 550 sonic dismembrator (Fisher), and insoluble material was removed by centrifugation at 16,000 ϫ g for 30 min. IPs were carried out as above.
Fractionation of Rat Brain-For each fractionation, whole fresh brain from one adult rat (Ͼ12 weeks) was homogenized in HB buffer (20 mM HEPES, pH 7.3, 40 mM potassium chloride, 5 mM EGTA, 5 mM magnesium chloride with protease inhibitors). Vesicle-enriched membrane fractions (P3) were prepared by sequential centrifugation at 900 ϫ g for 5 min (P1), 9,000 ϫ g for 10 min (P2), and 120,000 ϫ g avg for 90 min (P3), slightly modified from Ref. 36. The P3 pellet was resuspended in HB buffer containing 0.32 M sucrose and loaded on top of 11 ml of a 0.32-1.9 M continuous sucrose gradient in HB buffer. Gradients were centrifuged at 100,000 ϫ g avg in a SW41 rotor for 16 -18 h, and 24 ϫ 0.5-ml fractions were obtained from the top down. Proteins were precipitated with chloroform-methanol (37) and resuspended in equal volumes of 1ϫ SDS sample buffer before resolution by SDS-PAGE, followed by immunoblotting.
Cross-linking of Rat Brain Membrane Surface Proteins Followed by Immunoprecipitation-For cross-linking of cytoplasmically oriented proteins on P3 vesicle-enriched rat brain membranes, the P3 fraction was resuspended in HB buffer containing 0.32 M sucrose. The watersoluble cross-linker bis(sulfosuccinimidyl)suberate (Pierce) was added to 5 mM final concentration, samples were incubated for 30 min at room temperature, and the reaction was quenched by the addition of 50 mM 2 V. Ramesh, unpublished data.
Tris, pH 7.5, and incubation for 15 min at room temperature. Subsequently, NaCl, Nonidet P-40, deoxycholate, SDS, and protease inhibitors were added to the same concentrations as in RIPA buffer (see above), the indicated antibodies were added, and immunoprecipitations were carried out as described above, except that beads were washed three times with 0.1% Nonidet P-40.
Biotinylation of Rat Brain Membrane Surface Proteins Followed by Precipitation with Immobilized Streptavidin-Total rat brain was homogenized in HB buffer, as above, including 0.32 M sucrose. Membranes were fractionated by sequential centrifugation at 900 ϫ g for 5 min (P1), 9,000 ϫ g for 10 min (P2), and 120,000 ϫ g avg for 90 min (P3), as above. Pellets were resuspended in HB buffer including 0.32 M sucrose, and biotinylation with 0.5 mg/ml Sulfo-NHS-LC-biotin (Pierce) for 30 min at room temperature was carried out essentially as described (38). Reactions were quenched by the addition of 50 mM Tris, pH 9.0, and incubation at room temperature for 15 min. Subsequently, NaCl, Nonidet P-40, deoxycholate, SDS, and protease inhibitors were added to the same concentrations as in RIPA buffer (see above), and samples were incubated with immobilized streptavidin (Pierce) for 1 h. Except for incubation periods during biotinylation and quenching reactions, samples were kept at 4°C throughout. Beads were washed three times in 1% Nonidet P-40 buffer, and samples were resolved by SDS-PAGE, followed by immunoblotting. To evaluate glycosylation, endoglycosidase H or peptide-N 4 -(N-acetyl-␤-D-glucosaminyl) asparagine amidase F digestion of rat brain samples was performed as recommended by the manufacturer (New England Biolabs) prior to SDS-PAGE.
Primary Culture of Rat Cortical Neurons-Brains were obtained from day 17-18 Sprague-Dawley rat embryos, and primary cultures were prepared essentially as described (39) with the following modifications: coverslips were coated overnight in 0.1 mg/ml poly-D-lysine (Sigma); embryonic cortices were minced and dissociated in 2.5 units/ml dispase (Sigma); and dissociated neurons were plated in minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum, 50 units/ml penicillin, and 50 g/ml streptomycin. After 8 days in culture, coverslips were fixed with 4% paraformaldehyde in PBS for 20 min and subsequently processed for immunocytochemistry.
Immunofluorescence of Rat Cortical Neurons-For double labeling with two mAbs, fixed neurons were permeabilized with 0.1% Nonidet P-40 in PBS for 30 min, blocked for 1 h in 10% normal goat serum, and incubated with the first mAb in 1% bovine serum albumin in PBS for 1 h at room temperature, followed by three washes in PBS. Cells were then incubated with Alexa 594-conjugated goat anti-mouse antibody (1:1000) for 1 h, washed as above, incubated with the second mAb in bovine serum albumin/PBS for 1 h at room temperature, and washed as above, followed by incubation with Alexa 488-conjugated goat antimouse (1:2500) for 45 min, and again washed as above. Coverslips were mounted as above, and confocal images were captured using either a Nikon TE 300 microscope and a Bio-Rad MRC 100 laser confocal imaging system (see Fig. 4, A-O) or a Zeiss confocal imaging system, as above (see Fig. 4, P-R).

The KLC1 TPR Domain Binds Directly to the C Terminus of
TorsinA-To identify proteins that interact directly with wildtype human torsinA, amino acids 251-332 of the protein were used as a bait (Fig. 1A) to screen an adult human brain cDNA library using the LexA-based yeast two-hybrid system (26). This sequence contains the site of the GAG deletion and is predicted by alignment to comprise the small C-terminal helical subdomain of the AAA-fold (18). In a screen of 1 ϫ 10 6 independent transformants, 15 positive clones were identified, representing three groups. The 10 clones of one group all encoded overlapping fragments of human kinesin light chain 1 (KLC1), including two differentially spliced isoforms, KLC1-B and KLC1-C (Fig. 1, B and C). The three clones of another group encoded the nonmotor domain of an unconventional myosin, whereas a third group encoded two overlapping cDNA fragments of an uncharacterized protein. 3 The two human KLC1 isoforms were identical to the alternatively spliced KLC1 isoforms B and C recently identified in a human keratinocyte cDNA library (40). Compared with isoform B, isoform C contains an additional nine amino acids (VSMS-VEWNG) close to the C terminus (Fig. 1C). Of the 10 KLC1 clones, one encoded KLC1-B-(229 -547), and nine clones encoded KLC1-C-(112-256). Neither interacted with any of four different control bait plasmids (Table I), demonstrating a specific interaction between torsinA-(251-332) and both KLC isoforms. Since the KLC1-B cDNA fragment encoded only the major part of the TPR domain, but not the N-terminal heptad repeat (Fig. 1B), we conclude that the TPR domain and Cterminal region of KLC1 (amino acids 209 -547) are sufficient to mediate the interaction with the C-terminal part of torsinA (amino acids 251-332).
In yeast colony filter assays (Table I), we observed a difference in the strength of lacZ reporter gene activation and therefore of the interaction between KLC1 isoforms and wild-type or mutant (⌬GAG) forms of torsinA. This apparent difference in binding affinity was supported by yeast liquid culture ␤-galactosidase assays (Supplemental Fig. 1), but not by affinity precipitation assays (see below).
The interaction between endogenous torsinA and KLC1 was 3 C. Kamm, unpublished data.  (8). Bottom, torsinA fragment used as bait construct for the yeast two-hybrid screen. B, top, diagram of the domain structure of full-length human KLC1. Bottom, lines representing sequences of two human KLC1 isoforms, KLC1-C, and KLC1-B, corresponding to clones isolated from the yeast two-hybrid screen. C, diagram of alternatively spliced brain isoforms of KLC1 in the carboxyl terminus (40). further confirmed by in vitro affinity precipitation (GST pulldown) assays (Fig. 2). In order to examine whether endogenous torsin interacts with KLC1, GST-KLC1 fusion proteins were incubated with lysates of rat pheochromocytoma PC12 cells, which express torsin at high levels (12). GST-KLC1-B-(229 -547) and GST-KLC1-C-(112-556) both bound endogenous torsin ( Fig. 2A), with no significant binding of torsin to the GST control or beads alone. GST-KLC1 fusion proteins were also incubated with protein extracts from HEK 293T/17 cells transfected with full-length wild-type torsinA cDNAs. Wild-type torsinA bound to both GST-KLC1 isoforms but only negligibly to GST alone, with apparently somewhat more wild-type torsinA being retained by GST-KLC1-C-(112-556) than by GST-KLC1-B-(229 -547) (Fig. 2B).
In contrast to the yeast liquid culture ␤-galactosidase assays, however, we did not observe a difference in binding affinity of wild-type versus mutant torsinA to KLC1 in these affinity precipitation assays (Supplemental Fig. 2) using protein extracts from HEK 293T/17 cells transfected with full-length wild-type or mutant torsinA cDNAs.
Endogenous TorsinA and Kinesin-I Exist in a Complex in Vivo -To evaluate the association of endogenous torsin and KLC1, we performed immunoprecipitations from a variety of tissue sources (Fig. 3). In human brain (hippocampus), a polyclonal antibody against torsin, TAB1, specifically precipitated torsinA, KLC1, and kinesin heavy chain (KHC) but not KLC2 (Fig. 3A), indicating that these three proteins interact in vivo. Preimmune serum did not precipitate any of these proteins. We also observed co-precipitation of torsinA and KHC in adult rat brain (Fig. 3B), rat primary cortical neurons (Fig. 3C), embryonic (E16) rat brain (Supplemental Fig. 3A), and PC12 cells (Supplemental Fig. 3B). In the converse experiment, immunoprecipitation of kinesin in SH-SY5Y cells co-precipitated torsinA (Fig. 3D). Together, these results indicate that kinesin-I, which in mammals is a heterotetramer consisting of two heavy chains (KHCs) and two light chains (KLCs), is a physiologically relevant binding partner for wild-type torsinA. The fact that we were able to detect co-precipitated KLC only in human brain (Fig. 3A) and not in other tissue sources (Fig. 3,  B-D), whereas KHC co-precipitated with torsinA in all tissues and cells, is probably due to decreased efficiency in detection of KLC in Western blotting procedures (31,41).
TorsinA, KLC, and KHC Co-localize in Primary Neurons and Are Enriched at Neuronal Growth Cones-To investigate whether torsinA and kinesin-I co-localize to the same subcellular compartments in neurons, we performed immunocytochemistry on rat primary cortical neuronal cultures (Fig. 4). TorsinA and KLC showed a high extent of co-localization (Fig.  4, C and F) in cell bodies (Fig. 4, A-C) and neurites (Fig. 4, D-F) with a punctuate, apparently vesicular pattern. TorsinA and KHC also co-localized with a similar distribution (Fig. 4, J-L). Because two monoclonal antibodies and sequential incubation was used for double-labeling (see "Experimental Procedures"), we performed control labelings using only one primary monoclonal antibody (KLC-All (Fig. 4, G-I) or D-MG10 (Fig. 4, M-O)) with both anti-mouse secondary antibodies. No significant background signal was observed in these controls. We observed an enrichment of both torsinA and KLC in the proximal to central growth cone region (Fig. 4, P-R) relative to other segments of neuronal processes. TorsinA showed partial co-localization with GAP-43/neuromodulin in these neuronal cultures (data not shown). GAP-43 is known to be enriched in growth cones (42) and a cargo of kinesin-I (43,44).
Wild-type TorsinA Co-localizes with Endogenous Kinesin at the Tips of Processes, whereas Mutant TorsinA Remains Confined to the Cell Body-Next, we analyzed the subcellular localization of torsinA and FLAG KLC1-C-(112-256) when both proteins were overexpressed simultaneously in the mammalian neuronal-like cell line CAD (28), which has low levels of endogenous torsin (15). Transient co-transfection with both wild-type torsinA and FLAG KLC1-C-(112-256) cDNAs (Fig. 5, A-C) showed partial co-localization of both proteins in the cell body (arrowhead in C), with extensive co-localization in the processes, especially at the distal ends (arrows in C). In contrast, co-transfection with mutant torsinA and FLAG KLC1-C-(112-256) cDNAs (Fig. 5, D-F) yielded large inclusions, which were intensely labeled for torsinA, but not FLAG (Fig. 5, D and  E). TorsinA immunoreactivity was concentrated in the cell body with a reduced amount in the processes. Overexpressed FLAG-tagged KLC1-C-(112-256) was found throughout the cytoplasm in a fine punctuate pattern (Fig. 5E), typical of the distribution of endogenous KLC (45).
In CAD cells transiently transfected only with wild-type torsinA cDNA, torsinA and endogenous KLC also co-localized in the cell body and at the distal end of processes (Fig. 5, G-I), as above. In contrast, transient transfection with mutant torsinA alone again yielded torsinA-positive inclusions confined to the cell body, with loss of torsin enrichment at the distal end of processes (Fig. 5, J-L; arrow in L). In a majority of cells transfected with mutant torsinA (ϳ57% on average), a significant portion of endogenous KLC co-localized with torsinA in these inclusions (Fig. 5, M-O), as did endogenous KHC (data not shown), whereas in other transfected cells, this was not apparent (Fig. 5, J-L). The distribution of microtubules in CAD cells transfected with either wild-type or mutant torsinA, as assessed with an antibody against ␣-tubulin, was not apparently altered (data not shown), arguing against a nonspecific disruption of cytoskeletal elements.

TorsinA and KLC Co-fractionate in Sucrose Gradients of Rat Brain Vesicle-enriched Membranes and Can Be Cross-linked and Biotinylated on the Cytoplasmic Surface of Isolated
Membranes-To directly identify a membrane compartment containing both torsinA and KLC, subcellular fractionation of adult rat brain was carried out by differential centrifugation. Both torsinA and KLC were enriched in a P3 vesicle-enriched membrane fraction, although both were also present in the P1 fraction (Fig. 6A). This distribution of torsin was similar to that observed for another ER protein, BiP/Grp78. The P3 fraction a Symbols indicate degree of lacZ reporter gene activation, as judged by intensity of blue color in yeast colony filter assays. Ϫ, white; (ϩ), weak blue; ϩ, moderately intense blue; ϩϩ, intense blue; ϩϩϩ, very intense blue.
was further resolved on a continuous 0.32-1.9 M sucrose gradient, which revealed significant overlap in the distributions of torsinA, KLC, and the ER marker calnexin (Fig. 6B).
Kinesin is located at the cytoplasmic surface of vesicles (24), whereas torsinA when derived from cultured cells behaves as a luminal ER protein in protease protection assays (12,16). To determine whether a fraction of torsinA in mammalian brain exists on the cytoplasmic surface of vesicle-enriched membranes, we cross-linked cytoplasmically oriented proteins of rat brain P3 membranes using the water-soluble cross-linker bis(sulfosuccinimidyl)suberate, which does not cross membranes, followed by immunoprecipitation under stringent conditions (RIPA buffer). Antibodies against torsinA specifically precipitated higher molecular weight bands, which were recognized by antibodies against KHC, KLC, and torsinA (Fig. 7A), but not by a control antibody against protein-disulfide isomerase, an ER-resident protein (data not shown). The apparent molecular mass of the predominant band, ϳ220 -240 kDa, is consistent with a multiprotein complex containing torsinA (ϳ37 kDa), KLC1 (ϳ64 kDa), and KHC (ϳ130 kDa). Negative controls, which included mouse IgG (Fig. 7A), an unrelated antibody raised against tuberin (6E9), and an antibody against protein-disulfide isomerase (data not shown) did not precipitate these higher molecular weight bands.
To further confirm these findings by an independent method, subcellular fractionation of rat brain by differential centrifugation was carried out essentially as above (except that homogenization was carried out in HB buffer containing 0.32 M sucrose), and P1, P2, and P3 fractions were biotinylated using a water-soluble biotinylation reagent (Sulfo-NHS-LC-Biotin). Subsequent precipitation of biotinylated membrane surface proteins with immobilized streptavidin under stringent conditions precipitated KHC, as expected, and two lower molecular mass variants of immunoreactive torsinA (ϳ30 and 33 kDa), but not the 37-kDa form, from the P1 fraction, which contains a mixture of different types of organelles (Fig. 7B). These two torsinA variants were immunoreactive with two different antibodies against torsinA, D-M2A8 and TAB1. No precipitation of the entirely luminal protein BiP/Grp78 was observed, suggesting that membranes remained largely intact during fractionation. TorsinA (apparent mass 37 kDa) possesses two Nlinked glycosylation sites, matching the consensus sequence

FIG. 3. TorsinA exists in a complex with kinesin-I in vivo.
A, co-IP of endogenous torsinA, KLC and KHC from human brain (hippocampus) using TAB1 anti-torsin polyclonal antibody (IP TAB1). THL, total hippocampal lysate; IP con, control IP with preimmune serum. Western blot was probed with the following mAbs: H2 against KHC, 63-90 against KLC, or D-M2A8 against torsinA. B-D, co-IP of endogenous torsinA and KHC from adult rat brain (B), rat primary cortical neurons (C), or SH-SY5Y human neuroblastoma cells (D). TAB1 and appropriate controls were used for immunoprecipitations, as in A. In SH-SY5Y cells (D), additional reciprocal co-immunoprecipitation was performed using H2 anti-KHC mAb (IP H2). TL, total lysate; IP con M2, control IP with an unrelated monoclonal antibody, M2. Westerns blots were probed with H2, D-MG10, or D-M2A8 mAbs. NX(T/S), Asn 143 and Asn 158 , which have been demonstrated to be glycosylated in cultured cells (16). The two biotinylated torsinA variants identified in this study correspond in apparent mass to nonglycosylated forms of torsinA, with the 30-kDa variant presumably undergoing further proteolytic processing or possibly representing an alternative splice form. Consistent with this hypothesis, treatment with endoglycosidase H or peptide-N 4 -(N-acetyl-␤-D-glucosaminyl) asparagine amidase F reduced the 37-kDa immunoreactive form to a predominant band of ϳ33 kDa, corresponding in apparent mass to the biotinylated 33-kDa variant (Fig. 7C), whereas the 30-and 33-kDa variants were not further reduced in size. Taken together, these data indicate that some portion of torsinA in rat brain is exposed on the cytoplasmic surface of vesicle-enriched membranes and thereby capable of interacting with kinesin-I. DISCUSSION Our results demonstrate that the TPR domain of the KLC1 subunit of kinesin-I interacts directly with the carboxyl-terminal region of torsinA, a novel protein responsible for early onset torsion dystonia. This is the first reported binding partner for torsinA and hence provides critical insights into the function of this protein. KLC1 was initially identified as a binding partner for torsinA in a yeast two-hybrid screen, and the interaction between the two proteins was verified by affinity precipitation assays, co-immunoprecipitation, and cross-linking experiments. The association of these proteins was further confirmed by similar patterns of subcellular fractionation from rat brain tissue and co-localization in primary rat cortical neurons in cell bodies, neurites, and growth cones. Wild type torsinA and KLC1 were distributed together throughout the cytoplasm and processes of neural cells with enrichment at distal ends, whereas mutant torsinA was retained in the cell body in inclusions, frequently altering the distribution of endogenous kinesin by incorporation into these inclusions. These findings support the hypothesis that torsin participates in intracellular trafficking mediated by kinesin microtubule motors and predict a possible molecular etiology of early onset dystonia by interference with intracellular trafficking.
Direct Interaction of TorsinA and KLC-KLC1 is a subunit of kinesin-I, also named conventional kinesin, a plus-end-directed molecular motor protein involved in transport of membranous organelles along microtubules within cells (23). KLC is composed of two notable domains: an N-terminal heptad repeat region that mediates binding to the stalk and tail domains of KHC and a C-terminal domain of six imperfect TPR motifs (25). Two highly conserved domains in different subunits of kinesin-I have been proposed as cargo binding sites: the tail region of KHC (46) and the TPR domain of KLC. Recently, two different classes of proteins have been reported to bind directly to the TPR domain of KLC: the amyloid precursor protein (44) and the c-Jun N-terminal kinase-interacting proteins JIP-1, -2, and -3 (47,48). Here we demonstrate that the KLC1 TPR domain interacts directly with the C terminus of torsinA. TPR motifs occur in proteins as tandem repeat arrays, and TPR-containing proteins often form scaffolds that mediate protein-protein in-  teractions and assemble multiprotein complexes (49) by binding short peptides at the C terminus of their target proteins (50,51). The KLC TPR domain interacts with JIP-1 and JIP-2 through their extreme C termini (47), and its interaction with torsinA also involves the C-terminal region. In our two-hybrid screen from human brain, we isolated two distinct isoforms of KLC1. These KLC1 isoforms are identical to two alternatively spliced isoforms recently identified in a human keratinocyte cDNA library (40). The existence of multiple, differentially spliced KLC isoforms could reflect cell type-specific expression or targeting of kinesin to different classes of organelles.
Mutant Torsins and Mutant Kinesin Interfere with Intracellular Trafficking-Wild-type torsinA expressed in CAD cells co-localized with endogenous KLC at the distal end of processes, whereas CAD cells transfected with mutant torsinA cDNA formed large torsin-positive inclusions confined to the cell body, as previously described (15). These results are consistent with plus-end-directed transport of wild-type, but not mutant, torsinA along microtubules via kinesin and suggest that the binding of torsinA to KLC1 as a cargo is required for proper trafficking of torsinA within the cell. It remains to be determined whether mutant torsinA affects microtubule binding, motor activity, or cargo binding of kinesin.
If torsin binds kinesin, mutations in either might produce similar phenotypes. A torsin homolog in nematodes, OOC-5, is critical for nuclear-centrosome rotation preceding the establishment of spindle orientation and polarity in early cell division in Caenorhabditis elegans (52). Mutations in OOC-5 are maternal effect-lethal and prevent proper localization of the partitioning-defective proteins specifically at the two-cell stage. Mutants in UNC-116, the only KHC gene in C. elegans, also show maternal effect defects in both the first cell division and neuromuscular development (53,54). These overlapping phenotypes support the concept of torsins and kinesins participating in common molecular pathways.
Topographic Distribution of TorsinA-The orientation of torsinA with respect to different membrane compartments in cells remains an enigma. Experiments using cultured cell lines indicate that torsinA is membrane-associated (12,16,55,56) and resides primarily in the lumen of the ER but with little to none of the protein being exposed to the cytoplasmic surface, as assessed by protease protection assays (12,16). In contrast, the present study shows that in homogenates prepared from rat brain, a significant fraction of membrane-associated torsinA is exposed to the cytoplasm, as determined by cross-linking and biotinylation assays. It may be that torsinA exists in two states in nervous tissue, one within the lumen of the ER and one exposed on the cytoplasmic surface of the ER or vesicles derived from it. In the latter case, one would expect torsinA to be nonglycosylated, which is consistent with the size of immunoreactive torsinA observed to be exposed to the cytoplasm at its C-terminal domain. One potential hypothesis for insertion of this nonglycosylated form of torsinA into the ER membrane would be that the amino-terminal signal sequence initiates translocation and is cleaved, but that the following hydrophobic domain (i.e. amino acids 21-40) serves as a stop-transfer se-FIG. 6. TorsinA and KLC co-fractionate in a vesicle-enriched fraction of adult rat brain. A, fractionation profile of rat brain vesicles preparation resolved by differential centrifugation. 50 g of each fraction were analyzed by SDS-PAGE, followed by immunoblotting with the indicated antibodies. SN, total homogenate. Pellet (P1) and supernatant (S1) after 900 ϫ g centrifugation are shown as well as pellet (P2) and supernatant (S2) after 9,000 ϫ g centrifugation and pellet (P3) and supernatant (S3) after 120,000 ϫ g centrifugation. D-M28 specifically recognizes torsinA, whereas TAB1 recognizes both torsinA and torsinB (12). B, torsinA, KLC, and calnexin distributions overlap significantly in a linear 0.32-1.9 M sucrose gradient fractionation of P3 rat brain vesicles. Shown are immunoblots of equal volumes of the 24 collected fractions probed with antibodies against torsinA (D-M2A8), KLC (63-90) or calnexin (SPA-860). quence and transmembrane anchor for a traditional "type I" membrane protein, with the C-terminal domain exposed to the cytoplasm and thus accessible for binding to kinesin.
In these and other studies, it is apparent that torsinA undergoes anterograde transport along neuronal processes to terminals, presumably in association with the ER and/or vesicles (12,(57)(58)(59), apparently mediated by kinesin. Ultrastructural studies in primate brain demonstrated torsinA immunoreactivity associated with small clear vesicles in symmetric synapses in the striatum (59). Further, subcellular fractionation of human and monkey brain found enrichment of torsinA in particulate fractions including the P2 (crude synaptosomal membranes), P3 (light vesicles), and LP1 (synaptosomal membranes) fractions (59), the last of which does not contain detectable levels of the ER marker calnexin (60). Thus, whereas a large portion of torsinA is located in the ER lumen in cultured cells, at least some portion of it appears to be exposed on the cytoplasmic surface of other membranes in mammalian brain.
Does TorsinA Act as a Chaperone to Regulate the Activity of Kinesin-I or the Association of Kinesin-I with Cargo?-There is mounting evidence that the majority of kinesin in the cell is kept in an inactive ground state by binding of the KHC tail domain to the motor domain, which maintains kinesin in a folded conformation (61,62). Factors that have been proposed to regulate kinesin's ability to bind to and move along microtubules include conformational changes induced by the binding of cargo itself, changes in local pH, phosphorylation state, and modulation by chaperone proteins. At least one unidentified membrane-associated chaperone-type protein is required for release of kinesin from membranous cargo (63), and torsinA appears to be a likely candidate, since it can act as a chaperone protein (64,65). Further, blocking of a highly conserved epitope within the TPR domain releases kinesin from membranes and inhibits fast axonal transport (31), and our data indicate that torsinA binds to KLC1 through its TPR domain. Thus, our results raise the possibility that torsinA may act as a molecular chaperone for KLC1, potentially regulating the association of kinesin with cargo and/or activating kinesin by changing its conformation from a folded to an unfolded state.
The ER of most vertebrate cells is spread throughout the cell, FIG. 7. TorsinA and KLC can be cross-linked and biotinylated on the surface of isolated rat brain membranes. A, immunoprecipitation (IP) of a complex of torsinA, KLC, and KHC from isolated rat brain P3 vesicle-enriched membranes after cross-linking using water-soluble cross-linker bis(sulfosuccinimidyl)suberate in buffer HB, including 2 mM ATP and 3 mM EDTA. IPs were performed with either a combination of mAbs D-M2A8 and D-MG10 against torsinA or H2 mAb against KHC. Mouse IgG was used as a negative control for IP. Western blot was probed with H2 mAb for KHC, D-M2A8 mAb for torsinA, or 63-90 mAb for KLC. The lower molecular weight bands visible in the IP lanes most likely represent breakdown products of the cross-linked protein complex. B, precipitation of biotinylated torsinA from the surface of fractionated rat brain membranes. Following biotinylation with Sulfo-NHS-LC-biotin, biotinylated membrane surface proteins were precipitated with immobilized streptavidin. Western blots were probed with H2 mAb against KHC, PA1-104 polyclonal antibody against BiP/Grp78 as a control for an ER luminal protein, or D-M2A8 mAb against torsinA. TBH, total brain homogenate; open arrowhead, 37-kDa immunoreactive band; closed arrowheads, ϳ30and 33-kDa immunoreactive bands; prec, streptavidin-precipitated biotinylated samples. C, deglycosylation of torsinA derived from rat brain. Samples were incubated without deglycosidases (Ϫ) or with endoglycosidase H (Endo H) or peptide-N 4 -(N-acetyl-␤-D-glucosaminyl) asparagine amidase F (PNGase F) (ϩ). Western blot was probed with D-M2A8 mAb against torsinA. TBH, total brain homogenate; P1, pellet after 900 ϫ g centrifugation; open arrowhead, 37-kDa immunoreactive band; closed arrowheads, ϳ30and 33-kDa immunoreactive bands. The two biotinylated torsinA variants, precipitated from the same P1 membrane fraction with immobilized streptavidin (P1 prec) and resolved by SDS-PAGE on a separate gel, are shown for size comparison. including its processes, by kinesin-dependent transport along microtubules (66,67). Given the association of both torsin and kinesin with the ER and their direct interaction with each other, torsin could also be involved in kinesin-mediated anterograde spreading of tubulovesicular membranes within axons.
Significance of TorsinA Interaction with KLC1 in Early Onset Dystonia-Other neurological diseases caused by loss-offunction mutations in kinesin motor proteins include Charcot-Marie-Tooth disease type 2A, a hereditary neuropathy caused by a mutation in KIF1B (68), and one form of autosomal dominant hereditary spastic paraplegia, caused by a mutation in the neuronal KHC gene KIF5A (69). Further, amyloid precursor protein, which has been implicated in the pathogenesis of Alzheimer's disease, interacts directly with the KLC TPR domain (44) and is transported in the same axonal membrane compartment as kinesin-I, ␤-secretase, and presenilin-1 (36). Although torsinA is most highly expressed in dopaminergic nigrostriatal neurons, it is widely expressed elsewhere in brain and other tissues (7,9). However, abnormal movements and postures are the only clinical phenotype reported for patients carrying the DYT1 GAG deletion. This could result from the length of nigrostriatal fibers, which may make them especially vulnerable to deficits in cellular transport systems.
Carriers of the DYT1 GAG deletion who do not experience onset of symptoms by age 28 are very unlikely to develop dystonia later in life (70). The age of vulnerability is associated with a period of motor learning in humans and corresponds to a developmental stage of high synaptic plasticity in primate (71) and rodent (72) neostriatum. In rat brain, kinesin-I is enriched in elongating neurites and growth cones (73,74). Similarly, we found enrichment of both kinesin and torsinA in growth cones of cultured rat cortical neurons. It is tempting to speculate that the interaction and functional integrity of these two proteins could be required for active neurite outgrowth and/or synaptic plasticity, possibly associated with motor learning, in the brain. Development of early onset dystonia in carriers of the DYT1 GAG deletion may be a specific consequence of the altered interaction between mutant torsinA and kinesin in the context of other genetic or environmental factors affecting penetrance in the disease state.