Nuclear Aggregation of Polyglutamine-expanded Ataxin-3

Abstract

Expansion of a polymorphic polyglutamine segment is the common denominator of neurodegenerative polyglutamine diseases. The expanded proteins typically accumulate in large intranuclear inclusions and induce neurodegeneration. However, the mechanisms that determine the subcellular site and rate of inclusion formation are largely unknown. We found that the conserved putative nuclear localization sequence Arg-Lys-Arg-Arg, which is retained in a highly aggregation-prone fragment of ataxin-3, did not affect the site and degree of inclusion formation in a cell culture model of spinocerebellar ataxia type 3. Addition of synthetic nuclear export or import signals led to the expected localization of ataxin-3 and determined the subcellular site of aggregate formation. Triggering a cellular stress response by heat shock transcription factor ΔHSF1 coexpression abrogated aggregation in the cytoplasm but not in the nucleus. These findings indicate that native aggregation-prone fragments derived from expanded ataxin-3 may eventually escape the cytoplasmic quality control, resulting in aggregation in the nuclear compartment.

Introduction

Many human diseases are caused by the continuous expression of misfolded or misfolding-sensitive proteins (for review, see Ref. 1). The related polyglutamine (polyQ)³ family of neurodegenerative diseases is caused by expansion of a variable polyQ segment above a threshold of typically ∼40 residues. These expansions render the otherwise unrelated proteins susceptible to adopt non-native conformations that eventually become toxic or form toxic oligomers and aggregates. Large aggregates arise as intracellular inclusion bodies that are composed mainly of the mutant protein but also contain many other proteins, including transcription factors (2), molecular chaperones, and components of the ubiquitin-proteasome system (3). Several cellular pathways, including transcriptional dysregulation, inhibition of protein degradation, impairment of energy metabolism, and activation of cell death programs, have been implicated in the molecular mechanisms of the neurodegenerative processes caused by polyQ-expanded proteins (for review, see Ref. 4). These variations are likely determined by sequences outside the polyQ tract, which are specific for the individual disease proteins (5, 6). Even though toxicity has been experimentally dissociated from detectable inclusion body formation (e.g. Refs. 7 and 8), the hallmark of polyQ diseases is the characteristic formation of intraneuronal inclusions, most often in the nuclear compartment. Hence, it seems reasonable to argue that an imbalance in the production of toxic aggregation intermediates and their clearance constitutes the fundamental commonality of these diseases (9). Notably, the aggregation products accumulate, although cells are equipped with an elaborate network of molecular chaperones and cofactors with sufficient capacity to maintain a balanced protein homeostasis essential to cellular function (9, 10). Increasing evidence suggests that the nuclear compartment is the primary site of cellular toxicity in polyQ diseases (11–13). Employing a highly aggregation-prone fragment of the ATXN3 (ataxin-3) protein (14, 15), we investigated whether the nuclear compartment of neuroblastoma-derived cells (N2a) constitutes an environment that is more sensitive to polyQ protein aggregation than the corresponding cytoplasm. We found that heat shock protein-assisted clearance of aggregation-prone polyQ is by far more efficient in the cytoplasm than in the nucleus.

We further show that the basic stretch of amino acids (Arg-Lys-Arg-Arg) retained in the ATXN3 fragment and predicted to serve as a nuclear localization sequence (NLS) is not functional. Instead, other motifs must be responsible for the observed physiological and pathophysiological distribution of truncated ATXN3 in the cell. Preventing mutant ATXN3 fragments from entering the nuclear compartment might be sufficient to ameliorate disease symptoms, as these fragments are more efficiently removed in the cytoplasm.

EXPERIMENTAL PROCEDURES

Expression Constructs

FLAG- and Myc-tagged ATXN3 and truncated C-terminal Myc-tagged expression constructs were described previously (15). β-Galactosidase was expressed with pcDNA3.1-lacZ (Invitrogen); the heat shock element (HSE)-containing luciferase plasmid and the mutant heat shock transcription factor HSF1-encoding plasmids are described elsewhere (25). Mutation or deletion of the putative NLS motif Arg-Lys-Arg-Arg was carried out by site-directed mutagenesis following standard procedures. A synthetic nuclear import signal was cloned using oligonucleotides 5′-GAT CCA TGC CAA AAA AGA AGA GAA AGG TAA-3′ and 5′-GAT CTT ACC TTT CTC TTC TTT TTT GGC ATG-3′, coding for amino acid sequence PPKKKRKVDPKKKRKV derived from the SV40 large T antigen. A nuclear export signal (NES) was cloned using oligonucleotide 5′-CCC GGG ATG TTA GCT TTG AAA TTA GCC GGA CTA GAC ATC GGA TCC ATG GAC TAC-3′, corresponding to amino acid sequence LALKLAGLDI from protein kinase inhibitor-α. DNA was routinely prepared from Escherichia coli SURE cells (Stratagene).

Antibodies

Polyclonal antiserum against rat ATXN3 was raised in rabbit by standard procedures and affinity-purified with glutathione S-transferase-ATXN3Q22 (human sequence) cross-linked to glutathione-agarose as described previously (15). Anti-Myc monoclonal antibody (clone 9E10) and anti-histone H3 polyclonal antibody were purchased from Santa Cruz Biotechnology and Chemicon, respectively. Anti-Hsp70 monoclonal antibodies against the constitutive Hsc73 and inducible Hsp72 proteins (clones N27 and C92, respectively) were from Stressgen. Anti-p62 polyclonal antibody was purchased from Sigma.

Cell Culture

Mouse neuroblastoma N2a and human embryonic kidney HEK293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 μg/ml streptomycin. For transient transfections, 70% confluent cells on 35-mm dishes were transfected with 2 μg of DNA using Lipofectamine PLUS reagents (Invitrogen). For the preparation of lysates for Western blotting, cells were washed once with phosphate-buffered saline, scraped in lysis buffer (1× Tris-buffered saline containing 0.2% Triton X-100, 10 mm MgCl₂, and protease inhibitor mixture (Complete, Roche Applied Science)), and incubated with Benzonase (75 units/3-cm plate; Roche Applied Science) for 1 h at 4 °C. The protein concentration was determined with the Bio-Rad protein assay, and lysates were adjusted to 1 μg/μl total protein.

Filter Retardation Assays and Western Blot Analyses

These were described previously (16). In brief, cell lysates were filtered through a cellulose acetate membrane (0.2-μm pores; Schleicher & Schüll). Aggregates retained on the membrane were detected with anti-Myc antibody. Western blotting was performed by standard procedures. Before probing membranes with different antibodies, bound antibodies were removed by incubation in stripping buffer (63 mm Tris-HCl (pH 6.7), 2% SDS, and 100 mm β-mercaptoethanol) for 1 h at 50 °C, followed by extensive washing with Tris-buffered saline.

Preparation of Nuclear Fractions and Post-nuclear Supernatants

Confluently grown N2a or HEK293 cells were washed with ice-cold 25 mm HEPES/KOH (pH 7.6) and incubated with 200 mm sucrose for 20 min on ice. After removing the sucrose solution, the swollen cells were collected in 250 mm sucrose and 3 mm imidazole (pH 7.4) and disrupted with a Dounce homogenizer (Glas-Col). Residual intact cells and nuclei were separated by centrifugation at 1000 × g for 10 min at 4 °C. The remaining post-nuclear supernatant was adjusted to 20 mm Tris-HCl (pH 8) and 100 mm KCl.

RESULTS

The Stretch of Basic Amino Acids (Arg-Lys-Arg-Arg) Is Not a Nuclear Localization Signal for ATXN3

The truncated ATXN3 protein 257cQ71 (see Fig. 1 for a schematic overview of all constructs used in this study) robustly aggregated upon expression in cultured cells (Fig. 2, a and b) (15). Therefore, we employed this property to analyze the functionality of the putative NLS. Mutation or deletion of this motif should switch the site of aggregation from the nucleus to the cytoplasm. As shown in Fig. 3, transient overexpression of 257cQ71 in N2a cells resulted in nuclear aggregation in the vast majority of the cells, whereas only a minor fraction (∼20%) displayed exclusive cytoplasmic staining. This fraction may represent cells with very high ATXN3 expression and fast cytoplasmic aggregate nucleation that would prevent soluble species from entering the nuclear compartment.

FIGURE 1.

Schematic overview of the ATXN3 constructs used in this study. Synthetic nuclear export (NES) and nuclear import (NLS) motifs, respectively, and Josephin, ubiquitin-interacting motif (UIM), and polyQ (Q_n) domains are highlighted. The putative nuclear localization motif Arg-Lys-Arg-Arg that was deleted or replaced with a His-Asn-His-His motif in some constructs is also shown. All constructs contain a C-terminal Myc epitope. Full-length constructs have an N-terminal FLAG tag in addition.

FIGURE 2.

The cellular site of aggregation is directed by synthetic localization signals but not the intrinsic putative NLS. a, transiently transfected N2a cells were processed for immunofluorescence staining with anti-Myc antibody 48 h after transfection. Nuclei were stained with 4′,6-diamidino-2-phenylindole. b, cell lysates of transiently transfected N2a cells were analyzed for aggregation by filter retardation assay with anti-Myc antibody 48 h after transfection. Equal amounts of total lysate were applied.

FIGURE 3.

The putative nuclear localization motif Arg-Lys-Arg-Arg does not determine the subcellular site of inclusion formation by truncated ATXN3. N2a cells were transiently transfected with pcDNA3.1–257cQ71, pcDNA3.1–257cQ71-His-Asn-His-His, or pcDNA3.1–257cQ71ΔArg-Lys-Arg-Arg (a deletion mutant) and processed for immunofluorescence staining with anti-Myc antibody 48 h after transfection. About 70 cells per construct were analyzed for the site of aggregate formation by visual inspection and subdivided into classes. The Nucleus class contains cells with predominant nuclear aggregates; the Cytoplasm class contains cells with predominant cytoplasmic aggregates; and the Both class represents cells with aggregates in both compartments. The bars represent the relative distribution of aggregates among the three classes.

It was initially surprising that neither mutation of the NLS motif Arg-Lys-Arg-Arg to His-Asn-His-His nor its deletion changed this distinctive distribution pattern of the aggregates significantly, indicating that this motif is not involved in nuclear targeting of the truncated ATXN3 protein. For full-length ATXN3, a series of serine phosphorylations within the Josephin domain and the ubiquitin-interacting motifs are critically involved in the subcellular distribution (17, 18). The truncated ATXN3 protein used in this study is derived from isoform 2 and contains neither the Josephin domain nor ubiquitin-interacting motifs. It may thus contain yet unidentified targeting motifs or may undergo random shuttling and is eventually trapped in the nuclear compartment.

Synthetic Localization Sequences Determine the Subcellular Site of Inclusion Formation by Truncated ATXN3

We next fused synthetic nuclear export and nuclear import transport signals to the N terminus of the ATXN3 fragment to direct the proteins into the respective compartments (Fig. 1). As shown by immunofluorescence staining (Fig. 2a), this led to almost quantitative inclusion formation in the expected compartment. A filter retardation assay showed that nuclear and cytoplasmic aggregates were biochemically similar in nature, i.e. they showed a comparable SDS insolubility consistent with the formation of amyloid-like fibrils (Fig. 2b), the aggregate species that are trapped on filter membranes (19). Thus, the aggregates formed by overexpressed ATXN3 fragments in the nucleus and cytoplasm are biochemically similar, yet it remains unclear why the natural ATXN3 fragment aggregates almost quantitatively in the nucleus, although the presumed NLS is not functional.

Heat Shock Protein Induction Has a Differential Effect on Cytoplasmic Versus Nuclear ATXN3 Fragments

We therefore asked whether activation of molecular chaperones would have a differential effect on cytoplasmic versus nuclear ATXN3 that could explain the nuclear distribution of the natural ATXN3 fragment. Increased solubility in the cytoplasm would potentially allow fragments to escape into the nuclear compartment for aggregation. Similarly, increased sensitivity to degradation in the cytoplasmic compartment could explain the selective accumulation and aggregation of ATXN3 fragments in the nuclear compartment.

Overexpression of the constitutively active HSF1 variant ΔHSF1 (20) induced HSE-controlled luciferase expression many hundredfold in N2a cells compared with β-galactosidase expression used as a transfection control (Fig. 4a). Although the levels of the constitutively expressed heat shock protein Hsc73 did not change significantly, a strong induction of the stress-related protein Hsp72 was observed, confirming the induction of a cellular stress response (Fig. 4b, middle panels).

FIGURE 4.

ΔHSF1-mediated heat shock response efficiently inhibits aggregation of ATXN3 in the cytoplasm but not in the nucleus. a, N2a cells were transiently transfected with a 1- or 2-fold excess of the constitutively active heat shock transcription factor ΔHSF1 plasmid over the nuclear and cytoplasmic aggregation-prone ATXN3 constructs NLS-257cQ71 and NES-257cQ71, respectively. A β-galactosidase plasmid (1× β-gal) served as a transfection control. All cells were cotransfected with a luciferase-encoding plasmid under the control of an HSF1-responsive HSE. 24 h after transfection, luciferase activity in the cell lysates was quantified. Notice the axis break. AU, arbitrary units. b, N2a cells were transfected as described for a, followed by lysate preparation, Western blotting, and filter retardation analysis 48 h post-transfection. c, N2a cells were treated 19 h after transfection for additional 5 h with 10 μm lactacystin (LC), and lysates were prepared for filter retardation analysis. d, N2a cells were either cotransfected with ΔHSF1 or heat-stressed for 15 min at 43 °C in a water bath. 6 h later, nuclear lysates and post-nuclear supernatants were analyzed by Western blotting with the indicated antibodies. The asterisk indicates a cross-reacting nuclear protein of unknown identity. C, cytoplasm; N, nucleus; Ac, acetylated.

Coexpression of ΔHSF1 strongly reduced the steady-state levels of the NES-257cQ71 protein, whereas the levels of NLS-257cQ71 were not affected (Fig. 4b, upper panel). ATXN3 RNA levels were not impaired by ΔHSF1 under these conditions (supplemental Fig. S1). At the same time, the formation of SDS-insoluble fibrillar aggregates of the polyQ protein targeted to the cytoplasm was almost quantitatively inhibited, whereas nuclear targeted polyQ protein was not affected (Fig. 4b, lower panel). Thus, the evoked cellular stress response likely facilitated selective degradation of the polyQ protein in the cytoplasm. To confirm this hypothesis, we analyzed whether the substantial decrease in the levels of soluble NES-257cQ71 under induced cellular stress (Fig. 4b, upper panel) was due to proteasome-dependent degradation in the cytoplasm. As shown in Fig. 4c, proteasome inhibition by lactacystin restored the levels of SDS-insoluble polyQ protein in the cytoplasm, indicating that the strong reduction of cytoplasmic polyQ protein aggregation upon ΔHSF1 coexpression was due to stress response-related degradation of aggregate precursors.

Steady-state levels and aggregation of the nuclear targeted polyQ protein were not affected by ΔHSF1, although Hsp72 protein levels increased in the nuclear compartment upon ΔHSF1 coexpression or thermal stress (Fig. 4d), as has been shown previously (21). This indicates that the quality control mechanism that degrades cytoplasmic polyQ protein in cells under stress was not functional in the nuclear compartment. To further characterize which proteolytic system degrades ATXN3 fragments under induced stress, we determined the level of p62, a key regulatory protein of mammalian autophagy (22). Compared with β-galactosidase coexpression, p62 levels did not change with ΔHSF1 coexpression, whereas N2a cells adapted p62 levels upon inhibition of autophagy by 3-methyladenine or proteasome inhibition by lactacystin (supplemental Fig. S2). This indicates that ΔHSF1 stimulates mainly proteasome-dependent degradation of the ATXN3 fragments in the cytoplasm.

Targeting polyQ-expanded ATXN3 Fragments to the Cytoplasm Induces a Mild Heat Shock Response

Upon cotransfection of the NES-257cQ71 construct with the β-galactosidase control plasmid (Fig. 4a), we noticed a mild but significant activation of HSE-driven luciferase expression compared with the NLS-257cQ71 construct and analyzed this phenomenon in more detail. To measure the cellular stress response induced by cytoplasmic versus nuclear localized polyQ protein over time, N2a cells were cotransfected with the HSE-luciferase construct and NLS-257cQ71 and NES-257cQ71, respectively. Luciferase activity was measured every 24 h up to 72 h. As shown in Fig. 5a, nuclear targeted polyQ protein did not induce significant luciferase expression at any time point measured. In contrast, polyQ protein targeted to the cytoplasm induced robust HSE-driven luciferase expression 24 h after transfection as observed previously (compare Fig. 4a), which peaked at 48 h and then dropped 24 h later. This indicates that cytoplasmic localization of the polyQ protein is required to induce a heat shock response. This “endogenous” stress response is apparently sufficient neither to decrease the steady-state levels of the highly overexpressed cytoplasmic polyQ protein (Figs. 4b and 5b) nor to decrease the amount of aggregates (Fig. 2, a and b) as observed with ΔHSF1 co-overexpression.

FIGURE 5.

Truncated ATXN3 in the cytoplasm induces a mild heat shock response per se and is subject to lactacystin-related degradation. a, N2a cells were transiently cotransfected with the indicated constructs and a firefly luciferase-encoding plasmid under the control of an HSF1-responsive HSE. Luciferase activity was measured at the indicated time points after transfection. AU, arbitrary units. b, N2a cells were cotransfected with the indicated ATXN3 constructs and ΔHSF1 or β-galactosidase (β-Gal) plasmids as shown. 24 h post-transfection, cell lysates were analyzed by Western blotting with anti-ATXN3 antibody. mATXN3, mouse ATXN3.

On the basis of these data, we propose that, under more physiological conditions, where the amount of fragmented ATXN3 is likely to be lower, these fragments are efficiently removed by the ubiquitin-proteasome system but eventually escape this quality control mechanism into the nucleus. Post-mitotic neurons generally display an attenuated stress response and may thus have a limited capacity to remove polyQ fragments.

DISCUSSION

The nuclear compartment appears to be the major site of pathology in polyglutamine diseases (23), including spinocerebellar ataxia type 3 (SCA3) (13). The SCA3 disease protein ATXN3 contains a putative NLS (Arg-Lys-Arg-Arg) adjacent to its polyQ segment that is retained in highly aggregation-prone C-terminal fragments of the protein (14, 15). We therefore asked to what extent this motif is involved in the nuclear aggregation process in neuronal SCA3 cells. We observed that this NLS motif was not functional. Targeting the polyQ protein with a synthetic localization sequence to the nuclear compartment rendered it more resistant to stress-induced degradation than the corresponding protein targeted to the cytoplasm. These results can help explain why in SCA3 aggregates are found almost exclusively in the nuclear compartment.

Generally, a cell will respond to any disturbance of the protein-folding environment (e.g. accumulation of aberrant protein species) with activation of stress response pathways that increase protein refolding and degradation (24). In several chronic neurodegenerative diseases, however, this response seems to be attenuated. Scrapie-infected neuroblastoma cells, for example, fail to induce the expression of Hsp72 upon various stresses at all (25). Similarly, Hsp27, which has been shown to be activated and neuroprotective in acute models of neurodegeneration, is not induced in the chronic Huntington disease R6/2 mouse model (26). Most likely, stress responses are not only spatialized but also temporally restricted by cells to compensate for harmful consequences of persistent activation of stress pathways (27).

It has been suggested that intranuclear inclusion formation of polyQ-expanded ATXN3 is promoted by the nuclear environment (11), and in fact, heat-induced denaturation of proteins is more rapid in the nuclear compartment compared with the cytoplasm, as demonstrated by biophysical (28) and biochemical (29, 30) experiments. On the other hand, evidence has been provided that nuclear inclusions are active proteolytic centers (31). The latter finding would support the idea that inclusion formation in the nucleus is a rather physiological response of cells to the accumulation of misfolded proteins. Moreover, the inducible form of Hsp70 is rapidly produced and relocated to the nuclear compartment upon thermal and other stresses (Fig. 4d) (21), and the Hsp40/Hsp70 chaperone machinery has been shown to be functional in protecting nuclear luciferase from thermal denaturation in the nuclear compartment in vivo (32). Additionally, recent biophysical evidence refuses the hypothesis that increased macromolecular crowding in the nucleoplasm compared with the cytoplasm would contribute to inclusion formation. It is rather the cytoplasm that is slightly more crowded than the nucleoplasm (33). These findings would argue against the idea that the nuclear compartment is simply an unfavorable environment for polyQ-expanded proteins.

Our data suggest that nuclear aggregation of polyQ-expanded ATXN3 is not simply due to an NLS-mediated accumulation in an unfavorable nuclear environment, as the endogenous NLS motif is not active. Rather, the natural polyQ fragment shuffles in and out of the nucleus but is efficiently degraded by the ubiquitin-proteasome system only in the cytoplasm (Fig. 5b), as has been demonstrated for other proteasomal substrates (34). In support of this hypothesis, the steady-state levels of synthetic NLS-257cQ71, in contrast to NES-257cQ71, were not affected by a strong thermal or ΔHSF1-mediated increase in Hsp72 expression. This indicates that, in our SCA3 model, molecular chaperone-stimulated degradation, as observed for the androgen receptor (35), is functional in the cytoplasm but not in the nuclear compartment. In support of this conclusion, 257cQ71-expressing cells accumulated aggregates also in the cytoplasm upon proteasome inhibition (data not shown).

Our data show that cells can, in principle, better handle even largely overexpressed amounts of polyQ protein in the cytoplasm compared with the nucleus. Selective inhibition of nuclear import of expanded polyQ may have the therapeutic potential to improve pathology.