Spastin, the protein mutated in autosomal dominant hereditary spastic paraplegia, is involved in microtubule dynamics

Abstract

Hereditary spastic paraplegia (HSP) is characterized by progressive weakness and spasticity of the lower limbs, caused by the specific degeneration of the corticospinal tracts, the longest axons in humans. Most cases of the autosomal dominant form of the disease are due to mutations in the SPG4 gene, which encodes spastin, an ATPase belonging to the AAA family. The cellular pathways in which spastin operates and its role in causing degeneration of motor axons are currently unknown. By expressing wild-type or ATPase-defective spastin in several cell types, we now show that spastin interacts dynamically with microtubules. Spastin association with the microtubule cytoskeleton is mediated by the N-terminal region of the protein, and is regulated through the ATPase activity of the AAA domain. Expression of all the missense mutations into the AAA domain, which were previously identified in patients, leads to constitutive binding to microtubules in transfected cells and induces the disappearance of the aster and the formation of thick perinuclear bundles, suggesting a role of spastin in microtubule dynamics. Consistently, wild-type spastin promotes microtubule disassembly in transfected cells. These data suggest that spastin may be involved in microtubule dynamics similarly to the highly homologous microtubule-severing protein, katanin. Impairment of fine regulation of the microtubule cytoskeleton in long axons, due to spastin mutations, may underlie pathogenesis of HSP.

Received September 18, 2001; Revised and Accepted November 16, 2001.

INTRODUCTION

Hereditary spastic paraplegia (HSP) comprises a genetically heterogeneous group of inherited diseases, characterized by weakness and spasticity of the lower limbs (1,2). Spastic paraplegia may be an isolated feature (pure HSP), or be accompanied by additional neurological symptoms (complicated HSP). The neuropathological correlate of this clinical picture is a progressive retrograde degeneration of the corticospinal axons, the longest axons in the human body (3,4). Although several HSP genes have been mapped, only a few of them have been cloned so far. These genes were found to be involved in corticospinal axon development (L1CAM) (5), in glia to neuron signaling (PLP) (6), and in mitochondrial function (paraplegin) (7), indicating that the mechanisms that may contribute to the degeneration of motor axons are manifold (for review see 8).

The pathogenesis of the most frequent form of autosomal dominant spastic paraplegia is still completely unknown. This form is due to mutations in the SPG4 gene, which maps to chromosome 2p21–p22 (9,10) and encodes a 616 amino acid protein, named spastin (11). Similarly to paraplegin, spastin belongs to the AAA (ATPases associated with various cellular activities) family, which is characterized by a conserved domain of 230 amino acids with ATPase activity (12). Several papers have reported the spectrum of SPG4 mutations in HSP patients (13–18). According to these studies, SPG4 mutations account for at least 40% of all autosomal dominant HSP families. Missense, nonsense and splice-site mutations as well as deletions or insertions have all been observed in the spastin gene. Notably, all the missense mutations fall into the AAA domain, with the exception of the S44L substitution that appears to be disease causing only in the homozygous state (14), underlying the functional significance of this domain. The other mutations are scattered along the coding region of the gene and lead to premature termination codons, and mRNA instability, suggesting that haploinsufficiency is the molecular cause of the disease (18).

AAA proteins are involved in a wide variety of cellular processes, such as cell cycle, vesicular transport, mitochondrial function, peroxisome biogenesis and proteolysis (19–21). Based on sequence homology and phylogenetic analysis, spastin belongs to the subfamily-7 (12) or meiotic group (22) of AAA proteins. The best-characterized proteins of this subgroup, p60 katanin and Vps4/SKD1, are implicated in completely divergent cellular processes, such as microtubule severing (23,24) and endosomal morphology and trafficking (25,26), hampering the prediction of spastin function based on homologies. Although computer programs predict a nuclear import signal (11), spastin subcellular localization is still unknown, and we have no clue of the mechanism by which spastin mutations may lead to degeneration of corticospinal axons.

We now present data suggesting that spastin interacts with microtubules via its N-terminal region, and that this association is regulated through the ATPase activity of the AAA domain. Furthermore, we provide preliminary evidence suggesting that, like katanin, spastin may be involved in some aspects of microtubule disassembly. Finally, we show that all the spastin missense mutations located in the AAA domain, previously identified in HSP patients, bind constitutively to microtubules and lead to a redistribution of the microtubule cytoskeleton. These data suggest that HSP due to SPG4 mutations may depend on an impairment of microtubule dynamics in long axons.

RESULTS

ATPase-defective spastin associates with microtubules

In order to get an insight on the biological role of spastin, we investigated its subcellular distribution by transiently transfecting different epitope-tagged spastin constructs in Cos-7 cells (Fig. 1). Immunoprecipitation experiments after subcellular fractionation showed a band of the expected size (68 kDa) in the cytoplasmic fraction (Fig. 2). Immunofluorescence studies showed that spastin localizes to discrete punctate structures, which display a perinuclear distribution (Fig. 3A). These structures tend to increase in size with a longer period of expression and ultimately fill up the cytoplasm. Double immunofluorescence experiments ruled out the possibility that these compartments may correspond to known organelles, such as mitochondria, peroxisomes, early and late endosomes or lysosomes (data not shown). We have observed the same pattern of expression in other cell types such as HeLa (Fig. 3B), U2OS and GN11, a LHRH neuronal cell line. Furthermore, the use of different promoters and epitopes (HA versus myc or GFP) either positioned at the N- or C-terminus did not change spastin localization.

We then designed transfection experiments in order to monitor spastin localization in cells that are just beginning to express the construct (Materials and Methods). Under these conditions, specific spastin fluorescence labeled only one perinuclear region (Fig. 3C). This region is located near the Golgi apparatus (Fig. 3D), and corresponds to the center of microtubule asters, as assessed by double labeling with anti-α-tubulin antibody (Fig. 3E). This domain of expression is close to the centrosome revealed by γ-tubulin staining (Fig. 3F and G). These data suggest that the onset of spastin expression may be in the microtubule-organizing center and that, upon longer periods of expression, spastin may accumulate in cytoplasmic aggregates.

Other members of the AAA family, such as SKD1, p60 katanin and NSF bind and release protein substrates in a nucleotide-dependent manner (24,27–29). This may lead to transient association in vivo with their targets. However, this association can be revealed by expressing dominant-negative mutants unable to bind or hydrolyze ATP. To test whether this is true also for spastin, we overexpressed a well characterized mutation (K388R), which falls into the Walker A motif or P loop of the AAA domain, and is known to abrogate ATP binding in other members of the family (27,28,30). This mutant has been found in a patient with HSP and therefore represents a pathogenic form of spastin (13). When spastin^K388R was transfected in Cos-7 cells, at expression onset it was observed in a single perinuclear domain, similarly to the wild-type protein (Fig. 4A). However, with longer periods of expression, a filamentous pattern, reminiscent of association with the cytoskeleton, was detected (Fig. 4B). In some transfected cells a combination of filaments and of cytoplasmic aggregates was observed.

In order to assess the nature of the filaments, we used monoclonal antibodies against α-tubulin and vimentin in co-localization experiments, to detect microtubules and intermediate filaments, respectively. In addition, falloidin staining was employed to investigate a possible association with actin. We showed that the filaments labeled in spastin^K388R-transfected cells co-localize with α-tubulin (Fig. 4C and D), and not with vimentin (Fig. 4E and F) or actin (data not shown). To further confirm binding of spastin^K388R to microtubules, we repeated these experiments after treating the cells with nocodazole, a drug capable of inducing microtubule depolymerization by inhibiting addition of tubulin monomers. In all cases, nocodazole was effective in obtaining dispersion of mutant spastin and tubulin monomers within the cytoplasm in most cells (Fig. 4G and H).

Co-localization of spastin^K388R with microtubules shows peculiar features. First, although all filaments labeled by mutant spastin seem to co-localize with microtubules, not all the microtubules present in the cell are labeled by spastin (Fig. 4C and D). This data suggests that the mutant may bind a subset of microtubules. Secondly, microtubule distribution in transfected cells seems to be altered, with the disappearance of the aster and the formation of thick and long perinuclear bundles (Fig. 4C and D). Consistent with a redistribution of microtubules, dispersal of intermediate filaments also appears affected in transfected cells (Fig. 4E and F). Thirdly, in some cases these bundles appear to be more resistant to nocodazole treatment, suggesting that spastin may protect them from depolymerization (Fig. 4H).

Spastin associates with microtubules via its N-terminal region

The findings that spastin expression begins in the microtubule organizing center and that an ATPase-defective mutant associates with microtubules in vivo prompted us to investigate whether spastin may bind taxol-stabilized microtubules in vitro. Extracts from spastin-transfected cells were supplemented with taxol and centrifuged to sediment microtubules and associated proteins. In these conditions, full-length spastin was enriched in the microtubule fraction (Fig. 5A). In order to map a putative microtubule-binding domain, we transfected two artificial mutants in which either the AAA domain (spastin^Δ-AAA) or the N-terminal part of spastin were deleted (spastin^Δ-N) (Fig. 1). We show that the spastin^Δ-AAA retained the ability to co-sediment with polymerized microtubules in the in vitro binding assay, whereas spastin^Δ-N lost it (Fig. 5A). The same results were obtained in immunofluorescence experiments, where the spastin^Δ-AAA construct showed a filamentous pattern of expression susceptible to nocodazole treatment (Fig. 5B and C), while spastin^Δ-N exhibited a diffuse cytoplasmic staining (Fig. 5D). Taken together, these data indicate that spastin interacts with microtubules through its N-terminal region. The discrepancy between the data obtained in vivo and in vitro with wild-type spastin can be reconciled by hypothesizing that binding to microtubules is transient in vivo and is regulated by the activity of the AAA domain. Both a deletion of the entire AAA domain and the K388R missense mutation altering its ability to bind ATP are able to entrap the protein in a microtubule-bound state. On the contrary, the AAA domain per se is unable to associate with microtubules and to form aggregates.

Spastin disassembles microtubules in vivo

Spastin ability to bind microtubules in an ATP-dependent manner and to alter microtubule distribution and morphology led us to postulate that it could be implicated in some aspects of microtubule dynamics. Another AAA protein of the same subfamily, p60 katanin, has been largely studied for its microtubule severing activity at the centrosome (23,24,31). To assess a putative microtubule severing activity of spastin in vivo, we employed an assay previously used to measure in vivo microtubule severing of katanin and of its Caenorhabditis elegans homolog MEI-1 (30,32). In this assay, the microtubule cytoskeleton is examined by anti-tubulin immunofluorescence in Cos-7 cells expressing spastin–GFP and compared to neighboring untransfected cells. When wild-type spastin was expressed in Cos-7 cells, we observed a dramatic reduction in the intensity of tubulin immunofluorescence in ∼50% of transfected cells (Table 1; Fig. 6A, B, D and E). Reduced intensity of tubulin staining was never observed in cells transfected with the spastin^K388R mutant or with GFP alone. Furthermore, this effect is specific for microtubules, and does not affect intermediate filaments. Overexposure of the spastin-transfected cells showed that microtubules are broken, with the end-ends of the filament recognizable in a few cases (Fig. 6C and F). The remaining filaments are still emanating from the centrosome, but the dimension of the aster is greatly reduced (Fig. 6C and F).

Functional characterization of spastin missense mutations observed in HSP patients

To date, all the spastin missense mutations found in HSP patients are located into the AAA domain. The only exception is a serine to leucine substitution in position 44 that was reported to occur in the homozygous state (14). We have already shown that the K388R mutant associates with microtubules. As a first step to investigate the functional role of all the missense mutations identified in HSP patients, we looked at their subcellular localization. Several spastin missense mutations found in HSP patients were therefore introduced in appropriate expression vectors, by site-directed mutagenesis (Table 2). Based on the knowledge available on other members of the AAA family, missense mutations into the AAA domain are predicted to interfere either with ATP binding or hydrolysis.

All the mutants showed a filamentous pattern, which corresponded to association with microtubules, as demonstrated after nocodazole treatment of the cells. Microtubule binding showed the same characteristics that were described above for the K388R mutation, i.e. formation of thick perinuclear bundles, disappearance of the aster, and, in some cases, increased resistance to nocodazole treatment. The only two exceptions were the S44L substitution, and the S362C mutation that behaved like the wild-type protein. The first mutation lies outside the AAA, whereas the second affects an amino acid just at the beginning of the domain.

AAA proteins perform their functions in homo- or hetero-oligomeric complexes. In order to determine whether the spastin missense mutants could act as dominant-negative and interfere with expression of wild-type spastin, we co-transfected spastin^K388R–GFP or spastin^L426V–GFP with HA-spastin and looked at subcellular localization in double-transfected cells. We found that in all cases of co-transfection, the wild-type and the mutant proteins always co-localize (Fig. 7). In some co-transfected cells the wild-type protein labeled microtubules (Fig. 7E). In other cases, both the mutant and wild-type spastin localized to the cytoplasmic spots (Fig. 7A and B). This different behavior may reflect different rates of expression of the mutants relative to the wild-type protein. This result is consistent with the action of spastin being reliant on the formation of oligomers, and further experiments are necessary to confirm this hypothesis.

DISCUSSION

Microtubules are highly dynamic polymers of α- and β-tubulin subunits, which provide architectural support to eukaryotic cells, organize membranous organelles and act as railways along which cytoplasmic constituents are transported. Microtubules undergo rapid reorganization of their configuration during the cell cycle, and can switch abruptly between elongation and shortening. Several proteins have been identified which control changes of the microtubule cytoskeleton in vivo (32–35). We now present data suggesting that spastin may be involved in microtubule dynamics, opening new perspectives for understanding the pathogenesis of HSP.

We find that spastin associates with microtubules in vitro and that this interaction is mediated by its N-terminal region. In vivo, microtubule association can be detected only by expressing mutants lacking the entire AAA domain or defective in ATP binding or hydrolysis, indicating that binding to microtubules is transient and regulated through the nucleotide-binding state of the AAA domain. Other members of the AAA family display a transient ATP-dependent oligomerization in the presence of the protein substrate, and, upon ATP hydrolysis, undergo a conformational change that leads to release from the substrate (24,27). Thus, for these proteins, binding to the protein targets is transient in vivo. p60 katanin binds microtubules in vitro but was never shown to label microtubules in vivo (24,30,36). Similarly, association of Vps4/SKD1 to the endosomal membranes in transfected cells was found to be dependent on the nucleotide-binding state of the AAA domain (25,26).

Several lines of evidence indicate that spastin might be involved in regulating microtubule dynamics. First, overexpression of wild-type spastin in both Cos-7 and HeLa cells results in a microtubule-disassembly phenotype. This phenotype is never observed in cells transfected with a spastin mutant or with the empty vector, and is specific for microtubules. Secondly, overexpression of spastin mutants leading to constitutive binding to microtubules induces a redistribution of the microtubule array, with a disappearance of the aster and the formation of thick perinuclear bundles, which in some cases were resistant to nocodazole treatment. The spastin-coated microtubules may remain long and misplaced because of a perturbation of their dynamics due to the mutant protein. Thirdly, spastin is highly homologous within the AAA domain to p60 katanin and MEI-1, two microtubule-stimulated ATPases, which require ATP hydrolysis to disassemble microtubules (23,24,31,32,37). Katanin is a p60/p80 heterodimer, found at centrosomes and mitotic spindle poles, and thought to be responsible for releasing microtubules from their centrosomal attachment points (36,38). A similar role, albeit in meiosis, has been described for the MEI-1/MEI-2 complex in C.elegans (32). Both p60 katanin and MEI-1 require the WD-40-containing proteins p80 and MEI-2, respectively, to target to centrosomes and to potentiate their microtubule-severing activity (23,30,32). Although the N-terminal regions of spastin, p60 katanin and MEI-1 are not overtly conserved, in all cases a microtubule-binding domain has been mapped to this region. Similarly, protein targeting by the best-studied AAA family member, NSF, involves its N-terminal domain (39).

Although spastin may share functional properties with the known microtubule-severing proteins, several issues are still unresolved. We could not establish whether spastin binds microtubules at the centrosomes or along their length in vivo, and if this interaction is direct or mediated by association with another protein. Although transfections carried out in order to achieve very low levels of expression suggest a possible localization of spastin in the microtubule-organizing center, this data should be interpreted with caution and studies with specific antibodies to detect the endogenous protein will be needed. If binding to microtubules also occurs outside the centrosome, it appears to be anyway restricted to a subset of them, indicating that this association is dynamic and probably associated to a remodeling of the cytoskeleton. Finally, the microtubule-disassembly phenotype induced by spastin overexpression in transfected cells could derive either from promoting catastrophe by shrinking of microtubules from their ends or by the generation of internal breaks within a filament with a true severing mechanism. Our experiments do not allow to distinguish between these two possibilities and microtubule severing in vitro assays using a recombinant purified spastin protein are required before accounting spastin in the number of the known microtubule-severing proteins.

We showed that all spastin missense mutations, which fall into the AAA domain, behaved as ATPase-defective mutants, altering the ability of spastin to regulate interaction with the target and thus leading to constitutive binding to microtubules. The only exception is the S362C substitution, which is located at the immediate beginning of the domain, suggesting that this mutation impairs spastin function with a different mechanism. Co-transfection experiments of wild-type and mutant spastin are consistent with a complex being assembled containing a mixture of the two forms of the proteins. Whether this is the physiological situation or whether spastin is part of a hetero-oligomeric complex with another partner, similar to p60 katanin and MEI-1, needs to be clarified. Previous studies of genotype–phenotype correlation have shown that there is no difference in the severity of spastic paraplegia between patients with spastin missense mutations and those harboring mutations leading to a premature protein termination (13). Whenever the level of spastin mRNA has been tested in tissues from patients with this last kind of mutations, it has been found to be uniformly drastically reduced. These findings are consistent with mRNA instability, and have suggested that haploinsufficiency is the molecular cause of the disease (18). However, our data open up the possibility that a dominant-negative pathogenetic mechanism could be involved in patients with spastin missense mutations and justify more detailed studies to evaluate this possibility.

Spastin is a widely expressed protein (11), but the disease phenotype is remarkably restricted to corticospinal axons. Microtubules are essential for axonal growth and maintenance. The current model concerning the mechanism by which the microtubule array of a growing axon is established is that microtubules are nucleated at the centrosomes and then actively transported along axons and dendrites (40). Experiments in which functional-blocking antibodies against p60 katanin were microinjected in cultured neurons demonstrated that katanin has an essential role for releasing microtubules from the neuronal centrosome, but also for regulating their length after releasing (41). These experiments suggested that, in neurons, microtubule severing could occur also far from the centrosome, as a mechanism to keep microtubules sufficiently short to be efficiently transported into axons and dendrites. It is tempting to speculate that spastin is a novel protein involved in microtubule dynamics in neuronal cells. A fine regulation of spastin function may be crucial in the long processes of cortical motoneurons, where the presence of half the normal dosage of the protein could already be detrimental.

In conclusion, we provide evidence that spastin is involved in microtubule dynamics by binding transiently to microtubules in an ATPase-dependent manner and promoting their disassembly in vivo. Although future studies will address if spastin is a true microtubule-severing protein, this result opens new perspectives for understanding pathogenesis of HSP due to spastin mutations.

MATERIALS AND METHODS

Generation of spastin constructs

The coding region of spastin cDNA was amplified by PCR using PfuI polymerase (Promega), HeLa cDNA as template and specific primers designed on an available sequence (11). The resulting fragment was then subcloned in frame into the following eukaryotic expression vectors: pMT21-myc, pcDNA3-myc–GFP and pcDNA3-HA. In the resulting constructs, the myc and GFP tags are positioned C-terminal to the spastin-coding region, whereas the HA tag is N-terminal to it (Fig. 1).

The spastinΔ^-AAA construct (corresponding to the nonsense mutation C932G) was generated by PCR amplification with specific primers, using PfuI as polymerase, and subcloned into the pMT21-myc vector. The spastin^Δ-N construct, lacking the first 241 amino acids of spastin, was generated by digesting the spastin-myc construct with EcoRI–SpeI, followed by Klenow I treatment at room temperature for 15 min and self-ligation. The ATG is reconstituted after self-ligation. In each case, the integrity of the clones was determined by direct sequencing.

In vitro mutagenesis

To introduce the point mutations, S44L, S362C, G370R, F831C, N386K, K388R, L426V, C448Y, R460L, R499C and A556V, into both the spastin-myc and/or spastin-mycGFP constructs, in vitro mutagenesis was performed using the Quickchange site-directed mutagenesis kit (Stratagene), according the manufacturer’s instructions. The presence of the point mutation was then confirmed by DNA sequencing. The following primers were used: S44L, 5′-GCCCCTCCGCCCGAGTTGCCGCATAAGCGGAAC-3′; S362C, 5′-GTTATTCTTCCTTGTCTGAGGCCTGAG-3′; G370R, 5′-CCTGAGTTGTTCACAAGGCTTAGAGCTCCTG-3′; F831C, 5′-GGCTGTTACTCTGTGGTCCACCTGG-3′; N386K, 5′-GTCCACCTGGGAAGGGGAAGACAATGC-3′; K388R, 5′-CTGGGAATGGGAGGACAATGCTGGC-3′; L426V, 5′-GAAATTGGTGAGGGCTGTTTTTGCTGTGGCTCG-3′; C448Y, 5′-GTTGATAGCCTTTTGTATGAAAGAAGAGAAGGG-3′; R460L, 5′-GCACGATGCTAGTAGACTCCTAAAAACTGAATTTC-3′; R499C, 5′-GGCTGTTCTCAGGTGTTTCATCAAACG-3′; A556V, 5′-GCTTTGGCAAAAGATGTAGCACTGGGTCCTATC-3′.

Cell lines and antibodies

Monkey kidney Cos-7 cells, HeLa cells, GN11 (42) (kindly provided by R.Maggi, Milan) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Highclone) supplemented with 10% fetal bovine serum (Highclone). Monoclonal anti-myc antibody was produced from 9E10 hybridoma cells (43). Monoclonal anti-HA antibody was purchased from Babco, monoclonal anti-α-tubulin (1:250) from Molecular Probes. Anti-γ-tubulin (GTU88; 1:50) and anti-58K protein (1:100) were from Sigma. All the secondary antibodies for immunofluorescence were obtained from Dako, while horseradish peroxidase (HRP)-conjugated antibodies were from Amersham.

Cell transfection and immunofluorescence

All constructs were transfected using lipofectamine according to instructions provided by the manufacturer (Life Technologies). Transfections to obtain a low level of expression were performed by incubating cells with the DNA–lipofectamine mixture for 4 h and fixing cells 2 h later. Cells were grown either on coverslips or on multiwell chamberslides (Nunc) in DMEM, 10% FBS, transfected as described, and fixed in 4% paraformaldehyde/PBS at room temperature. For γ-tubulin immunofluorescence, cells were fixed in 100% methanol at –20°C for 15 min. After fixation, cells were permeabilized in 0.2% Triton X-100/PBS for 10 min, and blocked in 10% pig serum for 30 min. After blocking, cells were incubated with the primary antibodies for 2 h and appropriate secondary antibodies (1:100) for 1 h. To detect nuclei, cells were stained using the DNA-specific stain Hoechst (Sigma). When requested, cells were treated with 20 µM nocodazole (Calbiochem) for 2.5 h at 37°C. Cells were mounted with Vectashield (Dako) and examined with a Axioplan microscope (Zeiss) equipped with an Axiocam CCD camera and Axiovision digital imaging software (Zeiss). Images were then processed using PhotoShop 5.5 software (Adobe).

Microtubule disassembly in transfected Cos-7 cells

Cells were transfected with spastin–GFP, spastin^K388R–GFP, or pcDNA3–GFP, fixed 24 h post-transfection and analyzed by double immunofluorescence using GFP fluorescence and a monoclonal antibody against α-tubulin. At least three independent transfection experiments were performed for each construct. For each experiment, 100 transfected cells were examined to evaluate the intensity of anti-α-tubulin staining. We scored a cell as having a microtubule-disassembly phenotype only when a dramatic reduction in tubulin fluorescence was appreciated compared to neighboring untransfected cells. To analyze the integrity of microtubules in cells showing a dramatic decrease in the intensity of α-tubulin staining, images were acquired with a longer time of exposure.

Subcellular fractionation and immunoprecipitation

Transfected cells were collected 48 h post-transfection, resuspended in buffer A (10 mM HEPES pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT, 0.5 mM PMSF), incubated on ice for 10 min and homogenized. Samples were centrifuged at 510 g for 10 min at 4°C, the supernatant was collected and stored as a cytoplasmic fraction. The resulting pellet was resuspended in buffer B (20 mM HEPES pH 7.9, 1.5 mM MgCl₂, 0.42 M NaCl, 25% glycerol, 0.2 mM EDTA, 0.5 mM DTT, 0.5 mM PMSF), homogenized and incubated on ice for 30 min. The sample was then centrifuged at 12 900 g for 30 min at 4°C and the supernatant was analyzed as a nuclear fraction. Both nuclear and cytoplasmic fractions were concentrated by immunoprecipitation of the protein the 9E10 antibody. Samples from immunoprecipitation were fractionated on an 8% SDS–polyacrylamide gel, blotted on a PVDF membrane and revealed by immunoblot analysis.

Microtubule-binding assay

Transfected cells were collected 48 h post-transfection and lysed in PEM–DNNA buffer (80 mM PIPES pH 6.8, 1 mM EGTA, 1 mM MgCl₂, 0.5 mM DTT, 150 mM NaCl, 1% NP-40) supplemented with protease inhibitors, at 4°C for 1 h. Lysates were centrifuged at 610 g for 10 min at 4°C. Cytosol was then purified by successive centrifugations at 10 000 g for 10 min, at 21 000 g for 20 min and at 100 000 g for 1 h at 4°C. Each supernatant was then supplemented with 1 mM GTP (Boeringher) and 40 µM taxol (Molecular Probes) and incubated at 37°C for 30 min. Corresponding samples without taxol were also prepared. Each sample was layered over a 15% sucrose cushion and centrifuged at 30 000 g for 30 min at 30°C to sediment polymerized microtubules. The resulting supernatants were saved and pellets were resuspended in an equal volume of sample buffer 1× for electrophoresis and immunoblot analysis.

ACKNOWLEDGEMENTS

We wish to thank Germana Meroni and Vittoria Schiaffino for helpful discussions. This work has been supported by an Italian Telethon grant to E.I.R. and by the National Institutes of Health (1R01NS38713-01 to A.B.).

+

To whom correspondence should be addressed at: Telethon Institute of Genetics and Medicine, Via P. Castellino 111, 80131, Naples, Italy. Tel: [phone]; Fax: [phone]; Email: [e-mail]

Figure 1. Schematic representation of spastin constructs used in this study. The AAA domain is indicated by the grey bar. The N-terminal or C-terminal position of the different tags are shown. Numbers denote the residues present at the beginning and the end of each construct and of the AAA domain.

Figure 2. Subcellular fractionation indicates that spastin has a cytoplasmic localization. Cos-7 cells were transfected either with the pMT21-myc empty vector or with the spastin-myc construct. Forty-eight hours post-transfection samples were processed to make a cytoplasmic (C) and a nuclear (N) fraction. Both fractions were then concentrated by immunoprecipitation, subjected to electrophoresis on a 10% acrylamide gel followed by immunoblotting with the 9E10 antibody.

Figure 3. Immunofluorescence analysis of transiently expressed wild-type spastin. Spastin-myc shows a discrete punctate cytoplasmic localization both in Cos-7 (A) and HeLa cells (B) after 24 h of expression. When low levels of expression were achieved (Materials and Methods), spastin localizes to a discrete perinuclear area (C). Double immunofluorescence analysis using spastin–GFP (green signal) construct and appropriate markers (red signal), revealed that this expression domain is in the proximity to the Golgi apparatus (D), to the microtubule aster (E), and the centrosome (F and G). In (G), arrows point to the centrosomes. Monoclonal antibodies against 58 kDa Golgi protein, α-tubulin and GTU88 were used to detect the Golgi apparatus, microtubules and the centrosome, respectively.

Figure 4. Spastin^K388R localizes to microtubules. In all cases spastin^K388R was expressed as GFP fusion in Cos-7 cells and is visible as a green signal; nuclei are stained with Hoechst, blue signal; microtubules and intermediate filaments were revealed using monoclonal antibodies against α-tubulin and vimentin, respectively, red signal. Co-localization is indicated by yellow signal in merged image (A). Spastin^K388R expression begins in correspondence of the microtubule aster (A) and proceeds with accumulation in filamentous structures (B). These filaments correspond to a subset of microtubules (C and D) and do not co-localize with vimentin (E and F). Microtubule distribution is altered in transfected cells with a disappearance of the aster, and the formation of thick perinuclear bundles (D). Also, intermediate filaments distribution is affected (F). Microtubule association was confirmed by nocodazole treatment, which disrupts both the filamentous pattern of expression of spastin (G) and the microtubule network (H). In some cases, the spastin-labeled microtubule bundles were protected from nocodazole effect [arrows in (G and H)]. Expression of the spastin^K388R construct in other cell lines gave essentially identical results (data not shown).

Figure 5. Spastin interacts with microtubules via its N-terminus. (A) Microtubule-binding assay. Cos-7 cells were transfected with spastin-myc, or spastin^Δ-AAA or spastin^Δ-N constructs. Lysates (L) from transfected cells were either not supplemented (– taxol) or supplemented (+ taxol) with 40 µM taxol. After sedimentation on sucrose cushion, supernatant (S) and pellet (P) fractions were assayed for spastin, spastin^Δ-AAA, spastin^Δ-N and α-tubulin detection, using appropriate antibodies. When lysates are supplemented with taxol, in order to stabilize polymerized microtubules, both full-length spastin and spastin^Δ-AAA proteins are detected in the pellet together with polymerized microtubules, while spastin^Δ-N remains in the supernatant. As control, when samples are not treated with taxol, all the proteins are detected in the supernatant fraction after the sedimentation. Immunofluorescence analysis of spastin^Δ-AAA shows a filamentous pattern of expression (B), which is disrupted by nocodazole treatment (C). On the contrary, the spastin^Δ-N construct shows a diffuse cytoplasmic localization (D), confirming that the spastin^Δ-N construct is not able to bind microtubules both in vivo and in vitro.

Figure 6. Overexpression of wild-type spastin promotes microtubule disassembly. In all cases spastin was expressed for 24 h as GFP fusion in Cos-7 cells and is visible as a green signal; nuclei are stained with Hoechst, blue signal; microtubules were revealed using monoclonal antibodies against α-tubulin, red signal. Cells overexpressing spastin (A and D) show a lower intensity of α-tubulin staining (B and E) with respect to non-transfected cells. Moreover, overexposure of the transfected cells in (B) and (E) shows that microtubules are disassembled, and appear fragmented (C and F, enlarged views).

Figure 7. Wild-type and mutant spastin form complexes in transfected cells. Either spastin^L426V–GFP (A) or spastin^K388R–GFP (D) were co-transfected with wild-type HA-spastin in Cos-7 cells. Both mutated proteins are visible as green signal; nuclei are stained with Hoechst, blue signal; wild-type HA-spastin was revealed using monoclonal anti-HA antibody, red signal (B and E). Co-localization is indicated by yellow signal in merged images (C and F). In some co-transfected cells, both the mutant and wild-type spastin localized to the cytoplasmic spots (A–C). In other cases the wild-type protein is incorporated in wild-type/mutant heterodimers resulting in a filamentous pattern of expression (D–F).

Table 1.

Effects of overexpression of spastin wild-type on microtuble disassembly

Transfected plasmid	% GFP-positive cells with reduced microtubule staining	Cells counted/transfection	Number of transfections
pcDNA3-myc–GFP	0.7 ± 1.1	100	3
Spastin–GFP	50.5 ± 3.4	100	4
SpastinK388R–GFP	2.4 ± 2.5	100	3

Table 2.

Subcellular localization of transiently expressed spastin missense mutations reported in HSP patients

Spastin mutation	Subcellular localization	Sensitivity to nocodazole treatment
S44L	Cytosolic bodies	–
S362C	Cytosolic bodies	–
G370R	Filaments and cytosolic bodies	+
F831C	Filaments and cytosolic bodies	+
N386K	Filaments and few cytosolic bodies	+
K388R	Filaments and few cytosolic bodies	+
L426V	Filaments and sporadic cytosolic bodies	+
C448Y	Filaments and few cytosolic bodies	+
R460L	Filaments and sporadic cytosolic bodies	+
R499C	Filaments and few cytosolic bodies	+
A556V	Filaments and few cytosolic bodies	+