Golgi Fragmentation Occurs in the Cells with Prefibrillar α-Synuclein Aggregates and Precedes the Formation of Fibrillar Inclusion*

Amyloid-like fibrillar aggregates of intracellular proteins are common pathological features of human neurodegenerative diseases. However, the nature of pathogenic aggregates and the biological consequences of their formation remain elusive. Here, we describe (i) a model cellular system in which prefibrillar α-synuclein aggregates and fibrillar inclusions are naturally formed in the cytoplasm with distinctive kinetics and (ii) a tight correlation between the presence of prefibrillar aggregates and the Golgi fragmentation. Consistent with the structural abnormality of Golgi apparatus, trafficking and maturation of dopamine transporter through the biosynthetic pathway were impaired in the presence of α-synuclein aggregates. Reduction in cell viability was also observed in the prefibrillar aggregate-forming condition and before the inclusion formation. The fibrillar inclusions, on the other hand, showed no correlation with Golgi fragmentation and were preceded by these events. Furthermore, at the early stage of inclusion formation, active lysosomes and mitochondria were enriched in the juxtanuclear area and co-aggregate into a compact inclusion body, suggesting that the fibrillar inclusions might be the consequence of an attempt of the cell to remove abnormal protein aggregates and damaged organelles. These results support the hypothesis that prefibrillar α-synuclein aggregates are the pathogenic species and suggest that Golgi fragmentation and subsequent trafficking impairment are the specific consequence of α-synuclein aggregation.

2). ␣-Synuclein is a 140-amino acid protein that is enriched in presynaptic terminals of neurons (3). In the test tube, this protein forms fibrils, which resemble the ones isolated from postmortem brains with LB diseases (4 -6). A causative role of fibrillar ␣-synuclein inclusions in neurodegeneration has been suggested in a mouse model in which formation of intracytoplasmic ␣-synuclein inclusions coincided with the severe motor impairment (7). On the other hand, a study in a transgenic fly model that express human ␣-synuclein has shown that coexpression of molecular chaperone hsp70 alleviated the neurodegenerative phenotype but did not reduce the formation of ␣-synuclein-positive inclusion bodies (8). This result raised a question as to whether the fibrillar inclusions play a causative role in neurodegenerative process. Furthermore, recent in vitro studies revealed various non-fibrillar species during the course of fibrillation and suggested a possibility that these metastable intermediate species, not the fibrils themselves, might elicit cytotoxicity (9,10). Elucidating which particular aggregate species possess the principal cytotoxic effect holds the key to understanding the etiologic role of protein aggregation in the disease pathogenesis. Study of this problem, however, has been hampered by the lack of an experimental system in which intermediates of the endogenous fibrillation process can be biochemically defined and analyzed. Here, we have established such a system and have assessed the effects of prefibrillar intermediates that are formed naturally in the cytoplasm. In this report, we refer to prefibrillar aggregates as non-fibrillar oligomeric assemblies that precede the formation of fibrils and to inclusions as large deposits of aggregates that are usually found in juxtanuclear location.

MATERIALS AND METHODS
Antibodies-Monoclonal anti-␣-synuclein antibody, LB509, was purchased from Zymed Laboratories (South San Francisco, CA), and a polyclonal anti-␣-synuclein serum, 7071, was provided by Peter Lansbury (Brigham and Women's Hospital, Boston, MA). Antibodies against GM130, TGN46, and calnexin were obtained from BD Biosciences (San Diego, CA), Serotec (Oxford, UK), and StressGen Biotechnologies Corp. (Victoria, BC, Canada), respectively. Antibodies for mannosidase II and DAT were purchased from Chemicon (Temecula, CA). All the fluorescently labeled secondary antibodies were purchased from Jackson Im-munoResearch Laboratories (West Grove, PA). 125 I-Labeled anti-mouse IgG antibody was obtained from Amersham Biosciences (Piscataway, NJ). Goat anti-mouse IgG antibody, conjugated with a 10-nm gold particle, was obtained from Ted Pella Inc. (Redding, CA).
Aggregation Analysis at Different Multiplicity of Infection-Different amounts (see Fig. 1 legend) of adeno/␣-syn (11) and empty viral vector were mixed to make the final number of viral particle be 1.5 ϫ 10 8 . COS-7 cells (1.5 ϫ 10 6 cells) were infected with these mixtures as described previously (11). At day 3, the cells were extracted in phosphate-buffered saline (PBS) with 1% Triton X-100 and protease inhibitor mixture (Sigma), and the extracts were centrifuged at 16,000 ϫ g for 5 min to separate detergent-soluble (supernatant) and insoluble (pellet) fractions. The pellets were resuspended in half the volume of 1ϫ Laemmli sample buffer. Ten micrograms of supernatant and the equal volume of pellet were applied onto 12% SDS-polyacrylamide gel and subjected to Western blotting (12). For the quantitative analysis, the proteins were visualized using 125 I-labeled secondary antibody, and the monomers and the aggregate smears (from 60 kDa to the top of the gel) were quantified by computer-assisted densitometry using ImageQuaNT software (Molecular Dynamics) under equal light and power settings. Three independent experiments were performed.
For the Golgi analysis and cell viability assay, the cells were split onto cover-glasses into 12-well plates and into 35-mm dishes, respectively, at day 1 (the next day of infection) and incubated until the time of analysis, usually at day 3 or 4. We have noticed that the time course of the aggregation process changes slightly depending on the surface area of the tissue culture dish; i.e. aggregation is faster in 100-mm dishes than in 35-mm dishes. To normalize the effects of viral vector itself, the total amount of viral particles was adjusted to be equivalent with empty viral vector in all experiments.
Extraction and Separation of Prefibrillar Aggregates and Fibrillar Inclusions-COS-7 cells were infected with adeno/␣-syn at m.o.i. 75, and the cells were split into 35-mm dishes at day 1 and incubated until days 3 or 4. On the day of extraction, buffer T (0.25 M sucrose, 25 mM KCl, 5 mM MgCl 2 , 20 mM Tris, pH 7.5, 1% Triton X-100, protease inhibitor mixture) was added gently onto each dish and incubated at room temperature for 5 min. The Triton X-soluble supernatant was removed carefully and then buffer N (0.1 M Na 2 CO 3 , pH 11.5, with 1% Triton X-100 and protease inhibitor mixture) was added to each dish, and extract was obtained by scraping dishes and repeated pipetting. After incubation on ice for 5 min, the extract was centrifuged at 80 ϫ g for 5 min, at which condition only the fibrillar inclusions sedimented to form a pellet (43). The prefibrillar aggregates in the supernatant were collected by additional centrifugation at 16,000 ϫ g for 10 min.
Fluorescence Microscopy-Fluorescence staining, including nuclear staining with Hoechst 33258, was performed according to the procedures in Lee et al. (11). For the staining of lysosomes and mitochondria, the cells were incubated with 100 nM Lysotracker (Molecular Probes) and 200 nM Mitotracker (Molecular Probes), respectively, in the growth medium for 30 min, and then fixed and permeabilized. Morphology of GA was visualized using an antibody against GM130, TGN46, or mannosidase II. All the fluorescence images in this study were obtained with a laser scanning confocal microscope (LSM PASCAL, Zeiss). For quantitative analysis, images were obtained by the "tile-scan" of area of 0.48 mm 2 , and the number of cells with fragmented Golgi (Fig. 3) or fibrillar inclusions (Fig. 2) was counted in three random areas. Each image contained 140 cells on average, and the experiment was repeated more than three times. Golgi fragmentation was defined as the dispersion of small Golgi-immunoreactive foci. Only the cells with completely scattered Golgi fragments were counted, whereas the cells with long tubular Golgi staining were not. Thus, the percentage obtained in this study is likely to under-represent the actual degree of Golgi fragmentation.
Electron Microscopy-For electron microscopy (EM), ␣-synuclein was expressed in COS-7 cells at m.o.i. 75 for 4 days. The sections were prepared as described in Bouley et al. (13). Briefly, the cells were fixed in 2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.2, on ice for 35 min. After the post-fix in 1% osmium tetroxide at room temperature for 30 min, the cells were stained with 1% uranyl acetate for 1 h and then dehydrated with a series of different concentrations of ethanol. After the samples were embedded in gelatin capsule, the sections were prepared on carbon-coated nickel grids and stained with 1% uranyl acetate. For the immunolabeling, the sections were incubated in 1% gelatin/ PBS, followed by 0.03 M glycine/PBS. After a rinse with 2% bovine serum albumin (BSA)/PBS, the sections were incubated with LB509 antibody (1/250 dilution) in 2% BSA/PBS, and then with 10-nm goldconjugated goat anti-mouse IgG antibody. The sections were then stained with 1% uranyl acetate and lead citrate and observed with a Philips CM 12 electron microscope.
Biotinylation and Isolation of Cell Surface Proteins-Cells were infected with various ratios of ␣-synuclein and empty viral vectors as described before and incubated until day 3. For the expression of dopamine transporter (DAT), the infected cells were transfected with hDAT/pCDNA3.1(ϩ) on the next day. Biotinylation and isolation of cell surface proteins were carried out according to Melikian et al. (14), and all the procedures were performed on ice except where indicated otherwise. Briefly, the cells were rinsed twice with PBS 2ϩ (PBS with 0.1 mM CaCl 2 and 1.0 mM MgCl 2 ) and incubated with 1.0 mg/ml NHS-SS-biotin (Pierce, Rockford, IL) in PBS 2ϩ for 20 min twice. Remaining reactive NHS-SS-biotin was quenched twice with 0.1 M glycine for 20 min, and then the cells were extracted in radioimmune precipitation assay buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 1.0 mM EDTA, 0.1% SDS, 1% Triton X-100, 0.5% sodium deoxycholate, protease inhibitor mixture (Sigma)). The cell extracts with the same amounts of protein were incubated with Immunopure streptavidin beads (Pierce) at room temperature for 45 min with constant rotation. The bound fractions were washed four times with radioimmune precipitation assay buffer and eluted with 1ϫ Laemmli sample buffer. The unbound fractions were precipitated with 5% trichloroacetic acid and dissolved in 1ϫ Laemmli sample buffer. The samples were loaded onto SDS-PAGE and analyzed by Western blotting. Due to the abundance of cytoplasmic DAT relative to the cell surface DAT, one-hundredth equivalents of unbound samples (cytoplasmic proteins) were loaded.
Cell Viability Assay-The cells were trypsinized, and one-tenth of cells were mixed with the equal volume of 0.4% trypan blue solution (Sigma). The number of cells that exclude the dye was counted using a hemacytometer in triplicate. The data in Fig. 10 were obtained from two independent experiments. The rest of the trypsinized cells were extracted and fractionated as described above, and the detergent-insoluble fractions were subjected to Western blotting with LB509 antibody.

RESULTS
Prefibrillar ␣-Synuclein Aggregates in the Cytoplasm-To characterize the aggregation process of ␣-synuclein in cells, we introduced various amounts of the cDNA into COS-7 cells using recombinant adenoviral vector (adeno/␣-syn) and quantitatively analyzed the changes in the amounts of the monomers and aggregates. The level of monomers increased linearly with the increase of multiplicity of infection (m.o.i.) before reaching a plateau (ϳ1.7% of total cellular protein). Once the monomer level reached the plateau, the levels of aggregates elevated in response to increased m.o.i. (Fig. 1A). This non-linear relationship between the monomer level and the aggregation allows us to selectively investigate the effects of the monomers or aggregates by choosing the appropriate m.o.i. range. Two different types of ␣-synuclein aggregates were found by immunofluorescence staining: small punctate aggregates that are dispersed throughout the cytoplasm and large juxtanuclear inclusion bodies (43). The large inclusions were strongly stained with thioflavin S, a fluorescent dye specific to amyloid-like structure, whereas the small punctate aggregates were not (43). Electron microscopy (EM) shows that the inclusion bodies are filled with filaments with the width of 8 -10 nm, and they were labeled with ␣-synuclein antibody (Fig. 1C). The presence of ␣-synuclein fibrils in the inclusions was further confirmed by immunogold labeling of individual fibrils that were isolated from the inclusion preparations. Immuno-EM study also verified the non-fibrillar nature of small punctate aggregates upon isolation using the procedure described below (43).
A small amount of aggregates appeared at day 2 post-infection and increased drastically at day 3 with a small further increase at day 4, whereas the monomer levels remained constant throughout this period ( Fig. 2A). Until day 3, most of the aggregates were small, non-fibrillar ones, and less than 1% of cells had larger fibrillar inclusion bodies. However, during the 4th day, formation of fibrillar inclusion increased greatly (Fig.  2B). To confirm the time courses for the different aggregate species, we have recently established a biochemical method by which small non-fibrillar aggregates and fibrillar inclusions can be separated by their sizes and quantitated. After a series of extraction steps using detergent and basic pH, small nonfibrillar aggregates and fibrillar inclusions are separated into the supernatant and pellet, respectively, by a centrifugation at 80 ϫ g (43). When this procedure was applied to the timecourse study, we found that small aggregates in the supernatant appeared at day 3 and sustained until day 4, whereas the inclusions in the pellet became apparent no sooner than day 4 ( Fig. 2C), confirming that small non-fibrillar aggregates appear before the formation of fibrillar inclusions. Given the dye-binding property and the kinetic behavior, we propose that the small punctate aggregates are the cytoplasmic equivalents of prefibrillar intermediates of fibrillation.
Golgi Fragmentation by Prefibrillar ␣-Synuclein Aggregates-To investigate cellular consequences specific to ␣-synuclein aggregation, we have examined the morphology of Golgi apparatus (GA) for two reasons: (i) ␣-synuclein aggregates may target membranous organelles, because aggregates in our cell system appear initially in the membrane fraction, 2 consistent with the earlier finding in cell-free assay that membrane-bound ␣-synuclein forms aggregates more efficiently than the cytosolic ␣-synuclein (15), (ii) fragmentation of GA has been found in several neurodegenerative diseases, including Alzheimer's disease (AD), amyotrophic lateral sclerosis (ALS), Creutzfeldt-Jakob disease (CJD), and multiple system atrophy (MSA) (16 -18). Especially, the pathology of MSA is characterized by ␣-synuclein-positive inclusion bodies of neurons and oligodendrocytes (19 -21). Thus, the morphological integrity of GA was analyzed with respect to the ␣-synuclein aggregation states using an immunofluorescence microscopy with the antibody against a cis-Golgi-specific matrix protein GM130. In cells with diffuse ␣-synuclein staining, the GA has normal compact morphology near the nucleus (Fig. 3A). However, the cells with prefibrillar aggregates frequently show fragmentation and dispersion of the GA (Fig. 3, B and C). To determine if Golgi fragmentation is an aggregation-induced event, cells with fragmented Golgi were counted at different m.o.i. (Fig. 3D). Despite the presence of a high level of soluble ␣-synuclein, Golgi fragmentation at a m.o.i. of 20 was not different from the basal level. On the other hand, the Golgi fragmentation was significantly increased when prefibrillar aggregates formed with stable monomer levels (m.o.i. 75) (Fig. 3D), suggesting that increase of ␣-synuclein aggregates, rather than the monomer, might be the cause of Golgi fragmentation. To confirm the correlation between ␣-synuclein aggregation and Golgi fragmentation, an independent experiment was carried out to score the Golgi fragmentation in a population of cells that contained small punctate aggregates. Golgi fragmentation was found far more frequently in the population with the aggregates (88.9% at m.o.i. 75 and 66.7% at m.o.i. 20) than in the general population (11.5% at m.o.i. 75 and 4.3% at m.o.i. 20) (Fig. 3E). In contrast, expression of green fluorescent protein to ϳ9.1% of total cellular protein did not cause dramatic changes in the Golgi morphology (data not shown), suggesting that Golgi fragmentation is not a general consequence of overproduction of soluble proteins.
To determine the effects of ␣-synuclein aggregates on other Golgi compartments, we examine the morphologies of transand mid-Golgi using antibodies against TGN46, an integral component of trans-Golgi membrane, and mannosidase II, a mid-Golgi resident enzyme. Dual immunofluorescence staining shows that the aggregate-induced Golgi fragments contain both GM130 and TGN46. Furthermore, these proteins stayed segregated to the opposite poles in these fragments as in the normal compact GA (Fig. 4A), indicating that the cis-trans polarity was maintained even after the fragmentation. When the Golgi fragments were co-stained for GM130 and mannosidase II, the segregation of these proteins was also evident within the single fragments (Fig. 4B). However, the segregation was not as obvious as the co-staining of GM130 and TGN46, probably due to their relative distance between the compartments. These results suggest that ␣-synuclein aggregate-induced Golgi fragmentation occurs in the entire Golgi complex without obvious disturbance of the cis-trans polarity.
Reduction in Cell Surface Level of Dopamine Transporter-While passing through the Golgi stacks, secretory and membrane proteins undergo a series of covalent modifications, such as glycosylation and deglycosylation, and the fully mature proteins are sorted to their final destinations. To assess the functional consequence of the aggregate-induced Golgi fragmentation, we measured the changes in the cell surface level of a plasma membrane protein dopamine transporter (DAT). The cells were infected with different amounts of adeno/␣-syn and then transfected with DAT expression plasmids on the following day. After 2 more days of incubation, the cell surface proteins were covalently labeled with biotin and separated from the cytoplasmic proteins (non-biotinylated) using streptavidin resin. Western analysis of these fractions shows that cell surface level of DAT was reduced in the prefibrillar aggregateproducing condition (m.o.i. 75), compared with the empty vector alone (m.o.i. 0) (Fig. 5). However, increase of monomeric ␣-synuclein level did not result in any change in the cell surface expression of DAT (Fig. 5, see m.o.i. 7.5 and 25). Reduction in cell surface DAT was accompanied by the molecular mass shift of cytoplasmic DAT from 90 -100 kDa to 56 kDa (Fig. 5). The former corresponds to the endoglycosidase H (endoH)-resistant mature protein, and the latter is the endoH-sensitive biosynthetic intermediate (14). Because the endoH resistance is obtained in mid-Golgi compartment, these data imply that the newly synthesized proteins cannot reach the mid-Golgi in an aggregate-producing condition. In contrast, accumulation of the endoH-sensitive form indicates that the overall protein synthesis, and the initial glycosylation may not be compromised in this condition. These results suggest that prefibrillar ␣-synuclein aggregates cause the impaired protein trafficking and maturation through the biosynthetic pathway, probably by disrupting Golgi transport.
Consistent with the finding that the biosynthesis of endoHsensitive form of DAT is not affected by ␣-synuclein aggregation, the gross morphology and distribution of endoplasmic reticulum (ER) seem to remain unchanged in the presence of ␣-synuclein aggregates. Normally, the ER shows a dispersed reticular pattern throughout the cytoplasm. This pattern was maintained regardless of the ␣-synuclein aggregation state or structural integrity of GA in the cytoplasm (Fig. 6), suggesting that the ␣-synuclein aggregates have little effect on the ER morphology.
Inclusion Bodies May be the Response of the Cell to the Toxic Protein Aggregates-In contrast to the prefibrillar aggregates, no clear association was found between the fibrillar inclusions and the Golgi fragmentation. First, an increase in Golgi fragmentation occurs before the appearance of fibrillar inclusions (Figs. 2B and 3F). Second, the size of the cell population that contains fragmented GA does not increase during the fourth day when a sudden increase in the number of fibrillar inclusions takes place (Figs. 2B and 3F). Finally, fragmented GA was rarely observed in the cells with fibrillar inclusions. Instead, two types of GA morphology were associated with fibrillar inclusions: in most cases, Golgi components were confined and dispersed inside the fibrillar inclusions (Fig. 7A), and occasionally, some cells maintained the compact, although FIG. 4. ␣-Synuclein aggregates induce the Golgi fragmentation in the entire Golgi complex. Aggregates were induced as in Fig. 3, and the cis-and trans-Golgi (A) or cis-and mid-compartments (B) were stained with the antibodies against GM130 (red) and TGN46 (green) (A), or GM130 (green) and mannosidase II (red) (B), respectively. The lower images in A are the high magnification views of normal (1) and fragmented Golgi (2). Note that the Golgi fragments contain all the markers that represent individual cisternae and still maintain the cis-trans polarity. Scale bars, 10 m. slightly distorted, morphology of GA that surrounded the juxtanuclear fibrillar inclusions (Fig. 7B). The distorted compact Golgi that is adjacent to inclusion has also been reported previously in a study using a GFP-250 (N-terminal 252 amino acids of p115) chimera protein (22). These two types of Golgi morphology might be explained by the difference in the rate of conversion of prefibrillar intermediates to fibrillar inclusions; fast sequestration of the prefibrillar aggregates into the inclusions before they damage the GA might lead to the distorted compact Golgi adjacent to the inclusion body. Alternatively, inclusions could be formed from two different prefibrillar species, only one of which is capable of Golgi fragmentation. In any case, the facts that Golgi fragmentation occurs before the formation of inclusions and that some cells maintain the compact Golgi structure despite the presence of inclusions suggest that the inclusions themselves may not be the direct cause of the morphological disruption of the GA. Rather, co-staining of the fibrillar inclusions with ␣-synuclein and Golgi marker suggests that the inclusions might be a consequence of an attempt by the cell to remove abnormal protein aggregates and impaired organelles from the cytoplasm.
Consistent with this interpretation, intact lysosomes and mitochondria accumulated in the juxtanuclear region at what appears to be an early stage of inclusion formation (Fig. 8, A  and C) and are also present in the compact inclusions (Fig. 8, B  and D). Lysosomes and mitochondria, which normally tend to be dispersed in the cytoplasm, showed a rather localized pattern in the juxtanuclear region in the presence of the prefibrillar ␣-synuclein aggregates (Fig. 8, A and C). These organelles showed no sign of fragmentation and appear to be functionally intact, because the fluorescent dyes used in this study selectively accumulate in cellular compartment with low internal pH (Lysotracker) or with the membrane potential (Mitotracker). EM study also showed the accumulation of fragmented Golgi (Fig. 9, box 1) and intact mitochondria in the juxtanuclear region (Fig. 9). In addition, autophagosomes and lysosomes can be found in this area at the early stage of inclusion formation (Fig. 9, box 2). Proteasomes and chaperone molecules were also accumulated near the nucleus in the cells with ␣-synuclein aggregates (43). It has been suggested that the juxtanuclear pericentriolar region serves as the main location in which both the autophagic/lysosomal system and ubiquitin-proteasome system execute their degradation function to remove abnormal proteins and damaged organelles (23)(24)(25). Thus, the localization of functional lysosomes and mitochondria in the juxtanuclear region may be the manifestation of defense mechanisms of the cell against the protein aggregates. Together, these results suggest that the formation of ␣-synuclein aggregates triggers the action of the cellular defense system, which is represented by the accumulation of these aggregates and impaired organelles in the juxtanuclear FIG. 6. No gross changes in ER morphology and distribution. ␣-Synuclein was expressed at m.o.i. 75 for 3 days. A, co-staining of ␣-synuclein (red) and ER (green). ER was stained with anti-calnexin antibody. A cell with aggregates is indicated with an arrow. No changes in ER morphology can be recognized in the presence of aggregates, compared with the cells with diffuse ␣-synuclein staining. B, co-staining of calnexin (red) and GM130 (green). The cells with fragmented GA are marked with arrows. ER morphology and distribution seem unchanged regardless of the state of GA. Scale bars, 10 m GM130 is dispersed within the inclusion body. The inclusion body shown here is present on a different optical plane from the majority of other cellular components. This type of inclusion is much more frequently found than the type shown in B. B, distorted GA adjacent to a fibrillar inclusion (arrow). Scale bars, 10 m. region along with the mitochondria, autophagosomes/lysosomes, and proteasomes.
Cytotoxicity of Prefibrillar ␣-Synuclein Aggregates-A cytotoxic effect of ␣-synuclein has been reported in several mammalian cell systems (26 -29). However, the nature of toxic species has not been determined. To assess the cytotoxic effect of ␣-synuclein aggregates, we measured cell viability as a function of the occurrence of the aggregates. An increase of monomer, without forming aggregates, did not affect the viability (m.o.i. 0 -20) (Figs. 1A and Fig. 10A, day 3). In contrast, cell viability was significantly reduced when aggregates formed without increasing the monomer level (Fig. 10, A and B, day 3). Tight correlation between reduction in cell viability and ␣-synuclein aggregation was also found in the time-course study. Aggregates were formed at slower rates at lower m.o.i. values (m.o.i. 20 and 50), and when they became apparent on the fourth day, the viability was reduced correspondingly (Fig.  10, A and B). Like the Golgi fragmentation, reduction in cell viability occurred before the appearance of the fibrillar inclusions (Fig. 10A, day 3 at the m.o.i. of 70, and see also Fig. 2, B and C), implicating that the cytotoxic effect might be conferred by the prefibrillar aggregates. These results suggest that the cytotoxic effect of ␣-synuclein depends on its ability to form aggregates, especially the prefibrillar intermediates. DISCUSSION In this study, we report two biological effects of cytoplasmic ␣-synuclein aggregates; fragmentation of the GA and cell death. The fact that these effects correlate only with the formation of aggregates, and not with the monomer level, suggests that occurrence of these effects depends on the formation of higher order, quaternary structures. The prefibrillar intermediates seem to be responsible for these effects, because both of the effects occur in the presence of small prefibrillar aggregates, and before the formation of fibrillar inclusions. We were able to distinguish specific effects of prefibrillar intermediates from those of monomers and fibrillar inclusions by taking advantage of the following features of our experimental system. First, once the monomer level reaches a plateau, only aggregates increase in response to an increasing amount of ␣-synuclein cDNA without changing the monomer level. This phenomenon provides an effective means to distinguish the effects of aggregates from those of monomers. Second, the ability to define naturally occurring prefibrillar aggregates and fibrillar inclusions biochemically, allowed kinetic analysis of each species. This study showed that the aggregation process of ␣-synuclein in cells, similar to the findings in test tube, involves prefibrillar intermediates in the course of forming fibrillar inclusions (43). More importantly, such a kinetic delay of the inclusion formation enabled us to distinguish the cytophysiological effects of prefibrillar intermediates and fibrillar inclusions.
Earlier examinations of human brain tissues and animal models have shown that the Golgi fragmentation was associated with the neurodegenerative phenotypes. Fragmentation of the GA has been found in human neurodegenerative diseases, including AD, ALS, CJD, and MSA (16 -18). Moreover, transgenic mice expressing ALS-linked mutant superoxide dismutase 1 showed Golgi fragmentation in spinal cord motor FIG. 8. Accumulation of lysosomes and mitochondria at the inclusionforming site in aggregate-containing cells. ␣-Synuclein was expressed at m.o.i. 75 for 4 days. Cells with prefibrillar aggregates are indicated with arrows. A, lysosomes with internally low pH were labeled with Lysotracker. Lysosomes and ␣-synuclein aggregates were found in the juxtanuclear area that appears to be a "pre-inclusion" state. B, Lysotracker also stains the juxtanuclear inclusion body (arrowhead). C, mitochondria with intact membrane potential were labeled with Mitotracker. Functional mitochondria and ␣-synuclein were accumulated in the juxtanuclear region. D, mitochondria are also accumulated within the inclusion body (arrowhead). Scale bars, 10 m.
neurons in an early, preclinical stage, implicating the role of Golgi fragmentation in the early stage of neurodegeneration (30). Our present study links this well-documented pathological feature of neurodegenerative diseases, the Golgi fragmentation, to the prefibrillar aggregate form of ␣-synuclein. Although our results imply the correlation between Golgi fragmentation and cell death, they are not sufficient to suggest the cause/effect relationship between these two. Nevertheless, given the importance of the GA in maturation and trafficking of essential proteins and lipids, we speculate that synaptic sites might be particularly vulnerable to Golgi dysfunction. It is, therefore, noteworthy that transgenic mice producing non-fibrillar ␣-synuclein aggregates in neurons suffer from presynaptic degeneration in nigrostriatal system (31).
Whether inclusion bodies are toxic entities or harmless byproducts or even a part of a protective mechanism has been one of the central issues in amyloid-associated disorders. Our present study shows that (i) Golgi fragmentation and cell death are observed before the formation of inclusion bodies in mixed cell populations and that (ii) ␣-synuclein aggregates are pulled into the "pre-inclusion" area in the juxtanuclear region along with cellular defense organelles, such as autophagic/lysosomal vesicles and mitochondria. These results are consistent with the notion that inclusion bodies per se might not be directly harmful to cells, rather, they are consequences of the efforts to remove abnormal protein aggregates and damaged organelles from the cytoplasm. In a recent study using a Drosophila model in which exogenous expression of human ␣-synuclein induced Lewy body-like inclusions and dopaminergic neuronal loss, overproduction of a molecular chaperone hsp70 alleviated the neurodegenerative phenotype without reducing the number of inclusion bodies (8). This study also argues against the direct role of ␣-synuclein-positive inclusion body in the degenerative process. The non-causal relationship of inclusion bodies and cell death has also been proposed in other protein aggregation systems. A degenerative phenotype and the formation of detectable inclusion bodies have been dissociated in cell and animal models that overexpress expanded polyglutamine-containing proteins (32)(33)(34)(35)(36). In another study using a transgenic mouse model that expresses familial ALS-associated mutant superoxide dismutase 1, small detergent-insoluble protein complex occurs before the onset of disease phenotype, whereas the inclusion bodies can be detected once the clinical symptom is evident (37). Therefore, regardless of the constituent protein, the presence of inclusion bodies itself does not seem to be directly responsible for the cell death. However, these results do not exclude the possibility that the "process" of inclusion body formation might damage the cells; here, we refer to the process as the cellular events related to the inclusion formation after the diffusion-limited peripheral aggregation. For example, sequestration and depletion of vital proteins and organelles, such as proteasomes, mitochondria, and lysosomes, from the cytoplasm, as seen in the early stage of ␣-synuclein inclusion formation, could make the cells vulnerable to any secondary stresses. The sequestration, thus the reduced cellular activity, of the ubiquitin-proteasome system has indeed FIG. 9. EM analysis of the pre-inclusion state. The cells were infected with adeno/␣-syn at m.o.i. 75, incubated for 4 days, and processed for EM as described under "Materials and Methods." Most inclusions have compact structure with clear boundary. Occasionally, we observe what appears to be an early stage of inclusion formation. In this pre-inclusion stage, mitochondria (M), autophagosome precursors (arrows, box 2), lysosomes (arrowhead, box 2), and fragmented GA (fGA, box 1) are accumulated in the juxtanuclear region, and hardly detected outside this region. Neither the pre-inclusion stage nor the compact inclusion body could be found in the cells that were infected with empty viral vector. Scale bars, 5.6 m for the main image and 0.5 m for the box images. been proposed to be a general consequence of protein aggregation that can lead to cell death (38).
Interfering with the microtubule-mediated transport could also lead to the Golgi fragmentation. For example, microtubuledisrupting agents, such as nocodazole, cause a redistribution of Golgi proteins to ER exit sites (39). Thus, the Golgi fragmentation observed in our study could be a secondary consequence of overloading the microtubule-dependent transport system with the aggregates or of halting the trafficking at the negative ends of microtubules due to the accumulation of ␣-synuclein aggregates, rather than the direct effect of the ␣-synuclein aggregates. We do not have conclusive evidence to support either of these possibilities, but the architecture of the fragmented Golgi indicates that more complex mechanism may underlie the ␣-synuclein aggregate-induced Golgi fragmentation than the interference of microtubule-dependent transport. In case of the aggregate-induced Golgi fragmentation, components of different Golgi compartments seem to stay together in the Golgi fragments with the intact cis-trans polarity. This finding disagrees with what was observed during the nocodazole-induced Golgi fragmentation, in which trans-Golgi proteins were scattered more rapidly than mid-Golgi proteins, suggesting the separation of Golgi subcompartments from each other prior to the scattering (40).
Unlike the small punctate aggregates, juxtanuclear inclusion bodies contain ␣-synuclein fibrils. No evidence has been obtained for the presence of fibrillar ␣-synuclein aggregates outside the juxtanuclear inclusions. These findings implicate that, although the mechanism is not understood at present, the structural transition of prefibrillar intermediates into thermodynamically more stable fibrils occurs after they are transported and deposited in the inclusion-forming site (43). A recent study showed that ␣-synuclein fibrils were far less efficient in disrupting bilayer membranes than protofibrils (41), suggesting an inert nature of fibrils. In other studies using recombinant proteins or synthetic peptides that are not implicated in any disease, fibrils were non-toxic when treated to cultured cells, whereas the same treatment with non-fibrillar aggregates, which precede formation of fibrils, resulted in cell death (42). Therefore, formation of fibrillar inclusion bodies might protect the cells from the cytotoxicity of aggregates, not only by removing the prefibrillar aggregates from the cytoplasm but also by providing the right environment for converting them to inert fibrils.
In conclusion, the results presented in this report suggest that the pathogenic property of ␣-synuclein may not stem from its primary structure but from its ability to form toxic aggregates, particularly the prefibrillar aggregates. Importantly, fibrillar inclusions may not be directly toxic, rather, they may represent the outcome of the natural way the cell handles abnormal protein aggregates that are formed randomly throughout the cytoplasm. Thus, both preventing the formation of aggregates and accelerating the conversion of prefibrillar intermediates to fibrillar inclusions might be logical strategies for intervening the pathogenic progress of LB diseases.