Parkin is associated with cellular vesicles

Authors


Address correspondence and reprint requests to N Hattori at Department of Neurology, Juntendo University School of Medicine, 2–1-1 Hongo, Bunkyo, Tokyo 113–8421, Japan. E-mail: nhattori@med.juntendo.ac.jp

Abstract

We recently identified a novel gene, parkin, as a pathogenic gene for autosomal recessive juvenile parkinsonism. Parkin encodes a 52-kDa protein with a ubiquitin-like domain and two RING-finger motifs. To provide a insight into the function of parkin, we have examined its intracellular distribution in cultured cells. We found that parkin was localized in the trans-Golgi network and the secretory vesicles in U-373MG or SH-SY5Y cells by immunocytochemical analyses. In the subsequent subcellular fractionation studies of rat brain, we showed that parkin was copurified with the synaptic vesicles (SVs) when we used low ionic conditions throughout the procedure. An immunoelectromicroscopic analysis indicated that parkin was present on the SV membrane. Parkin was readily released from SVs into the soluble phase by increasing ionic strength at neutral pH, but not by a non-ionic detergent. To elucidate its responsible region for membrane association, we transfected with green fluorescent protein-tagged deletion mutants of parkin into COS-1 cells followed by subcellular fractionation. We demonstrated the ability of parkin to bind to the membranes through a broad region except for the ubiquitin-like domain. The significance of SV localization of parkin is discussed.

Abbreviations used
AR-JP

autosomal recessive juvenile parkinsonism

DMEM

Dulbecco's modified Eagle's medium

FITC

fluorescein isothiocyanate

GAP

growth-associated protein

GFP

green fluorescent protein

NGF

nerve growth factor

PBS

phosphate-buffered saline

PD

Parkinson's disease

SDG

linear sucrose density gradient

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

SN

substantia nigra

SNAP

synaptosomal-associated protein

SV

synaptic vesicle

TGN

trans-Golgi network

TRITC

tetramethylrhodamine

Ubl

ubiquitin-like

WGA

wheat-germ agglutinin.

Genetic factors undoubtedly contribute to the pathogenesis of Parkinson's disease (PD) and distinct genetic loci for different forms of familial PD have been mapped to particular chromosomal regions (Polymeropoulos et al. 1996; Gasser et al. 1998; Leroy et al. 1998b; Farrer et al. 1999). Although there is no direct evidence that any of these genetic factors have a direct role in the etiology of the common sporadic form of PD, elucidation of the pathogenic mechanisms of familial PD should provide crucial tools for better understanding of the sporadic PD.

The gene encoding α-synuclein was identified as a pathogenic gene (PARK1) for an autosomal dominant familial form of PD (Krüger et al. 1998; Polymeropoulos et al. 1997), whereas a novel gene, parkin (PARK2), was identified as a pathogenic gene for the autosomal recessive juvenile parkinsonism (AR-JP) (Kitada et al. 1998). AR-JP is a distinct clinical and genetic entity characterized by early onset, mild dystonia, diurnal fluctuation, transient improvement of parkinsonism after a sleep or nap, a good response to levodopa, and less frequent resting tremor (Yamamura et al. 1973). The pathologic changes in AR-JP include selective degeneration of pigmented neurons in the substantia nigra (SN) and the locus coeruleus, and lack of Lewy bodies (Takahashi et al. 1994) which are the hallmark of PD (Forno 1996).

The parkin gene consists of 12 exons spanning over 1.5 megabases and encodes a novel protein of 465 amino acids with a molecular weight of ∼52 kDa. The parkin protein has a ubiquitin-like (Ubl) domain at the N-terminal portion and two RING-finger motifs at the C-terminal portion. We and colleagues have identified a variety of mutations in AR-JP patients of different ethnic origin including Japanese, Turkish, French, Algerian, Italian, German, Portuguese and Greek (Hattori et al. 1998a,b; Kitada et al. 1998; Leroy et al. 1998a; Lücking et al. 1998; Abbas et al. 1999), indicating world-wide distribution of AR-JP.

We have shown that parkin is localized in the Golgi complex of neuronal cells (Shimura et al. 1999). The Golgi complex is involved in the transport, processing and sorting of newly synthesized proteins. There are many proteins associated with Golgi complex, among which the Golgi-resident enzymes involved in proteolytic process and oligosaccharide-processing of transported proteins are all transmembrane proteins with their major domains disposed to the lumen (Kornfeld and Kornfeld 1985). In contrast, there are also several proteins that are soluble and cytoplasmically disposed-membrane proteins that are involved in the vesicular transport through the Golgi complex (Rothman and Orci 1992; Rothman 1994). The distribution of parkin in the Golgi complex and the cytosol (Shimura et al. 1999) suggests that it is likely a peripheral membrane protein. To investigate the precise intracellular localization of parkin, we performed immunocytochemical, biochemical, and cell biological analyses. Here, we demonstrate that parkin is localized on trans-Golgi network (TGN) and cytoplasmic surface of synaptic vesicles (SVs) of neuronal cells.

Materials and methods

Antibodies

Anti-parkin antibody was prepared as described previously (Shimura et al. 1999). Anti-γ-adaptin, anti-synaptotagmin I, anti-synaptophysin, anti-synapsin I, anti-α-synuclein and anti-green fluorescent protein (GFP) antibodies were purchased from Sigma (St Louis, MO, USA), Wako Pure Chemical Industries (Osaka, Japan), Boehringer Mannheim (Mannheim, Germany), Chemicon (Temecula, CA, USA), Transduction Laboratories (Lexington, KY, USA) and Clontech (Palo Alto, CA, USA), respectively.

Immunocytochemistry

U-373MG (human glioblastoma) and SH-SY5Y (human neuroblastoma) cells were grown at 37°C (5% CO2) in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) fetal bovine serum, penicillin (100 IU/mL), and streptomycin (100 mg/mL). To induce cell differentiation, SH-SY5Y cells were incubated in complete medium plus 100 µm retinoic acid (Sigma) for at least 2 weeks. Cells were fixed for 10 min in 4% paraformaldehyde in 0.01 m phosphate-buffered saline (PBS), permeabilized with PBS containing 0.2% (w/v) Triton X-100, and incubated in normal goat serum for 20 min at room temperature. Cells were then incubated overnight with polyclonal anti-parkin antibody, followed by incubation with fluorescein isothiocyanate (FITC)-conjugated anti-rabbit secondary antibody for 1 h at room temperature. In the double labeling experiments, monoclonal antibody against γ-adaptin or against synaptotagmin I and polyclonal anti-parkin antibody were used to study the relative locations of these proteins. FITC-conjugated anti-rabbit secondary antibody was used for parkin and tetramethylrhodamine (TRITC)-conjugated anti-mouse secondary antibody was used for γ-adaptin or synaptotagmin I. After five washes with PBS, the cells were mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA). Fluorescence images were obtained with a Bio-Rad Laser Confocal system (MRC-1024, Bio-Rad Laboratories, Hercules, CA, USA) equipped with an ECLIPSE TE 300 microscope (Nikon, Tokyo, Japan) and an argon/krypton laser.

Preparation of synaptic vesicles from rat brain

The experimental protocol was approved by the Ethics Review Committee for Animal Experimentation of Juntendo University School of Medicine. The SVs were prepared as described previously (Hell and Jahn 1998), with minor modifications. After collecting the brain tissues, all steps were carried out on ice or at 4°C. Five rats were killed by cervical dislocation followed by decapitation. The brains were carefully dissected out, avoiding myelin-rich areas such as corpus callosum or medulla oblongata. The harvested tissue was placed into 45 mL ice-cold homogenization buffer (320 mm sucrose and 4 mm HEPES–NaOH, pH 7.4) and homogenized using a loose-fitting glass-Teflon homogenizer (12 up-and-down strokes, 900 r.p.m.) in the presence of a mixture of protease inhibitors (Complete EDTA-free, Boehringer Mannheim). The subsequent steps of differential centrifugation were performed in 50-mL polycarbonate tubes in a RPR20–2 rotor (Hitachi, Tokyo, Japan). The homogenate was centrifuged for 10 min at 1000 g. The resulting pellet (P1) was discarded, while the supernatant (S1) was collected and centrifuged for 15 min at 12 000 g. The supernatant (S2) was removed, and the pellet (P2) was washed by resuspension in 30 mL of homogenization buffer and recentrifuged for 15 min at 13 000 g to yield a supernatant, S2′, and a pellet, P2′. The latter pellet represents a crude synaptosomal fraction, which was subsequently resuspended in homogenization buffer to yield a final volume of 3 mL, then transferred to a glass-Teflon homogenizer. In the next step, 27 mL of ice-cold water containing protease inhibitors (Complete EDTA-free, Boehringer Mannheim) was added, and the whole suspension was immediately subjected to five up-and-down strokes at 1200 r.p.m. (hypotonic lysis). This was immediately followed by the addition of 250 µL of 1 m HEPES–NaOH buffer (pH 7.4), and centrifugation for 20 min at 33 000 g to yield the lysate pellet (LP1) and a lysate supernatant (LS1). The supernatant was centrifuged for 2 h at 260 000 g in a P70AT rotor (Hitachi). The supernatant (LS2) was discarded, and the pellet (LP2) was resuspended in 1.5 mL of 40 mm sucrose, using a tight-fitting glass-Teflon homogenizer (10 up-and-down strokes, 1200 r.p.m.), followed by extruding the suspension consecutively through a 23- and a 26-gauge hypodermic needles attached to a 10-mL disposable syringe. This suspension was used as the SV fraction.

Linear sucrose density gradient (SDG) of synaptic vesicles

SV fraction was layered on top of SDG ranging from 0.2 m sucrose to 2.0 m sucrose formed in 4 mm HEPES–NaOH buffer (pH 7.4). The gradient was centrifuged at 46 500 g for 13 h in a P40ST rotor (Hitachi). Fractions were collected starting from the top of the tube and equal volumes of each fraction were subjected to sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) followed by immunoblotting.

Immunoperoxidase labeling of frozen sections of crude synaptosomal fractions and electron microscopy

Frozen sections (5 µm thick) of the crude synaptosomal fraction were cut by a cryostat and fixed for 30 min in 4% paraformaldehyde in PBS at 4°C. After incubation with 2% bovine serum albumin in PBS for 30 min to saturate the non-specific binding sites for proteins, the sections were incubated overnight at 4°C in a polyclonal anti-parkin, monoclonal anti-synaptotagmin I or monoclonal anti-synaptophysin antibodies, followed by incubation with horseradish peroxidase-conjugated anti-rabbit or anti-mouse secondary antibody for 2 h at room temperature. The bound antibodies were fixed with 0.5% glutaraldehyde in PBS for 5 min at 4°C. The sections were then immersed in 0.55 mm 3,3′-diaminobenzidine tetrahydrochloride for 30 min at room temperature. The peroxidase enzymatic reaction was carried out by the addition of H2O2. Sections were postfixed in 2% OsO4 in PBS for 1 h at room temperature, then dehydrated in graded series of ethanol solutions and embedded in Epon 812. Thin sections were cut on a Sorvall Porter-Blum MT2-B ultra-microtome, stained with uranyl acetate and lead citrate and examined under a JEM-1200EX electron microscope operated at 80 kV.

Extraction of parkin

Aliquots of the SV fraction were treated for 30 min on ice in the absence or presence of the indicated concentrations of various salts or 1% (w/v) Triton X-100. After incubation, the suspensions were centrifuged at 100 000 g for 1 h to obtain supernatants and membrane pellets. Each fraction was adjusted to the same volume and analyzed by SDS–PAGE followed by immunoblotting.

Generation of parkin–GFP fusions

Standard molecular cloning techniques were used in these experiments. To obtain a full-length human parkin cDNA, reverse transcription polymerase chain reaction (RT-PCR) was performed. Poly(A) + RNA (1 µg/µL) from human brain (OriGene Technologies, Inc., Rockville, MD, USA) was primed to 1 µL (50 pmol/µL) random 9 mers at 70°C for 10 min, and transcribed to a cDNA copy, using 1 µL (5 U/µL) of AMV Reverse Transcriptase XL (Takara, Shiga, Japan). Then, a 2.5-µL cDNA aliquot was used as a template for PCR amplification. The following primer set was used for amplification of 1545-bp fragment containing the entire coding sequence of parkin cDNA: forward primer 5′-CCGGGAGGATTACCCAGGAGA-3′ and reverse primer 5′-GGGTATGCTCCCCCAGGATGT-3′. The resulting parkin cDNA cloned into pGEM®-T Easy Vector (Promega, Madison, WI, USA) was used as a template for the following PCR. To generate a chimeric gene composed of wild-type parkin gene fused with GFP, a full-length human parkin cDNA was amplified by PCR using primer set with XbaI linkers: forward primer 5′-GCTCTAGAATAGTGTTTGTCAGGTTCAAC-3′ and reverse primer 5′-GCTCTAGACACGTCGAACCAGTGGTCCC-3′. The PCR product was digested with XbaI, and then subcloned into pQBI 25 (Takara), which had been previously digested with NheI. This ensured the in-frame fusion of wild-type parkin at the N-terminus of the GFP coding sequence, under the control of the constitutive immediate early promoter of the cytomegalovirus. The constructs discussed in this paper are summarized in Fig. 5(a). Deletion mutants of parkin, ΔEx3–5, and Ubl-RING2 were generated by a double PCR. Ubl alone, ΔEx4–12, ΔEx5–12, ΔEx6–12, and ΔEx1–5 mutants were amplified by a single PCR. The PCR products of these deletion mutants were subcloned into NheI site of pQBI 25. Missense mutants fused with GFP were generated using QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA, USA), according to the protocol provided by the manufacturer. All constructs were confirmed by DNA sequencing.

Cell transfection and fluorescence microscopy

Cell lines that were used for transfection experiments included COS-1 (monkey kidney) and PC12 (rat pheochromocytoma) cells. COS-1 cells were cultured at 37°C (5% CO2) in DMEM supplemented with 10% (v/v) fetal bovine serum, penicillin (100 IU/mL), and streptomycin (100 mg/mL). PC12 cells were grown at 37°C (5% CO2) in complete medium consisting of 85% RPMI 1640, 10% (v/v) heat-inactivated horse serum, 5% (v/v) fetal bovine serum, penicillin (87 IU/mL), and streptomycin (87 mg/mL). To induce cell differentiation, PC12 cells were incubated in complete medium plus 100 ng/mL 2.5S nerve growth factor (NGF) (Promega) for 7 days. Prior to plating PC12 cells, the culture surfaces to be used for experiments were coated with poly-l-lysine. Both cells were transiently transfected with plasmid DNA using Lipofectamine (Gibco-BRL, Long Island, NY, USA), according to the protocol provided by the manufacturer. The localization of parkin-GFP fusions in live and fixed cells was examined by fluorescence microscopy between 24 and 48 h after transfection. For staining of Golgi complex, cells were fixed for 10 min in 4% paraformaldehyde in PBS at room temperature, permeabilized with PBS containing 0.2% (w/v) Triton X-100, and incubated with 4 µg/mL TRITC-conjugated wheat germ agglutinin (WGA, Sigma). After five washes with PBS, the cells were mounted with Vectashield (Vector Laboratories, Burlingame, CA, USA) and examined with a fluorescence microscope.

Cell fractionation

For cell fractionation studies, transiently transfected COS-1 cells were rinsed with ice-cold PBS, scraped into the same buffer, and centrifuged at 600 g for 5 min. Cell pellets were resuspended in homogenization buffer (10 mm Tris-HCl, pH 7.4, 0.25 m sucrose, and 1 mm zinc acetate) in the presence of a mixture of protease inhibitors (Complete EDTA-free, Boehringer Mannheim), sonicated at 4°C (10 s, three times), and nuclei and unbroken cells were pelleted by centrifugation at 1000 g for 10 min. The postnuclear supernatant was centrifuged at 100 000 g for 1 h at 4°C to separate cytosolic and membrane fractions. The relative amount of parkin-GFP fusion was analyzed by SDS–PAGE followed by immunoblotting with anti-GFP antibody.

Results

Localization of parkin on the secretory vesicles and the TGN

When U-373MG cells were treated with anti-parkin antibody, we found a positive staining on the perinuclear region. Parkin was colocalized with γ-adaptin, a marker specific for TGN (Figs 1a–c) (Ahle et al. 1988). In contrast, differentiated SH-SY5Y cells with properties of dopaminergic neurons (Biedler et al. 1978) showed punctate distribution of parkin in the cell bodies and cell processes, in addition to the perinuclear staining (Figs 1d–f). This clear punctate appearance is indicative of localization of parkin to secretory vesicles. To confirm this observation, we performed double-labeling experiments and confirmed that parkin was colocalized with synaptotagmin I, a major membrane protein of secretory vesicles budding from the TGN (Figs 1d–f) (Orci et al. 1987; Tooze et al. 1987; Tooze and Huttner 1990; Jung and Scheller 1991; Regnier-Vigouroux et al. 1991; Walch-Solimena et al. 1993; Bauerfeind et al. 1994). In these cells, anti-parkin antibody recognized a protein of approximately 52 kDa, which disappeared when the antibody was pre-incubated with antigen (Fig. 1g). These results convincingly indicated that the antibody was useful for detection of parkin in U-373MG and differentiated SH-SY5Y cells.

Figure 1.

 Colocalization of parkin with γ-adaptin and synaptotagmin I in U-373MG and differentiated SH-SY5Y cells, respectively. U-373MG and differentiated SH-SY5Y cells were stained with anti-parkin antibody (a and d, respectively), and then stained with anti-γ-adaptin (b) and anti-synaptotagmin I (e) antibodies. The samples were examined with a MRC-1024 confocal microscope and superimpositions were produced to compare the staining pattern (c and f). Scale bars = 50 µm. To determine the specificity of anti-parkin antibody, an identical blot of each cell lysate (5 µg) was incubated with anti-parkin antibody (g, left-hand panel) or with the antibody pre-absorbed with excess antigen (aa293–306; g, right-hand panel).

Association of parkin with SVs in rat brain

First, SVs were fractionated from rat brain extracts by differential fractionation and the presence of parkin was analyzed by immunoblotting. In each fraction, anti-parkin antibody recognized a protein of approximately 52 kDa that disappeared when the antibody was pre-incubated with antigen (Fig. 2a). These results strongly indicated that the antibody was useful for the detection of parkin in rat brain extracts. As the purification of SVs proceeded, parkin was concentrated together with two integral SV membrane proteins, synaptotagmin I and synaptophysin, which were known to be transported from the TGN to the nerve terminal (Bauerfeind et al. 1994; Orci et al. 1987; Tooze et al. 1987; Tooze and Huttner 1990; Jung and Scheller 1991; Regnier-Vigouroux et al. 1991). This finding strongly suggested the localization of parkin on the same membrane organelles as these SV membrane proteins (Fig. 2a). However, parkin was readily released into the supernatant after high-speed centrifugation of the crude synaptosomal lysate (LS2: see Materials), whereas synaptotagmin I and synaptophysin were not released into the supernatant (Fig. 2a). This finding indicates a weaker membrane-association of parkin compared with the two integral membrane proteins. Synapsin I, known as a peripheral SV membrane protein (Huttner et al. 1983), exhibited a distribution pattern similar to that of parkin (Fig. 2a). On the other hand, α-synuclein, which is abundantly present in presynaptic terminals, was not copurified with SV fraction, but was concentrated in LS2 (Fig. 2a), as reported previously (Kahle et al. 2000). We also performed further fractionation of SVs by SDG centrifugation. As seen in Fig. 2(b), synaptotagmin I, synaptophysin and synapsin I were recovered as two broad peaks (fractions 3–5 and fractions 10–12), corresponding to the small clear-core SVs and the large dense-core vesicles, respectively (Bauerfeind et al. 1994; Efthimiopoulos et al. 1998). Parkin was concentrated in these two fractions, strongly suggesting its association with the SVs. The presence of parkin, similar to synapsin I but not synaptotagmin I/synaptophysin, in the top fractions of gradient (fractions 1 and 2) again indicates the weak association of parkin with membranes.

Figure 2.

 (a) Parkin concentrations in various subcellular fractions in the rat brain. Subcellular fractions obtained are designated as described in Materials and methods. Aliquots of fractions, containing equal amounts of protein, were subjected to SDS–PAGE followed by immunoblotting with anti-parkin, anti-synaptotagmin I, anti-synaptophysin, anti-synapsin I and anti-α-synuclein antibodies. To determine the specificity of anti-parkin antibody against rat brain tissue, an identical blot of each fraction was incubated with the antibody pre-absorbed with excess antigen (aa293–306; pre-absorbed anti-parkin). (b) Distribution of parkin, synaptotagmin I, synaptophysin and synapsin I in SV fraction by SDG. Equal volumes of each fraction were analyzed by SDS–PAGE followed by immunoblotting with anti-parkin, anti-synaptotagmin I, anti-synaptophysin and anti-synapsin I antibodies. (c) Electron micrograph illustrating immunoperoxidase staining for parkin, synaptotagmin I and synaptophysin in frozen sections of fixed crude synaptosomal fraction. Note that the reaction products for parkin are associated with vesicles of various sizes (arrows) and the lack of any product in the mitochondrion. Very similar staining patterns were observed for synaptotagmin I and synaptophysin. The section was also stained with rabbit pre-immune serum as a negative control. Mt, mitochondrion; arrowheads, postsynaptic densities. Scale bar = 250 nm.

To gain direct evidence for the association of parkin with SVs, immunoelectron microscopy was performed on frozen sections of the crude synaptosomal fraction (P2′: see Materials). Immunoreactivity for parkin was detected on vesicular structures that were also stained with synaptotagmin I and synaptophysin, indicating that parkin was localized on the SVs (Fig. 2c).

Release of parkin from SV membranes

Parkin has no obvious amino acid sequences for targeting signal and transmembrane domains based on computer analysis (Kyte and Doolittle 1982), yet it is associated with SV membranes. How could this happen? To address this question, we examined the effects of ionic strength on the association of parkin with SVs (Fig. 3a). Under low-salt conditions at neutral pH, parkin was recovered mostly in the pellets. Increasing the salt concentrations, even to 150 mm KCl, resulted in a considerable loss of parkin from the SVs, whereas synaptophysin remained in the pellets. These results indicate that parkin is weakly associated with the cytoplasmic surface of SV membranes. To our surprise, parkin was not released from the SVs by a non-ionic detergent Triton X-100, whereas synaptophysin was readily solubilized (Fig. 3b). This manner of parkin release from SV membranes was quite similar to that of synapsin I, again indicating that parkin was peripherally localized on the SVs (Figs 3a and b).

Figure 3.

 Effects of various salts and non-ionic detergent on the solubilization of parkin, synaptophysin and synapsin I from SVs. (a) Equal amounts of SV fraction were treated with the indicated final concentrations (mM) of KCl, NaCl, or MgCl2. (b) Equal amounts of SV fraction were incubated in the absence (Control) or presence of 1% (w/v) Triton X-100. After separation of soluble and insoluble material by centrifugation, equal volumes of each fraction were subjected to SDS–PAGE followed by immunoblotting with anti-parkin, anti-synaptophysin and anti-synapsin I antibodies.

Subcellular localization of wild-type parkin-GFP in transfected cells

Although the above results indicate that parkin is associated with SVs, its binding domain is not known. To further study this issue, we transfected COS-1 cells and NGF-treated PC12 cells with a parkin-GFP fusion gene and determined the subcellular localization by fluorescence microscopy. In COS-1 cells, intense green fluorescence was detected in the perinuclear region together with red fluorescence of WGA, a marker for the Golgi region (Figs 4a–c). In NGF-treated PC12 cells, parkin-GFP was detected as a punctate distribution along the process and in the cell body, most likely secretory vesicles, in addition to perinuclear localization (Fig. 4d). The distribution of parkin-GFP in COS-1 and NGF-treated PC12 cells was quite similar to the profile of authentic parkin in glioblastoma U-373MG cells and differentiated neuroblastoma SH-SY5Y cells, respectively (Fig. 1). These results also indicated that tagging of parkin with GFP did not alter its intracellular localization in transfected cells.

Figure 4.

 Subcellular localization of wild-type parkin-GFP. COS-1 cells were transfected with parkin-GFP (a) and stained with a Golgi marker, WGA (b). Parkin-GFP was colocalized with WGA (c). In NGF-treated PC12 cell, parkin-GFP was visualized on punctate structures in process and perinuclear region (d). Scale bars = 50 µm.

Binding ability of mutant parkins to membranes

To determine the domains of parkin required for membrane association, we constructed a series of mutant parkin-GFP fusion genes and performed transfection studies for expression in COS-1 cells (Fig. 5a1). This was followed by fluorescence microscopy and subcellular fractionation, then by immunoblotting. Mutant parkins of ΔEx6–12 or ΔEx1–5 were localized in the perinuclear region of COS-1 cells in a pattern similar to the wild-type parkin-GFP (Fig. 5b). Such perinuclear fluorescence was colocalized with WGA. This was in contrast to the diffuse fluorescence pattern seen with mutant parkin of Ubl-domain alone (Fig. 5b). These results indicate that the membrane-binding domains reside in the amino acid moieties 77–206 (Ex 3–5) and 207–465 (Ex 6–12). To narrow down the responsible region, three additional mutant parkins were tested. Again, mutant parkin of Ubl-domain alone was not associated with the membrane, whereas mutant parkin of ΔEx4–12 was partially located in the perinuclear region and mutant parkin of ΔEx5–12 was more densely accumulated in the Golgi region (Fig. 5b). These results indicate the presence of binding domains at the amino acid moieties 77–138 (Ex 3) and 139–178 (Ex 4). Mutant parkin of Ubl-RING2 showed weak association with the perinuclear region, consistent with the presence of the binding ability in the amino acid moieties 418–465 (RING2), whereas mutant parkin of ΔEx1–5 more intensely accumulated in the Golgi region, indicating that the amino acid moieties 207–417 (Ex 6–11) were also capable of targeting the membrane (Fig. 5b). Subcellular fractionation showed that wild-type parkin and mutant parkins of ΔEx6–12, ΔEx1–5 and ΔEx5–12 were abundant in the high-speed membrane fractions, whereas mutant parkin of Ubl-domain was only found in the cytosol (Fig. 5c). Mutant parkins of ΔEx4–12 and Ubl-RING2 were equally collected in both the high-speed pellets and supernatants (Fig. 5c). The results of fractionation studies are in agreement with the data of fluorescence studies on transfected cells (Fig. 5b). We therefore conclude that parkin can associate with the membranes through a broad region except the Ubl-domain of the molecule.

Figure 5.

 (a) Schematic illustration of parkin-GFP fusions various types of truncated mutants (1) and mutants identified in AR-JP patients (2). Dotted lines indicate deleted regions. Arrows show the sites of point mutations. Ubl, ubiquitin-like domain; RING, RING-finger motif; IBR, in-between-RINGs. (b and d) Visualization of various mutants of parkin-GFP. COS-1 cells were transfected with parkin-GFP fusions (top panels) and stained with a Golgi marker, WGA (middle panels) and superimposed images were produced to compare the staining pattern (bottom panels). Scale bars = 50 µm. (c and e) Cell fractionation of various mutants of parkin-GFP. COS-1 cells were transfected with parkin-GFP fusions, and lysates were separated into supernatants (S) and pellets (P). Equal volumes of each fraction were subjected to SDS–PAGE followed by immunoblotting with anti-GFP antibodies.

Finally, we examined various mutant parkins identified in AR-JP patients for their abilities to associate with the membrane (Fig. 5a2). All AR-JP mutant parkin-GFP fusions were found mainly in the Golgi region in COS-1 cells (Fig. 5d). Subcellular fractionation also showed that all of these mutant parkin-GFP fusions were mainly collected in the membrane fractions (Fig. 5e). These results indicate that mutant parkins produced in AR-JP patients appear to retain their abilities to associate with the membranes.

Discussion

We previously postulated that parkin may be involved in the vesicular transport system based on its subcellular localization (Shimura et al. 1999). Here, we identified precise intracellular localization of parkin using cultured cells and rat brain extracts. We demonstrated a perinuclear localization and punctate appearance in neuronal processes of parkin by immunocytochemical analysis on the cultured neuroblastoma SH-SY5Y cells indicating its localization in the Golgi complex, particularly TGN, and the secretory vesicles. In general, transporting vesicles travel a long distance to their destinations via membrane-trafficking in neuronal axons, because the axon lacks a protein synthesis machinery. Thus, these observations strongly suggest that parkin is linked to the vesicular transport system of neuronal cells.

Next, we examined whether or not parkin was associated with SVs in the brain; SVs are equivalent to the secretory vesicles in cultured cells (Bauerfeind et al. 1994). In this regard, fresh brain tissue is essential for isolation of nerve terminals containing SVs (Hell and Jahn 1998), and thus we used rat brain to purify synaptosomes and found that parkin was localized on the cytoplasmic surface of the SVs. Co-purification of parkin with SVs required the use of low-salt conditions throughout the experimental procedure. Even under physiological ionic strength, parkin was released from the SVs. It is possible that the association of parkin with SVs was artefactual. To test this possibility, we used synapsin I (Huttner et al. 1983) and α-synuclein (Kahle et al. 2000), both of which are abundantly present in presynaptic terminals, for the same analyses. Synapsin I, specifically localized in the SVs, also required low-salt conditions for its SV association and was released even at physiological ionic strength in a manner quite similar to parkin. Moreover, α-synuclein was not collected in SV fraction but in soluble fraction of synaptosomes, although α-synuclein has been reported to be localized on the SVs in immunohistochemical studies (Iwai et al. 1995). The release of these proteins, even under low-salt conditions, might be caused by non-physiological dilutions of the organelles such as hypotonic lysis of the synaptosomes. Taken together, it seems unlikely that parkin associates with the SV membranes artefactually. Moreover, we confirmed the localization of parkin on SVs using the synaptosomal fraction by immuno-electron microscopic analysis. Immunoreactivity for parkin was detected on vesicular structures of various sizes. In comparison, purified SVs reported previously (Hell and Jahn 1998) were of a uniform size. The reason for the heterogeneous morphology is the use of crude synaptosomal fraction as a sample instead of chromatography purified fraction. Because immunoreactivities for synaptotagmin I and synaptophysin were also seen on vesicles of various sizes in our synaptosomal fractions, part of these vesicles might correspond to SV precursors (Okada et al. 1995). Indeed, recent studies have shown that synaptophysin is localized on vesicular organelles of various sizes and shapes in axon terminals of mouse dorsal root ganglion cells (Nakata et al. 1998). Thus, the relatively large-sized organelles in the synaptosomes might correspond to tubulovesicular organelles that mediate the transport of newly synthesized proteins from the TGN to the presynaptoc terminals (Nakata et al. 1998). Based on these findings, we speculate that parkin is synthesized on free ribosomes, anchored to the surface of the TGN, then transported as a component of SVs to the synaptic terminal. This transport mechanism is similar to other neuronal/peripheral membrane proteins such as growth-associated protein (GAP)-43 and synaptosomal-associated protein (SNAP)-25 (Liu et al. 1994; Veit et al. 1996).

How does parkin attach to SV membranes? We showed that parkin was easily dissociated from SV membranes by increasing ionic strength and, hence, we thought parkin association with membrane was not via lipid moiety (palmitoylation) as was reported for GAP-43 and SNAP-25 (Skene and Virag 1989). Membrane binding via palmitoylation is resistant to high-salt extraction. Thus, it is likely that the interaction of parkin with SV membranes is electrostatic in nature, which is in fact known for synapsin I, a phosphoprotein specifically localized on the SVs (Huttner et al. 1983). The association of synapsin I with SV membrane is dynamically regulated by phosphorylation (Hosaka et al. 1999). In this respect, parkin also has several phosphorylation sites, and it is possible that membrane binding of parkin may be regulated by the phosphorylation–dephosphorylation mechanism.

Recently, a number of RING finger-containing proteins have been implicated in the ubiquitin-proteasome pathway acting as a ubiquitin-protein ligase (Deshaies 1999; Harper and Elledge 1999; Joazeiro et al. 1999; Lorick et al. 1999; Martinez-Noel et al. 1999; Moynihan et al. 1999; Xie and Varshavsky 1999). In fact, in a separate study, we identified that parkin functions as a ubiquitin-protein ligase (Shimura et al. 2000). To our knowledge, parkin is the first recognized SV protein involved in the ubiquitin-proteasome pathway. In AR-JP patients, the absence of Lewy bodies is a unique pathologic feature, suggesting that parkin is essential for the formation of Lewy bodies. In contrast, in sporadic PD patients, parkin is detected in Lewy bodies (Shimura et al. 1999) together with abnormally accumulated proteins such as α-synuclein (Spillantini et al. 1997), which is degraded by the ubiquitin-proteasome pathway (Bennett et al. 1999). Interestingly, α-synuclein is associated with SVs (Iwai et al. 1995) and carried by the vesicle-moving fast component of axonal transport (Jensen et al. 1998). Moreover, similarities exist in the expression patterns of brain mRNAs (Solano et al. 2000) and intracellular distribution of parkin and α-synuclein, suggesting that these two proteins might be implicated in a common pathway leading to neurodegeneration of the SN. The insolubility of parkin in non-ionic detergent suggests that it may form a large protein complex or interact with cytoskeletal elements such as tau protein. Interestingly, the presence of neurofibrillary tangles with tau protein in the brain of a AR-JP patient have been reported (Mori et al. 1998). Thus, parkin may interact with not only α-synuclein but also with tau protein.

Membrane fractionation experiments of cells transfected with a series of mutant parkin constructs demonstrated that a broad region of parkin, excluding Ubl-domain, is responsible for membrane-binding. As a ubiquitin-protein ligase, parkin recruits ubiquitin-conjugating enzymes, UbcH7 and UbcH8, through its C-terminal RING-finger motifs (Shimura et al. 2000; Zhang et al. 2000) and most ubiquitin-conjugating enzymes are soluble proteins (Biederer et al. 1996). It seems unlikely that parkin-bound ubiquitin-conjugating enzymes mediate the membrane association, although the exact nature remains to be elucidated.

We also showed that several parkin mutants present in AR-JP patients were equally capable of membrane-association. In these patients, the loss of ligase activity of parkin might underlie the pathologic consequence of the mutations. The genotype–phenotype correlations so far not found in the AR-JP patients remains as further challenging investigation (Hattori et al. 1998a; Hattori et al. 1998b; Abbas et al. 1999).

There is a growing evidence to support that abnormal accumulation of the ubiquitinated proteins is common in many neurodegenerative disorders such as Alzheimer's disease, amyotrophic lateral sclerosis, the prion disorders and PD (Alves-Rodrigues et al. 1998). Interestingly, parkin is directly involved in the ubiquitin-proteasome pathway, in which it plays a crucial role as a ubiquitin-protein ligase to find and trap a specific target protein(s). Thus, the loss-of-function of parkin should lead to accumulation of non-ubiquitinated target protein(s). Since parkin is present in SVs, it is reasonable to assume that a target protein(s) could be located in the synaptic terminals. We propose that synaptic accumulation of parkin-specific target protein(s) for ubiquitination might result in the degeneration of nigral neurons. The isolation of target protein(s) is crucial for elucidation of the molecular mechanisms of nigral neuronal cell death in conjunction with the role of the intriguing protein ‘parkin’.

Acknowledgements

The authors thank Drs Kimie Murayama and Noriko Shindo for technical help. This study was in part supported by: Grant-in-Aid for Scientific Research on Priority Areas and Grant-in-Aid for High Technology Research Center from Ministry of Education, Science, Sports, Technology, and Culture, Japan; Grant-in-Aid for Health Science Promotion; Grant-in-Aid for Neurodegenerative disorders from Ministry of Health and Welfare, Japan; and a ‘Center of Excellence’ Grant from the National Parkinson Foundation, Miami, Florida, USA.

Ancillary