Phosphorylation of Parkin by the Cyclin-dependent Kinase 5 at the Linker Region Modulates Its Ubiquitin-Ligase Activity and Aggregation*

Mutations in Parkin are responsible for a large percentage of autosomal recessive juvenile parkinsonism cases. Parkin displays ubiquitin-ligase activity and protects against cell death promoted by several insults. Therefore, regulation of Parkin activities is important for understanding the dopaminergic cell death observed in Parkinson disease. We now report that cyclin-dependent kinase 5 (Cdk5) phosphorylates Parkin both in vitro and in vivo. We found that highly specific Cdk5 inhibitors and a dominant negative Cdk5 construct inhibited Parkin phosphorylation, suggesting that a significant portion of Parkin is phosphorylated by Cdk5. Parkin interacts with Cdk5 as observed by co-immunoprecipitation experiments of transfected cells and rat brains. Phosphorylation by Cdk5 decreased the auto-ubiquitylation of Parkin both in vitro and in vivo. We identified Ser-131 located at the linker region of Parkin as the major Cdk5 phosphorylation site. The Cdk5 phosphorylation-deficient S131A Parkin mutant displayed a higher auto-ubiquitylation level and increased ubiquitylation activity toward its substrates synphilin-1 and p38. Additionally, the S131A Parkin mutant more significantly accumulated into inclusions in human dopaminergic cells when compared with the wild-type Parkin. Furthermore, S131A Parkin mutant increased the formation of synphilin-1/α-synuclein inclusions, suggesting that the levels of Parkin phosphorylation and ubiquitylation may modulate the formation of inclusion bodies relevant to the disease. The data indicate that Cdk5 is a new regulator of the Parkin ubiquitin-ligase activity and modulates its ability to accumulate into and modify inclusions. Phosphorylation by Cdk5 may contribute to the accumulation of toxic Parkin substrates and decrease the ability of dopaminergic cells to cope with toxic insults in Parkinson disease.

Parkinson disease (PD) 2 is one of the most common neurodegenerative diseases. It is characterized by a progressive degeneration of dopaminergic neurons in the substantia nigra and the presence of cytoplasmic inclusions called Lewy bodies in surviving neurons (1).
The large majority of PD cases are sporadic, but six genes have been found to be mutated in familial forms of PD (1)(2)(3). Mutations in Parkin are the most frequent cause of familial PD and are responsible for a large percent of autosomal recessive juvenile parkinsonism (AR-JP) (4). This juvenile form of the disease is characterized by the death of dopaminergic neurons in the substantia nigra with the absence of Lewy bodies in surviving neurons, raising the possibility that Parkin may be essential for the formation of Lewy bodies (1,5). Parkin may also play a role in sporadic PD, as several studies demonstrated that Parkin immunoreactivity is present in Lewy bodies (6 -8).
Parkin has a ubiquitin-like domain at the N terminus connected to the C-terminal RING finger domains by a linker region (1). The presence of the RING finger domains confers to Parkin the ability to work as an E3 ubiquitin-ligase enzyme (9,10). On the other hand, little is known about the function of the linker region.
Drosophila Parkin null mutants display mitochondria dysfunction (18), indicating that Parkin may be important in maintaining mitochondrial function. Parkin also protects cells against toxicity elicited by many insults, including ␣-synuclein, Pael-R and p38 overexpression, PD-linked stressors, H 2 O 2 , and unfolded protein stress response (12, 19 -22). Although the protective function of Parkin may rely on its E3 ubiquitin-ligase activity (14,22), some Parkin disease mutants that retain the E3 ubiquitin-ligase activity lack neuroprotective activity (16). Therefore, it is important to clarify the connection between the E3 ubiquitin-ligase activity of Parkin and its ability to protect against cell death. On the other hand, the neuroprotective function of Parkin also depends on its solubility, as some PD-linked stressors induce Parkin aggregation, which seems to decrease its neuroprotective activity (20).
Phosphorylation has been known as an important regulator of E3 ubiquitin-ligases (23,24). Cdk5 activity has been implicated in the pathogenesis of PD. Cdk5 is present in Lewy bodies (25)(26)(27) and is required for 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-elicited dopaminergic cell death in mice models (28). In an attempt to better understand the mechanisms that regulate Parkin activity, we sought to investigate whether the protein kinase Cdk5 phosphorylates Parkin and whether this phosphorylation influences Parkin function.
We report that Parkin is phosphorylated in vitro and in vivo by Cdk5. We show that phosphorylation by Cdk5 regulates the catalytic activity of Parkin as measured by auto-ubiquitylation and the ability to ubiquitylate its substrates synphilin-1 and p38. We also identified serine 131 as the major Cdk5 phosphorylation site in Parkin. The Cdk5 phosphorylation-deficient S131A Parkin mutant displays increased auto-ubiquitylation and ubiquitylates more efficiently synphilin-1 and p38 proteins. Furthermore, S131A Parkin mutant is more prone to aggregation upon proteasome inhibition and also increases the formation of synphilin-1/␣-synuclein inclusions. Our results suggest that phosphorylation of Parkin by Cdk5 decreases its E3 ubiquitin-ligase activity and regulates the formation of cytosolic inclusions relevant to PD.

EXPERIMENTAL PROCEDURES
Materials-[ 32 P]Orthophosphate and [ 32 P]ATP were purchased from Amersham Biosciences. Roscovitine and other protein kinase inhibitors were purchased from Sigma.
Site-directed Mutagenesis-Putative Cdk5 serine/threonine sites in Parkin were predicted by NetPhos and Motif Scan programs. Full-length Parkin constructs mutated at predicted Cdk5 or additional kinase sites were generated by PCR using primers that contained alanine codons instead of serine/threonine ones. Mutations of Cdk5 sites were confirmed by doublestrand sequencing.
Cell Culture and Transfections-HEK 293 and SH-SY5Y cells (kindly provided by K. L. Lim) were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum in a 5% CO 2 atmosphere. Cells were transiently transfected with N-terminally tagged pRK5 and pFLAG-CMV-2 plasmids utilizing Lipofectamine 2000 (Invitrogen) and were processed after 36 h.
For experiments using siRNA, HEK 293 cells were transfected with 100 nM siRNAs using Lipofectamine 2000 and processed after 72 h. Silencer validated siRNA to Cdk5 (ID number 1466), and negative control siRNA 1 was obtained from Ambion.
For co-immunoprecipitations of endogenous proteins, rat brains were homogenized in buffer used for the co-immunoprecipitation experiments of transfected cells. Brain homogenates were clarified by centrifugation at 13,000 ϫ g for 5 min. Anti-Cdk5 antibody was coupled to protein G beads (30) and incubated for 8 h with brain homogenate (2 mg/ml). Immunoprecipitates were washed with lysis buffer containing 500 mM NaCl and were detected by Western blot.
In Vitro Kinase Assays-HEK 293 cells were transfected with HA-Parkin cDNAs. Thirty six hours after transfection, the cells were lysed as described above for the co-immunoprecipitation experiments. HA-Parkin was immunoprecipitated with anti-HA antibody and washed with lysis buffer containing 500 mM NaCl. Immunoprecipitated Parkin was incubated with recombinant Cdk5/p25 (Upstate) at 37°C for 1 h in buffer containing 40 mM Tris (pH 7.6), 2 mM dithiothreitol, 5 mM MgCl 2 , 2 g/ml soybean trypsin inhibitor, 0.05 mM unlabeled ATP, and 0.25 mCi/ml [␥-32 P]ATP. Reactions were stopped by adding SDS-sample buffer and analyzed by SDS-PAGE. The amount of 32 P-labeled Parkin was quantified by PhosphorImager analysis. Loading of immunoprecipitated HA-Parkin was determined by Western blot using anti-HA antibody.
In Vivo Phosphorylation Assays-After overnight serum starvation in phosphate-free medium, transfected HEK 293 cells were incubated for 3-6 h at 37°C with serum-free/phosphatefree medium containing 200 -400 Ci/ml [ 32 P]orthophosphate. Cells were harvested and lysed in buffer containing 50 mM Tris-HCl (pH 7.4), 140 mM NaCl, 1% Triton X-100, 0.1% SDS, 20 mM NaF, 2 mM Na 3 VO 4 , 10 mM PP i , 20 mM ␤-glycerol phosphate, 30 M MG132 and protease inhibitor mixture (Complete, Roche Applied Science). Immunoprecipitation of HA-Parkin was carried out with anti-HA antibody for 4 h at 4°C. Beads were washed with lysis buffer supplemented with 500 mM NaCl and analyzed by 8% SDS-PAGE. Densitometric quantification of radiolabeled HA-Parkin was carried out by PhosphorImager analysis. Loading of immunoprecipitated HA-Parkin was determined by Western blot using anti-HA antibody. For the phosphorylation experiments using transfected Parkin with myc-Cdk5/His-p25 or addition of 1 M okadaic acid, one 35-mm plate was used for each condition. For the experiments with transfected Parkin but in the absence of Cdk5 or okadaic acid, one 100-mm plate was used for each condition.
For the phosphorylation experiments of endogenous Parkin, untransfected HEK 293 cells were serum-starved overnight in phosphate-free medium and then incubated in regular medium for an additional 6 h, in the absence or the presence of roscovitine. During the last hour of incubation, 10 M A23187 was added to the medium. Two 100-mm plates were used for each condition. Cells were harvested and lysed in the buffer described above for the in vivo phosphorylation experiments using transfected cells. Immunoprecipitation of Parkin was car-ried out with anti-Parkin antibody for 4 h at 4°C. Beads were washed with lysis buffer supplemented with 500 mM NaCl and analyzed by 8% SDS-PAGE. Phosphorylated Parkin was detected by Western blot using a selective anti-phosphoserine antibody according to manufacturer's instructions (Qiagen). Equal loading of immunoprecipitated Parkin was determined by Western blot using anti-Parkin antibody.
In Vitro Ubiquitylation Assays-For the in vitro ubiquitylation assays of pre-phosphorylated Parkin, HA-Parkin was cotransfected with myc-Cdk5 and His-p25 into HEK 293 cells. After 36 h, cells were lysed by sonication in buffer containing 50 mM Tris-HCl (pH 7.4), 140 mM NaCl, 1% Triton X-100, 0.1% SDS, 20 mM NaF, 2 mM Na 3 VO 4 , 30 M MG132 and protease inhibitor mixture (Complete, Roche Applied Science). Such harsh lysis conditions (sonication in the presence of Triton X-100 and SDS) were employed to prevent co-immunoprecipitation of Cdk5 with Parkin, which could interfere with the other components of the ubiquitin system. Immunoprecipitation of HA-Parkin was carried out with anti-HA antibody for 4 h at 4°C. Beads were extensively washed with lysis buffer supplemented with 500 mM NaCl. Immunoprecipitated Parkin was incubated in reaction medium containing 40 mM Tris (pH 7.6), 5 mM MgCl 2 , 2 mM dithiothreitol, 1 mM ATP␥S, 7.5 g of ubiquitin, 1 M ubiquitin aldehyde, 100 ng of ubiquitin-activating enzyme and 200 ng of UbcH7. The samples were incubated at 37°C for 1 h, and ubiquitylated Parkin was determined by Western blot using anti-HA antibody.
For the in vitro ubiquitylation assays of Parkin substrate, synphilin-1 and p38 were translated using TNT wheat germ in vitro translation kit (Promega) using [ 35 S]methionine (Amersham Biosciences). In vitro translated proteins were incubated with the immunoprecipitated HA-Parkin in the reaction medium described above. Reactions were incubated at 37°C for 1 h, terminated by the addition of SDS sample buffer, and resolved on SDS-polyacrylamide gels. 35 S-Synphilin-1 and 35 S-p38 were determined by PhosphorImager analysis.
Immunocytochemistry Assays-Transfected HEK 293 and SH-SY5Y cells were incubated in the absence or presence of 10 M lactacystin for 8 -16 h, fixed with 4% paraformaldehyde for 15 min, and blocked in phosphate-buffered saline containing 0.2% Triton X-100 and 5% normal goat serum. Cells were stained with anti-HA (Covance) and anti-Myc (Santa Cruz Biotechnology) as described (29). Immunolabeling was detected using fluorescein isothiocyanate-and Cy 3 -labeled secondary antibodies (The Jackson Laboratories). The percent of cells containing cytosolic inclusions was counted by an investigator unaware of the treatment groups. Statistics of the number of inclusion-containing cells were analyzed by paired two-tailed Student's t test.

RESULTS
Phosphorylation of Parkin by Cdk5-We found that Parkin has four putative Cdk5 phosphorylation sites according to the motif (S/T)PX(K/R) (31), all located in the linker region that connects the ubiquitin-like domain and the C-terminal RING fingers. Thus, we sought to investigate whether Parkin is a physiologically relevant target of Cdk5. We first carried out in vitro phosphorylation experiments where recombinant GST-Parkin was incubated with His-Cdk5 and its activator p25 (32). We found that Parkin is directly phosphorylated by Cdk5 (Fig.  1A). We also tested if Parkin is phosphorylated by GSK3␤, another protein kinase involved in neurodegenerative diseases (33). Different from Cdk5, we observed little phosphorylation of Parkin by GSK3␤ (Fig. 1A).
We next carried out in vivo phosphorylation assays in which HEK 293 cells were co-transfected with full-length HA-Parkin, myc-Cdk5, and its activator His-p25. Accordingly, Cdk5 increased by 238 Ϯ 40% the levels of Parkin phosphorylation (Fig. 1B). In agreement with the in vitro phosphorylation experiments, GSK3␤ promoted only a modest phosphorylation of Parkin.
We next investigated if endogenous Cdk5 could phosphorylate Parkin. For this, we first checked the levels of Parkin phosphorylation in the absence and presence of the phosphatase inhibitor okadaic acid, which was shown to increase Parkin phosphorylation levels (34). We found that the incubation with okadaic acid robustly increases the steady-state levels of Parkin phosphorylation in HEK 293 cells (Fig. 1C). We then carried out in vivo phosphorylation experiments in the presence of okadaic acid, and we found that the highly selective Cdk5 inhibitors roscovitine and olomoucine decreased by ϳ30% the phosphorylation of Parkin (Fig. 1D), suggesting that Parkin is phosphorylated in vivo by Cdk5.
To ensure that the effects we observed were specific for Cdk5, we carried out phosphorylation experiments in which Parkin was co-transfected with the dominant negative of Cdk5 (35). Accordingly, the dominant negative of Cdk5 promoted about a 30% decrease in Parkin phosphorylation levels in the presence of okadaic acid (Fig. 1E), confirming that Parkin is specifically phosphorylated by Cdk5.
We carried out additional in vivo phosphorylation experiments to determine whether Parkin is phosphorylated by Cdk5 using endogenous proteins and in the absence of okadaic acid. To determine whether endogenous Cdk5 phosphorylates Parkin, we transfected SH-SY5Y cells with HA-Parkin and treated with roscovitine in the absence of okadaic acid. Although the phosphorylation levels without okadaic acid are lower (Fig. 1C), they can be easily detected at longer exposure times (Fig. 2). We found that the addition of roscovitine significantly decreased the phosphorylation of Parkin ( Fig. 2A). We then carried out phosphorylation experiments in the absence of okadaic acid using untransfected HEK 293 cells, which were previously shown to express detectable levels of endogenous Parkin (34). We incubated the untransfected cells with the Ca 2ϩ ionophore A23187 in order to increase the pool of activated Cdk5 (36). We found that roscovitine decreased the phosphorylation of endogenous Parkin by about 30% (Fig. 2B). In addition, siRNA to Cdk5 promoted about a 35% decrease in the phosphorylation of endogenous Parkin compared with control (Fig. 2C). Together, these results suggest that Cdk5 phosphorylates endogenous Parkin.
The ability of Cdk5 to interact with Parkin was also investigated by co-immunoprecipitation experiments using HEK 293 cells co-transfected with HA-Parkin and myc-Cdk5. When anti-HA was used to immunoprecipitate HA-Parkin, myc-Cdk5 was found associated with HA-Parkin (Fig. 3A). The interaction of Cdk5 with Parkin is specific because the control FKBP12 did not co-immunoprecipitate with Parkin under the same experimental conditions (Fig. 3A). Co-transfection of the Cdk5 activator p25 did not disrupt the interaction between Parkin and Cdk5 (data not shown). Furthermore, addition of increasing concentrations of roscovitine did not alter the extent of co-immunoprecipitation of Cdk5 with Parkin, suggesting that Parkin phosphorylation does not modulate the physical interaction between Cdk5 and Parkin (Fig. 3B). In addition, Parkin specifically co-immunoprecipitated with Cdk5 from rat brain tissue (Fig. 3C) but not with control beads, indicating that these two proteins interact in vivo. Under our experimental conditions, we immunoprecipitated about 30% of the total Cdk5 present in brain lysate (data not shown). We estimate that about 4.5% of Parkin co-immunoprecipitates with Cdk5 in rat brain (Fig. 3C). This amount of co-immunoprecipitation is compatible with the transient nature of interaction between protein kinases and their protein substrates and with the fact that both Cdk5 and Parkin interact with several other proteins.  We next sought to identify the residues of Parkin that are phosphorylated by Cdk5. We generated a series of full-length HA-Parkin constructs with point mutations at the four putative Cdk5 sites located at the linker region. We co-transfected HEK 293 cells with myc-Cdk5, His-p25, and either wild-type or Parkin mutants. We found that the S131A mutant had 40 Ϯ 2% lower phosphorylation levels in vivo than wild-type Parkin or the other Cdk5 mutants (Fig. 4A). This indicates that serine 131 is a major Cdk5 phosphorylation site within Parkin.
Yamamoto et al. (34) have previously reported that Parkin is phosphorylated at the serine residues 101, 131, 136, 296, and 378, as determined by mass spectrometry analysis and in vitro phosphorylation assays. To determine the in vivo specificity of Ser-131 to Cdk5, we generated a Parkin construct where the serines 101, 136, 296, and 378 were mutated to alanine but not serine 131. This Parkin construct is referred to as Parkin ⌬Ser. We transfected HEK 293 cells with Parkin ⌬Ser and carried the in vivo phosphorylation experiments in the presence of different protein kinase inhibitors. Among the several protein kinase inhibitors utilized, only the Cdk5 inhibitor, roscovitine, was able to significantly inhibit Ser-131 phosphorylation (Fig. 4B). This inhibition was of 33.5 Ϯ 4.5%, whereas inhibitors of protein kinase C (bisindolylmaleimide I), protein kinase A (H-89), casein kinase I (IC261), and extracellular signal-regulated kinase (ERK) kinase (PD9805) promoted either no inhibition or very minor inhibition of Ser-131, indicating that Ser-131 of Parkin is preferentially phosphorylated by Cdk5 in cells.
Regulation of Parkin Auto-ubiquitylation by Cdk5-Parkin was shown to function as an E3 ubiquitin-ligase (9, 10). To investigate whether phosphorylation by Cdk5 regulates Parkin ubiquitin-ligase activity, we first carried out in vitro ubiquitylation experiments where immunoprecipitated Parkin was assayed for auto-ubiquitylation by incubation with purified components of the ubiquitin system. To avoid interference of Cdk5/p25 with the components of the ubiquitin system, we used pre-phosphorylated Parkin, in which CdK5/p25 was not present in the ubiquitylation assay. For this, we co-transfected HEK 293 cells with HA-Parkin, myc-Cdk5, and His-p25, and we immunoprecipitated Parkin in the presence of phosphatase inhibitors and under harsh conditions to avoid co-immunoprecipitation of Cdk5 (see "Experimental Procedures"). We observed that Parkin pre-phosphorylated by Cdk5 displays significantly less auto-ubiquitylation, indicating that Cdk5 modulates the ligase activity of Parkin (Fig. 5A).
We next examined if the phosphorylation by Cdk5 interferes with the in vivo auto-ubiquitylation of Parkin. For this, ubiquitylation of Parkin by FLAG-ubiquitin was detected by immunoprecipitating HA-Parkin from HEK 293 cells and probing with anti-FLAG. In accordance with the in vitro experiments, Co-immunoprecipitation of myc-Cdk5 was checked using an anti-Myc antibody (middle panel). Equal levels of HA-Parkin immunoprecipitation were checked using anti-HA antibody (lower panel). C, Parkin co-immunoprecipitates with endogenous Cdk5. Cdk5 was immunoprecipitated from rat brain lysate using mouse anti-Cdk5 antibody, and detection of co-immunoprecipitation was carried out using rabbit anti-Parkin antibody. Parkin was detected in Cdk5-immunoprecipitate but not in beads alone. The figures are representative of three (A and C) or two (B) independent experiments. we found that the in vivo auto-ubiquitylation of Parkin was significantly decreased by Cdk5/p25 overexpression (Fig. 5B).
To verify the involvement of the endogenous Cdk5 in Parkin auto-ubiquitylation, we tested the effect of the highly selective Cdk5 inhibitor roscovitine without Cdk5 overexpression. Accordingly, addition of increasing concentrations of roscovitine promoted a strong increase in Parkin auto-ubiquitylation in HEK 293 cells (Fig. 5C). Thus, the endogenous Cdk5 in HEK 293 cells diminishes Parkin auto-ubiquitylation.
Because the Cdk5 inhibitor roscovitine alters the phosphorylation state of several cellular proteins other than Parkin, we employed the S131A Parkin mutant as a tool to specifically evaluate the role of Cdk5 in Parkin function. To confirm the involvement of Cdk5 in Parkin auto-ubiquitylation, we monitored the ubiquitylation state of the Cdk5 phosphorylation-deficient S131A mutant in HEK 293 cells. The auto-ubiquitylation of wild-type and S131A Parkin mutant was monitored by immunoprecipitating HA-Parkin from HEK 293 cells co-transfected with FLAG-ubiquitin, and probing the blot with anti-FLAG. In agreement with the increase in the in vivo auto-ubiquitylation promoted by roscovitine, the autoubiquitylation of the Parkin S131A mutant was significantly higher than that of the wild-type Parkin (Fig. 5D).
Regulation of Parkin E3 Ubiquitin-Ligase Activity by Cdk5-To determine whether phosphorylation by Cdk5 also changes the Parkin E3 ubiquitin-ligase activity toward its substrates, we monitored the ubiquitylation of synphilin-1 and p38 (11,12,17). We first carried out in vitro ubiquitylation experiments where we immunoprecipitated Parkin, and we assayed for synphilin-1 and p38 ubiquitylation by incubating the immunoprecipitates with in vitro translated synphilin-1 or p38 and purified components of the ubiquitin system. In addition to being more auto-ubiquitylated (Fig. 5D), we found that the phosphorylation-deficient S131A mutant is more efficient in ubiquitylating in vitro synphilin-1 and p38 than the wild-type protein (Fig. 6, A  and B). Similar to the in vitro experiments, the in vivo ubiquitylation of p38 promoted by the S131A mutant was also higher than that observed with wild-type Parkin (Fig. 6C). This suggests that the ability of Parkin to ubiquitylate synphilin-1 and p38 correlates with its degree of auto-ubiquitylation, and both processes are modulated by Cdk5.
Role of Cdk5 Phosphorylation in Parkin Inclusion Formation-It has been shown that decreased proteasomal activity causes the formation of inclusions within the cytoplasm of cells overexpressing Parkin (37)(38)(39). Moreover, Parkin associates with and ubiquitylates synphilin-1/␣-synuclein inclusions (11,40). Therefore, the changes we found in Parkin ubiquitylation by Cdk5 phosphorylation prompted us to compare the formation of inclusion bodies between the phosphorylation-deficient S131A Parkin mutant with that of wild-type Parkin. Human dopaminergic SH-SY5Y cells were transfected either with wildtype or S131A Parkin mutant, treated with the proteasome inhibitor lactacystin, and the amount of Parkin inclusion bodies was determined by immunocytochemistry. We found that transfection of the S131A Parkin mutant elicited significantly more inclusions than the wild-type protein (Fig. 7, A and B). This suggests that the higher auto-ubiquitylation levels of the S131A mutant facilitate its aggregation into inclusions bodies.
We next investigated whether phosphorylation of Ser-131 of Parkin may also modulate synphilin-1/␣-synuclein inclusion body formation. It has been shown that Parkin ubiquitylates (11) and increases the formation of synphilin-1/␣-synuclein inclusion bodies (41). We now found that co-transfection of S131A Parkin mutant with synphilin-1 and ␣-synuclein significantly increased the formation of synphilin-1/␣-synuclein HA-Parkin was immunoprecipitated using an anti-HA antibody. Immunoprecipitates were analyzed for ubiquitylation using an anti-FLAG antibody (upper panel). D, phosphorylation-deficient S131A Parkin mutant displays higher auto-ubiquitylation levels. HEK 293 cells were co-transfected with HA-Parkin (wild-type or S131A mutant) and FLAG-ubiquitin and treated for 12 h with 10 M lactacystin. HA-Parkin was immunoprecipitated using an anti-HA antibody. Immunoprecipitates were analyzed for ubiquitylation using an anti-FLAG antibody (upper panel). The figures are representative of three independent experiments.
inclusions when compared with cells co-transfected with wildtype Parkin (Fig. 8, A and B). This suggests that phosphorylation of Parkin by Cdk5 modulates the amount of synphilin-1/ ␣-synuclein inclusion body formation.

DISCUSSION
Mutations in Parkin are responsible for the death of dopaminergic neurons in AR-JP (4). Parkin has also been implicated in sporadic PD because of its connection with the ubiquitin proteasome system (9, 10) and its presence in Lewy bodies of patients with sporadic PD (6 -8,38). In this paper we described a new mechanism that modulates Parkin E3 ubiquitin-ligase activity and auto-ubiquitylation. Our results shed light on the regulation of Parkin function by phosphorylation at the linker region, with implications for the pathogenesis of PD.
We demonstrated that Parkin is phosphorylated in vitro and in vivo by Cdk5 (Figs. 1 and 2), and that this phosphorylation robustly decreases the E3 ubiquitin-ligase activity of Parkin (Fig. 5). We also identified Ser-131 located at the linker region as the major site for Parkin phosphorylation by Cdk5 (Fig. 4). Accordingly, the Cdk5 phosphorylation-deficient S131A mutant is more auto-ubiquitylated and displays higher E3 ubiquitin-ligase activity toward synphilin-1 and p38 (Figs. 5 and 6). We also found that the S131A mutant is more prone to form inclusion in a human dopaminergic cell line and increases the formation of synphilin-1/␣-synuclein inclusions, indicating that Cdk5 regulates several aspects of Parkin function.
Cdk5 has been implicated as an essential modulator of synaptic transmission and neurodegenerative diseases (42). Cdk5 phosphorylates Tau, and hyperphosphorylated Tau aggregates into paired helical filaments characteristic of Alzheimer disease (43,44). Several groups reported the presence of Cdk5 in the Lewy bodies of sporadic PD (25)(26)(27). Cdk5 knock-out mice are less sensitive to dopaminergic cell death promoted by 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine, indicating a possible role of Cdk5 also in PD (28). The functional interaction between Cdk5 and Parkin we now describe further connects Cdk5 with PD. Studies that investigate Cdk5 activity in the sub-FIGURE 6. Cdk5 modulates the Parkin E3 ubiquitin-ligase activity toward synphilin-1 and p38. A, increased in vitro ubiquitylation of synphilin-1 by Parkin S131A mutant. HEK 293 cells were transfected with HA-Parkin wildtype (WT) or S131A mutant. In vitro translated synphilin-1 was incubated with immunoprecipitated (IP) HA-Parkin and the indicated components of the ubiquitin system. The levels of ubiquitylated synphilin-1 were visualized by autoradiography. B, increased in vitro ubiquitylation of p38 by Parkin S131A mutant. HEK 293 cells were transfected with HA-Parkin wild-type or S131A mutant. In vitro translated p38 was incubated with immunoprecipitated HA-Parkin and the indicated components of the ubiquitin system. The levels of ubiquitylated p38 were visualized by autoradiography. C, increased in vivo ubiquitylation of p38 by Parkin S131A mutant. HEK 293 cells were co-transfected with HA-p38, myc-Parkin (wild-type or S131A mutant), and FLAG-ubiquitin and treated for 12 h with 10 M lactacystin. HA-p38 was immunoprecipitated using an anti-HA antibody. Immunoprecipitates were analyzed for ubiquitylation using an anti-FLAG antibody (upper panel). The figure panels are representative of three (A and B) or four (C) independent experiments.

FIGURE 7. Cdk5 modulates Parkin inclusion formation in human dopaminergic cells.
A, SH-SY5Y cells were transfected with HA-Parkin (wild-type or S131A mutant) and incubated with 10 M lactacystin for 12 h. Immunocytochemistry was carried out using anti-HA (red) antibody. Nuclei were stained with TOPRO-3. Scale bar, 20 m. B, quantification of the percent of inclusion body formation in SH-SY5Y cells transfected with HA-Parkin (wild-type (WT) or S131A mutant) and treated with 10 M lactacystin. Error bars represent standard error of four independent experiments. ***, significantly different from control at p Ͻ 0.001.

FIGURE 8. Phosphorylation of Parkin at Ser-131 modulates synphilin-1/␣-synuclein inclusion body formation in human dopaminergic cells.
A, SH-SY5Y cells were co-transfected with myc-synphilin-1, HA-␣synuclein and Parkin, (wild-type (WT) or S131A mutant). Intracellular inclusions were visualized by immunocytochemistry using anti-Myc (green) and anti-HA (red) antibodies. Nuclei were stained with TOPRO-3. Scale bar, 20 m. B, quantification of the percent of inclusion body formation in SH-SY5Y cotransfected cells with myc-synphilin-1, HA-␣-synuclein, and Parkin (wild-type or S131A mutant). Error bars represent standard error of three independent experiments. ***, significantly different from control at p Ͻ 0.001. stantia nigra of PD patients will be important to corroborate the importance of Cdk5 in the pathogenesis of the disease.
We now show that a significant portion of Parkin is phosphorylated by Cdk5 because Cdk5 inhibitors and a dominant negative of Cdk5 inhibited by 30 -40% the in vivo phosphorylation of Parkin. We also found that endogenous Cdk5 phosphorylates endogenous Parkin, confirming the relevance of Cdk5 phosphorylation of Parkin in the absence of okadaic acid (Fig.  2B). In addition, Cdk5 phosphorylation decreased the E3 ubiquitin-ligase activity of Parkin, raising the possibility that the toxic roles attributed to Cdk5 in a PD model (42) could be mediated in part by a decrease in Parkin function and accumulation of toxic Parkin substrates.
Parkin was shown to be phosphorylated in vivo at serines 101, 131, 136, and 296 by mass spectroscopy analysis (34). Ser-131 was found by the same authors to be phosphorylated in vitro by cAMP-dependent protein kinase. Although we cannot discard the contribution of cAMP-dependent protein kinase or other unknown kinases to Ser-131 phosphorylation, we now present data in vivo suggesting that Ser-131 is preferentially phosphorylated by Cdk5 (Fig. 4B).
The finding of Ser-131 as the Cdk5 phosphorylation site in Parkin (Fig. 4) allowed us to analyze the effect of Cdk5 phosphorylation on the ability of Parkin to ubiquitylate itself and some of its substrates, synphilin-1 and p38 (Figs. 5 and 6). We found that the phosphorylation-deficient S131A Parkin mutant had increased ability to ubiquitylate itself as well as synphilin-1 and the p38, indicating that the auto-ubiquitylation of Parkin may be directly related to its ability to ubiquitylate relevant substrates. In agreement, S-nitrosylation of Parkin promotes a concomitant decrease of Parkin auto-ubiquitylation as well as Parkin-mediated ubiquitylation of synphilin-1 (45,46). On the other hand, some PD-linked Parkin mutants retain their selfubiquitylation activity but display decreased ability to ubiquitylate p38 and synphilin-1 (16). This suggests that some Parkin disease mutations may promote a dissociation between the auto-ubiquitylation activity of Parkin and its ability to ubiquitylate substrates. Thus, additional studies will be necessary to determine how strict is the connection between the auto-ubiquitylation and the ubiquitylation of substrates. Nonetheless, because p38 is a toxic protein that accumulates in PD brain tissues (12,17), our data raise the possibility that phosphorylation of Parkin by Cdk5 decrease p38 degradation and may contribute to its accumulation in PD brain tissues and therefore to the death of dopaminergic neurons.
The main site of Cdk5 phosphorylation (Ser-131) is located in the linker region of Parkin, which connects the N-terminal ubiquitin-like domain with the C-terminal RING fingers. Our data indicate that the linker domain has a regulatory role on the Parkin E3 ubiquitin ligase activity. In agreement with our data, Tanaka and co-workers (47) recently reported that the 14-3-3 protein specifically binds to the linker region of Parkin and strongly inhibits Parkin self-ubiquitylation and E3 ubiquitin ligase activity toward the substrate synphilin-1.
The use of the Cdk5 phosphorylation-deficient mutant also allowed us to correlate the increase of Parkin auto-ubiquitylation with the increase of inclusion body formation by Parkin in human dopaminergic cells (Figs. 5 and 7). Thus, Parkin autoubiquitylation may directly affect its ability to accumulate into inclusions. In accordance with our data, it was reported that PD-linked Parkin mutants that have increased auto-ubiquitylation, such as R275W and C431F, are more prone to aggregation into inclusion bodies (16). The accumulation of ubiquitylated proteins into inclusion bodies has also been proposed for other PD-related proteins present in Lewy bodies, such as synphilin isoforms and UCH-L1 (29, 48 -50). We have shown previously that the ubiquitylation of synphilin-1 is essential for its aggregation into inclusion bodies and that GSK3␤ modulates the degree of synphilin-1 ubiquitylation and inclusion formation (24,29). The data now presented correlating the degree of Parkin ubiquitylation and inclusion body formation strengthen the possibility of a more general connection between ubiquitylation of proteins involved in neurodegenerative diseases and inclusion body formation.
Parkin was shown to ubiquitylate and to increase the formation of synphilin-1/␣-synuclein inclusions (11,41). We now provide evidence that phosphorylation of Parkin at Ser-131 modulates this process. Based on the finding that synphilin-1/ ␣-synuclein inclusions were reported to be cytoprotective (51), we raise the possibility that Cdk5 may increase cell toxicity in pathological conditions by preventing the formation of these inclusions.
Parkin protects cells against the toxicity caused by a wide range of cellular insults (20 -22). Thus, it is conceivable that Cdk5 activation in PD may contribute to the accumulation of toxic Parkin substrates, such as p38, by decreasing the E3 ubiquitin ligase of Parkin. In sum, our findings shed new light on the regulation of Parkin function, with implications for the formation of intracellular inclusions and cell death mechanisms in PD.