Abstract
Mutations in the ubiquitously expressed gene PTEN-induced kinase 1 (Pink1) cause autosomal recessive Parkinson's disease. Pink1 encodes a putative serine/threonine kinase with an N-terminal mitochondrial targeting sequence. The mechanism that leads to selective degeneration of dopaminergic neurons via Pink1 mutations is unknown. A full-length pre-protein (66 kDa) and an N-terminally truncated mature form (55 kDa) have been described in human brain. Here, we report that the endogenous 66 kDa and 55 kDa Pink1 forms in cultured cells are not exclusive to mitochondria but also occur in cytosolic and microsome-rich fractions. Pink1 66 kDa is the predominant isoform in cultured cells. Using unbiased analyses of immunoisolated Pink1 complexes by mass spectrometry, co-immunoprecipitation and Hsp90 inhibitor studies, we identify Pink1 as a novel Cdc37/Hsp90 client kinase. This chaperone system influences both the subcellular distribution and the 66/55 kDa protein ratio of Pink1. PD-causing Pink1 mutations decrease whereas Parkin expression increases the Pink1 66/55 kDa protein ratio, biochemically linking Pink1 and Parkin and highlighting the potential relevance of this ratio for PD pathogenesis. Finally, we document the influence of Parkin on Pink1 subcellular distribution, providing further evidence for a common pathogenic pathway in recessive PD.
INTRODUCTION
Parkinson's disease (PD) is the most common neurodegenerative movement disorder, affecting more than 3% of the population over age 65. In PD, the formation of cytoplasmic protein inclusions (Lewy bodies) in dopaminergic neurons of the substantia nigra that project axons to the striatum is accompanied by progressive loss of these neurons. Abnormal protein aggregation, oxidative stress and, in particular, mitochondrial dysfunction have all been proposed to contribute to the pathogenesis of PD. The identification in recent years of causative genes for rare familial forms of PD that share clinical and neuropathological features with the more common, idiopathic form has greatly accelerated approaches to dissecting the molecular basis of PD ( 1 , 2 ). Understanding the normal and abnormal biology of these gene products is crucial to the development of disease-modifying therapeutic agents for this complex disorder.
PTEN-induced kinase 1 (Pink1) is one of three genes known to cause autosomal recessive PD (ARPD), and haploinsufficiency of Pink1 could also represent a susceptibility factor for some cases of idiopathic PD ( 2–4 ). Thus, understanding the properties of wild-type and mutant Pink1 should not only explain this rare familial ARPD but help illuminate the much more common idiopathic form. Human Pink1 encodes a 581-residue protein with a predicted N-terminal mitochondrial targeting sequence (MTS) and a large serine/threonine kinase domain ( 3 ). The confirmed mitochondrial localization and autophosphorylation activity of Pink1 help link mitochondrial signaling pathways to PD ( 3 , 5 ). Taken together, PD-causing mutations in the Pink1 kinase domain that reduce phosphorylation activity ( 5–7 ), the autosomal recessive inheritance pattern for Pink1 mutations ( 3 , 8 ) and the dopaminergic cell loss found in Pink1 knock-out flies ( 9 , 10 ) all strongly suggest that loss of Pink1 kinase activity can precipitate PD.
The physiological function of Pink1 is unknown. Pink1-deficient flies have alterations in mitochondrial morphology ( 9 , 11 ), and neuronal SH-SY5Y cells and non-neuronal HeLa cells lacking Pink1 are reported to have increased susceptibility to apoptotic cell death ( 12 , 13 ). Pink1 impairment can affect the apoptotic cascade upstream of cytochrome C release ( 14 ). Over-expression of Pink1, but not ARPD-linked mutant Pink1, protects cells against several stressors, including proteosomal inhibition and oxidative stress ( 3 , 14 ). Very recently, the mitochondrial chaperone TRAP1 has been identified as the first Pink1 substrate ( 15 ). Recent findings also link Pink1 to the two other known ARPD genes, Parkin and DJ-1 ( 9 , 11 , 16 ). Of special importance in this regard are the observations that Parkin and Pink1 knock-out flies have very similar phenotypes and that Parkin rescues Pink1 knock-out phenotypes such as muscle degeneration and mitochondrial fragmentation, suggesting that Pink1 and Parkin share a common pathway, with Parkin apparently functioning downstream of Pink1 ( 9 , 11 ).
Two major Pink1 isoforms with apparent molecular weights of ∼66 and ∼55 kDa, have been detected in human brain extracts and upon cellular Pink1 overexpression ( 5 , 17 ). Both 66 and 55 kDa Pink1 have been localized to mitochondria, mainly to the inner and outer membrane fraction, but some evidence for an additional, non-mitochondrial Pink1 localization has been provided ( 5 , 6 , 14 , 17 ). Pink1 55 kDa appears to represent a mature form derived from Pink1 66 kDa upon removal of the leader sequence ( 5 ). Proteasomal inhibition alters the 66/55 kDa ratio by increasing 55 kDa Pink1, and elevated levels of Pink1 55 kDa have been found in the substantia nigra and cerebrellum of PD patients, suggesting that the 66/55 kDa ratio might be implicated in Pink1 pathobiology ( 18 ). It has been speculated that certain PD-linked stressors can increase levels of 55 kDa over 66 kDa Pink1 ( 18 ). Nevertheless, it is unclear if such an increase represents a neuroprotective response or rather contributes actively to PD pathogenesis. With regard to a potentially non-mitochondrial locus of Pink1, it is notable that kinases often have cell- and compartment-specific functions and sometimes show opposing functions in different compartments. For all these reasons, it is important to elucidate factors that regulate the generation of the 66 and 55 kDa Pink1 forms and their subcellular localizations, as such factors could become therapeutic targets for increasing the neuroprotective pool of Pink1.
In the present study, we searched in an unbiased fashion by mass spectrometry for Pink1-interacting proteins. We identify Pink1 as a novel Hsp90/Cdc37 client kinase. Further, we confirm an additional, non-mitochondrial subcellular locus for endogenous Pink1 and then demonstrate the influence of the Hsp90/Cdc37 chaperone system on both the subcellular distribution and the 66/55 kDa protein ratio of Pink1. Finally, additional experiments reveal the potential importance of the Pink1 66/55 kDa ratio for PD and suggest a biochemical link between Parkin and Pink1.
RESULTS
Solubility and subcellular localization of Pink1 in mammalian cultured cells
The putative N-terminal mitochondrial leader sequence of Pink1 suggests that the protein is localized to the mitochondrion. However, several laboratories have observed that Pink1 is not exclusively found in mitochondria, although it has been speculated that this could be due to artificial overexpression ( 5 ). Further, a large pool of cellular Pink1 has been reported to be insoluble in detergents such as Triton X-100 or CHAPS ( 6 , 18 ). Therefore, we compared the soluble (0.25% TX-100 extractable) and insoluble (1% SDS extractable) pools of Pink1 using Western-blot analysis after either transient overexpression in COS-7 cells (Fig. 1 A) or stable lentiviral-mediated expression in HEK-293FT cells (Fig. 1 B). Both full-length Pink1 at ∼66 kDa and processed Pink1 at ∼55 kDa were present in the soluble and the insoluble fractions. After transient expression in COS-7 cells, Pink1 55 kDa levels were consistently higher in the insoluble than soluble fraction, whereas Pink1 66 kDa levels were similar in the two fractions. In contrast, stably expressed Pink1 in HEK-293FT and COS-7 cells occurred mostly in the soluble fraction (Fig. 1 B, and data not shown).
Pink1 subcellular distribution and solubility. Human Pink1 cDNA was ( A ) transiently transfected into COS-7 cells or ( B ) stably expressed in HEK-293FT cells using lentivirus. Soluble (0.25% TX-100) and insoluble (1% SDS) extracts were then probed for Pink1 by infrared immunoblotting. Subcellular fractionation of COS-7 cells transiently expressing Pink1 ( C ) or of HEK-293FT cells stably expressing Pink1 ( D ) using differential centrifugation. Fractions were probed for Pink1 and marker proteins using infrared immunoblotting. ( E ) Ability of BC100-494 Pink1 antibody (Novus) to detect endogenous Pink1 in HEK-293FT using Western blotting. Two distinct shRNAs directed against human Pink1 or a control shRNA were lentiviral expressed in HEK-293 cells. Soluble extracts were probed for Pink1 by Western blot, revealing endogenous Pink1 66 kDa and Pink1 55 kDa that run at the size of exogenous Pink1 isoforms and slightly faster than the exogenous Pink1-FLAG isoforms. Because 200-fold less total protein of exogenous Pink1 lysate samples was loaded, actin was not detected in these samples (for complete blots, see Supplementary Material, Figure S1a and b). ( F ) Subcellular fractionation of HEK-293FT cells using differential centrifugation. Fractions were probed for endogenous Pink1 and marker proteins by immunoblotting.
Next, we determined the subcellular localization of exogenous and then endogenous Pink1 using differential centrifugation and subsequent immunoblotting for Pink1 and for compartment marker proteins such as actin (cytosol), VDAC (mitochondria) and calnexin (ER). Both the 66 and 55 kDa species were detected in the 100 000 g supernatant (S100g), the 10 000 g pellet (P10k) and the 100 000 g pellet (P100k) of COS-7 cells transiently expressing Pink1 (Fig. 1 C). These findings are in agreement with a previous subcellular fractionation study of exogenous Pink1 ( 5 ). In the subsequent sections, we refer to S100g as the cytosolic fraction, P10k as the mitochondria-rich fraction and P100k as the microsome-rich fraction. These designations are based on the distribution of compartment marker proteins and sedimentation velocity, but the limitation of these designations is discussed below. Quantitative analysis using infrared immunoblotting (Odyssey infrared imaging system, Li-COR Bioscience) showed that the 66 kDa form had a predominantly microsomal localization (Cyt: 11.9 ± 0.8%, Mito: 36.7 ± 1.2%, Micro: 51.3 ± 1.7%; n = 3, means ± SEMs), whereas the 55 kDa Pink1 occurred mostly in the mitochondria-rich fraction (Cyt: 30.1 ± 1.5%, Mito: 46.7 ± 0.7%, Micro: 22.8 ± 1.7%; n = 3). Similar results were obtained for the HEK-293FT cells transduced with lentivirus expressing Pink1 (Fig. 1 D). However, in this cell line, both Pink1 forms were predominantly in the microsome-rich fraction (66 kDa: Cyt: 15 ± 1.7%, Mito: 20.4 ± 4.8%, Micro: 64.5 ± 6.5%; 55 kDa: Cyt: 41.2 ± 4.9%. Mito: 13.1 ± 5%, Micro 45.6 ± 7%: n = 3). In order to support these subcellular fractionation studies, we performed confocal light microscopy, also revealing a partially non-mitochodrial Pink1 localization (Supplementary Material, Figure S2).
To address the possibility that this non-exclusively mitochondrial localization was due to overexpression, we next analyzed the subcellular distribution of endogenous Pink1 in untransfected HEK-293FT. We first used RNA interference to validate the ability of a commercial Pink1 antibody (BC100-494 from Novus) to detect endogenous Pink1 by Western blotting. Two distinct shRNAs directed against human Pink1 and a negative control shRNA were expressed in HEK-293FT cells using the lentiviral system, and lysates were probed for Pink1 (Fig. 1 E). Proteins at the sizes of exogenous Pink1 66 and 55 kDa were specifically downregulated by both human-specific Pink1 shRNAs, confirming the ability of the 494 antibody to detect endogenous Pink1 66 and 55 kDa isoforms. Of note, the antibody detected several proteins non-specifically, and Pink1 55 kDa was only detectable after prolonged Western blot exposure, indicating its low cellular level (Supplementary Material, Figures S1a and b for complete blots). We also addressed the specificity of the two shRNAs by expressing a Pink1 rescue construct (mutated at the shRNA target sequence) in the presence of Pink1 shRNA. As expected, these Pink1 shRNA constructs were not able to silence the rescue construct (Supplementary Material, Figure S1d). Endogenous Pink1 could also be detected in HeLa cells and, to a lesser extent, in SH-SY5Y human neuroblastoma cells (Supplementary Material, Figure S1c and data not shown). The levels of endogenous Pink1 were too low for infrared quantitative immunoblotting and for microscopy. Importantly, analysis of HEK-293FT subcellular fractions probed for endogenous Pink1 (Fig. 1 F) revealed the same relative distribution of endogenous Pink1 across cytosolic, mitochondria-rich and microsome-rich fractions as we observed for exogenous Pink1 in HEK-293FT cells (Fig. 1 D). We conclude that the cytosolic and microsomal localizations of Pink1 are not due to overexpression and that a substantial fraction of both endogenous and over-expressed Pink1 occurs outside of mitochondria at steady state. These results also indicate that exogenous Pink1 expression is a useful (and necessary, given available antibodies) approach to studying the regulation of Pink1 subcellular localization.
Isolation and MS/MS analysis of Pink1 complexes
Because the identification of cellular proteins physically interacting with Pink1 could reveal important information about Pink1 and its function, we immunoisolated Pink1 complexes from HEK-293FT cells that were transduced with lentivirus expressing human Pink1 plus a C-terminal FLAG epitope tag. These complexes were then analyzed by MS/MS sequencing. As a negative control, a mock isolation was performed through all of the immunoisolation steps (Fig. 2 A) from sister cultures of HEK-293FT cells transduced with mock lentivirus (See Materials and Methods and Supplementary Data for details of isolation and MS/MS analysis). The MS/MS analysis revealed that aside from Pink1, three other proteins were present in the Pink1-FLAG-specific eluate that were extremely low or undetectable in the mock control eluate: Cdc37 and Hsp90 α and ß (Fig. 2 B). Several other proteins were present in both samples in equal amounts, e.g. kinesin (Fig. 2 B). Western blotting using antibodies specific for Hsp90 (α/ß) or Cdc37 confirmed the MS/MS data by showing that both endogenous Hsp90 (α/ß) and endogenous Cdc37 are detectable only in Pink1-FLAG, not control (mock lentivirus), immunoprecipitates of HEK-293FT cell lysates (Fig. 2 C).
Identification of Cdc37 and Hsp90 as Pink1 binding proteins. ( A ) Scheme illustrating the Pink1-FLAG pull down ( B ) MS/MS sequencing results of Pink1-FLAG pull down. Relative peak area comparison between Pink1-FLAG (PF) and negative control reveals Cdc37, Hsp90 α and Hsp90 β as Pink1 binding partners. As exemplified by kinesin, several proteins were detected in both samples to similar levels. ( C ) Western blot analysis M2 anti-FLAG pull down of Pink1-FLAG by probing for Cdc37 and Hsp90 confirms the MS/MS sequencing results.
Hsp90 and Cdc37 are known to form a kinase-specific chaperone system ( 19 ). Its client proteins include many oncogenic kinases such as Akt, Raf-1 and Src and also the PD-linked kinase, LRRK2. Precisely what the interaction with the chaperone heterodimer actually does to the client kinase remains unclear, but it is generally accepted that the chaperone system stabilizes the kinase, and in the case of Akt, it has been shown that the system prevents its dephosphorylation ( 20 ). Even though Hsp90 is a relatively abundant and essential protein, it is currently a major drug target in cancer research because its inhibition destabilizes a wide variety of oncogenic kinases (reviewed in 19 ). Moreover, the Hsp90 inhibitor, Geldanamycin, protects dopaminergic neurons against α-synuclein cytotoxicity in Drosophila melanogaster ( 21 ). In light of the mitochondrial localization sequence of Pink1, it is noteworthy that Hsp90 is involved in the delivery of some mitochondrial preproteins to the outer mitochondrial membrane translocation machinery ( 22 ).
Pink1 is a bona fide Hsp90/Cdc37 client kinase
The ATPase activity of Hsp90 is required for its chaperone activity ( 23 ). Inhibitors such as the naturally occurring benzoquinone, Geldanamycin, and its synthetic derivative, 17AAG, specifically bind to and interfere with the ATP/ADP binding site of Hsp90. This prevents formation of the chaperone/client protein complex and often leads to destabilization of its client proteins ( 24 ). To assess whether Pink1 stability depends upon Hsp90 chaperone activity and Pink1 is therefore a true Cdc37/Hsp90 client, we probed exogenous and endogenous Pink1 protein levels in HEK-293FT cells after treatment with increasing doses of Geldanamycin. HEK-293FT cells stably expressing a construct encoding Pink1 plus a downstream internal ribosome reentry site (IRES) and GFP reporter were treated for 2 h with Geldanamycin. Exogenous Pink1 66 and 55 kDa protein levels were decreased in a dose-dependent manner after Geldanamycin treatment (Fig. 3 A). On the other hand, GFP and actin levels were not reduced, indicating that the Pink1 reduction is not due to altered transcription or other non-specific effects of Geldanamycin. We quantified the ratio of the 66 to 55 kDa Pink1 protein levels by quantitative infrared immunoblotting and found that the ratio decreased with increasing amounts of Geldanamycin (Fig. 3 B), revealing that the Pink1 66 kDa form is more sensitive to Hsp90 inhibition than Pink1 55 kDa. To exclude that Pink1 only serves as a Cdc37/Hsp90 client upon its overexpression, we examined endogenous Pink1 levels in untransfected HEK-293FT cells after Geldanamycin treatment and probed for the binding of endogenous Pink1 to Cdc37. Both the 66 and 55 kDa endogenous Pink1 proteins were destabilized in a dose-dependent manner after 16 h Geldanamycin treatment, and the 66 kDa form was again more sensitive (Fig. 3 C). We noticed that endogenous Pink1 was less sensitive to Hsp90 inhibition than exogenous Pink1, as indicated by the prolonged incubation time required to achieve marked downregulation in the untransfected cells. In order to probe for binding of endogenous Pink1 to Cdc37, we transiently transfected HEK-293FT cells with Cdc37-FLAG and probed FLAG immunoprecipitates for endogenous Pink1. Figure 3 D (representative of four independent experiments) shows that endogenous Pink1 66 kDa is highly enriched in FLAG immunoprecipitates upon Cdc37 FLAG expression, compared to mock treated HEK-293FT cells. Taken together, our results suggest that Pink1 is a novel, bona fide Cdc37/Hsp90 client kinase.
Hsp90 inhibition destabilizes Pink1. ( A ) HEK-293FT cells stably expressing Pink1 IRES GFP were treated for 2 h with increasing concentrations of the Hsp90 inhibitor, Geldanamycin. Soluble extracts were probed for Pink1, GFP and actin using infrared immunoblotting. ( B ) The ratio of Pink1 66/55 kDa proteins decreased with increased Geldanamycin concentrations ( n = 3). ( C ) Untransfected HEK-293 cells were treated for 16 h with increasing concentrations of Geldanamycin. Soluble extracts were blotted for endogenous Pink1, Hsp90 and actin. ( D ) Cdc37-FLAG or mock vector was transiently expressed in HEK-293FT. Cell extracts were probed for Cdc37 and endogenous Pink1 66 kDa (Input, lower panel) and then subjected to immunoprecipitation with an anti-FLAG antibody. Subsequent Western blotting for Pink1 shows specific enrichment of endogenous Pink1 66 kDa in Cdc37-FLAG precipitate. Pink1 shRNA was used to confirm the correct size of the precipitated protein.
Cdc37 1–200 and Cdc37 181–376 show increased Pink1 binding and alter Pink1 processing and solubility
The ability of Hsp90 to interact with and stabilize its client kinases depends upon its co-chaperone, Cdc37. To determine if Cdc37 preferentially binds in vitro to one of the two major Pink1 isoforms (66 or 55 kDa), we transiently transfected untagged Pink1 and FLAG tagged Cdc37 into COS-7 cells and probed FLAG immunoprecipitates for Pink1. Both Pink1 isoforms were present in Cdc37 immunoprecipitates at a similar ratio to that observed in the straight lysates (Fig. 4 A, lane 2), suggesting that there is no preference for either of the two major isoforms. No Pink1 isoform was detected in immunoprecipitates in the absence of FLAG-Cdc37 expression, demonstrating the specificity of the co-immunoprecipitation of Pink1 (Fig. 4 A, lane 1). Next, we probed for the interaction of Pink1 with several different truncated FLAG-Cdc37 proteins ( 25 ), using the same system as described above for wild-type (wt) Cdc37. All of the truncated Cdc37 proteins were able to bind Pink1 to some extent, suggesting that Cdc37 may use more than one binding site to interact with its client, Pink1 (Fig. 4 A, lanes 3–7). Cdc37 truncations 1–180, 1–276 and 201–376 each bound to Pink1 to a similar extent as did full-length (wt) Cdc37 (Fig. 4 A, lanes 3, 5 and 7). Surprisingly, two truncations, Cdc37 1–200 and 181–376, consistently bound more Pink1 than did full-length Cdc37 (Fig. 4 A, lanes 4 and 6). These two truncations share an exposed hydrophobic peptide, amino acids 180–200, found at the C- or N-terminus, respectively, and this motif has previously been identified as an apparent initial client-binding site in Cdc37 ( 25 ). The increased binding of Cdc37 1–200 and Cdc37 181–376 to Pink1 was not unique, as we found that both these truncations showed increased binding to the known Cdc37 client kinase, Akt, and to a non-client kinase, ERK (data not shown).
Effects of wt Cdc37 and various Cdc37 truncations on the Pink1 66/55 kDa ratio. ( A ) Pink1 and full-length FLAG-Cdc37 or each indicated FLAG-Cdc37 truncation construct were transiently co-expressed in COS-7 cells. Cell extracts were probed for FLAG-Cdc37, Hsp90 and Pink1 and then subjected to immunoprecipitation with anti-FLAG antibody, followed by Western blotting for Pink1 and Hsp90. ( B ) Soluble or ( C ) insoluble extracts of COS-7 cells co-expressing Pink1 and FLAG-Cdc37 or each indicated FLAG-Cdc37 truncations were probed for Pink1 via infrared immunoblotting. Pink1 66/55 kDa ratios were quantified in three independent experiments (means ± SEM). ( D ) Pink1 IRES GFP and FLAG-Cdc37, FLAG-Cdc37 1–200 or FLAG-Cdc37 181-376 were transiently co-expressed in COS-7 cells. Western blotting confirmed the presence of the expressed proteins. Quantification via infrared immunoblotting of ( E ) Pink1 66 kDa/GFP ratio and ( F ) Pink1 55 kDa/GFP ratio reveals significant increases of soluble Pink1 55 kDa after Cdc37 1–200 and Cdc37 181–376 expression but no significant change of Pink1 66 kDa levels ( n = 3, means ± SEM). Significant changes are indicated (** P < 0.01 and *** P < 0.001; One-way ANOVA with Newman–Keuls multiple comparison test).
Pink1 55 kDa levels are reported to be increased in Pink1 ARPD and idiopathic PD brain tissue ( 18 ), and thus it is important to understand how Pink1 processing is regulated and to search for factors that alter Pink1 isoform levels. We noticed an apparent decrease of the 66 kDa and increase of the 55 kDa form of exogenous Pink1 in COS-7 soluble cell lysates upon overexpression of either truncated Cdc37 1–200 or 181–376 (Fig. 4 A, topmost panel). As analyzed by quantitative infrared immunoblotting, the 66/55 kDa ratio in the soluble fraction decreased markedly and significantly from a value of 4.3 ± 0.3 for wt Cdc37 to 1.1 ± 0.1 for Cdc37 1–200 and 1.9 ± 0.2 for Cdc37 181–376 (mean ± SEM, n = 3), whereas the other three Cdc37 truncations did not significantly change the ratio (Fig. 4 B). To determine whether this decreased ratio in the soluble fraction resulted from asymmetrical shifts of Pink1 isoforms between the soluble and insoluble Pink1 pools or from altered processing of the 66 kDa form within the soluble pool, we determined the ratio of 66 to 55 kDa Pink1 in the insoluble fraction upon co-expression with wt Cdc37 or each of the five Cdc37 truncations. We found that co-expressing Cdc37 1–200 or Cdc37 181–376 with Pink1 led to a significant increase of the 66/55 kDa ratio in the insoluble fraction (Fig. 4 C). Quantitatively, the ratio increased from 1.7 ± 0.2 for wt Cdc37 to 2.4 ± 0.2 for Cdc37 1–200 and 2.4 ± 0.1 for Cdc37 181–376 (means ± SEM, n = 3). Indeed, normalization of the two Pink1 isoforms to the internally expressed control protein, IRES GFP, confirmed an increase in Pink1 55 kDa as the basis for the ratio change in the soluble fraction (Figs. 4 D–F), supporting a shift of Pink1 55 kDa from the insoluble to the soluble fraction. Because the Cdc37 1–200 and Cdc37 181–376 proteins did not co-immunoprecipitate with Hsp90 (Fig. 4 A, lanes 4 and 6), the observed effect is Hsp90-independent. In this regard, Hsp90-independent Cdc37 functions have been described before ( 26 ). Of interest, we observed a trend that co-expression of wt Cdc37 with Pink1 led to an increase of the 66/55 kDa ratio in both the soluble and insoluble fractions (Figs. 4 B and C). Taken together, our data suggest that formation of proper Hsp90/Cdc37/Pink1 complexes might be involved in regulating the relative solubilities of the mature and processed Pink1 isoforms.
Wt and 1–200 Cdc37 expression alters Pink1 subcellular localization
The significant shift in the Pink1 66/55 kDa ratio upon co-expression of Cdc37 truncations having residues 180–200 exposed on their N- or C-terminus, respectively, led us to hypothesize that Cdc37 might be involved in the mechanism that controls Pink1’s subcellular localization. Accordingly, we asked whether the distribution of both Pink1 isoforms across the cytosolic, mitochondria-rich and microsome-rich fractions changes upon co-expression of Pink1 with wt Cdc37 or Cdc37 1–200 in COS-7 cells. We calculated the percentage distribution of the Pink1 66 and 55 kDa proteins by applying quantitative infrared immunblotting after subcellular fractionation by differential centrifugation, as described above. Expression of wt Cdc37 and Cdc37 1–200 was confirmed by Western blotting (data not shown). First, we compared the distribution of Pink1 66 kDa within the three subcellular fractions. Quantification based on three independent experiments revealed that co-expression of either Cdc37 construct significantly decreased the cytosolic pool of Pink1 66 kDa from 11.9 ± 0.8% (mock) to 4.9 ± 1.5% for wt Cdc37 and 4.9 ± 0.2% for Cdc37 1–200 (Fig. 5 A). Whereas co-expressing wt Cdc37 led only to a shift of 66 kDa protein from the cytosolic to the microsome-rich fraction, co-expressing Cdc37 1–200 dramatically increased the mitochondrial 66 kDa pool and significantly reduced both the cytosolic and microsomal pools of 66 kDa. Thus, wt Cdc37 (which can bind Hsp90) and Cdc37 1–200 (which cannot) have distinct effects on Pink1 66 kDa subcellular distribution, probably due to the absence of Hsp90 in the Pink1-Cdc37 1–200 complex. With regard to Pink1 55 kDa, both wt Cdc37 and Cdc37 1–200 had similar effects (Fig. 5 B): the cytosolic 55 kDa form decreased significantly from 30.1 ± 1.5% (mock) to 16.0 ± 2.3% (wt Cdc37) and 9.4 ± 1.4% (Cdc37 1–200). Conversely, the mitochondrial pool increased from 46.7 ± 0.7% (mock) to 54.7 ± 4.1% (wt Cdc37) and 64.1 ± 6.5% (Cdc37 1–200), while no significant change was observed in the microsome-rich fraction. These results suggest that Cdc37 is involved in the mechanism that leads to translocation of the Pink1 55 kDa protein from the cytosol to the mitochondria-rich fraction but not the microsome-rich fraction. Quantification of the 66/55 kDa ratio in the different subcellular fractions revealed a significant increase without any Cdc37 over-expression of this ratio in the microsome-rich fraction versus the cytosolic and mitochondria-rich fractions (Fig. 5 C). Overexpressing wt Cdc37 significantly increased the mitochondrial and microsomal 66/55 kDa ratio, and overexpressing Cdc37 1–200 significantly decreased this ratio in the microsome-rich fraction, again indicating distinct effects of wt Cdc37 and Cdc37 1–200 on Pink1 distribution (Fig. 5 C).
Cdc37 wt and Cdc37 1–200 alter the subcellular distribution of Pink1 66 kDa and 55 kDa forms and Pink1 112–581. Quantification of subcellular distributions among cytosolic (S100k), mitochondria-rich (P10k) and microsome-rich (P100k) fractions of exogenous ( A ) Pink1 66 kDa and ( B ) Pink1 55 kDa after mock, Cdc37 or Cdc37 1–200 co-expression. ( C ) Pink1 66/55 kDa protein ratios in cytosolic (S100k), mitochondria-rich (P10k) and microsome-rich (P100k) after mock, Cdc37 or Cdc37 1–200 co-expression. Values represent at least three independent experiments (means ± SEM). Significant changes are indicated (* P < 0.05, ** P < 0.01 and *** P < 0.001, One-way ANOVA with Newman–Keuls multiple comparison test).
ARPD and kinase-dead Pink1 mutant proteins are associated with an altered 66/55 kDa ratio
Next, we asked whether an ARPD-causing mutant form of Pink1 and an engineered kinase-dead (KD) Pink1 protein show an altered 66/55 kDa ratio. Most ARPD missense mutations of Pink1 lie within the canonical serine/threonine kinase domain of Pink1 and presumably affect its enzymatic activity ( 5 , 7 ). A few mutations are found in or adjacent to the MTS ( 3 , 27 ). We quantified 66/55 kDa ratios in polyclonal COS-7 cells stably expressing an ARPD-causing Pink1 mutant protein (G309D) or else an engineered KD (K219A/D362A/D384A) with an internally expressed control protein (IRES GFP). The G309D site is within the kinase domain, and reduced autophosphorylation and neuroprotective activity have been reported for G309D ( 4 , 5 ). Loss of phosphorylation activity for Pink1 KD has been verified in vitro ( 5 , 15 ). Probing for Pink1 in soluble extracts of the distinct COS-7 stable cell lines and quantification of four independent experiments revealed that the Pink1 ARPD mutation and the KD mutation lead to a significant decrease of the 66/55 kDa ratio of soluble Pink1 (Fig. 6 A and B). Similar results were found for ARPD Pink1 C92F and H271Q mutant proteins (data not shown). Interestingly, normalization of the two Pink1 isoforms of Pink1 G309D and Pink1 KD to the internal IRES GFP reveals that both isoform levels are significantly reduced compared to Pink1 wt isoforms (Fig. 6 A, C and D). The larger decrease of the Pink1 66 kDa isoforms explains the decreased Pink1 66/55 kDa ratio we detected. Importantly, transiently expressed Pink1 ARPD mutant proteins did not display a ratio change, and stably expressed Pink1 is associated with a higher Pink1 66/55 kDa ratio than transiently expressed Pink1 (data not shown and Figs 4 B, 6 B, 8 B and Supplementary Material, Figure S4b). Taken together, the results in this section suggest that the 66/55 kDa ratio may be important with regard to Pink1-linked ARPD and also sporadic PD.
Pink1 G309D and KD mutants display decreased Pink1 66/55 kDa protein ratios. Pink1, Pink1 G309D or a Pink1 KD was stably expressed in COS-7 cells with an internally expressed control protein (IRES GFP). ( A ) Lysates were probed for Pink1 and GFP expression using infrared immunoblotting. ( B ) Pink1 66/55 kDa ratios quantified in soluble extracts using infrared immunoblotting show decreased Pink1 66/55 kDa ratios for both Pink1 mutant proteins ( n ≥ 4, means ± SEM). Quantification via infrared immunoblotting of ( D ) Pink1 66 kDa/GFP ratio and ( E ) Pink1 55 kDa/GFP ratio show decreased levels of Pink1 66 kDa and Pink1 55 kDa mutant proteins. ( n ≥ 4, means ± SEM). Significant changes are indicated (** P < 0.01 and *** P < 0.001, One-way ANOVA with Newman–Keuls multiple comparison test).
Influence of Parkin and DJ-1 on the Pink1 66/55 kDa ratio and Pink1 subcellular distribution
Recent studies point to a possible convergence of all three ARPD genes—Pink1, Parkin and DJ-1—in a single pathogenic pathway ( 8 , 28 ), but more work is required to confirm this hypothesis. Therefore, we first asked whether Parkin or DJ-1 could influence the PD-relevant ratio of Pink1 66 to 55 kDa isoforms. To do this, we generated polyclonal COS-7 cells stably expressing empty vector, Parkin or DJ-1 using lentivirus and then transiently transfected these stable cell lines with Pink1 cDNA. Analysis of the soluble and insoluble extracts of these cell lines by probing for Pink1 revealed more than a doubling of the 66/55 kDa Pink1 ratio in both extracts upon Parkin overexpression but no change upon DJ-1 overexpression (Fig. 7 A, not shown). Notably, Parkin has an effect opposite to that of Pink1 ARPD mutations on the Pink1 66/55 kDa ratio. Normalization of the two Pink1 isoforms to the internally expressed control protein (IRES GFP) then showed that the increase of the Pink1 66/55 ratio after Parkin expression was due mostly to an increase of Pink1 66 kDa (Fig. 7 B–D). To address whether the effect of Parkin is linked to its PD relevant activity, we analyzed the Pink1 66/55 kDa ratio upon overexpression of two Parkin ARPD mutations. As shown in Figure 7 A, Parkin ARPD R42P and W453OPA mutant proteins effect the Pink1 66/55 ratio to a significantly lesser degree than wt Parkin. To exclude that overexpression of functional E3 ligases in general causes an increase of the Pink1 66/55 kDa ratio, we assessed the Pink1 66 and 55 kDa forms after co-expression of two other E3 ligases, namely c-Cbl and HHARI. As shown in Supplementary Material, Figure S3, these ligases did not alter the ratio, supporting the specificity of the Parkin effect. Together, these results show a specific effect on the Pink1 66/55 kDa ratio and, in particular, on the Pink1 66 kDa protein levels by Parkin but not two other E3 ligases, thereby biochemically linking the two ARPD genes.
Parkin overexpression increases Pink1 66/55 kDa ratio and the Pink1 mitochondrial pool. ( A ) Quantification via infrared immunoblotting of Pink1 66/55 kDa ratio in soluble extracts of COS-7 cells that transiently express Pink1 and stably express mock, Parkin, Parkin R42P, Parkin W453OPA or DJ-1 ( n ≥ 3, means ± SEM) reveals an increase of Pink1 66/55 kDa ratio after Parkin wt and mutant protein expression. The increase is significantly higher after wt Parkin expression than after Parkin R42P and W453OPA expression. Significant changes are indicated (** P < 0.01 and *** P < 0.001, One-way ANOVA with Newman–Keuls multiple comparison test). ( B ) Pink1 IRES GFP transiently expressed in stably mock or Parkin expressing COS-7 cells. Western blotting confirmed the presence of the expressed proteins. Quantification via infrared immunoblotting of ( C ) Pink1 66 kDa/GFP ratio and ( D ) Pink1 55 kDa/GFP ratio reveals large significant increase of Pink1 66 kDa after Parkin co-expression but only a small change of Pink1 55 kDa levels ( n = 3, means ± SEM). Significant changes are indicated (* P < 0.05 and *** P < 0.001; Student's t -test). Quantification of relative subcellular distributions among cytosolic (S100k), mitochondria-rich (P10k) and microsome-rich (P100k) fractions of ( E ) exogenous Pink1 66 kDa and ( F ) exogenous Pink1 55 kDa after mock, or stable Parkin expression ( n = 4, means ± SEM) (* P < 0.05 and *** P < 0.001, One-way ANOVA with Newman–Keuls multiple comparison test).
Next, we asked whether the distribution of transiently expressed Pink1 changes among cytosolic, mitochondria-rich and microsome-rich fractions in COS-7 cells stably expressing Parkin. We calculated the percentage distribution of the Pink1 66 kDa and 55 kDa proteins by applying quantitative infrared immunblotting after subcellular fractionation by differential centrifugation, as described earlier. First, we compared the distribution of Pink1 66 kDa in mock and Parkin expressing cells among the three subcellular fractions (Fig. 7 E). The mitochondria-rich Pink1 66 kDa pool significantly increased from 37.9 ± 3% (mock) to 56.1 ± 0.8% upon Parkin expression (means ± SEM, n = 4). Conversely, Parkin expression led to a significant decrease of Pink 66 kDa in the microsome-rich fraction [55.8 ± 2.4% (mock) to 39.7 ± 1.3%, means ± SEM, n = 4]. A slight but insignificant decrease was observed for the cytosolic Pink1 66 kDa pool upon Parkin expression. With regard to the Pink1 55 kDa subcellular distribution, Parkin expression had a small effect (Fig. 7 F). The mitochondria-rich Pink1 55 kDa pool increased significantly from 55 ± 0.8% (mock) to 66.4 ± 1%, whereas 55 kDa significantly decreased both in the cytosolic fraction from 16.2 ± 2.5% (mock) to 10.8 ± 1.5% and in the microsome-rich fraction from 28.6 ± 1.3% (mock) to 22.4 ± 1.5% (means ± SEM, n = 4). These results demonstrate a relatively increased mitochondrial pool of Pink1 in the presence of overexpressed Parkin and thereby suggest a link between Parkin levels and the mitochondrial targeting of Pink1.
Effect of Parkin knockdown on Pink1 66/55 kDa ratio in human neuroblastoma cells
Last, we sought to determine whether Parkin knockdown produces a reciprocal effect to that of Parkin overexpression on the Pink1 isoform levels in human SH-SY5Y neuroblastoma cells that express a reasonable amount of endogenous Parkin. First, we established that Parkin overexpression dramatically increases Pink1 66 kDa levels in SH-SY5Y cells and thereby increases their 66/55 kDa ratio in a closely similar fashion to the COS-7 cells (Supplementary Material, Figure S4). Then we used lentivirus infection to generate polyclonal SH-SY5Y cells stably co-expressing Parkin shRNA (or control shRNA) plus Pink1 IRES GFP. Probing their lysates for Parkin and Pink1 showed an efficient knockdown of endogenous Parkin and an apparent decrease of the Pink1 66/55 kDa ratio in Parkin shRNA but not control shRNA infected SH-SY5Y cells (Fig. 8 A). Analysis of three independent experiments using infrared immunoblotting for Pink1 and GFP revealed that Parkin knockdown has a significant reciprocal effect to that of Parkin overexpression on the Pink1 66/55 kDa ratio and Pink1 66 kDa levels (Figs. 8 B and C). Quantitatively, Pink1 66/55 kDa ratio decreases from 5.14 ± 0.3 to 3.27 ± 0.1 (Fig. 8 B) and the Pink1 66 kDa levels decline by 42 ± 9% (Fig. 8 C) after Parkin silencing. Parkin overexpression in SH-SY5Y cells had resulted in a ∼8-fold increase of Pink1 66 levels but only a ∼3-fold increase of Pink1 55 levels (Supplementary Material, Figure S4c and d). Therefore, it is not surprising that Pink1 55 kDa levels do not significantly change after Parkin knockdown (Fig. 8 D). Together, these results further support an upstream effect of Parkin on the Pink1 66/55 kDa ratio via a change in the levels of the 66 kDa form.
Stable Parkin silencing in neuroblastoma SH-SY5Y cells decreases Pink1 66 kDa levels and the 66/55 kDa ratio. ( A ) Parkin shRNA (or control shRNA) and Pink1 IRES GFP expressing stable polyclonal SH-SY5Y were generated. Lysates were probed for Pink1, GFP, Actin and endogenous Parkin, revealing efficient Parkin knockdown and an effect on Pink1 66/55 kDa ratio after Parkin shRNA expression. ( B ) Pink1 66/55 kDa ratios quantified in soluble extracts using infrared immunoblotting show a significant decreased Pink1 66/55 kDa ratio after Parkin silencing ( n = 3, means ± SEM). Quantification via infrared immunoblotting of ( C ) Pink1 66 kDa/GFP ratio and ( D ) Pink1 55 kDa/GFP ratio reveals significant decrease of Pink1 66 kDa but no significant change of Pink1 55 kDa levels ( n = 3, means ± SEM). Significant changes are indicated (** P < 0.001; Student's t -test).
DISCUSSION
Pink1 is not solely a mitochondrial kinase
The confirmed mitochondrial localization of Pink1 as a gene product that causes ARPD has strengthened the link between PD pathogenesis and mitochondrial dysfunction ( 3 , 5 ). Nevertheless, a diffuse Pink1 staining pattern in astrocytes but not in neurons provides evidence for extra-mitochondrial Pink1 locations ( 17 ). Also previous subcellular localization studies of exogenous Pink1 suggested the existence of non-mitochondrial Pink1 pools ( 5 ). In the current study, we show in cultured cells that both endogenous, full-length 66 kDa Pink1 and the processed 55 kDa form are present not only in mitochondria-rich but also in cytosolic and microsome-rich fractions. Our consistent detection of endogenous and exogenous cytosolic Pink1 that is soluble in detergent-free buffer is not consistent with the proposal that Pink1 is an integral membrane protein ( 6 , 17 ). Evidence for a small, apparently microsomal pool of Pink1 in rat brain has been published, but the authors explained this by mitochondrial contamination in their microsomal subcellular fraction ( 17 ). We exclude mitochondrial contamination as the source of endogenous Pink1 in the microsome-rich fraction based on our documenting high amounts of endogenous (and overexpressed) Pink1 in microsome-rich fractions that contain very low levels of the mitochondrial marker protein VDAC but substantial levels of the microsomal protein, Calnexin.
The abundance of the immature precursor compared to the mature form that we found for both endogenous and exogenous Pink1 (Fig. 1 C, D and F) is surprising. Overloaded translocase complexes could explain accumulation of exogenous Pink1 66 kDa precursor but not of its endogenous counterpart. Detection of endogenous mature Pink1 55 kDa in the cytosol also seems surprising but has been reported before for exogenous Pink1 ( 5 ). Future experiments should reveal whether mature (55 kDa) Pink1 is released from mitochondria or microsomes, or whether it is generated in the cytosol by a processing that is alternative to mitochondrial processing. How both Pink1 isoforms end up in microsome-rich fractions that normally consist of fragments of plasma membrane and the endoplasmic reticulum is not explicable by an obvious targeting motif within the Pink1 sequence. We cannot exclude that our observed alterations in the subcellular distribution of Pink1 are partially due to changes in the solubility state of Pink1. Therefore, our designation of the fractions could potentially be partly misleading. But the similar distributions of endogenous and exogenous Pink1 in HEK-293FT cells show that the reported distribution is not due to overexpression and is, therefore, physiologically relevant. Further experiments are required to explore these unanticipated subcellular pools of Pink1 and determine their precise nature.
Pink1 is a client kinase of the Hsp90/Cdc37 chaperone system
We employed an unbiased Pink1 pull-down and subsequent MS/MS sequencing to identify Cdc37 and Hsp90 as Pink1 interacting proteins. The interactions were then confirmed via reverse co-immunprecipitation in a heterologous system. Cdc37 is a molecular co-chaperone that functions with Hsp90 to promote folding of many kinases ( 19 ). Hsp90 inhibition leads to instability of its client proteins. In this regard, we show that endogenous and exogenous Pink1 protein levels are dose-dependently sensitive to the Hsp90 inhibitor Geldanamycin, providing support for Pink1 being a Cdc37/Hsp90 client kinase. Pink1 co-immunoprecipitates with non-overlapping N-terminally and C-terminally truncated Cdc37, suggesting more than one interaction site between these proteins. Interaction of Cdc37 with protein kinases has been reported to be mediated principally by its N-terminal region ( 29 ). Hsp90 by itself has previously been implicated in mitochondrial protein targeting ( 22 ). To our knowledge, the present study provides the first evidence that the Cdc37/Hsp90 chaperone system can influence the subcellular distribution of a client protein. We determined the percentage subcellular distribution of both Pink1 isoforms using subcellular fractionation by differential centrifugation and subsequent quantitative immunoblotting. This approach allowed us to detect small changes in the percentage subcellular distribution. Relative cytosolic Pink1 levels (both 66 and 55 kDa forms) were significantly reduced upon Cdc37 overexpression, suggesting that the chaperone complex participates in targeting mechanisms. Rescue experiments in yeast with a Cdc37 mutant protein that is incapable of binding Hsp90 as well as in vitro studies showing Cdc37 chaperone activity independent of Hsp90 have provided strong evidence that Cdc37 can function independently of Hsp90 (26). Indeed, we found that Cdc37 1–200, which does not bind Hsp90 but displays an increased Pink1 binding efficiency, dramatically shifts 66 kDa but not 55 kDa Pink1 from the microsomal-rich to the mitochondrial-rich fraction. These findings suggest that the mitochondrial localization mechanisms of 66 kDa and 55 kDa Pink1 are distinct and differently regulated. We propose that the Cdc37/Hsp90/Pink1 complex is destined for a translocation that leads to Pink1 processing (to 55 kDa), whereas in the absence of Hsp90 in the complex, Pink1 might be attached to mitochondria as a full-length precursor (66 kDa). The first 34 amino acids of Pink1 are sufficient for mitochondrial targeting ( 6 ). We found that Pink1 112–581 runs on SDS–PAGE at the size of mature Pink1 55 kDa, suggesting that the processing occurs around residue 112 (data not shown). Thus, it is well possible that Pink1 residues 34–112 contain other information than that required for targeting. It is well established that signal sequences have functions beyond targeting ( 30 ). The abundance of endogenous Pink1 66 kDa compared to mature 55 kDa in cell culture, the tight regulation of the Pink1 66/55 kDa protein ratio and the apparently distinct subcellular localizations all suggest individual functions for Pink1 66 kDa and Pink1 55 kDa. In this regard, the OPA1 yeast homologue Mgm1 is an example of a mitochondrial protein whose distinct processed isoforms have individual functions, and the ratio of these isoforms affects mitochondrial dynamics ( 31 ).
Potential importance of the Pink1 66/55 kDa ratio in ARPD
We demonstrate that ARPD-linked loss-of-function mutations in Pink1 as well as an engineered KD version lead to significant decreases in the Pink1 66/55 kDa ratio. To detect these effects, we found that steady-state levels of Pink1 in stable cell lines were required. Our use of polyclonal stable cell lines excludes clonal variation as a cause of the decreased ratios we observed. Since both Pink1 G309D and Pink1 KD showed lower protein steady state levels than did wild-type Pink1, it will be important to elucidate the mechanism of how these mutations apparently affect the level of Pink1 protein. Of special interest is the finding that overexpression of Parkin, which is known to be protective against several PD-associated insults, including dopamine and α-synuclein mediated neurotoxicity ( 32 , 33 ), strongly increased the 66/55 kDa ratio, whereas DJ-1 had no effect. These various results and the reciprocal effect of Parkin silencing lead us to propose that the Pink1 66/55 kDa ratio may turn out to be an important correlate of PD-related neuronal dysfunction and provide the first biochemical link between Parkin and Pink1 and, in particular, the Pink1 66/55 kDa ratio and Pink1 66 kDa protein levels. That ARPD-causing mutant Parkins have a lower 66/55 kDa elevating effect than wt Parkin strengthens the relevance of this ratio and the Parkin/Pink1 link to PD pathogenesis. Whether the Pink1 66/55 kDa ratio is decreased in PD and ARPD brains remains to be determined. Based on our data in Figs 6–8 , it is tempting to speculate that relative increases in Pink1 66 kDa levels are protective whereas relative increases in Pink1 55 kDa levels may be harmful. Our results in Fig. 6 also suggest that Pink1 kinase activity plays a role in the regulation of Pink1 isoform levels and the 66/55 kDa ratio.
Nevertheless, several other explanations need to be considered in light of our data. First, Parkin expression increased the relative mitochondrial pool of both Pink1 isoforms (Fig. 7 E and F). Conversely, we have recently detected a clear trend that loss of Pink1 kinase activity leads to decreased Pink1 mitochondrial localization (data not shown). Thus, the relative mitochondrial localization of Pink1 may be more relevant to PD pathogenic effects than the ratio alteration. Second, a decreased 66/55 kDa ratio, as displayed by Pink1 ARPD and Pink1 KD mutant proteins (Fig. 6 ) could represent respectively a compensation for reduced kinase activity.
The emerging Pink1–Parkin connection
Genetic rescue experiments in Drosophila strongly suggest that Pink1 acts upstream of Parkin in a common pathway ( 9 , 11 ). Our study now tightens the Pink1/Parkin connection on a biochemical level but also points to possible upstream effects of Parkin on Pink1. Parkin, but not DJ-1, overexpression doubled the Pink1 66/55 kDa ratio (Fig. 7 ), principally by increasing 66 kDa. This is in agreement with the opposite effects observed after Parkin silencing (Fig. 8 ). Our studies do not provide a mechanism for this phenomenon. It appears that wt Parkin, but not other E3 ligases, can stabilize the Pink1 66 kDa protein. Interestingly, Parkin overexpression increased the mitochondrial Pink1 pool in cells (Fig. 7 D and E), likewise supporting an upstream effect of Parkin on Pink1. Future experiments should clarify how much this Parkin effect on Pink1 mitochondrial localization contributes to Parkin-mediated neuroprotection. Comparisons of the effects of Parkin and Cdc37 on Pink1 suggest that changes in the Pink1 66/55 kDa ratio do not correlate directly with alterations seen in subcellular Pink1 distribution. It should be noted that the occurrence of apparent upstream as well as downstream effects of Parkin on Pink1 is not necessarily conflicting; a classical feedback mechanism between the proteins is a possible explanation. It will now be important to study any effects of Parkin on Pink1 in vivo .
Further experiments should determine whether the Pink1 66/55 kDa ratio and Pink1 subcellular distribution are modifying pathogenic factors for PD and may thus represent potential therapeutic targets. It will be important to probe other PD associated biochemical conditions for effects on this ratio and on the regulation of Pink1 subcellular distribution in neuronal cells.
MATERIALS AND METHODS
Plasmids for overexpression
To produce a Pink1 expression vector, human Pink1 was PCR amplified using Image clone 5214483 as a template and cloned into pcDNA3.1+ (Invitrogen). Primers were designed with a BamH1 site at the 5′ end and an EcoRI site after the stop codon at the 3′ site. Similarly, a C-terminal FLAG tagged Pink1 expression construct, pcDNA3.1 Pink-FLAG, was cloned by adding a FLAG epitope tag into the 3′ primer before a stop codon. A KD Pink1 expression vector was constructed as above using pcDNA-DEST47 Pink1 K219A/D362A/D384A (gift of M. Cookson) as the template. The Quikchange site-directed mutagenesis kit (Stratagene) was first used to repair a mutation (P209A) that is present in the template (Image clone 5214483) and then to introduce recessive PD-causing mutations, G309D, H271Q and C92F. In order to express GFP tagged wt Pink1, Pink1-GFP was PCR-amplified using pcDNA-DEST47 Pink1-GFP as template (gift of M. Cookson) and cloned as EcoRI fragment into pcDNA3.1. To produce a DJ-1 expression construct, human DJ-1 was PCR amplified using image clone BC008188 as template and cloned into pcDNA3.1HA as BamH1-EcorI fragment. Myc-HHARI and HA-c-Cbl expression vectors were gifts of H. Hardley and H. Band, respectively. Cdc37 expression constructs ( 25 ) and pcDNA3.1 Parkin constructs were described previously ( 34 ).
For construction of lentiviral expression plasmids, NheI-EcorI Pink1, Parkin and DJ1 fragments from the pcDNA3.1 based vectors mentioned above were cloned into pCDH/CMV7/IRES GFP that was constructed by replacing the SalI-BamH1-excised resistance gene of pCDH-MCS1-EF1-Puro (System Biosciences, Mountain View, CA, USA) with an IRES GFP. All constructs were verified by DNA sequencing.
Plasmids for Pink1 and Parkin shRNA expression
Human Pink1 U6 shRNA expression plasmids were constructed using the Block-iT U6 system (Invitrogen). Sense and antisense oligonucleotides (Pink1 shRNA 1, Top: CACCgccaacaggctcacagagaagtgttCGAAaacacttctctgtgagcctgttggc, bottom: AAAAgccaacaggctcacagagaagtgttTTCGaacacttctctgtgagcctgttggc; Pink1 shRNA 2, top: CACCggacgctgttcctcgttatgaagaaGAGAttcttcataacgaggaacagcgtcc, bottom: AAAAggacgctgttcctcgttatgaagaaTCTCttcttcataacgaggaacagcgtcc; Parkin shRNA, top: CACCggatcagcagagcattgttcaCGAAtgaacaatgctctgctgatcc, bottom: AAAAggatcagcagagcattgttcaTTCGtgaacaatgctctgctgatcc; shRNA control, top: CACCgcatcgaaatgc, bottom: AAAAgcatttcgatgc) were annealed and ligated into pENTR/U6 resulting in pENTR/U6/ Pink1(1), pENTR/U6/ Pink1(2) and pENTR/U6/empty, respectively. Subsequently, the U6 shRNA cassettes were transferred into pLenti6/BLOCK-iT-DEST using Gateway site-specific recombination resulting in pLenti6/Block-iT Pink1(1), pLenti6/Block-iT Pink1(2), pLenti6/Block-iT Parkin and pLenti6/Block-iT control, respectively.
Antibodies
Polyclonal rabbit Pink1 antibody (BC100-494) was purchased from Novus Biologicals (Littleton, CO, USA). The antibody was used at 1:1000 for the detection of endogenous Pink1 using ECL and for exogenous Pink1 using infrared immunoblotting (Odyssey infrared imaging system, Li-COR Bioscience). The antibody detects exogenous Pink1 at a 1:10 000 dilution using ECL (not shown). Following antibodies were used at dilutions recommended by the manufacturer: Monoclonal mouse M2 anti-FLAG antibody from Sigma Aldrich (St Louis, MO, USA), monoclonal mouse anti-Cdc37 antibody (C1) and Actin (8226) from Abcam (Cambridge, MA, USA), rabbit polyclonal HSP 90 α/β and Parkin antibody from Cell Signaling Technology (Danvers, MA, USA), polyclonal rabbit VDAC (PA1-954A) from Affinity Bioreagents (Golden, CO, USA) and mouse monoclonal Calnexin from Stressgene (San Diego, CA, USA), mouse monoclonal anti-Myc (9E10) from Santa Cruz Biotechnology (Santa Cruz, CA, USA) and rat monoclonal anti-HA (3F10) from Roche (Indianapolis, IN, USA).
Cell culture and transient transfections
COS-7, SH-SY5Y cells (both from ATCG) and HEK-293FT cells (from Invitrogen) were grown adherently and maintained in DMEM containing 10% fetal bovine serum, penicillin (100 U/ml), Streptomycin (100 µg/ml) and L-Glutamine (2 m m ) at 37°C in 5% CO 2 /95% air. COS-7 and SH-SY5Y cells were transiently transfected using Fugene6 (Roche) or Lipofectamine 2000 (Invitrogen), respectively, according to the manufacturer's guidelines. Cells were harvested 24 h after transfection.
Lentiviral production and stable cell lines
Nine microgram Packaging vector (ViraPower packaging mix; Invitrogen) and 3 µg lentiviral expression plasmid (pLenti6/Block-iT or pCDH/CMV7/IG based) were co-transfected into HEK-293FT packaging cells with Lipofectamine 2000 (Invitrogen). After 60 h, the culture supernatants containing lentivirus were filtered (0.45 µM), aliquoted and stored until use at −80°C. Cells at 50% confluency in 6-well dishes were transduced in presence of 4 µg/ml Polybrene (Sigma Aldrich; St Louis, MO, USA). Transduced cells are designated polyclonal stable cell lines. Pink1, Parkin and control shRNA stable cell lines were selected in 4 µg/ml Blasticidin after transduction for 2 weeks. Cells were used for experiments up to five passages.
Subcellular fractionation using differentialcentrifugation
Cytosolic, mitochondrial and microsomal subcellular fractions were isolated by cell disruption followed by differential centrifugation and washing. Transiently transfected or polyclonal transduced cells were disrupted in Isolation Buffer (250 m m Sucrose, 10 m m Tris/MOPS pH7.4 0.1 m m EGTA) using a combination of Dounce homogenization followed by 10 expulsions through a 27-gauge needle. Homogenates were centrifuged at 200 g for 5 min at 4°C, and then the supernatant was spun at 10 000 g for 10 min at 4°C. The resultant pellet (mitochondria) was saved after washing in isolation buffer. The post mitochondrial supernatant was spun at 100 000 g for 60 min. The high-spin pellet was designated after washing in isolation buffer the microsome-rich fraction, and the supernatant was designated the cytosolic fraction. Probing for actin (Cytosol), VDAC (Mitochondria) and calnexin (ER) using immunoblotting was used to verify the subcellular fractionation.
Soluble and insoluble cell extracts, immunoprecipitation and immunoblotting
Cells were harvested in PBS and lysed for 30 min on ice in 0.25% TX-100, PBS, 1 m m EDTA, plus protease inhibitors and phosphatase inhibitors. Lysates were centrifuged at 20 000 ×g for 30 min at 4°C. The supernatant was designated the soluble (TX-100 extractable) fraction. The resulting insoluble pellet was re-suspend and sonicated in PBS 1%SDS (SDS-extractable fraction). Protein concentrations of soluble cellular fractions were determined using the BCA protein assay kit (Pierce). To prepare protein normalized samples of the insoluble fractions, samples were adjusted according to concentrations of soluble fractions.
Standard methods were used for co-immunoprecipitation. FLAG-tagged proteins were immunoprecipitated from soluble lysates using Anti-FLAG agarose beads purchased from Sigma Aldrich (St Louis, MO, USA) and beads were washed in 50 m m Tris pH 7.6, 300 m m NaCl, 0.05% Protease-free BSA, 0.5% Triton-100 and 0.01% SDS.
Equal amounts of protein extracts or agarose beads heated at 65°C for 10 min in Laemmli sample buffer were loaded on Tris/Glycine gels (Invitrogen Life Technologies). Following semi-dry transfer, the Immobilon-P (Milipore) membrane was blocked with 5% non-fat dry milk in PBS containing 0.1% Tween 20 (polyoxyethylene 20-Sorbitan monolaurate; Fisher Scientific) for 1 h at RT and then incubated overnight at 4°C with the indicated antibodies. After washing with PBS 0.1% Tween 20, membranes were incubated with peroxidase-conjugated secondary antibodies against mouse, rat or rabbit IgG, and reactions revealed after washing with Enhanced Chemiluminescence Western blotting detection reagent (Amersham Corp.).
For quantitative Western blot analysis by Odyssey infrared immunoblotting, a slightly different protocol was used. Proteins were transferred to Immobilon FL (Milipore) membranes and blocked with 5% non-fat dry milk in PBS. Membranes were then incubated overnight with primary antibodies in 5% non-fat milk in PBS-T and washed in PBS-T. After incubation with infrared-labeled secondary antibodies, blots were imaged with a dual-channel infrared imager (Odyssey infrared imager, LiCOR Bioscience). Immunoreactive proteins were quantified using Odyssey 1.2 software. Statistical analysis were carried out applying one-way analysis of variance (ANOVA) followed by the post hoc Newman–Keuls test or Student's t -test using Prism4 (GraphPad) software, and a value of P < 0.05 was considered significant.
Immunoisolation and MS/MS analysis of Pink1complexes
HEK-293FT cells (∼5×10 7 ) transduced with lentivirus expressing Pink1-FLAG were washed once in ice-cold PBS prior to lysis in 20 ml 0.25% TX-100 PBS 1 m m EDTA (lysis buffer) including protease and phosphatase Inhibitors. The lysate was centrifuged twice at 20 000 g for 30 min, and the supernatant was filtered through a 0.22 µM cartridge and then pre-cleared with mouse IgG agarose for 3 h at 4°C. Pre-cleared lysate (20 mg protein) was then incubated by end-over-end rotation for 12 h at 4°C with 800 µl anti-FLAG M2 affinity resin (Sigma Aldrich, St Louis) that had been pre-equilibrated in lysis buffer. The resin was washed once in 50 ml lysis buffer, once in 50 ml 0.5% NP-40 PBS, once in 50 ml 0.1% NP-40 PBS and finally re-suspended in 1 ml 0.1% NP-40 PBS. Liquid was spun down in Handee Spin Cups (Pierce), and resin was immediately re-suspended in 500 µl elution buffer (0.3 µg/ml FLAG peptide, 0.1% NP-40 PBS). After incubation for 60 min at 4°C, elute was spun down and lyophilized in 1.5 ml tubes. As a negative control, the same procedure was done simultaneously with an equal number of cells expressing only mock vector. Lyophilized samples were analysed using MS/MS sequencing as described in supplementary material.
Confocal microscopy
COS-7 cells were transiently transfected with Pink1-GFP or Pink1 using Fugene6 (Roche). Transfected cells were stained with Mitotracker (Invitrogen) and fixed in 4% paraformaldehyde. For immunostaining, cells were permeabilized with ice-cold acetone and then probed with Pink1 antibody (1:2000 dilution) and Alexa 488 fluorescently conjugated anti-rabbit secondary antibody. Cells were captured using a Zeiss LSM510 confocal microscope.
SUPPLEMENTARY MATERIAL
Supplementary Material is available at HMG Online.
FUNDING
This work was supported by a NIH grant (NS 038375) to D.J.S. A.W was supported by a long-term EMBO fellowship and by the Swiss National Foundation. Funding to pay Open Access publication charges for this article was provided by an NIH grant.
ACKNOWLEDGEMENTS
We thank Matt Hemming for cloning of pCDH/CMV7/IG, Bryan Krastins (HPCGG, Partners) for MS/MS sequencing and Mark Cookson (NIH) for providing Pink1-GFP and Pink1 KD expression vectors. We are grateful to Matt LaVoie, Matt Hemming and Irit Rappley for critical reading of the manuscript and helpful suggestions.
Conflict of Interest statement . None declared.
