Address correspondence and reprint requests to Jean-Christophe Rochet, Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University, 575 Stadium Mall Drive, RHPH 410A, West Lafayette, Indiana 47907-2091, USA. E-mail: firstname.lastname@example.org
Mitochondrial dysfunction, proteasome inhibition, and α-synuclein aggregation are thought to play important roles in the pathogenesis of Parkinson’s disease (PD). Rare cases of early-onset PD have been linked to mutations in the gene encoding DJ-1, a protein with antioxidant and chaperone functions. In this study, we examined whether DJ-1 protects against various stresses involved in PD, and we investigated the underlying mechanisms. Expression of wild-type DJ-1 rescued primary dopaminergic neurons from toxicity elicited by rotenone, proteasome inhibitors, and mutant α-synuclein. Neurons with reduced levels of endogenous DJ-1 were sensitized to each of these insults, and DJ-1 mutants involved in familial PD exhibited decreased neuroprotective activity. DJ-1 alleviated rotenone toxicity by up-regulating total intracellular glutathione. In contrast, inhibition of α-synuclein toxicity by DJ-1 correlated with up-regulation of the stress-inducible form of Hsp70. RNA interference studies revealed that this increase in Hsp70 levels was necessary for DJ-1-mediated suppression of α-synuclein aggregation, but not toxicity. Our findings suggest that DJ-1 acts as a versatile pro-survival factor in dopaminergic neurons, activating different protective mechanisms in response to a diverse range of PD-related insults.
Parkinson’s disease (PD) is a neurodegenerative disorder involving a loss of dopaminergic neurons in the substantia nigra (Dawson and Dawson 2003). Post-mortem studies of the brains of PD patients provide evidence for mitochondrial complex I impairment, increased oxidative stress, and proteasome dysfunction, suggesting that these phenomena may play a role in pathogenesis (Beal 2003; Jenner 2003; Olanow and McNaught 2006). A neuropathological hallmark of PD is the presence in some surviving neurons of Lewy bodies, cytosolic inclusions enriched with fibrillar α-synuclein (Spillantini et al. 1997). Mutations in the α-synuclein gene have been linked to early-onset, autosomal-dominant PD (Dawson and Dawson 2003). The A30P and A53T mutants form neurotoxic, prefibrillar oligomers (‘protofibrils’) more rapidly than the wild-type protein (Conway et al. 2000), suggesting that the increased risk of PD associated with these mutants may be related to accelerated protofibril formation.
DJ-1 is structurally similar to the chaperone Hsp31, a related member of the DJ-1/ThiJ/PfpI superfamily, and it inhibits protein aggregation in test-tube and cellular models (Lee et al. 2003; Shendelman et al. 2004; Zhou et al. 2006). The suppression of aggregate formation by DJ-1 occurs via a redox-dependent mechanism involving the oxidation of cysteine 53 and/or cysteine 106 (Shendelman et al. 2004; Zhou et al. 2006). In addition, DJ-1 has been shown to up-regulate Hsp70 in a dopaminergic cell line, and this up-regulation correlates with DJ-1-dependent suppression of α-synuclein toxicity (Zhou and Freed 2005).
These observations suggest that DJ-1 has multiple neuroprotective functions, including an antioxidant function involving up-regulation of GSH and a chaperone function involving up-regulation of Hsp70. However, it is unclear whether both activities are generally involved in neuroprotection by DJ-1, or whether a single function is sufficient for protection against specific PD-related insults. In this study, we examined the effects of wild-type DJ-1 and two familial mutants (M26I and E64D) on dopaminergic neurotoxicity induced by rotenone, proteasome inhibitors, and mutant α-synuclein, and we explored the underlying mechanisms. Our results indicate that wild-type DJ-1, but not the familial mutants, protects against these diverse insults by activating distinct cellular responses.
Materials and methods
The following antibodies were used in this study: mouse anti-β-actin (Sigma–Aldrich, St Louis, MO, USA); rabbit anti-DJ-1 PARK7 (Neuromics, Edina, MN, USA); mouse anti-GABA (clone 5A9, Chemicon, Temecula, CA, USA); rabbit anti-GCL (Lab Vision, Fremont, CA, USA); mouse anti-HO-1 (clone HO-1-1, Assay Designs, Ann Arbor, MI, USA); mouse anti-Hsp25/27 (MAB3842, Chemicon); mouse anti-Hsp40 (clone 20, Assay Designs); mouse anti-iHsp70 (clone C92F3A-5, Assay Designs); mouse anti-MAP2 (clone AP20, Chemicon); mouse anti-NQO-1 (clone A180, Invitrogen, Carlsbad, CA, USA); rabbit anti-Nrf2 (Santa Cruz Biotechnology, Santa Cruz, CA, USA); mouse anti-8-oxo-dG (Trevigen, Gaithersburg, MD, USA); mouse anti-α-synuclein LB509 (Invitrogen); mouse anti-α-synuclein Syn-1 (clone 42, BD PharMingen, San Diego, CA, USA); rabbit anti-tyrosine hydroxylase (TH) (Chemicon); mouse anti-V5 (Invitrogen); anti-mouse IgG-Alexa Fluor 488 and anti-rabbit IgG-Alexa Fluor 594 (Invitrogen); and anti-mouse IgG and anti-rabbit IgG conjugated to alkaline phosphatase (Promega, Madison, WI, USA).
Preparation of lentiviral constructs
The ViraPower Lentivirus Expression System (Invitrogen) was used to generate lentivirus encoding human α-synuclein (A53T), human DJ-1 with an N-terminal hexahistidine tag, human iHsp70, or β-galactosidase (LacZ) fused to the V5 epitope. The presence of a histidine tag on DJ-1 did not affect the results presented in this report. The preparation of A53T and LacZ lentiviruses was described previously (Cooper et al. 2006). A cDNA encoding DJ-1 (Open Biosystems, Huntsville, AL, USA) was amplified by PCR and subcloned into the Kpn I and Xho I sites of the vector pENTR1A to generate pENTR-DJ-1WT. The constructs pENTR-DJ-1M26I, pENTR-DJ-1E64D, and pENTR-DJ-1L166P were prepared from pENTR-DJ-1WT by site-directed mutagenesis using the Quikchange method (Stratagene, La Jolla, CA, USA). A cDNA encoding iHsp70 was subcloned into the BamH I and Xho I sites of pENTR1A to generate the construct pENTR-iHsp70. The insert from each pENTR1A construct was transferred into the pLENTI6/V5 DEST lentiviral expression vector via a recombination reaction and sequenced using an Applied Biosystems DNA sequencer (University of Wisconsin and Purdue University).
Lentiviral constructs encoding short hairpin RNAs (shRNAs) were used to down-regulate rat DJ-1 and iHsp70. An shRNA-encoding cassette spanning amino acid residues 20–26 of rat DJ-1 (Yokota et al. 2003) was ligated into the entry vector pENTR/U6 (Invitrogen) and subsequently transferred to the lentiviral vector pLenti6/BLOCK-iT DEST (Invitrogen) by recombination. Hsp70-specific lentiviral silencing constructs (Sigma–Aldrich, product number NM_005346) had shRNA-encoding cassettes derived from the region spanning nucleotides 1502-1520 or 872-890 of the rat Hsp70 1a/1b coding sequence. These constructs are referred to here as ‘shRNA #1’ and ‘shRNA #2,’ respectively. Because only shRNA #2 successfully down-regulated iHsp70, we used the construct encoding shRNA #1 as a negative control.
Primary midbrain cultures were prepared via dissection of day 17 embryos obtained from pregnant Sprague–Dawley rats (Harlan, Indianapolis, IN, USA) as described previously (Cooper et al. 2006). All of the procedures involving animal handling were approved by the Purdue Animal Care and Use Committee. The cells were plated on coverslips at a density of 1250 cells per mm2. Four days after plating, the cells were treated with cytosine arabinofuranoside (AraC) (20 μM, 48 h) to inhibit the growth of glial cells. After exposure to AraC, the cultures were incubated in fresh media for an additional 24 h. At this stage (i.e., 7 days in vitro), the glial cells accounted for approximately 50% of the total cell population, and the neurons appeared differentiated with extended processes.
Lentiviral transductions and treatments with pharmacological agents
Primary cultures (7 days in vitro) were untransduced or transduced with lentivirus in the presence of polybrene (6 μg/mL). Unless otherwise specified, the transductions were carried out for 72 h at a multiplicity of infection (MOI) of 10 in the case of lentivirus encoding DJ-1, A53T, DJ-1 shRNA, or Hsp70 shRNA, or an MOI of 8 in the case of iHsp70 lentivirus. In experiments aimed at determining the effects of DJ-1 expression, DJ-1 down-regulation, or iHsp70 expression on mitochondrial dysfunction or proteasome impairment, the cells were incubated with lentivirus for 72 h. The cells were then treated with fresh media containing rotenone, MG132, or lactacystin at a concentration of 100 nM, 10 μM, or 10 μM (respectively) for 24 h, unless otherwise specified. Control cells were incubated in the absence of virus and then treated with fresh media supplemented with vehicle (dimethyl sulfoxide, 0.002-1% [v/v]). In experiments aimed at determining the effects of DJ-1 expression, DJ-1 down-regulation, iHsp70 expression, or iHsp70 down-regulation on α-synuclein neurotoxicity or aggregation, the cells were incubated with lentiviruses for 72 h and treated with fresh media for an additional 24–48 h. Control experiments with LacZ lentivirus were carried out to confirm that the toxic effects of A53T lentivirus and the protective effects of DJ-1 or iHsp70 lentivirus were specific. In studies involving DJ-1 down-regulation, the specificity of the RNAi effect was verified by co-transducing primary cultures with shRNA lentivirus (MOI = 10) and lentivirus encoding human, wild-type or mutant DJ-1 (MOI = 3). The DJ-1 lentivirus was applied at an MOI of 3 in this case because this amount of virus was found suitable to express human DJ-1 at approximately the normal level of the endogenous rat protein.
To test the effects of N-acetyl-cysteine (NAC) on neurotoxicity elicited by complex I inhibition or proteasome dysfunction, primary midbrain cultures were treated with rotenone or MG132 in the absence or presence of NAC (final concentration, 1 mM unless otherwise specified). To assess whether DJ-1 and NAC protect against rotenone toxicity via a common mechanism, the cells were treated with rotenone and NAC after transduction with wild-type DJ-1 lentivirus. To test the effects of NAC on α-synuclein neurotoxicity, the cells were incubated with A53T lentivirus in the absence or presence of NAC. After a 72-h transduction period, the cells were treated with fresh media (either with or without NAC) for an additional 24 h prior to analysis.
Protein expression levels in the primary midbrain cultures were determined via western blot analysis. The cells were dislodged from the plate by trypsinization, collected by centrifugation, washed with phosphate-buffered saline (PBS) (136 mM NaCl, 0.268 mM KCl, 10 mM Na2HPO4, 1.76 mM KH2PO4, pH 7.4), and lysed in buffer L (20 mM Tris–HCl, pH 7.4, 25 mM KCl, 5 mM MgCl2, 0.25 M sucrose, 1% (v/v) Triton X-100, 0.5 mg/mL benzamidine, 2 μg/mL aprotinin, 3.6 μg/mL leupeptin, 0.75 mM phenylmethylsulfonyl fluoride, 700 units/mL DNase I). After centrifugation at 13 000 g, the detergent-soluble (supernatant) fraction was recovered (unless otherwise specified, only the soluble fraction was analyzed). In some experiments, the detergent-insoluble (pellet) fraction was retained and resolubilized in buffer L containing 10% (v/v) glycerol and 2% (w/v) sodium dodecyl sulfate (SDS). The protein concentration in the soluble or insoluble fraction was measured using the bicinchoninic acid Protein Assay Kit (Pierce Biotechnology, Rockford, IL, USA), and equal amounts of protein were separated via SDS–polyacrylamide gel electrophoresis on a 4–20% (w/v) polyacrylamide gel. The proteins were transferred to nitrocellulose or polyvinylidene difluoride, and the membrane was probed with a primary antibody specific for human α-synuclein (LB509, 1 : 250), mammalian α-synuclein (Syn-1, 1 : 1000), mammalian DJ-1 (PARK7, 1 : 2000), mammalian iHsp70 (1 : 3000), mammalian Nrf2 (1 : 500), or mammalian β-actin (1 : 10 000). The membrane was treated with a secondary anti-mouse or anti-rabbit, alkaline phosphatase-conjugated antibody (1 : 5000). Chemifluorescence images were obtained and analyzed using a Typhoon imaging system (GE Health Sciences, Piscataway, NJ, USA). Quantitative analysis of the western blotting data was carried out using ImageQuant software (GE Health Sciences). Chemifluorescent signals were normalized to the level of β-actin and subsequently plotted as a percentage of the control. Each western blot in this paper is a representative of at least two independent analyses.
Primary cells were fixed, permeabilized, and blocked as described (Cooper et al. 2006). After washing with PBS, the cells were treated overnight at 4°C with the following combinations of primary antibodies: (i) anti-MAP2 (1 : 500) and anti-TH (1 : 500), or anti-GABA (1 : 500) and anti-TH (1 : 500), to monitor relative dopaminergic cell viability; (ii) anti-8-oxo-dG (1 : 500) and anti-TH (1 : 500), to determine the extent of oxidative damage to DNA in dopaminergic neurons; or (iii) anti-iHsp70 (1 : 500) and anti-TH (1 : 500) or anti-MAP2 (1 : 500), to monitor the up-regulation of iHsp70 in dopaminergic neurons or total neurons, respectively. To determine the lentiviral transduction efficiency, untransduced cells or cells transduced with LacZ lentivirus (expressing β-galactosidase fused to the V5 epitope) were treated at 22°C for 1 h with the following combinations of primary antibodies: (i) anti-V5 (1 : 500) and anti-MAP2 (1 : 500), or (ii) anti-V5 (1 : 500) and anti-TH (1 : 500). The coverslips were treated with secondary antibodies and mounted onto slides as described (Cooper et al. 2006).
Measurement of primary neuron viability
MAP2- and TH-immunoreactive primary neurons were counted in 10 randomly chosen observation fields for each experimental condition using a Nikon TE2000-U inverted fluorescence microscope (Nikon Instruments, Melville, NY, USA) with a 20× objective. The data were expressed as the percentage of MAP2+ neurons that were also TH+ (this ratiometric approach was used to correct for variations in cell density). Alternatively, GABA- and TH-immunoreactive neurons were counted, and the data were expressed as the percentage of total neurons (sum of TH- and GABA-positive) that were TH-positive. Each experiment was repeated 2–4 times using embryonic neurons isolated from different pregnant rats.
Analysis of oxidative damage to DNA
Primary midbrain cultures co-stained with 8-oxo-dG- and TH-specific antibodies were examined using a Nikon TE2000-U inverted fluorescence microscope with a 20× objective. Images were taken, and the fluorescence intensity due to 8-oxo-dG staining was quantified in the TH+ neurons using Metamorph software (Molecular Devices, Sunnyvale, CA, USA). In total, 40–50 TH+ neurons were analyzed for each experimental condition. Alternatively, the experiment was carried out by co-staining the cultures with an 8-oxo-dG-specific antibody and the nuclear stain 4′,6-diamidino-2-phenylindole (DAPI). In this case, the fluorescence intensity of 8-oxo-dG staining was quantified in ∼ 100 DAPI-stained neurons per experimental condition.
Measurement of protein carbonyls
Primary midbrain cells were dislodged from the plate by trypsinization, collected by centrifugation, washed with PBS, and lysed in buffer L. The protein concentration in the soluble fraction was measured using the bicinchoninic acid Protein Assay Kit. Each protein sample (9 μg) was characterized with respect to protein carbonyl content via western blotting using the OxyBlot Oxidized Protein Detection Kit (Chemicon).
Measurement of intracellular GSH
Primary midbrain cells were dislodged from the plate by trypsinization and collected by centrifugation. The cells were washed with 0.4 M 2-(N-morpholino)ethane sulfonic acid, 0.1 M phosphate, 2 mM EDTA (pH 6.0) and lysed in the same buffer via sonication. The cell lysate (5 μg protein) was assayed for total intracellular GSH using the Glutathione Assay Kit (Cayman Chemical Company, Ann Arbor, MI, USA). Typically we detected 20-25 nmol of total GSH per mg of protein in the cell lysate. When determining the ratio of reduced glutathione to oxidized glutathione (i.e., ratio of GSH to GSSG), the sample was divided into two aliquots. One aliquot was treated with 2-vinylpyridine to derivatize GSH, and the GSSG level was measured via the kit protocol. The second aliquot was assayed for total GSH. The level of GSH (reduced) was then calculated as the difference between total GSH and GSSG.
Analysis of iHsp70 levels in individual cells
Primary midbrain cultures co-stained with antibodies specific for iHsp70 and TH or MAP2 were examined using a Nikon TE2000-U inverted fluorescence microscope with a 20× objective. Images were taken, and the fluorescence intensity of iHsp70 staining was quantified in the cell bodies of TH+ or MAP2+ neurons using Metamorph software. The iHsp70 staining intensity in each neuron was corrected by subtracting the signal associated with the underlying glial cell monolayer (this background signal was determined by measuring the fluorescence intensity in an area equivalent to that of the neuronal cell body immediately adjacent to the neuron). Because glial cells showed faint MAP2+ staining and were readily distinguished from neurons because of their morphology, they were also scored for their iHsp70 content in cultures co-stained for iHsp70 and MAP2. In total, 30–60 cells were analyzed for each experimental condition.
Lentiviral-mediated gene expression in primary neuronal cultures
The goal of this study was to characterize the neuroprotective function of DJ-1 against various PD-related insults in primary mesencephalic cultures. The relative number of dopaminergic neurons in the cultures was determined immunocytochemically by staining with antibodies specific for MAP2, a general neuronal marker, and TH, a specific marker of dopaminergic neurons (Fig. 1a). Of the total MAP2+ neurons in these cultures, approximately 5 ± 1% were TH+ (i.e., dopaminergic). To test whether a lentiviral-mediated DNA delivery system could be used to introduce genes efficiently in MAP2- and TH-positive neurons, the cells were treated with a control lentivirus encoding β-galactosidase fused to the V5 epitope (‘LacZ’ virus) (Fig. 1b). Approximately 90% of MAP2+ neurons and 80% of TH+ neurons were infected with the LacZ virus, suggesting that both types of neurons were transduced with high efficiency. From this result, we concluded that the lentiviral system could be used to examine the effects of gene expression on dopaminergic cell viability in the next part of our study.
DJ-1 protects primary dopaminergic neurons against rotenone neurotoxicity
To test the effect of mitochondrial dysfunction on the viability of dopaminergic neurons in our cellular model, primary midbrain cultures were treated with the complex I inhibitor rotenone and analyzed immunocytochemically. Rotenone induced a dose-dependent decrease in the relative number of TH+ neurons (Fig. 2a), consistent with earlier findings (Ahmadi et al. 2003). The nearly four-fold decrease in the relative number of TH+ neurons upon treatment with 100 nM rotenone resulted from a 10-fold decrease in the absolute number of TH+ neurons, compared to only a 2.5-fold decrease in the absolute number of MAP2+ neurons. Similar results were obtained when co-staining for TH and GABA: namely, rotenone induced a greater decrease in TH+ dopaminergic neurons compared to GABA+ non-dopaminergic neurons (Fig. 2b).
To address whether DJ-1 protected primary dopaminergic neurons from mitochondrial dysfunction, we first examined whether wild-type and mutant forms of the protein could be expressed from a lentiviral construct. The primary cultures were transduced with lentivirus encoding wild-type DJ-1, M26I, or E64D, at various MOI for various times. Cell lysates were analyzed via western blotting using a primary antibody that recognizes mammalian DJ-1 (Fig. 2c). The data indicated that wild-type DJ-1 and the familial mutants were expressed at similar levels in the primary cultures. Maximal expression was observed in cells transduced with lentivirus at an MOI of 10 for 72 h, and then incubated in fresh media for 24–48 h (data not shown). Human DJ-1 was ∼1.6-fold more abundant than the endogenous rat protein in cells transduced at an MOI of 10 (Fig. 2c). In contrast, the L166P mutant was not expressed at detectable levels and, therefore, was not characterized further in this study.
Next, we examined whether wild-type or mutant DJ-1 protected primary dopaminergic neurons from rotenone toxicity. Midbrain cultures were transduced with lentivirus encoding wild-type DJ-1, M26I, or E64D and subsequently treated with rotenone (100 nM). The relative number of TH+ neurons was greater in cultures transduced with wild-type DJ-1 virus compared to untransduced control cells or cells that expressed M26I, E64D, or LacZ (Fig. 2d). Because TH- and MAP2-positive neurons were transduced to nearly equal extents by the lentivirus, the apparent inhibitory effect of wild-type DJ-1 on preferential dopaminergic neurotoxicity did not merely result from the unequal infection of different subsets of neurons.
In other experiments, we showed that primary dopaminergic neurons were sensitized to toxicity induced by a lower dose of rotenone (5 nM) following a ∼ 55% reduction in endogenous DJ-1 levels achieved via RNAi (Fig. 2e and f). This sensitization was rescued by reintroducing human wild-type DJ-1 (but not M26I or E64D) at the normal level of the endogenous rat protein, demonstrating the specificity of the RNAi. These data suggested that wild-type but not mutant DJ-1 protected dopaminergic neurons from the toxic effects of rotenone in our primary cell-culture model.
DJ-1 suppresses oxidative damage and up-regulates GSH in rotenone-treated cells
Because rotenone causes dopaminergic cell death by triggering a buildup of reactive oxygen species (ROS) (Sherer et al. 2003), we hypothesized that DJ-1 protects against rotenone-induced neurodegeneration by suppressing intracellular oxidative stress. To address this hypothesis, the cells were assayed for 8-oxo-deoxyguanosine (8-oxo-dG) and protein carbonyls, two classic markers of oxidative damage. TH+ neurons exhibited significantly enhanced 8-oxo-dG staining in rotenone-treated versus untreated cultures (Fig. 3a). The increase in 8-oxo-dG levels was partially suppressed by transduction with lentivirus encoding wild-type DJ-1, but not LacZ (Fig. 3a). M26I and E64D had less pronounced inhibitory effects on 8-oxo-dG accumulation than wild-type DJ-1. Although these trends did not reach statistical significance in TH+ neurons, significant differences between wild-type and mutant DJ-1 were observed when we expanded the study to include non-dopaminergic neurons, due to the greater number of cells analyzed (Supplementary material Fig. S1a). Rotenone-treated cultures also had higher levels of protein carbonyls than untreated cells, and this increase was mitigated by transduction with lentivirus encoding wild-type DJ-1 but not M26I, E64D, or LacZ (Fig. 3b). Together, these data suggested that the wild-type protein (in contrast to familial mutant DJ-1) protects against rotenone neurotoxicity by inhibiting intracellular oxidative damage.
To determine whether the antioxidant function of DJ-1 was sufficient to prevent rotenone-induced dopaminergic neurotoxicity, we compared the protective effects of DJ-1 and the small-molecule antioxidant, N-acetyl-cysteine (NAC). NAC was chosen for these studies because it readily penetrates the cell membrane, exhibits similar redox properties as GSH (because of the presence of a sulfhydryl group), and enhances GSH biosynthesis by providing a supply of the precursor cysteine (Maher 2005). Accordingly, it would be expected to mimic antioxidant effects of DJ-1 involving GSH up-regulation (Zhou and Freed 2005). Similar numbers of TH+ neurons were observed in primary midbrain cultures treated simultaneously with rotenone and NAC (1 or 5 mM) compared to rotenone-treated cells transduced with DJ-1 lentivirus (Fig. 3c and data not shown). Dopaminergic cell viability was also similar in primary cultures treated with both DJ-1 lentivirus and NAC compared to cultures treated with either agent alone (Fig. 3c). The sensitization to rotenone neurotoxicity elicited by DJ-1-specific shRNA lentivirus was rescued to similar extents by NAC and virus encoding human wild-type DJ-1 (Fig. 3d). In addition, NAC suppressed rotenone-induced protein carbonyl formation with similar efficacy as wild-type DJ-1 (Fig. 3b). From these data, we concluded that wild-type DJ-1 protects against rotenone neurotoxicity primarily via its antioxidant function.
Next, we explored mechanisms by which DJ-1 suppresses rotenone-induced oxidative stress. We reasoned that the previously reported up-regulation of GSH or Hsp70 by DJ-1 (Zhou and Freed 2005) could account for protection against rotenone-induced oxidative damage. Namely, an increase in GSH levels would prevent a buildup of ROS, whereas the up-regulation of Hsp70 would suppress the formation of protein aggregates that cause mitochondrial dysfunction and oxidative stress (Rochet 2007). Rotenone-treated primary midbrain cultures had significantly lower levels of total, intracellular GSH (i.e., sum of oxidized and reduced forms of GSH) and a decreased ratio of reduced GSH to GSSG than untreated cells (Fig. 4a and b). The loss of total GSH and the decreased GSH/GSSG ratio were rescued in cells transduced with wild-type DJ-1 lentivirus, and this protective effect (similar to that of NAC) was less pronounced in cells expressing M26I. In contrast to its up-regulation of GSH, wild-type DJ-1 had no effect on iHsp70 levels in rotenone-treated primary cultures (Supplementary material Fig. S1b). These results suggest that wild-type DJ-1 protects against rotenone toxicity via an antioxidant mechanism involving up-regulation of total GSH and the GSH/GSSG ratio.
DJ-1 protects primary dopaminergic neurons against toxicity induced by proteasome inhibition
To test the effect of proteasome dysfunction on the viability of dopaminergic neurons, midbrain cultures were treated with MG132 or lactacystin and analyzed immunocytochemically. Treatment of the cultures with either inhibitor led to a decrease in the relative number of TH+ neurons (Fig. 5a), in agreement with previous reports (Petrucelli et al. 2002; Mytilineou et al. 2004). Similar results were obtained when co-staining for TH and GABA (Fig. 5b). Next, midbrain cultures were transduced with lentivirus encoding wild-type or mutant DJ-1 prior to treatment with proteasome inhibitor. The relative number of TH+ neurons was greater in MG132- or lactacystin-treated cultures that expressed wild-type DJ-1 compared to untransduced control cells or cells that expressed M26I, E64D, or LacZ (Fig. 5c and d). We also showed that primary dopaminergic neurons were sensitized to toxicity induced by MG132 (1 μM) via RNAi-mediated DJ-1 down-regulation (Fig. 5e). This sensitization was rescued by reintroducing human wild-type DJ-1, but not M26I or E64D (Fig. 5e). Together, these observations suggested that wild-type DJ-1 protected dopaminergic neurons from toxicity because of proteasome impairment in our cell-culture model.
DJ-1 up-regulates iHsp70 in MG132-treated cells
Our next objective was to explore mechanisms by which DJ-1 inhibits neurotoxicity elicited by proteasome dysfunction. Total intracellular GSH levels were unaffected by treatment with MG132, and wild-type DJ-1 did not induce up-regulation of GSH in cells exposed to the proteasome inhibitor (Fig. 6a). MG132 also had no effect on 8-oxo-dG levels in TH+ or MAP2+ neurons (data not shown). In contrast to DJ-1, NAC failed to suppress MG132-induced dopaminergic neurotoxicity (Fig. 6b). These findings suggested that MG132 did not trigger a significant increase in oxidative stress in our primary cell-culture model, and the antioxidant function of DJ-1 did not play a major role in protection against MG132 neurotoxicity. Previous studies have shown that proteasome dysfunction leads to protein aggregation coupled with neuronal cell death (Olanow and McNaught 2006), and molecular chaperones are known to mitigate neurotoxicity resulting from proteasome impairment (Muchowski and Wacker 2005). Accordingly, we hypothesized that DJ-1 inhibits MG132-induced neurodegeneration by causing an increase in Hsp70 levels. To address this hypothesis, we assayed iHsp70 in the detergent-soluble and -insoluble fractions of cell lysates via western blot analysis. Untransduced cells exposed to MG132 had ∼ 7.5-fold higher soluble iHsp70 levels and exhibited a pronounced increase in insoluble iHsp70 compared to cells incubated in the absence of proteasome inhibitor (Fig. 6c–f). A further ∼ 1.3-fold increase in soluble iHsp70 levels was evident in cells expressing wild-type DJ-1, but not in cells expressing M26I or E64D (Fig. 6c and d) or in cells treated with NAC (data not shown). The increase in soluble iHsp70 induced by wild-type DJ-1 correlated with a ∼ 2-fold decrease in insoluble iHsp70 (Fig. 6e and f). Immunocytochemical data indicated that the increase in iHsp70 levels occurred in TH+ and MAP2+ neurons and in glial cells (Fig. 6g; Supplementary material Fig. S2a and b). To investigate whether the up-regulation of iHsp70 was sufficient for the suppression of MG132 toxicity by DJ-1, we directly compared the neuroprotective effects of lentivirally expressed DJ-1 and iHsp70 in primary midbrain cultures. In these experiments, the MOI of the iHsp70 lentivirus was carefully adjusted to ensure that iHsp70 was expressed at the same level as in cells transduced with wild-type DJ-1 virus (Supplementary material Fig. S3a). MG132-treated cultures transduced with iHsp70 virus had higher numbers of TH+ neurons than untransduced cultures, but lower numbers of TH+ neurons than cultures expressing DJ-1 (Fig. 6h). These results suggest that the up-regulation of iHsp70 only accounts in part for DJ-1-mediated neuroprotection against proteasome impairment.
DJ-1 protects primary dopaminergic neurons against α-synuclein neurotoxicity
Next, we addressed whether α-synuclein induced preferential toxicity towards dopaminergic neurons in our primary cell-culture model. We focused on the A53T mutant rather than the wild-type protein because data obtained in our laboratory and by others (Petrucelli et al. 2002) indicated that wild-type α-synuclein is only weakly toxic to primary dopaminergic neurons. The midbrain cultures were transduced with lentivirus encoding A53T, and cell lysates were analyzed via western blotting with a primary antibody specific for human α-synuclein (Fig. 7a). The data indicated that A53T was expressed maximally in cultures transduced with lentivirus at an MOI of 10 for 72 h, followed by incubation in fresh media for an additional 24–48 h.
The relative number of TH+ neurons was lower in primary cultures infected with A53T lentivirus compared to untransduced control cells (Fig. 7b). The LacZ lentivirus had no effect on the viability of dopaminergic neurons (data not shown), indicating that the TH+ cell loss was due to the expression of A53T rather than a non-specific effect of lentiviral infection. These results were consistent with previous studies (Petrucelli et al. 2002; Zhou and Freed 2005). To determine whether DJ-1 protected dopaminergic neurons from A53T neurotoxicity, midbrain cultures were transduced with A53T lentivirus or co-transduced with viruses encoding A53T and wild-type DJ-1, M26I, E64D, or LacZ. The relative number of TH+ neurons was greater in primary cultures that co-expressed A53T and wild-type DJ-1 compared to cultures that expressed A53T alone, A53T plus mutant DJ-1, or A53T plus LacZ (Fig. 7c). We also showed that primary dopaminergic neurons were sensitized to toxicity induced by A53T lentivirus (MOI = 1) via RNAi-mediated DJ-1 down-regulation (Fig. 7d). This sensitization was rescued by reintroducing human wild-type DJ-1, but not M26I or E64D (Fig. 7d). Together, these findings suggested that wild-type DJ-1 protected dopaminergic neurons from toxicity induced by A53T in our cell-culture model.
DJ-1 up-regulates iHsp70 and suppresses α-synuclein aggregation in cells expressing A53T
Next, we explored mechanisms by which DJ-1 inhibits A53T neurotoxicity. Total intracellular GSH levels were unaffected by the expression of A53T alone or A53T plus wild-type DJ-1 (Supplementary material Fig. S4a). Moreover, NAC did not inhibit A53T-induced dopaminergic neurotoxicity, in contrast to DJ-1 (Fig. 8a). These findings suggested that mutant α-synuclein did not cause significant oxidative stress in our cell-culture model, and the antioxidant function of DJ-1 did not play a major role in mitigating A53T-induced neurodegeneration. Instead, we hypothesized that DJ-1 suppresses α-synuclein neurotoxicity by up-regulating Hsp70, as previously demonstrated in a dopaminergic cell line (Zhou and Freed 2005). Consistent with this idea, levels of detergent-soluble iHsp70 were significantly greater in cells co-expressing A53T and wild-type DJ-1 compared to cells expressing A53T alone (Fig. 8b and c). In these experiments, we only measured levels of soluble iHsp70 because the insoluble form was undetectable above the background. iHsp70 up-regulation was not induced by the expression of M26I or E64D or by treatment of the cells with NAC (Fig. 8c). Importantly, expression of iHsp70 at the same level as in cells transduced with wild-type DJ-1 lentivirus reproduced the protective effect of DJ-1 against A53T neurotoxicity (Fig. 8d; Supplementary material Fig. S3b). Strikingly, however, the ability of DJ-1 to protect against A53T neurotoxicity was unaffected by transducing the cells with lentivirus encoding an shRNA that silences iHsp70 (shRNA #2) (Fig. 8e and f).
Previous studies have shown that dopaminergic cell death induced by α-synuclein involves the formation of potentially toxic oligomers or aggregates (Zhou and Freed 2005; Periquet et al. 2007). Accordingly, we hypothesized that DJ-1 inhibits A53T neurotoxicity by interfering with α-synuclein self-assembly. To address this hypothesis, cell lysates were analyzed via western blotting with a primary antibody (Syn-1) that recognizes both human and rat α-synuclein. Cells expressing A53T had higher amounts of SDS-resistant α-synuclein oligomers than untransduced control cells (Fig. 9a–d). A dramatic reduction in α-synuclein oligomer levels was evident in cells co-expressing wild-type DJ-1, whereas this effect was less pronounced in cells treated with NAC (Fig. 9a) or in cells co-expressing M26I, E64D, or LacZ (Fig. 9b). Expression of iHsp70 at the same level as in cells transduced with wild-type DJ-1 lentivirus reproduced the inhibitory effect of DJ-1 on A53T aggregation (Fig. 9c). The DJ-1-mediated suppression of α-synuclein aggregation was ablated by down-regulation of iHsp70 with shRNA #2, whereas it was unaffected by the negative-control shRNA #1 (Fig. 9d). These results suggested that (i) wild-type DJ-1 has a greater propensity to suppress the formation of potentially toxic α-synuclein aggregates than M26I or E64D, and (ii) this inhibitory effect involves up-regulation of iHsp70.
The reduction in the α-synuclein oligomer signal induced by wild-type DJ-1 was accompanied by a small but reproducible decrease in the intensity of the band corresponding to monomeric α-synuclein. We verified that the reduction in monomeric α-synuclein does not involve repression of gene expression from the lentiviral construct (data not shown). Instead, the decrease in monomer levels may reflect an enhanced susceptibility of the protein to proteolysis following DJ-1-mediated inhibition of α-synuclein oligomerization. Importantly, we showed that wild-type DJ-1 causes a pronounced decrease in the relative amount of α-synuclein oligomer even when the oligomer signal is normalized to the monomer band intensity (Supplementary material Fig. S4b).
An advantage of the primary cell-culture model used in this study is that it consists of multiple cell types, including dopaminergic and non-dopaminergic neurons and glia, similar to the native environment of the midbrain. Moreover, it can be used to model the selective death of dopaminergic neurons in the substantia nigra elicited by various PD-related insults (Petrucelli et al. 2002; Ahmadi et al. 2003; Mytilineou et al. 2004; Zhou and Freed 2005; Cooper et al. 2006). In this study, midbrain cultures were analyzed immunocytochemically to monitor responses involving discrete subpopulations of cells (e.g., changes in levels of 8-oxo-dG in TH+ neurons or iHsp70 in TH+ neurons, MAP2+ neurons, and glia). In contrast, cell lysates were analyzed to measure responses involving the total cell population (e.g., changes in levels of protein carbonyls, GSH, iHsp70, and oligomeric α-synuclein). Our results indicate that wild-type DJ-1 suppresses dopaminergic neurotoxicity induced by mitochondrial dysfunction, proteasome inhibition, and α-synuclein expression via different cellular mechanisms.
Effect of DJ-1 on rotenone toxicity
Dopaminergic neurons that expressed human wild-type DJ-1 were relatively resistant to rotenone toxicity, whereas the down-regulation of rat DJ-1 sensitized the neurons to rotenone-induced cell death. The results suggest that wild-type DJ-1 protects dopaminergic neurons from toxicity associated with complex I impairment. Cells exposed to rotenone had high amounts of oxidatively damaged DNA and proteins, decreased levels of total GSH, and a decreased ratio of reduced GSH to GSSG, suggesting that the complex I inhibitor triggered neurodegeneration by causing a buildup of ROS (Sherer et al. 2003; Maher 2005). Wild-type DJ-1 suppressed rotenone-induced oxidative stress, and this inhibitory effect correlated with the up-regulation of total GSH and rescue of the GSH/GSSG ratio. Treatment of primary midbrain cultures with NAC had similar effects on GSH metabolism and prevented rotenone-induced dopaminergic neurotoxicity with similar efficacy as DJ-1. The combined protective effect of NAC plus DJ-1 was no greater than the effect of each individually, and NAC rescued the enhanced sensitivity to rotenone neurotoxicity elicited by DJ-1 knockdown. These observations suggest that an increase in GSH levels and/or the GSH/GSSG ratio is sufficient for DJ-1-mediated neuroprotection against mitochondrial dysfunction and oxidative stress. Our findings are consistent with an earlier report that DJ-1 up-regulates GSH in neuronal cells exposed to H2O2 by inducing an increase in GCL mRNA and enzyme activity (Zhou and Freed 2005).
A previous study revealed that DJ-1 stabilizes Nrf2, a transcriptional regulator of antioxidant response genes including the gene encoding GCL (Clements et al. 2006). We hypothesized that the DJ-1-mediated increase in total GSH in rotenone-treated cells might occur via Nrf2 stabilization. Consistent with this hypothesis, we showed that rotenone-treated cultures expressing wild-type DJ-1 accumulated significantly higher levels of a protein recognized by a Nrf2-specific antibody compared to untransduced control cells (Supplementary material Fig. S5a and b). The apparent molecular weight of this protein was 100 kDa, equivalent to the previously reported molecular weight of a Nrf2-actin complex recognized by the same antibody (Kang et al. 2002). The levels of the 100 kDa protein were significantly higher in rotenone-treated cells transduced with wild-type DJ-1 lentivirus compared to cells expressing M26I or E64D or cells co-treated with NAC. In contrast, wild-type DJ-1 did not induce up-regulation of the 100 kDa protein in cells exposed to MG132 or A53T lentivirus (Supplementary material Fig. S5c and d), in agreement with our observation that DJ-1-mediated neuroprotection against both of these insults does not involve an increase in total intracellular GSH. Unexpectedly, we failed to observe a DJ-1-dependent increase in the levels of three Nrf2-regulated proteins – GCL, heme oxygenase-1 (HO-1), and NAD(P)H:quinone oxidoreductase (NQO1) – via western blot analysis of lysates from rotenone-treated cells (data not shown). Additional studies are required to determine whether DJ-1 protects against complex I inhibition by up-regulating the Nrf2-GSH pathway and/or via other antioxidant mechanisms, including suppression of H2O2 accumulation in mitochondria (Andres-Mateos et al. 2007).
Effect of DJ-1 on toxicity induced by proteasome inhibitors
Dopaminergic neurons were partially rescued from toxicity elicited by MG132 or lactacystin via the expression of wild-type DJ-1, and a partial knockdown of rat DJ-1 sensitized the neurons to MG132-induced cell death. These findings are consistent with previous observations that DJ-1 inhibits proteasome inhibitor-induced toxicity (Yokota et al. 2003; Martinat et al. 2004), although they differ from other data suggesting that DJ-1 lacks this protective effect (Zhou and Freed 2005). Cells treated with MG132 had increased levels of soluble iHsp70, in agreement with previous findings (Rideout et al. 2005). MG132 treatment also led to an increase in insoluble iHsp70, presumably because the increased burden of aggregated protein in cells exposed to proteasome inhibitor results in trapping of the chaperone in the insoluble fraction. Expression of wild-type DJ-1 in MG132-treated cells resulted in an increase in soluble iHsp70 levels and a decrease in insoluble iHsp70 relative to cells treated with MG132 alone, implying that DJ-1 functions as a chaperone (Lee et al. 2003; Shendelman et al. 2004; Zhou et al. 2006) and partially relieves iHsp70 of its load of aggregated client proteins. However, the decrease in insoluble iHsp70 only partially accounted for the DJ-1-dependent up-regulation of soluble iHsp70, given that the chaperone was markedly less abundant in the insoluble fraction under all conditions. Part of the DJ-1-dependent up-regulation of iHsp70 occurred in TH+ neurons, suggesting that this response plays a critical role in mitigating selective dopaminergic neurotoxicity (Rideout et al. 2005). Because iHsp70 up-regulation is not sufficient for the full DJ-1 neuroprotective effect against proteasome dysfunction, additional mechanisms not explored in this study are likely to be involved, including anti-apoptotic activities and signaling through the PI3K-Akt pathway (Junn et al. 2005; Xu et al. 2005; Yang et al. 2005; Fan et al. 2008).
Effect of DJ-1 on α-synuclein neurotoxicity
The expression of wild-type DJ-1 alleviated α-synuclein neurotoxicity and aggregation in primary cultures transduced with A53T lentivirus, and down-regulation of rat DJ-1 sensitized dopaminergic neurons to A53T-induced cell death. The inhibitory effect of DJ-1 on A53T neurotoxicity correlated with a ∼ 1.5-fold increase in iHsp70 levels. These results support previous observations that wild-type DJ-1 protects neuronal cells from mutant α-synuclein toxicity, and that DJ-1 induces a ∼ 2-fold up-regulation of Hsp70 in a dopaminergic cell line expressing A53T (Xu et al. 2005; Zhou and Freed 2005). We also showed using cultures co-transduced with A53T- and iHsp70-encoding lentiviruses that up-regulation of the chaperone was sufficient for DJ-1-mediated inhibition of α-synuclein neurotoxicity and aggregation. This finding agrees with earlier observations that Hsp70 over-expression mitigates α-synuclein-induced dopaminergic cell death in cellular and animal models of PD (Auluck et al. 2002; Klucken et al. 2004).
We showed via RNA silencing that iHsp70 is not essential for DJ-1-mediated protection against A53T neurotoxicity, although it is necessary for the inhibition of α-synuclein aggregation. The observation that DJ-1 suppresses A53T-induced cell death without abrogating α-synuclein aggregation in cells depleted of iHsp70 suggests that DJ-1 protects against A53T neurotoxicity via redundant mechanisms. In addition to eliminating potentially toxic α-synuclein aggregates via up-regulation of iHsp70, DJ-1 may activate various anti-apoptotic signaling pathways (Junn et al. 2005; Xu et al. 2005; Yang et al. 2005; Fan et al. 2008). An alternative possibility is that the α-synuclein oligomers observed in this study are not involved in neurotoxicity, and, therefore, DJ-1 neuroprotection may not require a decrease in α-synuclein oligomer levels. Although we have shown that iHsp70 up-regulation is not essential for DJ-1-mediated inhibition of α-synuclein neurotoxicity in our cell-culture model, the same may not be true in vivo. Rather, we predict that multiple, DJ-1-dependent pro-survival mechanisms (including iHsp70 activation) may be necessary to protect neurons from chronic α-synuclein toxicity in the brains of PD patients.
We speculated that the mechanism of DJ-1-dependent iHsp70 up-regulation might involve activation of heat-shock factor 1 (HSF1), a transcription factor that regulates the expression of heat-shock proteins including Hsp70 (Rochet 2007). However, we failed to observe a DJ-1-dependent increase in the levels of two other HSF1 targets, Hsp25, and Hsp40 (data not shown), arguing against a classic HSF1-dependent mechanism. Additional studies are required to understand the molecular basis of DJ-1-mediated iHsp70 up-regulation.
Effect of mutant DJ-1 on PD-related insults
M26I and E64D had a decreased propensity to prevent dopaminergic neurotoxicity elicited by rotenone, MG132, and A53T, and this functional deficit correlated with a diminished ability to activate the different protective responses explored in this study. Accordingly, nigral neurons in the brains of patients with the M26I or E64D mutation may be highly vulnerable to PD-related insults that occur with age and/or exposure to environmental toxins. E64D exhibited a marginally greater propensity to carry out some neuroprotective functions (e.g., suppression of DNA oxidation, up-regulation of GSH) than M26I, consistent with our recent finding that recombinant M26I has a lower thermodynamic stability (Hulleman et al. 2007). Oxidative modifications destabilize DJ-1 to a similar extent as the M26I substitution (Hulleman et al. 2007), and DJ-1 becomes extensively oxidized in sporadic PD and during aging (Choi et al. 2006; Meulener et al. 2006). Therefore, a decrease in the antioxidant and chaperone functions of DJ-1 may contribute to neurodegeneration in familial and sporadic forms of PD.
A rationale for different protective mechanisms of DJ-1
The principal finding of this study is that DJ-1 activates distinct protective mechanisms in response to different PD-related insults. Previously it was shown that neuroprotection by DJ-1 against oxidative stress or A53T toxicity correlated with up-regulation of GSH or Hsp70, respectively (Zhou and Freed 2005). Here, we extend these findings by demonstrating that the antioxidant function of DJ-1 is sufficient to suppress neurotoxicity associated with impairment of mitochondrial complex I, whereas the up-regulation of chaperone activity is sufficient to inhibit neurodegeneration elicited by mutant α-synuclein. We also show that DJ-1 induces GSH up-regulation in cells treated with a complex I inhibitor, but not in response to proteasome dysfunction or A53T expression. In contrast, DJ-1 induces up-regulation of Hsp70 in cells treated with a proteasome inhibitor or A53T lentivirus, but not in response to impaired complex I activity. In rotenone-treated cells, the activation of an antioxidant response is reasonable because complex I inhibition elicits neurotoxicity by triggering a buildup of ROS (Sherer et al. 2003). A chaperone mechanism would be less protective in this case because protein misfolding occurs downstream of ROS accumulation, and macromolecules other than proteins are also subject to oxidative damage. In A53T-expressing cells, the activation of a chaperone mechanism is appropriate because dopaminergic neurodegeneration triggered by mutant α-synuclein likely involves the formation of toxic aggregates. An antioxidant mechanism would be less beneficial in this case because basal levels of intracellular oxidative stress play a less significant role in driving aggregate formation than the accumulation of A53T, a variant with a high intrinsic propensity to undergo self-assembly.
The mechanism by which DJ-1 activates distinct protective pathways in cells exposed to different insults is unclear. As one possibility, DJ-1 may undergo oxidation at cysteine 106 in response to oxidative stress but not proteasome impairment or α-synuclein expression. In turn, oxidized DJ-1 may activate GSH biosynthesis and reduction, whereas the unoxidized form of the protein may induce up-regulation of Hsp70 (Zhou and Freed 2005). Alternatively, different subsets of DJ-1-interacting proteins may be available in cells subjected to different insults, resulting in the formation of protein complexes with different functional outputs. Another unresolved question is whether the protective mechanisms of DJ-1 in our mixed cultures and in human brain are activated preferentially in neurons or glia, given that the protein is abundant in both cell types (Bandopadhyay et al. 2004; Olzmann et al. 2007). In this study, we showed via immunocytochemical staining that DJ-1-mediated up-regulation of iHsp70 in MG132-treated cultures occurs in dopaminergic and non-dopaminergic neurons and in glia. It remains to be determined whether protective responses identified in cell lysates (e.g., suppression of protein carbonyl formation or α-synuclein aggregation; up-regulation of GSH) must occur in neurons, glia, or both cell types to inhibit dopaminergic neurodegeneration induced by PD-related stresses.
In summary, our data indicate that DJ-1 is a functionally dynamic pro-survival factor, with the ability to protect against a diverse range of toxic phenomena by activating different cellular responses. A loss of DJ-1 function in patients with familial or sporadic PD likely involves the disruption of a number of critical pathways required for dopaminergic cell viability. Ultimately, therapies that target these multiple downstream pathways may prove beneficial in the treatment of PD.
This work was supported by NIH grants RO1 NS049221 and RO3 AG027123, grants from the American Parkinson Disease Association, the Parkinson’s Disease Foundation, the American Association of Colleges of Pharmacy, the Showalter Trust, and a Pilot Grant from the Purdue-UAB Botanicals Research Center (P50 AT000477-06). The research described herein was conducted in a facility constructed with support from Research Facilities Improvement Program Grants Number C06-14499 and C06-15480 from the National Center for Research Resources of the National Institutes of Health. We thank Dr Gary Isom, Joshua Lisinicchia, Susan Roy, Catherine Volk, and Dr Val Watts for advice and assistance in preparing primary mesencephalic cultures, Dr Jeremy Schieler for assistance with immunocytochemistry, and Dr Xiao-Jiang Li and Dr Paul Muchowski for providing the construct encoding iHsp70.