Parkinson Disease-associated DJ-1 Is Required for the Expression of the Glial Cell Line-derived Neurotrophic Factor Receptor RET in Human Neuroblastoma Cells*

Mutations in PARK7/DJ-1 are associated with autosomal recessive, early onset Parkinson disease (PD). DJ-1 is an atypical peroxiredoxin-like peroxidase that may act as a redox-dependent chaperone and a regulator of transcription. Here we show that DJ-1 plays an essential role in the expression of rearranged during transfection (RET), a receptor for the glial cell line-derived neurotrophic factor, a neuroprotective molecule for dopaminergic neurons, the main target of degeneration in PD. The inducible loss of DJ-1 triggers the establishment of hypoxia and the production of reactive oxygen species that stabilize the hypoxia-inducible factor-1α (HIF-1a). HIF-1a expression is required for RET down-regulation. This study establishes for the first time a molecular link between the lack of functional DJ-1 and the glial cell line-derived neurotrophic factor signaling pathway that may explain the adult-onset loss of dopaminergic neurons. Furthermore, it suggests that hypoxia may play an important role in PD.

Parkinson disease (PD) 2 is a progressive neurodegenerative disorder characterized by tremor, akinesia/bradykinesia, and rigidity. Its neuropathological hallmark is the selective degeneration of mesencephalic DA neurons in the substantia nigra that leads to the decrease of dopamine at their synapses in the striatum (1). Although the etiology of sporadic PD remains poorly understood and therapeutic treatments are only symptomatic, the analysis of post mortem brains (2) showed many biochemical and cellular changes including evidences of oxidative stress with an alteration of the intracellular redox equilibrium.
The identification of genetic loci (PARK1-14) and the molecular cloning of some of the genes linked to rare forms of familial PD have provided crucial insights into the mechanism of the pathogenesis (3). Autosomal recessive early onset PD has been associated to mutations in PARK7/DJ-1 (4). These patients present large homozygous genomic deletions as well as truncating, splice-site, and frame shifts mutations. In human post mortem brains of sporadic PD patients, DJ-1 protein is irreversibly oxidized and, thus, inactivated (5,6). Free radicals predominantly modify DJ-1 on the cysteine residue at position 106 (Cys-106) causing a shift in the isoelectric point (pI) from 6.2 to 5.8. This modification is also observed in PD animal models and in aged flies and mice (7). DJ-1 encodes for a highly conserved, ubiquitously expressed protein involved in multiple cellular processes including sperm maturation, fertilization, and oncogenesis. It may act as an atypical peroxiredoxin-like peroxidase, a redox-dependent chaperone, and a regulator of transcription and RNA metabolism (3).
To study the effects of the lack of a functional gene, DJ-1 knock-out (KO) mice and flies were generated. Although they did not show death of DA neurons, increased vulnerabilities to neurotoxic agents were observed (8 -13). Furthermore, embryonic stem cells deficient for DJ-1 expression and DA neurons derived from KO mice displayed increased sensitivity to oxidative stress and proteasome inhibition (14). Nonetheless, the chain of events that links the loss of functional DJ-1 to neuronal dysfunction is still unclear. The survival of developing and postnatal DA neurons is supported by neurotrophic factors, such as the glial cell line-derived neurotrophic factor (GDNF) and the brain-derived neurotrophic factor (15).
Endogenous neurotrophins regulate natural cell death during development and maintain target innervations and cell survival during postnatal life. Declining production of a neurotrophic factor or impaired signal transduction in aging neurons may contribute to neurodegeneration in selected brain areas (16). GDNF is a member of the GDNF family of neurotrophic factors that signals through a two-component receptor complex consisting of the glycosylphosphatidylinositol-linked GDNF family receptor ␣ and the RET receptor-tyrosine kinase, previously identified as a proto-oncogene (17).
RET is strongly expressed both at the mRNA and protein levels in adult mesencephalic DA neurons (18,19). Interestingly, DA neuron-specific ablation of RET in conditional KO mice caused progressive and adult-onset loss of DA neurons, degeneration of DA nerve terminals in the striatum, and pronounced glial activation (19). Therefore, loss of RET expression in DA neurons generates PD-like dysfunctions in mouse models, thus indicating its crucial role in mesencephalic DA cell survival.
Here we show that DJ-1 has an essential role in the regulation of RET expression levels in the human neuroblastoma SH-SY5Y cell line. The inducible loss of DJ-1 triggers the generation of a hypoxic state and the accumulation of free radical species that stabilize the hypoxia-inducible factor-1␣ (HIF-1a). This is required for the down-regulation of RET expression.
This study establishes for the first time a molecular link between the loss of functional DJ-1 and the GDNF signaling pathway that may explain the adult-onset loss of DA neurons. Furthermore, it highlights a possible role for hypoxia in PD pathogenesis.
Cell Culture and Transfections-Human neuroblastoma SH-SY5Y cells (ATCC) were maintained in culture as suggested by vendors. SH-SY5Y cells were transfected with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.
For preparation of SH-SY5Y-inducible cell lines, first we generated the acceptor cells by transfection of pcDNA6-TetR and selection with Blasticidin (3 g/ml) (InvivoGen). Individual clones were confirmed by Western blot analysis for tetracycline repressor (TetR) protein level.
We then generated SH-SY5Y-inducible stable cell lines by transfecting linearized pSuperior containing interference oli-gonucleotides against DJ-1 and selecting for positive clones in the presence of 300 g/ml G418 (Invitrogen) and 3 g/ml Blasticidin (InvivoGen). Individual clones were confirmed by Western blot analysis, immunofluorescence, and quantitative real time PCR (qPCR). Silencing induction was performed by adding 2.5 g/ml doxycycline hyclate (Sigma) every 48 h for 10 days. For HIF-1a silencing, cells were transfected with Oligofectamine (Invitrogen) according to the manufacturer's instructions.
RNA Isolation, Reverse Transcription, and qPCR-Total RNA was isolated using the TRIzol reagent (Invitrogen) following the manufacturer's instructions. Single-strand cDNA was obtained from 1 g of purified RNA using the iSCRIPT TM cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. qPCR was performed using SYBR Green PCR Master Mix (Bio-Rad) and iCycler IQ Real time PCR System (Bio-Rad). Expression of DJ-1, RET, CDC42, GRM8, FLNA, CAMK2B, and ITGB1 was analyzed using specific oligonucleotides (supplemental information).
Microarray Processing and Data Analysis-Total RNA was purified using the RNeasy mini kit (Qiagen). RNA quality was checked using a bioanalyzer (Agilent 2100; Agilent Technologies), and RNA quantity was measured with ND-1000 Nanodrop spectrophotometer. 10 g of RNA sample was used for microarray analysis on Affymetrix GeneChip Human U133A 2.0 Arrays (Affymetrix). Data processing was performed in the R computing environment using packages from the Bio-Conductor software project. Robust multi-array average (RMA) normalization was applied (23). Normalized data were then filtered based on the Affymetrix detection call so that only probes that had a Present call in at least one of the arrays were retained. Data were then imported in the MultiExperiment Viewer (MeV) software (24), and statistical analysis was performed with the SAM (Significance Analysis of Microarrays) module (25) to detect significantly differentially expressed genes. Microarray data have been deposited in the NCBI Gene Expression Omnibus (GEO) with accession number GSE17204. Differentially expressed genes were then specifically examined based on their Gene Ontology annotation (26) and through the use of Ingenuity Pathways Analysis (Ingenuity Systems) and DAVID Bioinformatics Resources (27).
Detection of Intracellular Hypoxia-Cellular hypoxia was assessed using Hypoxyprobe-1 solution (Hypoxyprobe TM -1 Plus, Chemicon). Hypoxyprobe-1 is an exogenous nitroaromatic compound that is metabolized in a stepwise reduction pathway by cellular nitro-reductase enzymes that are able to use the nitroaromatic compounds as alternative electron acceptors in conditions of low physiological pO 2 . The consequent fragmentation of the imidazole ring leads to formation of chemical adducts with several macromolecular components of cells that can be detected by a specific antibody.
In our experiments cell medium was supplemented with Hypoxyprobe solution (20 M) for 2 h before detection. Immunofluorescence was then performed with anti-Hypoxyprobe TM -1 antibody (Chemicon) 1:600.
Fluorescence-activated Cell Sorter Analysis-Scramble clones and siDJ-1s clones were grown in a 35-mm 2 dish to 80% confluency and treated with doxycycline (2.5 g/ml) for 10 days. Cells were washed twice with phosphate-buffered saline, harvested by trypsinization, fixed with ice-cold 70% ethanol, treated with RNase A (0.25 mg/ml), and stained with propidium iodide (0.02 mg/ml). Samples were further analyzed on a BD Biosciences FACScan. Data were analyzed by the BD CellQuest Software.
Quantification of Cellular ROS-Dihydroethidium (DHE) was used as a specific dye for O 2 . detection (28). SH-SY5Y cells were incubated with 10 M DHE for 20 min in serum-free medium. Dye-loaded cells were washed, mounted, and examined under confocal microscope. Statistical Analysis-All experiments were repeated in triplicate or more. Data represent the mean Ϯ S.E. When necessary, each group was compared individually with reference control group using Student's t test (Microsoft Excel software).

Generation of Inducible Stable Cell Lines Expressing RNA
Interference against DJ-1-To mimic the loss-of-function effects as seen in PD patients with DJ-1 mutations, we used siDJ-1 constructs to inhibit the synthesis of endogenous DJ-1 in the human SH-SY5Y neuroblastoma cell line.
Several attempts of constitutive expression of very efficient siDJ-1s failed for the positive selection of cells that maintained a low but relevant amount of endogenous DJ-1 protein. Therefore, we decided to take advantage of an inducible interference (RNAi) approach in which the expression of the interfering sequence is controlled by the TetR system upon incubation with doxycycline (Doxy). This system allows the selection of stable clones before RNAi induction.
To control off-target effects, we used two different RNAi molecules; that is, a first sequence (siDJ-1#1) targeting a region in the 3Ј-untranslated region of human DJ-1 mRNA and a second sequence (siDJ-1#2) mapping to the second exon. This last sequence was previously demonstrated to partially down-regulate DJ-1 expression when used in transient transfection exper-iments (29). As an additional control, we established inducible cell lines with a random sequence (scramble). For each of the cell lines we selected 24 single clones that were screened for Doxy-inducible down-regulation of DJ-1 expression level by qPCR. To exclude clonal effects, two independent clones for each cell line with maximum RNAi effect were selected and further analyzed. As shown in Fig. 1A, siDJ-1#1 caused a profound decrease in DJ-1 mRNA level, achieving a Ͼ90% silenc- , and siDJ-1#2 treated without or with Doxy (2.5 g/ml) for 10 days. Experiments were performed in triplicate. C, densitometric analysis is shown of DJ-1 protein levels as calculated from Western blot analysis in B. Values above each column were normalized to actin levels and compared with scramble clone a as control. JUNE 11, 2010 • VOLUME 285 • NUMBER 24

DJ-1 Regulates RET Expression
ing as compared with scramble controls (clones a and b). These results were consistent in the two independent clones A and B. siDJ-1#2 proved to be only partially effective, as previously published (29), with a 50% reduction in DJ-1 levels. Only one clone was obtained during the selection process for siDJ-1#2, as most of the clones did not show any variation in endogenous DJ-1 content (data not shown). Additional controls for the efficacy of DJ-1 interference were performed using parental SH-SY5Y cells with similar results (data not shown). To monitor the kinetics of silencing, we measured DJ-1 mRNA levels by qPCR at 3, 6, and 10 days after Doxy induction. A 90% reduction in the amount of DJ-1 mRNA was already present after 3 days of Doxy, and it persists till the tenth day (data not shown).
To assess the effects of DJ-1 mRNA silencing on the levels of DJ-1 protein, we monitored the down-regulation of DJ-1 by Western blot analysis. Scramble clones were used as controls. We found that the amount of residual DJ-1 protein was reduced to almost undetectable levels (Ͼ90% reduction) after 10 days of Doxy treatment (Fig. 1, B and C). Kinetic experiments showed that DJ-1 protein started to decrease at 6 days (70%) (data not shown), and 90% was knocked down after 10 days. As expected from qPCR data, siDJ-1#2 could reduce DJ-1 protein levels to only 50% (Fig. 1, B and C).
When comparing DJ-1 mRNA levels in un-induced and -induced siDJ-1#1, clones, we observed a partial leakage of Tet promoter activity in Doxy-untreated cells. The effect was much milder at the protein level (Fig. 1C). For consistency, all our experiments were carried out comparing DJ-1-interfered clones and scramble controls after 10 days of induction. In summary, we have obtained a human neuroblastoma cell model to study the effects of DJ-1 loss in vitro.
Gene Expression Analysis-To unveil the pathways altered after DJ-1 loss, we performed gene expression profiling on siDJ-1#1 clones displaying the maximum level of interference (clones A and B). Two scramble clones (clones a and b) were used as controls. Samples were hybridized on Affymetrix Gene-Chip Human Genome U133A 2.0 arrays. Data obtained from two independent experiments for each line were collected and analyzed.
To ensure that changes warranted further study, statistical analysis was performed with SAM (Significance Analysis of Microarrays (25)). Applying a false discovery rate of 10% and a 2-fold-change cutoff, we found 166 genes that were differentially regulated in both siDJ-1#1 clones versus the two scramble ones, with 102 down-regulated and 64 up-regulated transcripts. By Gene Ontology analysis, we clustered the differentially regulated genes to signal transduction, cell motility, developmental processes, cell adhesion, synaptic transmission, neurogenesis, and regulation of transcription. All the genes that were identified with a false discovery rate of 10% are shown in supplemental Table S1.
Fully stringent analysis (false discovery rate of 0%) was then applied, obtaining a list of 10 up-regulated and 16 down-regulated genes in DJ-1 RNAi cells (Table 1). As expected, DJ-1 was the most down-regulated gene.
To validate microarray data, we performed qPCR analysis on the top-three most up-regulated (GRM8, CDC42, CAMK2B) and down-regulated (ITGB1, FLNA, RET) genes. Changes in gene expression revealed by Affymetrix profiling were all confirmed by qPCR (Fig. 2, A and B). We found that the metabotropic glutamate receptor 8 (GRM8) was up-regulated more than 15-fold compared with scramble clones, whereas a 7-fold induction was observed for the GTPase CDC42 and 2-fold for the calcium/calmodulin-dependent protein kinase II beta (CAMK2B). A 2-fold reduction was indeed confirmed for integrin-␤1 (ITGB1) and filamin A  (FLNA). Importantly, a 9-fold decrease in mRNA expression was measured for RET.

DJ-1 Regulates RET Expression in Human Neuroblastoma
Cells-Because conditional genetic ablation of RET in DA neurons provokes PD-like dysfunctions and its ligand GDNF is a well known neurotrophic factor for DA cells, we decided to focus our attention on the effects of DJ-1 loss on RET expression.
To ascertain the specificity of DJ-1-mediated RET downregulation and to monitor the role of DJ-1 dosage, we analyzed RET mRNA levels in siDJ-1#2 clone (Fig. 3A). A strong down-regulation of RET expression (6-fold) was confirmed in this independent siDJ-1 clone, thus excluding any offtarget effects of sequence siDJ-1#1. Interestingly, we found that RET down-regulation was titrated with DJ-1 levels, further suggesting a causal relationship between DJ-1 and RET expression.
By Western blot analysis, RET protein resulted strongly reduced by loss of DJ-1 (Ͼ90% decrease) as shown by the almost complete disappearance of RET-specific immunostaining (Fig.  3, B and C). Compatible with a partial reduction of RET mRNA as measured with qPCR, RET protein levels in siDJ-1#2 clone were lowered, although to a lesser extent (ϳ60%).
To further confirm the specificity of DJ-1-mediated effects on RET expression, we performed rescue experiments in siDJ-1#1 clones. Scramble clones were used as controls. Because the siDJ-1#1 sequence was designed on the 3Ј-non-coding untranslated region, experiments were carried out with 2ϫFLAG-DJ-1 construct that contained only the open reading frame. The reintroduction of DJ-1 wild type (wt) protein restored the large majority of RET expression in siDJ-1#1 clones (80% of wt) (Fig. 3C). Interestingly, the effect of ectopic DJ-1 expression was also seen in scramble clones, where increased DJ-1 dosage further enhanced endogenous RET protein levels. Altogether, our results demonstrate that loss of DJ-1 decreases the expression of RET, the co-receptor for GDNF, in human neuroblastoma cells.

Loss of DJ-1 Stabilizes HIF-1a and Triggers
Hypoxia-The HIF-1a has been previously associated to a decrease in RET mRNA levels, thus representing a good candidate transcription factor for mediating the effects of DJ-1 loss (21). Because we did not detect any differences in HIF-1a transcripts by microarray analysis, we measured HIF-1a expression by Western blot. A specific up-regulation of HIF-1a protein levels was found in siDJ-1#1 as compared with scramble clones (Fig. 4A). Interestingly, HIF-1a protein was stabilized concomitantly to RET down-regulation in both siDJ-1#1 clones (Fig. 4A).
As a further support to HIF-1a protein stabilization in siDJ-1#1 cells, we found that some established HIF-1a target genes, like vascular endothelial growth factor A (VEGFA) and adrenomedullin (ADM), were up-regulated by DJ-1 loss (supplemental Table S1).
To verify whether HIF-1a and RET protein levels were sensitive to hypoxia in this in vitro system, SH-SY5Y neuroblastoma cells were treated with the hypoxia-mimic cobalt chloride (CoCl 2 ) for 6, 12, and 24 h. After 6 h of treatment, HIF-1a was strongly stabilized, whereas RET protein levels were reduced. After 24 h of treatment, RET protein was almost completely undetectable (Fig. 4B). To prove that the mechanism of RET down-regulation was strictly dependent on the stabilization of HIF-1a, we performed transient knockdown experiments for HIF-1a.
Wild type SH-SY5Y cells were transfected with siHIF-1a#1 and siHIF-1a#2 as well as with siRISC-FREE and water (mock) controls and treated with CoCl 2 for 24 h. RET and HIF-1a protein levels were then monitored by Western blot.
As shown in Fig. 4C, RET protein levels were consistently increased by HIF-1a silencing compared with controls in treated conditions. We then performed similar experiments in the inducible stable clones siDJ-1#1 A and B. Concomitant with silencing induction for DJ-1, we transiently transfected both siDJ-1#1 clones A and B with siHIF1a#1 and siHIF1a#2. RET and HIF-1a protein levels were then analyzed by Western blot. As shown in Fig. 4D, in both clones silenced for DJ-1 the reduction of HIF-1a was strictly correlated with an increase of RET expression. Therefore, RET protein levels were sensitive to loss of DJ-1 in an HIF-1a-dependent manner in human neuroblastoma cells.
We then asked whether the absence of DJ-1 was sufficient to cause a hypoxic state. By taking advantage of a fluorescent probe that quantifies cellular hypoxia (Hypoxyprobe), strong staining for hypoxic chemical adducts was revealed in DJ-1 siDJ-1#1 clones and was absent in scramble controls (Fig. 5A). These results were confirmed in siDJ-1#2 cells, proving the specificity of the effect ( supplemental Fig. S1).
To further prove that the establishment of the hypoxic state was strictly dependent on the loss of DJ-1 and to exclude offtargets of siDJ-1#1, we performed a rescue experiment transfecting 2ϫFLAG-DJ-1 in siDJ-1#1 clones. A construct encoding for an unrelated protein (DsRed2, red fluorescent protein) was used as an additional control. Immunofluorescence analysis revealed that overexpression of DJ-1, but not of red fluorescent protein, rescues the hypoxic status of neuroblastoma cells as compared with non-transfected cells (Fig. 5B). Scatter plot analysis of fluorescent signals was used to quantify the level of co-localization between Hypoxyprobe and DJ-1. Indeed, Pearson's coefficient indicated a negative correlation between the presence of hypoxic conditions and the expression of DJ-1. This effect was independent of cell densities and morphologies. Importantly, ectopic reintroduction of DJ-1 wt in siDJ-1#1 cells could reduce the levels of HIF-1a to 50% (Fig. 5C), compatible with the low transfection efficiency observed in neuroblastoma cells. Therefore, DJ-1 loss provoked a generalized hypoxic condition in SH-SY5Y neuroblastoma cells that stabilized HIF-1a transcription factor and inhibited RET expression level.

. Stabilization of HIF-1a protein in SH-SY5Y neuroblastoma cells is required for down-regulation of RET.
A, representative Western blot analysis of RET and HIF-1a protein levels is shown. Cell lysates have been extracted from inducible stable scramble clones a and b and siDJ-1#1 clones A and B, treated with Doxy (2.5 g/ml) for 10 days. B, SH-SY5Y neuroblastoma cells were untreated or treated with CoCl 2 for 6, 12, and 24 h. Protein levels of RET and HIF-1a were analyzed by Western blot. C, SH-SY5Y neuroblastoma cells were transfected with siHIF-1a#1, siHIF-1a#2, and siRISC-free or mock and treated with CoCl 2 for 24 h. RET and HIF-1a protein levels were analyzed by Western blot. D, shown is a Western blot of RET protein levels in siDJ-1#1 A and B clones transfected with siHIF-1a#1, siHIF-1a#2, siRISC-free or mock. All experiments were performed in triplicate.

Loss of DJ-1 Induces ROS Accumulation and Increases Cell
Death-Because hypoxia increases ROS production (30) and DJ-1 is involved in intracellular redox homeostasis (31), we asked whether in the absence of DJ-1, cellular ROS content was augmented. To this purpose we chose DHE as the fluorescent probe to monitor ROS production. After induction of DJ-1 interference, we could detect an increase of free radical species as compared with controls (Fig. 6).
We then asked whether ROS accumulation may impact cellular viability and proliferation. By fluorescence-activated cell sorter analysis, we did not observe any differences in siDJ-1#1 clones compared with scramble. However, a count of apoptotic nuclei revealed a small but significant increase in the number of dead cells as a consequence of DJ-1 interference. Increased apoptosis was further validated by the induction of poly(ADP-ribose)polymerase-1 cleavage (supplemental Fig. S2). In summary, DJ-1 loss in SH-SY5Y neuroblastoma cells provokes per se an accumulation of toxic ROS species and enhances cell death.
DJ-1 wt, but Not C106A or PD-linked L166P Mutants, Rescues RET Expression-Because a decrease in RET expression seems to be dependent on the establishment of hypoxia and ROS accumulation and Cys-106 of human DJ-1 has been proved to be essential for its ROS scavenging activity (31), we assessed whether Cys-106 is involved in DJ-1 control of RET expression. Rescue experiments were then carried out in siDJ-1#1 cells. As expected, the reintroduction of DJ-1 wt protein restored RET expression in siDJ-1#1 clones (Fig. 7). In contrast, overexpression of 2ϫFLAG-C106A mutant DJ-1 had no effect on RET levels, proving that DJ-1 ROS scavenging activity is required for the regulation of RET expression.
Finally, we asked whether RET down-regulation may also play a role in PD cases with DJ-1 missense mutations. To this purpose we transfected the PD-associated L166P mutant into siDJ-1#1 cells. L166P protein levels were lower than wt DJ-1, as expected (32). Interestingly, L166P mutant was not able to rescue RET expression as assayed by Western blot (Fig. 7). Therefore, Cys-106 is required for DJ-1-mediated regulation of RET expression, and FIGURE 5. Loss of DJ-1 triggers hypoxia. A, shown is analysis of hypoxia conditions in both scramble clones a and b and both siDJ-1#1 clones, A and B, treated with Doxy (2.5 g/ml) for 10 days. Hypoxyprobe was added into the medium of both scramble and siDJ-1#1 clones for 2 h, and then immunofluorescence was performed using a monoclonal anti-Hypoxyprobe antibody as described under "Experimental Procedures." Scale bar, 50 m. DAPI, 4,6-diamidino-2-phenylindole. B, shown is an analysis of hypoxia condition in both scramble clones a and b and both siDJ-1#1 clones, A and B, treated with Doxy (2.5 g/ml) for 10 days and transfected with 2ϫFLAG-DJ-1 (DJ-1) or red fluorescent protein (ctrl), respectively. Scale bar, 20 mm. A representative scatter plot of colocalization analysis of siDJ-1#1 clones transfected with 2ϫFLAG-DJ-1 (DJ-1) or red fluorescent protein (ctrl) is shown. Horizontal and vertical scales represent fluorescence intensity value of green and red confocal channels, respectively. Pearson's coefficient of colocalization is indicated. All experiments have been performed in triplicate. *, p Ͻ 0.05. C, shown is a representative Western blot analysis of HIF-1a protein levels in siDJ-1#1 and scramble clones treated with Doxy (2.5 g/ml) for 10 days and transfected with empty vector or 2ϫFLAG-DJ-1. Experiments have been performed in triplicate and on two different scramble clones a and b and siDJ-1#1 clones A and B.

DJ-1 Regulates RET Expression
PD-linked L166P mutation is unable to maintain the physiological level of RET.

DISCUSSION
Until now at least five independent mouse lines, four Drosophila models, one zebrafish, and one Caenorhabditis elegans mutants have been generated to study the effects of lack of DJ-1 on mesencephalic DA neurons as seen in PD pathogenesis (31,(33)(34)(35). Interestingly, no DA cell loss has been reported even in aging animals, suggesting the in vivo establishment of compensatory mechanisms for DJ-1 absence. Some alterations in dopamine levels, release, and uptake have been indeed observed that may explain some limited age-dependent behavioral changes (8). A number of neuronal and non-neuronal cell lines with constitutive interference of DJ-1 expression have also been reported. The only phenotype that appears to be common to the various animal and cellular models is the enhanced sensitivity to exogenous oxidative stressors, including MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine) and paraquat as well as proteasome inhibitors. Some of these effects have been associated with an alteration of the Nrf2-mediated antioxidant response both in vitro and in vivo (36). Overexpression data have provided complementary support to the hypothesis of DJ-1 as a major component of cellular redox homeostasis, proving its protective function.
Although no increase of endogenous oxidative stress has been reported in at least two different KO mouse strains, iso-lated mitochondria from another KO mouse model showed a 2-fold increase in H 2 O 2 content and a 60% reduction of mitochondrial aconitase activity (31). In older mice an increase of mitochondrial superoxidase dismutase and glutathione peroxidase was observed, leading to the hypothesis that DJ-1 acts as an atypical peroxiredoxin-like peroxidase scavenging H 2 O 2 through the Cys-106 residue.
Animal models do not, thus, seem to provide a satisfactory tool to unveil the molecular mechanisms altered by DJ-1 loss in vivo. Furthermore, the constitutive complete absence of DJ-1 expression in neuronal cells in culture has been difficult to achieve for the positive selection of residual DJ-1-expressing clones. Therefore, a full elucidation of the biological processes that are altered in absence of DJ-1 remains unanswered.
To address this important issue, we decided to generate human neuroblastoma SH-SY5Y cells with inducible downregulation of DJ-1 expression. This cell line has been widely adopted as in vitro model to study PD for its high dopamine-␤hydroxylase content (29,(37)(38)(39)(40)(41). The use of an inducible promoter allowed an unbiased selection of clones when the effects of down-regulation of DJ-1 expression were not yet manifested. With this strategy we could obtain cells in which DJ-1 protein levels were almost undetectable. Taking advantage of the clones showing the strongest down-regulation of DJ-1, we performed gene expression analysis to unveil molecular mechanisms altered by the loss of DJ-1. The strong down-regulation of DJ-1 was responsible for the alteration of 166 genes referring to various GO identifiers including oxidative stress response. Among them, the Nrf2-dependent NAD(P)H dehydrogenase, quinone 1 (NQO1), and SLC7A11, a subunit of the glutamate/cystineantiporter system x(c)(Ϫ) that regulates intracellular glutathione levels, were induced. These data are in agreement with previous gene expression analyses performed in constitutive DJ-1 knock-down cells that highlighted the loss of the antioxidant gene network dependent on the master regulator Nrf2, including its target NQO1 (36).
In addition, several elements of glutamatergic neurotransmission were altered, including an up-regulation of GMR8 and

DJ-1 Regulates RET Expression
of the ionotropic GRIA2 with a concomitant down-regulation of GMR7 (supplemental Table S1). A strong induction of CAMK2B was also observed, and this may be relevant for the role of this kinase in neuronal plasticity and its induction after amphetamine sensitization and schizophrenia.
Within the 26 genes identified with a more stringent statistical analysis, 3 seemed to implicate pathways involved in PD. A strong down-regulation of mitofusin1 may suggest the interplay between DJ-1 expression and the mitochondrial fusion/ fission machinery. Furthermore, the increase in the eukaryotic translation initiation factor 2␣ kinase 1 highlights the importance of a well tuned translational control, a crucial homeostatic response to cellular stress.
Among DJ-1-regulated genes, we decided to focus our attention on the GDNF receptor RET. RET expression level, both transcript and protein, was strongly decreased in the absence of DJ-1. Furthermore, we found that the observed RET down-regulation was strictly correlated with the residual quantity of DJ-1, as interference clones expressing 50% of DJ-1 showed a modest but consistent decrease of RET. It is noteworthy that only a strong downregulation of DJ-1 was able to completely inhibit the expression of RET receptor. This is important considering that only the homozygous deletion of DJ-1 has been associated to PD onset.
The down-regulation of RET has strong implications for PD as its ligand GDNF promoted survival and differentiation of midbrain DA cells (15). Importantly, conditional KO mice for this receptor showed a progressive and late degeneration of these neurons, as in PD (19). Furthermore, PD-mimicking toxins like MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine), rotenone, and 6-hydroxydopamine were able to downregulate RET both in cellular models and in mice (42). Therefore, down-regulation of RET expression could implicate that GDNF neurotrophic support to DA neurons may be impaired by DJ-1 loss preceding neuronal degeneration.
By compiling a list of all transcription factors that may regulate RET, we found that HIF-1a was the only transcription factor associated so far with a down-regulation of RET mRNA level (21). We then demonstrated that the post-transcriptional stabilization of HIF-1a subunit triggered by DJ-1 loss was required for a strong down-regulation of RET protein in human neuroblastoma cells.
By taking advantage of Hypoxyprobe we then demonstrated that in the absence of a functional DJ-1, human neuroblastoma cells suffered from hypoxic conditions. Rescue experiments indeed proved that the reintroduction of DJ-1 expression restored physiological O 2 concentration independently of cellular densities or morphologies. This is the first evidence that a lack of DJ-1 expression per se generates hypoxia, although a role of DJ-1 in this homeostatic response has been recently reported (43,44). When U2OS osteosarcoma cells and transformed mouse embryo fibroblasts were cultured at low O 2 concentration, the absence of DJ-1 hampered HIF-1a stabilization and a HIF-1a-dependent transcriptional program (43). This led to the hypothesis that DJ-1 is required for a correct cellular response to hypoxia. Although our data seem to suggest a different effect of DJ-1 absence, our work has been almost exclusively carried out in normoxia conditions. Interestingly, when Vasseur et al. (43) analyzed the effects of lack of DJ-1 in U2OS cells in normoxic conditions, an increase, albeit small, of HIF-1a protein levels was observed, as we reported here in neuroblastoma cells. Furthermore, the absence of DJ-1 triggered the early stages of autophagy, a well known cellular response to hypoxia (43). Therefore, future experiments will address whether the absence of DJ-1 exerts different effects in normoxia and hypoxia or if cell type-specific responses take place.
Although hypoxia as well as HIF-1a induction are not presently implicated in the pathogenesis of PD, vascular impairment has been described in several human post mortem brains of PD patients. Burke et al. (45) and Oo et al. (46) have previously shown that brief ischemic-anoxic insult in a 7-day-old pup provoked apoptosis in DA neurons of the substantia nigra, suggesting a differential vulnerability of these cells to low oxygen levels (47,48). It is worth noting that DJ-1 deficiency sensitized brains to ischemic damage in vivo and that its protective activity depended on Cys-106 (49). Furthermore, hypoxia has been recently been implicated in a variety of neurodegenerative diseases including Alzheimer and amyotrophic lateral sclerosis. Therefore, there is an urgent need for a thorough analysis of the presence of hypoxia in vivo in PD patients and animal models.
As for the neuroprotective role after ischemia, Cys-106 is essential in DJ-1-mediated control of hypoxia and RET expression. This is significant as Cys-106 was required for DJ-1 peroxiredoxin-like peroxidase activity, and a C106A mutation blocked oxidation-induced mitochondrial localization and protection against 1-methyl-4-phenylpyridinium (MPP ϩ ) toxicity in neuronal cells (50). Furthermore, Cys-106 has been found oxidized both in human post mortem brains as well as in animal models of PD. Because DJ-1 is not a substrate for sulfenic acid reductase, its oxidation is irreversible, leading to protein inactivation. These data further suggest important commonalities in the biochemical mechanisms of DJ-1 acting as a neuroprotective molecule to ROS damage and as a regulator of oxygen homeostasis.
Neurons have a high energy demand, mainly used for maintaining ion gradients for action potentials and for transportation of molecules. Because glycolysis provides less than 5% of energy supply, their ATP production relies on molecular oxygen to support oxidative phosphorylation. O 2 sensing and metabolisms, thus, play a fundamental role in cellular homeostasis as hyperoxia and hypoxia, both, could be detrimental to cellular physiology. It is well known that oxygen tension regulates mitochondrial DNA-encoded Complex I gene expression (48), and high oxygen concentration induces mitochondrial biogenesis (51). A high rate of energy production may then increase intracellular ROS. On the other hand, chronic low oxygen tension may provoke oxidative stress, triggering ROS production at the mitochondrial complex III (30). The lack of the mitochondrial protein DJ-1 may then be responsible for the impairment of the oxygen-sensing machinery and for an increase of free radical species leading to the alteration of several O 2 -dependent cellular processes including the expression of the GDNF receptor RET, an indispensable player for the survival of DA neurons.