Paraquat Neurotoxicity Is Mediated by a Bak-dependent Mechanism*

Paraquat (PQ) causes selective degeneration of dopaminergic neurons in the substantia nigra pars compacta, reproducing an important pathological feature of Parkinson disease. Oxidative stress, c-Jun N-terminal kinase activation, and α-synuclein aggregation are each induced by PQ, but details of the cell death mechanisms involved remain unclear. We have identified a Bak-dependent cell death mechanism that is required for PQ-induced neurotoxicity. PQ induced morphological and biochemical features that were consistent with apoptosis, including dose-dependent cytochrome c release, with subsequent caspase-3 and poly(ADP-ribose) polymerase cleavage. Changes in nuclear morphology and loss of viability were blocked by cycloheximide, caspase inhibitor, and Bcl-2 overexpression. Evaluation of Bcl-2 family members showed that PQ induced high levels of Bak, Bid, BNip3, and Noxa. Small interfering RNA-mediated knockdown of BNip3, Noxa, and Bak each protected cells from PQ, but Bax knockdown did not. Finally, we tested the sensitivity of Bak-deficient mice and found them to be resistant to PQ treatments that depleted tyrosine hydroxylase immuno-positive neurons in the substantia nigra pars compacta of wild-type mice.

The depletion of dopaminergic neurons in PD is thought to result from programmed cell death (13), a process that is critical for eliminating cells in development and homeostasis. The most common form of programmed cell death is Type I/apoptosis, which can be triggered through extrinsic and intrinsic cell death pathways (14). Activation of the intrinsic pathway centers on the release of apoptogenic factors from mitochondria, including cytochrome c (15). Cytochrome c normally shuttles electrons between complexes III and IV in intermembrane space of mitochondria, but once released into the cytoplasm it complexes with Apaf-1 and procaspase-9 to form an active apoptosome (16). Because apoptosome formation quickly leads to executioner caspase activation and apoptosis, the release of cytochrome c is tightly regulated. Pro-apoptotic members of the Bcl-2 family facilitate cytochrome c release by causing transient membrane disruptions, referred to as mitochondrial outer membrane permeabilization (MOMP; 17). Once unrestrained, Bcl-2 family members Bax and Bak trigger MOMP. In healthy cells Bax and Bak are blocked by anti-apoptotic members such as Bcl-2 and Bcl-xL (18). Apoptotic stimuli can disrupt these interactions to disinhibit Bax and/or Bak and cause MOMP. Pro-apoptotic BH3-only members of the family bind to antiapoptotic members, such as Bcl-2 and Bcl-xL, thereby freeing Bax and Bak to cause MOMP and release cytochrome c.
The cytotoxic actions of PQ clearly involve reactive oxygen species (ROS), but its specificity for dopaminergic neurons and possible involvement of the intrinsic cell death pathway are unclear (9). Although positively charged, PQ is actively transported across the blood-brain barrier by the same neutral amino acid transporter used by L-valine and L-dopa (12,20). Inside neurons, PQ induces oxidative stress by producing superoxide anions through redox cycling (11). Dopaminergic neurons may be particularly sensitive to high ROS because of dopamine metabolism that creates hydrogen peroxide and superoxide radicals, which in turn react with nitric oxide to form highly toxic peroxynitrite (21). High ROS production stretches metabolic resources as the cell removes misfolded proteins, replaces oxidized lipids, and repairs damaged DNA (22). The cells that cannot compensate for this oxidative stress undergo default apoptosis or necrosis under severe conditions, but PQ-treated cells can be protected by superoxide dismutase/ catalase mimetics (23)(24)(25). Although it is known that PQ-induced ROS activates JNK (24,26) and leads to caspase-3 cleavage, mechanisms acting in between have not been identified.
In the present study, we investigated PQ-induced cell death, which shows morphological and biochemical features that are consistent with apoptosis. PQ induced high levels of proapoptotic Bcl-2 family members. Loss-of-function approaches demonstrated that Bak knockdown protected cells from PQ, and Bak-deficient mice were resistant to PQ neurotoxicity. We found that PQ neurotoxicity is mediated by a Bak-dependent mechanism.

EXPERIMENTAL PROCEDURES
Reagents-Unless otherwise indicated all reagents were purchased from Sigma. Dulbecco's modified Eagle's medium: Ham's F-12 medium was from Biowhittaker, and fetal calf serum was from Hyclone (Logan, UT). Paraquat was from Chem Service (West Chester, PA), and z-VAD-fmk was from Enzyme Systems (Aurora, OH). Hoechst 33342, propidium iodide, CalceinAM, and TMRE were from Molecular Probes-Invitrogen (Eugene, OR).
SK-N-SH Cell Lines-Parental lines of SK-N-SH were obtained from ATCC maintained in Dulbecco's modified Eagle's medium:Ham's F-12 medium (1:1) supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. The cells were grown and treated in a humidified incubator with 5% CO 2 and 10% O 2 at 37°C. Thawed vials of SK-N-SH cells were used for no more than 15 passages.
Treatment Conditions-Typically, the cells were plated at 3.4 ϫ 10 4 /cm 2 (3.0 ϫ 10 5 /well) in each well of a 6-well plate and grown for 24 h prior to experimental procedures. Cell death was induced by incubation with 200 or 400 M paraquat/PQ (unless otherwise indicated) for 18 h. When cycloheximide (1 M) or z-VAD-fmk (50 M) were used, they were added to media at time 0. Staurosporine was used at 250 nM.
Assessment of Cell Viability and Apoptosis-To assay nuclei for apoptotic morphology, the cells were incubated in medium containing 25 M Hoechst 33342 for 5 min at room temperature, and the images were captured using a Nikon TE2000 inverted fluorescent microscope with a 20ϫ FLUOR objective and a Hamamatsu ORCA-AG 12-bit CCD camera using Image-Pro Software (3-6 images/well). For CalceinAM/propidium iodide viability assay, the cells were incubated in medium containing 5 M CalceinAM and 100 ng/ml propidium iodide for 15 min at 37°C, and then the images were captured with a fluorescent video microscope (3-6 images/well). For ⌬ cells were incubated in medium containing 50 nM TMRE for 20 min at 37°C and kept on ice until analysis with a BD FACScan.
Cytochrome c Release from Mitochondria-The cells were fractionated into cytosolic and heavy membrane fractions, as previously described (27). Briefly, the cells were lysed in cold homogenization buffer (20 mM HEPES-KOH, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM dithiothreitol) containing 250 mM sucrose and homogenized on ice with a Dounce homogenizer. The homogenates were centrifuged at 700 ϫ g for 10 min at 4°C, and the supernatants were centrifuged at 22,400 ϫ g for 15 min at 4°C. The resulting supernatants were the cytosolic fraction, and the pellets were mitochondria-rich. The pellets were resuspended in mitochondria buffer (50 mM HEPES-KOH, pH 7.4, 1 mM EDTA, 0.2 mM dithiothreitol, 1% Nonidet P-40) containing 10% glycerol and further centrifuged at 22,400 ϫ g for 15 min at 4°C. The resulting supernatants were heavy membrane/mitochondrial fractions. The fractions were separated and immunoblotted with cytochrome c-specific antibody (no. 556433; BD Biosciences).
Stable SK/Bcl-2 Lines-Stable Bcl-2-expressing SK-N-SH lines were made by transfecting SK-N-SH cells with His 6tagged human Bcl-2/pcDNA3 (gift of Dr. Doug Green) using Lipofectamine 2000, followed by selection for G418 resistance. Expression of the transgene was confirmed by immunoblotting whole cell extracts with anti-His antibody (sc-8036; Santa Cruz).
RNA Isolation and qPCR-The cells were rinsed with PBS, and total RNA was isolated with TRIzol (Invitrogen). Onetenth volume of chloroform was added, and the solutions were mixed by vortex. After centrifugation, the aqueous phase (containing RNA) was extracted and mixed with 1/10 volume of 5 M NaAC (pH 5.2). RNA was precipitated by mixing with equal volumes of isopropanol at room temperature for 30 min. Precipitated RNA was concentrated by centrifugation at 14,000 ϫ g (4°C) for 30 min and then rinsed with RNase-free 90% EtOH, followed by a second centrifugation. After removal of trace EtOH, the RNA pellet was allowed to air dry for 5-10 min and resuspended in RNase free 10 mM Tris. RNA was quantified by spectrophotometry (A 260 ), and 1 g of total RNA was used for each reverse transcriptase reaction with Superscript III (Invitrogen) and oligo(dT) primers, at 65°C for 5 min, on ice for 1 min, at 50°C for 60 min, and at 70°C for 15 min. Each qPCR contained 1/50 th of a single reverse transcription reaction (20 ng of initial template), primers, and qPCR master mix containing SYBR green (Abgene, UK). Primer sets that cross an intron-exon boundary were preferentially used to prevent background amplification of genomic DNA. Triplicate reactions for each oligonucleotide primer set were prepared and run simultaneously in a 96-well plate using an ABI PRISM 7700 sequence detection system with the following conditions: 15 s at 95°C and 1 min at 60°C for 40 cycles. Glyceraldehyde-3-phosphate dehydrogenase and actin primers served as internal controls.
siRNA-Small interfering RNA (siRNA) for human Bak, Bax, and BNip3 were designed with the Whitehead Institute siRNA selection program (28). The sequences for siRNA were: human BNip3 sense, 5Ј-GAA GAU GGG CAG AUC AUG UUU; human BNip3 antisense, 5Ј-ACA UGA UCU GCC CAU CUU CUU; human Bak sense, 5Ј-CCG ACG CUA UGA CUC AGA CUU; human Bak antisense, 5Ј-CUC UGA GUC AUA GCG UCG GUU; human Bax sense, 5Ј-CCG CCU CAC UCA CCA UCU GUU; and human Bax antisense, 5Ј-UUG GCG GAG UGA GUG GUA GAC. Those siRNA were synthesized at the DNA/RNA Peptide Synthesis Facility in the Beckman Research Institute at the City of Hope Medical Center (Duarte, CA). Noxa siRNA were purchased (sc-37305; Santa Cruz Biotechnology). Duplexed siRNA was transfected into SK-N-SH cells with HiPerFect (Qiagen) as per the manufacturer's instructions. Knockdown was confirmed with immunoblots, using whole cell lysates.
Immunostaining-Wild-type C57BL/6 mice were obtained from Charles River (Hollister, CA). They received two intraperitoneal injections of either saline or 10 mg/kg PQ separated by a 1-week interval and were sacrificed 2 days after the second injection. This time point was chosen because it coincides with the onset of cell loss in the SNpc (46). The mice were deeply anesthetized and intracardially perfused with heparinized saline, followed by 4% paraformaldehyde. All of the animal studies were done according to National Institutes of Health and Institutional Animal Care and Use Committee guidelines. The brains were excised and cryoprotected by immersion in sucrose solution prior to sectioning on a cryostat. Coronal sections through the entire midbrain were made at 40 M and stored in preservative at Ϫ20°C until use. Free floating sections were rinsed in PBS and antigen recovered by immersing the sections in 80°C 10 mM citrate (pH 8.5) for 20 min. Endogenous peroxidase activity was then depleted by immersing sections in 3% H 2 O 2 for 10 min and rinsing three times in PBS. The sections were blocked by incubation in 10% normal goat serum in PBS for 1 h, and the sections were incubated with primary antibody in PBST (1% normal goat serum) at 4°C for 48 h. Primary antibodies used were anti-Noxa (IMG-451 at 1:200; Imgenex), anti-Bak (TC-102 at 1:200; Calbiochem), anti-BNip3 (B7931 at 1:100; Sigma), and anti-TH (T-2928 at 1:800; Sigma). The sections were rinsed three times with PBST for 10 min each, then incubated in appropriate biotinylated secondary antibodies (anti-mouse BA-9200 or anti-rabbit BA-1000, both from Vector Labs) for 2 h at room temperature, and processed with ABC Vectastain as per the manufacturer's specifications (Vector Labs). The sections were developed with 600 M diaminobenzidine in PBS for 5-20 min, with 1.5 ϫ 10 Ϫ3 % H 2 O 2 . Midbrain sections from six PQ-treated and six saline-treated mice were used for immunostaining. SNpc in each section was traced using a mouse brain stereotaxic atlas as a guide (29), and immuno-positive cells were counted within those boundaries. Cell counts were done for BNip3 on six sections, Noxa on five sections, and Bak on four sections. The statistical significance was established from using Student's t test S.E.
Bak Knock-out Mice-Eight-week old mice that were deficient in Bak (strain B6-129-Bak1 tm1Thsn /J) and the appropriate control strain (C57Bl/6J) were obtained from Jackson Laboratories (Bar Harbor, ME). The animals received three intraperitoneal injections of either PQ or saline spaced 1 week apart. The mice were sacrificed 1 week after the last injection. A block containing the midbrain was immersion-fixed in 4% paraformaldehyde overnight at 4°C prior to cryoprotection in graded sucrose solutions. The tissue block was snap frozen in cold isopentane and sectioned into 40 M sections on a cryostat for stereological cell counting.
Stereology Cell Counting-For stereological cell counting, a series containing every sixth section throughout the entire substantia nigra was immunostained with an antibody against tyrosine hydroxylase (1:800; Pel Freez Biologicals, Rogers, AR) and counterstained with 0.5% Cresyl violet as previously described (11,30). The substantia nigra was delineated at low magnification and subsequently sampled at high magnification (StereoInvestigator; MBF Bioscience, Williston, VT). Stereological estimates were made of both the number of TH-positive cells in the substantia nigra and the total number of nigral neurons using the optical fractionator method to ensure that loss of TH immunoreactivity was due to neuronal loss and not a downregulation of the marker (11). The co-efficient of error ranged from 0.07 to 0.09 (19).

RESULTS
PQ Induces Apoptosis-Apoptotic cells usually display a condensed brightly stained nucleus (Fig. 1A). PQ significantly increased the percentage of apoptotic nuclei in cultures of SK-N-SH neuroblastoma cells treated for 18 h (Fig. 1B). This cell death could be blocked by the ribosome inhibitor cycloheximide or the caspase inhibitor z-VAD-fmk. Another feature of apoptotic cells is the activation of caspase 3 and the cleavage of caspase substrates such as poly(ADP-ribose) polymerase, and both occurred in PQ-treated cells (Fig. 1C). These findings were consistent over three separate experiments and in studies involving longer time points (not shown). A key step in the intrinsic cell death pathway is the release of cytochrome c from the intermembrane space of mitochondria. Subcellular fractionation showed dose-dependent increases of cytochrome c in cytosolic extracts, indicative of release (Fig. 1D). Cytochrome c release in response to PQ could be blocked by cycloheximide, but not z-VAD-fmk, indicating caspase-independent release. Together, these findings demonstrate that PQ neurotoxicity induces typical apoptosis, which requires new protein synthesis and caspase activity.
To determine whether cycloheximide and z-VAD-fmk blocked cell death and not simply the appearance of apoptotic nuclei, we also evaluated cells for viability using calceinAM ( Fig.  2A and supplemental Fig. S1) and YOPRO3 (supplemental Fig.  S1) assays. Cultures treated with PQ showed a significantly lower percentage of viable cells compared with controls. Cycloheximide and z-VAD-fmk provided significant protection Paraquat Neurotoxicity through Bak FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 3359 against PQ-induced death, at least until the time of assay. To determine whether MOMP may be relevant for PQ-induced apoptosis, we generated SK-N-SH cell lines that stably express anti-apoptotic Bcl-2 (supplemental Fig. S1). SK/Bcl-2 clones were highly resistant to PQ neurotoxicity ( Fig. 2A). The resistance of SK/Bcl-2 clones to PQ was consistent over three experiments. These findings establish that PQ-induced apoptosis can be blocked by at least one anti-apoptotic member of the Bcl-2 family, suggesting that the mechanism may involve MOMP.
To confirm that PQ-induced cell death is apoptotic, as opposed to necrotic or caspase-independent, we evaluated inner mitochondrial membrane potential (⌬) using the fluorescent tracer TMRE. During necrotic cell death, permeability transitions cause the matrix to swell and burst, disrupting ⌬ and releasing cytochrome c simultaneously. In contrast, during apoptosis, the structural integrity of the inner mitochondrial membrane is maintained through MOMP and cytochrome c release, until active caspases cleave critical proteins and dissipate ⌬ (31). FACS analysis of TMRE stained SK-N-SH cells showed that PQ reduced the percentage of cells with ⌬, compared with controls (Fig. 2B), and that this effect could be blocked by cycloheximide or by Bcl-2 overexpression in SK/Bcl2 cells. Significantly, z-VAD-fmk prevented the loss of ⌬ in cells treated with PQ ( Fig. 2B and supplemental Fig. S2), indicating that PQ neurotoxicity does not directly disrupt inner mitochondrial membrane integrity but that loss of ⌬ is a downstream, caspase-dependent process.
Cycloheximide-mediated protection established that PQ neurotoxicity requires new protein synthesis that may include pro-apoptotic members of the Bcl-2 family. We evaluated expression patterns of key Bcl-2 family members using qPCR and immunoblots. Bak and Bax are key players in MOMP as bak Ϫ/Ϫ /bax Ϫ/Ϫ cells are resistant to DNA damage, endoplasmic reticulum stress, and many other apoptotic stimuli (32,33). PQ induced bak mRNA (3x) (Fig. 3A) and Bak protein (1.5x) (Fig. 3B). Interestingly, PQ did not induce bax mRNA (Fig. 3A) or Bax protein (Fig.  3B). Bak is constitutively localized to the outer mitochondrial membrane where it is inhibited by Bcl-xL and Mcl-1 (34,35). PQ induced bcl-xL and bfl-1/A1 expression (Fig.  3A) but not mcl-1. Protein levels of Bcl-xL and Bfl-1 were also higher (Fig. 3B). Anti-apoptotic members use direct binding to prevent Bak and Bax from causing MOMP and permitting cytochrome c release. This inhibition can be disrupted by BH3only members that bind to anti-apoptotic members, freeing Bak and Bax to cause MOMP. PQ induced the BH3-only members bnip3, noxa, and bid (Fig. 3C). Interestingly, BNip3 and Noxa bind to Bcl-xL and  Mcl-1, respectively, two important inhibitors of Bak. BNip3 (1.6x) and Noxa (2.5x) protein levels were also higher after PQ treatment. Full-length Bid is cytoplasmic until cleaved by caspases-8, -10, or -2 to form tBID, which then translocates to mitochondria and facilitates both Bak and Bax activation. Although PQ potently induced Bid expression in SK-N-SH cells, it is noteworthy that SH-SY5Y cells, the parental line of SK-N-SH cells, do not express caspase-8 because of DNA methylation (36,37). Taken together, these findings demonstrate that PQ induces the pro-apoptotic Bcl-2 family member Bak and two of its activators, BNip3 and Noxa.
Bak, BNip3, and Noxa Knockdown Protect against PQ Neurotoxicity-To determine the physiological significance of BNip3, Noxa, and Bak in PQ neurotoxicity, we knocked down the expression of each one using siRNA. Bak siRNA protected SK-N-SH cells completely at 200 M and partially at 400 M of   (Fig. 4A). Significant knockdown of Bak protein under those conditions was confirmed by immunoblotting (Fig. 4B). Bax is critical from MPTP-induced nigrostriatal degeneration (38) but was not strongly induced by PQ, so we used siRNA to knockdown Bax as a control. Knocking down Bax protein levels did not protect SK-N-SH cells from PQ at 200 or 400 M (Fig.  4C). Immunoblotting confirmed knockdown of Bax under these conditions (Fig. 4D). These findings suggested that PQ does not have a generalized effect on Bcl-2 members but depends specifically on Bak for MOMP. BNip3 knockdown also protected SK-N-SH cells from PQ at 200 and 400 M (Fig. 4E). Immunoblotting revealed significant knockdown of BNip3 protein in cells treated with siRNA, but PQ induction of BNip3 was still noticeable even in extracts from knockdown cultures (Fig.  4F). SiRNA-mediated knockdown of Noxa protected SK-N-SH cells from the toxic effects of PQ at 200 and 400 M (Fig. 4G). Protection from 400 M PQ was higher with Noxa siRNA than with Bak or BNip3 knockdown. Immunoblots confirmed Noxa knockdown under the conditions tested (Fig. 4H). Taken together, lowering protein levels of Bak, BNip3, or Noxa each provided protection against PQ neurotoxicity.
Bak Knock-out Mice Are Resistant to PQ Neurotoxicity-We immunostained midbrain slices from PQ-treated mice and found significantly more Bak and BNip3 immuno-positive cells in the SNpc than in PBS-treated controls (Table 1 and supplemental Figs. S3 and S4). To conclusively demonstrate the significance of Bak in PQ neurotoxicity in vivo, we tested Bakdeficient (bak Ϫ/Ϫ ) and control C57BL6/J mice for their sensitivity to PQ-induced depletion of TH ϩ neurons in the SNpc. Bak-deficient mice have no overt phenotype. Wild-type mice injected with saline had 5925 Ϯ 54 TH ϩ neurons/side in cross-sections at the level of the third cranial nerve. However, PQ-treated mice had only 4110 Ϯ 116 TH ϩ neurons (Fig. 5A), which was a statistically significant decrease (p Ͻ 0.005). In contrast, bak Ϫ/Ϫ mice injected with saline had 5671 Ϯ 111 TH ϩ cells/side, and PQ-treated mice had 5435 Ϯ 411, a difference that was not statistically significant. Nissl-stained neuron counts in the SNpc of these mice showed similar results. Wildtype mice injected with saline had 7122 Ϯ 197 Nissl-stained neurons/side, but PQ-treated wild-type mice had 5136 Ϯ 151 (Fig. 5B), a statistically significant decrease (p Ͻ 0.005). However, bak Ϫ/Ϫ mice injected with saline had 6688 Ϯ 215 neurons in the SNpc/side, whereas PQ-treated knock-outs had 6508 Ϯ 432, which was not significantly different. Together, these data indicate that PQ induces the production of new proteins that activate Bak, resulting in biochemical and morphological changes that are consistent with apoptotic cell death.

DISCUSSION
Since the discovery that MPTP can induce selective nigrostriatal neurodegeneration, there has been a concerted effort to identify environmental factors that contribute to sporadic PD. The discovery of several PD-linked loci has provided important insights into mechanisms that can kill nigrostriatal neurons, yet a growing body of work continues to show that late onset PD has a significant environmental component. A more thorough understanding of gene-toxicant interactions is required to determine how environmental triggers exacerbate genetic predisposition in some individuals but not others. The present study adds to this knowledge by establishing the significance of a key cell death pathway in the neurotoxicity of a widely used herbicide that has been implicated in sporadic PD.
Overproduction of ROS has been implicated in the pathophysiology of PD, Alzheimer disease, polyglutamine diseases, amyotrophic lateral sclerosis, and other neurodegenerative disorders (1). ROS modify proteins causing misfolding and aggregation, as well as the activation of cell death pathways, which are common features in neurodegeneration. ROS are incontestably involved in the mechanisms by which PQ kills dopaminergic neurons (26,39); however, molecular links between PQinduced oxidative stress and caspase activation remain unresolved. In the present study, we found that PQ triggers apoptosis through the intrinsic cell death pathway, which includes Bak-dependent MOMP, cytochrome c release, caspase-3 activation, and ultimately apoptotic cell death.
Although PQ shares some structural similarities with MPP ϩ (40, 41), early suggestions that they share a common mecha-  Wild-type (wt) mice treated with PQ showed a statistically significant (***, p Ͻ 0.005) depletion of neurons in the SNpc, but there was no significant difference between total neurons in bak Ϫ/Ϫ mice treated with saline or PQ. Mice of each genotype received saline or serial PQ injections, as previously described (11). The mice were coded and perfused, and their brains were sectioned for immunostaining with anti-TH. Blinded stereological counting of TH ϩ and total neurons in the SNpc was done for three sections from three mice in each group. The error bars represent S.E.
nism have not been supported. Unlike MPP ϩ , PQ entry into neurons does not rely on DAT (42), and it has a lower binding affinity for complex I (43). The unresolved ability of PQ to inhibit mitochondrial complex I (44) has led to speculation that it may kill neurons through ATP depletion leading to necrotic death. We found that PQ neurotoxicity does not lead to the direct loss of ⌬ through mitochondrial dysfunction, because caspase activity was required to compromise the inner mitochondrial membrane. PQ-induced apoptosis requires new protein synthesis to initiate apoptosis, and our findings suggest that BNip3 and Noxa are among the newly made proteins that may drive this process. Further divergence between MPTP/ MPP ϩ and PQ effects on the oxidation status of thioredoxin have also been shown to impact caspase-3 cleavage (39). MPP ϩ activates cell death through JNK, p53, and ultimately Bax-mediated release of cytochrome c. Although PQ also activates JNK and its downstream target c-Jun (26), in knockdown studies we found that Bax is not required for PQ-induced cell death. Links between ROS generation/JNK activation and caspase-3 activation remain a black box that may involve components, such as transcription factors, mitochondrial dysfunction, chaperones, the proteasome, and others. Our findings suggest that these upstream influences are integrated by Bcl-2 family member expression profiles. PQ neurotoxicity acts through a Bak-dependent mechanism that may rely on BNip3 and Noxa (Fig. 6). Our finding that Bak-deficient mice are resistant to PQ provides conclusive evidence that Bak is necessary for PQ neurotoxicity in vivo. Bak is constitutively present on the surface of mitochondria where it is inhibited by Mcl-1 and Bcl-xL but not Bcl-2 (34,45). Huang and colleagues (34,35) have reported that Bak-mediated cytochrome c release requires a double hit to neutralize both anti-apoptotic Bcl-xL and Mcl-1. Although PQ raised protein levels of Bcl-xL slightly, it strongly induced the expression of BH3-only members that interfere with Bcl-xL and Mcl-1 function. BNip3 specifically interacts with Bcl-xL and blocks its binding to Bak, whereas Noxa binds strongly to Mcl-1 and blocks its inhibition of Bak. PQ increased the number of Bak-positive cells in the SNpc and provided the double-hit necessary to neutralize the anti-apoptotic effects of Bcl-xL and Mcl-1. We found that PQ induced higher BNip3 and Noxa expression in vitro and that knocking down these proteins, or Bak, made cells more resistant to PQ-induced apoptosis. Basal expression of Noxa was detected as immuno-positive neurons in the SNpc of untreated B6 mice, and the decrease of Noxa-positive cells in PQ-treated mice may simply reflect depletion of neurons in this nucleus. Our findings show that Bak is a necessary component of PQ-induced cell death in vitro and in vivo and provides a novel apoptotic stimulus that works through a Bak-specific mechanism, perhaps independently of Bax.
The involvement of BNip3 and Noxa raises the possibility that another transcription factor, hypoxia-induced factor-1␣, may also be an important upstream component because both genes are transcriptional targets for hypoxia-induced factor-1␣ (40,46,47). Because hypoxia-induced factor-1␣ expression increases under ischemic conditions in the brain and heart (48), it is possible that hypoxia or transient ischemia may further elevate BNip3 and Noxa and thereby enhance PQ-dependent neurotoxicity. Interestingly, head trauma often causes ischemic injuries, and head injury has been linked with an increased risk for PD (19,49).
In summary, our in vitro and in vivo findings indicate that PQ-induced apoptosis results from Bak-mediated MOMP, with cytochrome c release and caspase-3 activation. This report reduces the number of unknown steps between ROS/JNK and caspase-3 activation and demonstrates that these steps act on a relatively small proportion of Bcl-2 family members that regulate Bak. Findings in this report have potential implications for human neurodegenerative processes and may help identify mechanisms by which environmental toxicants and other risk factors contribute to PD pathogenesis. Noxa specifically binds to Mcl-1 and BNip3 binds to Bcl-xL (34). Thus, PQ induces the expression of two BH3-only members that block two major inhibitors of Bak, Mcl-1 and Bcl-xL. Binding of Noxa to Mcl-1 and BNip3 to Bcl-xL causes disinhibition of Bak, making it available to create MOMP. MOMP releases cytochrome c from the intermembrane space into the cytoplasm where it can interact with Apaf-1 and procaspase-9 to create an apoptosome. The fully active apoptosome processes and activates executioner caspase-3 triggering apoptosis.