Oxidative stress in the aging substantia nigra and the etiology of Parkinson's disease

Parkinson's disease (PD) is the most common neurodegenerative movement disorder. The global prevalence of PD is predicted to double by 2,040 (Dorsey & Bloem, 2018 ), making it the fastest growing neurodegenerative disorder ahead of Alzheimer's disease (Feigin et al., 2017 ). For perspective, if PD were transmissible, it would now be considered a global pandemic (Dorsey & Bloem, 2018 ). Movement dysfunction in PD results from the progressive death of dopaminergic neurons in the substantia nigra pars compacta (SNc), and accumulating evidence implicates oxidative stress as a key driver of the complex degenerating cascade underlying dopaminergic neurodegeneration in all forms of PD (Blesa, Trigo‐Damas, Quiroga‐Varela, & Jackson‐Lewis, 2015 ; Dias, Junn, & Mouradian, 2013 ). Oxidative stress arises from dysregulation of cellular redox activity, where production of reactive oxygen species (ROS; Figure 1 ) outweighs clearance by endogenous antioxidant enzymes and molecular chaperones. Oxidative stress in itself is therefore not pathological; rather, ROS accumulation following cellular redox imbalance mediates neuronal damage. Although ROS constitute important signaling molecules regulating physiological gene transcription and protein interactions (Schieber & Chandel, 2014 ), ROS accumulation can result in oxidative damage to lipids, proteins, DNA, and RNA depending on the subcellular location of ROS production, compromising neuronal function and structural integrity (Schieber & Chandel, 2014 ). Importantly, data collected from early‐stage PD patients demonstrate that elevated oxidative stress is a robust feature of initial disease stages, occurring prior to significant neuron loss (Ferrer, Martinez, Blanco, Dalfo, & Carmona, 2011 ). This implicates uncontrolled ROS generation as a potential causative factor in dopaminergic neuron death, rather than being a secondary response to progressive neurodegeneration. A better understanding of the complex role oxidative stress plays in the etiology of PD may therefore reveal new targets for therapeutic modification and preclinical diagnosis. Figure 1 Common reactive oxygen species, their production, and clearance. Incomplete reduction of molecular oxygen (O 2 ) produces superoxide radicals (O 2 − ), which may be converted to hydroxyl radicals ( ● OH) via Haber–Weiss chemistry, or to hydrogen peroxide (H 2 O 2 ) through the action of enzymes or molecules with superoxide dismutase (SOD) activity. Hydrogen peroxide is also a substrate for hydroxyl radical production via Fenton chemistry, catalyzed by labile ferrous iron. Hydrogen peroxide decomposition to water and oxygen is mediated by the enzymatic action of glutathione peroxidase (GPx) coupled to redox cycling of reduced (GSH) and oxidized (GSSG) glutathione, and also by catalase. Unpaired electrons are highlighted in red

Mitochondria are a primary intracellular source of ROS during healthy aging (Brand et al., 2004 ). Mitochondrial ATP production powers neural activity and maintains cellular homeostasis, which is achieved through oxidative phosphorylation in the mitochondrial electron transport chain (ETC; Figure 2 ). Premature electron transfer from complexes I and III of the ETC to O 2 occurs naturally in intact mammalian mitochondria (Drose & Brandt, 2008 ; Kussmaul & Hirst, 2006 ) and generates superoxide radicals (O 2 − ) as a physiological by‐product of energy production. These ROS can trigger the formation of hydroxyl radicals ( ● OH), which are thought to mediate primary neuronal oxidative damage both within and outside of mitochondria following their diffusion out of mitochondria (Weidinger & Kozlov, 2015 ). As a counterbalance, mitochondria contain two of the three eukaryotic superoxide dismutases (SOD), which detoxify O 2 − into less harmful hydrogen peroxide (H 2 O 2 ; Ruszkiewicz & Albrecht, 2015 ). Manganese superoxide dismutase (SOD2) is localized to the mitochondrial matrix and inner membrane (Karnati, Luers, Pfreimer, & Baumgart‐Vogt, 2013 ), whereas copper/zinc superoxide dismutase (SOD1) exists within the mitochondrial intermembrane space, cytosol, and many other cellular compartments (Kawamata & Manfredi, 2010 ). Mitochondrial H 2 O 2 produced by SOD1/2 is decomposed to innocuous O 2 and H 2 O via specific mitochondrial glutathione peroxidases (GPx1/4; Brigelius‐Flohe & Maiorino, 2013 ) and peroxiredoxins (PRx3/5; Ruszkiewicz & Albrecht, 2015 ; Figure 2 ). While H 2 O 2 likely represents less of a threat for oxidative damage compared with more redox‐active ROS, such as ● OH, the longer half‐life of H 2 O 2 and greater rate of diffusion from mitochondria into other cellular compartments allow it to act as an effective redox signaling molecule (Collins et al., 2012 ; Murphy, 2009 ). This involves reversible oxidative modification of proteins, especially thiol groups of cysteine residues (Eaton, 2006 ), which act as a redox switch by altering physiological protein functions, promoting alternative protein functions, or facilitating secondary interactions (D'Autreaux & Toledano, 2007 ; Eaton, 2006 ; Murphy, 2009 ). Effective regulation of mitochondrial H 2 O 2 by endogenous antioxidant pathways therefore constitutes an essential mechanism for maintaining physiological redox signaling and homeostasis. Figure 2 Reactive oxygen species are an inherent by‐product of oxidative phosphorylation in the mitochondrial ETC. (a) Electrons generated by the tricarboxylic acid cycle in the mitochondrial matrix are shuttled to ETC complexes I and II by NADPH and FADH 2 , respectively. They are then transferred to complex IV of the ETC with the help of inner mitochondrial membrane (IMM) electron shuttles (Q, coenzyme Q; C, cytochrome c) where they reduce molecular oxygen to water, a process which simultaneously drives ATP production by ATP synthase (ETC complex V). A small amount of premature electron leakage occurs naturally during oxidative phosphorylation, whereby electrons bound within ETC complexes I and III diffuse into both the mitochondrial matrix and intermembrane space (IMS). Here, they may cause incomplete reduction of molecular oxygen (O 2 ), generating superoxide radicals (O 2 − ) that may subsequently be converted to hydrogen peroxide (H 2 O 2 ) through the action of superoxide dismutase 1 or 2 (SOD1/2). Electron leakage from the electron transport chain is worsened during healthy aging, or by pathogenic factors such as genetic mutations ( SNCA , SOD1 ), environmental toxins (MPTP, rotenone), or misfolded proteins (α‐synuclein, SOD1). (b) Mitochondrial H 2 O 2 levels are regulated by glutathione peroxidase (GPx) and peroxiredoxin (PRx), coupled to the redox cycling of glutathione (GSH/GSSG) and thioredoxin (TRx SH /TRx SS ), respectively. While the oxidation component of each cycle is mediated by GPx and PRx, glutathione reductase (GR) and thioredoxin reductase (TRxR) drive NADPH‐dependent glutathione and thioredoxin reduction, respectively, to complete the redox loop 3.1 Energy production and pacemaker activity in the substantia nigra pars compacta The inherent production of ROS during ATP synthesis is greater within specific neuronal populations with higher energy demands, including dopaminergic neurons of the SNc. The large and unmyelinated axonal arbor of these neurons, whose size and complexity are orders of magnitude greater than other classes of dopamine neurons and other types of neurons in the brain (Pissadaki & Bolam, 2013 ), necessitates higher rates of ATP production to maintain resting membrane potential, propagate action potentials, and enable synaptic transmission. Unlike the majority of neurons throughout the brain, adult SNc dopamine neurons are also autonomously active. Regular action potentials (2–4 Hz) are generated in the absence of synaptic input (Gra

Significant amounts of ROS are generated within SNc dopaminergic neurons and surrounding glia by the oxidative metabolism of dopamine (Westlund, Denney, Rose, & Abell, 1988 ; Figure 3 ). Oxidative deamination of dopamine by monoamine oxidases generates H 2 O 2 as a by‐product, whereas enzymatic oxidation of dopamine's electron‐rich catechol moiety by cyclooxygenases, tyrosinase, and other enzymes produces O 2 − (Muñoz, Huenchuguala, Paris, & Segura‐Aguilar, 2012 ). Auto‐oxidation of dopamine may also occur via interactions with labile iron and other biometals (Hare & Double, 2016 ), generating ROS (H 2 O 2 , O 2 − , ● OH), pro‐oxidant dopamine‐o‐quinones (DAQ) and a raft of other neurotoxins (Figure 3 ), comprehensively reviewed elsewhere (Hare & Double, 2016 ; Sun, Pham, Hare, & Waite, 2018 ). Importantly, specific pathways within this complex network of iron–dopamine chemistry are seemingly favored by different chemical environments in vitro. Acidosis accelerates dopamine oxidation and promotes aminochrome formation, suggesting clinically measured brain tissue acidosis in PD may promote iron–dopamine redox chemistry and the accumulation of pro‐oxidant aminochrome in the SNc in this disorder (Sun et al., 2018 ). While these investigations have yet to be translated into complex cellular systems, they may have the potential to underlie the progressive and worsening nature of cell loss in PD. Figure 3 Dopamine metabolism and ROS production. Enzymatic decomposition of dopamine to homovanillic acid is mediated by monoamine oxidase (MAO), catechol‐o‐methyl transferase (COMT), and aldehyde dehydrogenase (ALDH). Conversely, dopamine may be oxidized to dopamine‐o‐quinone by tyrosinase (Tyr), cyclooxygenase (COX), or labile ferric iron (Fe 3+ ). Dopamine‐o‐quinones are reactive intermediates for the generation of more damaging compounds, including 6‐hydroxydopamine and R‐Salsolinol. Endogenous detoxification of dopamine‐o‐quinones involves cyclization to produce aminochrome, and subsequent oxidation and polymerization to generate neuromelanin. Hydrogen peroxide (H 2 O 2 ) produced by MAO, dopamine oxidation and dopamine‐o‐quinone oxidation, can participate in Fenton chemistry and react with labile ferrous iron (Fe 2+ ) to generate damaging hydroxyl radicals ( ● OH) Of the iron–dopamine metabolites, DAQs constitute particularly versatile intermediates in pathways producing harmful pro‐oxidant dopamine derivatives, aside from their own capacity to alkylate protein thiol and amine groups and promote protein oxidation in the presence of ROS (Meiser, Weindl, & Hiller, 2013 ). The tetrahydroisoquinoline salsolinol is one such DAQ derivative, which enhances oxidative stress and mitochondrial damage by inhibiting ETC function (Su et al., 2013 ). Salsolinol also disrupts clearance of dopamine by monoamine oxidases (Napolitano, Manini, & d'Ischia, 2011 ), shifting dopamine metabolism toward more damaging metabolic pathways which produce DAQs. Another derivative, 6‐hydroxydopamine (6‐OHDA), generates substantial amounts of O 2 − by inhibiting mitochondrial ETC complexes I and IV (Puspita et al., 2017 ). Given the enormous neurotoxic potential of DAQs, and the remarkably slow conversion of DAQs to neuromelanin (NM), it is likely endogenous mechanisms of DAQ detoxification exist. Potential roles for SOD1, glutathione transferase (by way of glutathione conjugation), and macrophage migration inhibitory factor have been suggested via evidence of direct interactions with DAQs (Emdadul Haque et al., 2003 ). 4.1 Iron accumulation Pro‐oxidant interactions between iron and dopamine are suggested to be enhanced in the aging SNc because of a preferential accumulation of labile iron in this brain region (Hare & Double, 2016 ). This is perhaps associated with age‐dependent ferritin dysfunction documented in Caenorhabditis elegans , whereby reactive ferrous iron (Fe 2+ ) is no longer efficiently oxidized to more chemically stable ferric iron (Fe 3+ ) for storage (James et al., 2015 ). Similar experiments have not been performed in human postmortem tissues, owing to difficulties in preserving iron redox state during tissue collection procedures, as well as our inability to accurately assay ferritin iron loading in vivo. Further, it will be important to determine the relative contributions of microglia and astrocytes to iron accumulation in the aging SNc, as these non‐neuronal cell types typically store approximately three times the quantity of iron compared with neurons without exhibiting signs of iron‐mediated toxicity (Bishop, Dang, Dringen, & Robinson, 2011 ). Iron accumulation in PD is significantly enhanced compared with healthy aging; iron levels are elevated twofold in the postmortem SNc compared with age‐matched controls (Dexter et al., 1989 ; Genoud et al., 2017 ). In vivo MRI imaging of SNc proton transverse relaxation states, known to be correlated with regional iron content, in early PD patients

Neuromelanin is a complex biopigment composed of 35% lipids (primarily dolichol; Fedorow, Pickford, et al., 2006 ), 15% covalently bound peptides, and numerous products of catecholamine metabolism, explaining its spatial distribution within select populations of catecholamine neurons within the brain (Double et al., 2008 ). Neuromelanin accumulates in the cytoplasm as large amorphous granules of an inconsistent size (Fedorow, Pickford, et al., 2006 ), contrast to the regular spherical macrostructure of peripheral melanins. Optical density and area measurements of unstained NM in the postmortem SNc identify three developmental phases, beginning with the initiation of pigmentation at approximately 3 years of age (Fedorow, Halliday, et al., 2006 ). Increases in pigment granule volume and pigment density within granules are subsequently observed until age 20; however, after this point, increases in pigment density within granules occur without substantial growth in pigment volume, suggesting regulation of NM production and turnover. The exact mechanism of NM production in the human SNc is unclear; however, NM is considered a complex biopolymer associated with dopamine autoxidation, rather than enzymatic catalysis. Tyrosinase catalyzes the rate‐limiting step of peripheral melanin synthesis; however, despite tyrosinase mRNA expression in the human SNc (Xu et al., 1997 ), no tyrosinase protein has yet been identified in this region (Tribl, Arzberger, Riederer, & Gerlach, 2007 ). Conversely, in vitro oxidation of dopamine produces dopamine melanin (DAM), which exhibits a moderate degree of chemical similarity to human NM (Double et al., 2000 ). Artificial synthesis of DAM in mouse SNc cell cultures is heavily reliant on nonvesicular dopamine, as well as ferric iron (Sulzer et al., 2000 ). This may be explained by the comparative susceptibility of nonvesicular dopamine to oxidation, which is catalyzed by ferric iron, producing DAQs and aminochrome. The identification of these particular products of dopamine oxidation in human NM suggests a degree of translation of these findings to human NM formation (Smythies, 1996 ; Figure 3 ). Despite similarities between NM and DAM, however, human NM possesses a substantially more complex structure and biochemical composition (Double et al., 2000 ), suggesting oxidized neurotransmitter is not the sole component of human NM. Further, not all cells that produce dopamine contain NM (Gaspar et al., 1983 ), suggesting that NM production may either be induced or inhibited, or that a mechanism of NM clearance exists. Regulation of NM production and/or degradation is consistent with the constant volume of NM within mature SNc dopamine neurons past the age of 20 (Fedorow, Halliday, et al., 2006 ), as uncontrolled autoxidation of dopamine over subsequent decades would be expected to result in a linear increase in NM volume. Functionally, NM is believed to bind, store, protect, and release free dopamine, regulate redox‐active iron to minimize pro‐oxidant Fenton chemistry, and sequester a range of potentially toxic metal cations (zinc, copper, manganese, chromium, cobalt, mercury, lead, and cadmium) and chemicals (derivatives of paraquat, salsolinol, MPTP; Haining & Achat‐Mendes, 2017 ). Neuromelanin therefore bears closer resemblance to a protective cellular scaffold, rather than merely an aggregate of metabolic products, which is capable of sequestering toxic chemicals away from cellular compartments where they can participate in damaging biochemical reactions. In this way, NM may play a key role in redox homeostasis in the healthy SNc, especially with regard to the mitigation of iron–dopamine redox chemistry. Indeed, human‐derived NM prevents iron‐mediated ROS generation and antioxidant depletion in vitro (Zecca, Casella, et al., 2008 ). 5.1 Neuromelanin in PD: a loss or gain of function? Contrast to its protective role in the healthy SNc, alterations to NM density and composition are thought to exacerbate ROS generation, iron accumulation, and α‐synuclein aggregation in the PD SNc (Faucheux et al., 2003 ; Halliday et al., 2005 ). An early increase in NM density within pigment granules is reported within morphologically normal SNc dopamine neurons, which is associated with increased NM oxidation and iron loading (Faucheux et al., 2003 ). Both of these factors promote the concentration of α‐synuclein to the lipid component of NM at the expense of cholesterol (Halliday et al., 2005 ), and iron loading is also shown to potentiate peroxidation of human‐derived NM in vitro (Zecca, Casella, et al., 2008 ). The accumulation of α‐synuclein pathology on NM is associated with a significant reduction in NM density within SNc dopamine neurons (Faucheux et al., 2003 ; Halliday et al., 2005 ), suggesting early α‐synuclein redistribution to NM promotes its decomposition, which is likely to impair the neuroprotective function of NM in PD. Indeed, reductions in NM density are associated with elev

It is clear that ROS accumulation is a primary feature of numerous damaging molecular pathways present during early‐stage PD, prior to initiation of neuron death. Excessive ROS accumulation can promote, or directly trigger, numerous cell death pathways including apoptosis (intrinsic and extrinsic), cytoplasmic cell death (parthanatos and necroptosis), and autophagic cell death (Morris, Walker, Berk, Maes, & Puri, 2018 ; Redza‐Dutordoir & Averill‐Bates, 2016 ), although technological limitations prevent us from determining which are activated in the SNc of PD patients in vivo. Further, the relatively slow rate of SNc dopamine neuron loss and rapid clearance of dead dopamine neurons by phagocytosis makes postmortem detection of cell death markers difficult and often unreliable. TUNEL staining, which labels DNA fragmentation as an indication of apoptosis, generates variable results in the PD SNc ranging from no TUNEL‐positive neurons to a high percentage of TUNEL‐positive neurons in PD and age‐matched controls (Levy, Malagelada, & Greene, 2009 ). More indirect approaches to identify and quantify the expression of cell death pathways in postmortem PD tissues have evaluated the expression of pathway constituents, including Bcl‐2 family proteins, caspases, Fas, and p55 (Jellinger, 2000 ; Mogi et al., 2000 ), with similar variable results and reproducibility. Compounding these issues, evidence of other cell death signaling pathways in PD patient tissues, such as markers of autophagic cell death, is difficult to interpret as they can signal processes responsible for either suppressing or promoting cell death (Levy et al., 2009 ). For these reasons, most research into which cell death pathways are activated in the PD SNc, and what their primary triggers are, has utilized dopaminergic cell cultures and animal models of PD. In these models, neuronal cell death associated with mutations in PD‐linked genes ( SNCA, LRRK2 , DJ‐1, PARK2 , PINK‐1) involves varying degrees of activation of intrinsic (Iaccarino et al., 2007 ; Yamada, Iwatsubo, Mizuno, & Mochizuki, 2004 ) and extrinsic (Ho, Rideout, Ribe, Troy, & Dauer, 2009 ) apoptosis, autophagic cell death (Venderova & Park, 2012 ), and parthanatos (Kam et al., 2018 ). Similarly, intrinsic (Clayton, Clark, & Sharpe, 2005 ; Perier et al., 2005 ) and extrinsic (Hayley et al., 2004 ) apoptosis, necroptosis (Callizot, Combes, Henriques, & Poindron, 2019 ), and parthanatos (David, Andrabi, Dawson, & Dawson, 2009 ) are implicated in dopamine neuron death resulting from the administration of PD‐mimetic toxins MPTP, rotenone, and 6‐OHDA, albeit to differing extents. Discerning the role of ROS in activating and/or accelerating these pathways adds another layer of complexity, requiring a clear relationship between a ROS levels and pathway activation to be established. Given the relatively new discovery of pathways such as parthanatos, or the recent evolution of pathways such as necroptosis from the coalescence of existing cell death mechanisms, there has been little investigation of a specific role for ROS in dopamine neuron death mediated by these pathways. Increased ROS production accompanied neuron death in MPTP‐, rotenone‐, and 6‐OHDA‐treated primary cultures of rat mesencephalic dopamine neurons, where there was also evidence of significant parthanatos (MPTP, rotenone, 6‐OHDA) and necroptosis (MPTP, rotenone; Callizot et al., 2019 ). While ROS accumulation is indeed known to trigger both of these pathways (Morris et al., 2018 ), it is unknown whether antioxidant‐mediated attenuation of ROS levels reduces the activation of these pathways following toxin administration; the final step required to confirm causality. It would also be a valuable extension to see such an experiment repeated in a more complex mammalian model system, where the influences of other non‐neuronal cell types implicated in the toxicity of PD‐linked agents would be present. Contrast to parthanatos and necroptosis, a strong body of data implicates ROS accumulation in the priming, initiation and acceleration of apoptosis in SNc dopamine neurons in cell and animal models of PD. An abundance of evidence implicates mitochondrially driven (intrinsic) apoptosis, in particular, to ROS‐driven cell death in PD (Venderova & Park, 2012 ), likely linked to well‐documented mitochondrial dysfunction reported in degenerating brain regions in PD. 7.1 Mitochondria‐dependent (intrinsic) apoptosis The intrinsic (mitochondrial) cell death pathway is activated by cellular stressors, including excessive ROS accumulation, which trigger conformational change and mitochondrial localization of pro‐apoptotic Bcl‐2 family proteins (Bax, Bak, Bad, Bim). These proteins oligomerize in the outer mitochondrial membrane and form pores that alter mitochondrial membrane permeability, facilitating the release of pro‐apoptotic proteins from the mitochondrial intermembrane space, including cytochrome c, second

Despite a clear contribution of ROS accumulation to the initiation and/or acceleration of cell death in PD, numerous large and well‐conducted clinical trials targeting oxidative stress in this disorder have shown little improvement in clinical outcome for patients (Table 1 ). Approximately eleven double‐blind, placebo‐controlled randomized clinical trials of antioxidants have been completed in PD patients who were within 5 years of diagnosis or less, targeting multiple biochemical pathways for durations ranging between 1 month and 8 years. No evidence of clinical benefits, as measured by improvements in unified Parkinson's disease rating scale (UPDRS) scores, were observed despite these wide and varied approaches. However, rather than immediately adopt the pessimistic perspective that therapies targeting oxidative stress in PD are inherently flawed in concept, we should first consider some of the reasons this may be so (Murphy, 2014 ). Table 1 Summary of antioxidant clinical trials for Parkinson's disease Authors Study design Participants Intervention/duration Redox modulation Outcome measures Status/results The Parkinson Study Group and DATATOP study ( 1993 ) DPRCT 800 early PD patients, within 5 years of diagnosis 2‐year intervention Randomization Deprenyl (10 mg/day) α‐tocopherol (2,000 IU/day) Deprenyl/α‐tocopherol Placebo Deprenyl—reduced toxin‐induced ● OH − formation (Wu, Chiueh, Pert, & Murphy, 1993 ), and iron‐induced oxidative stress (Budni et al., 2007 ). α‐Tocopherol—lipophilic ROS scavenger, reduced lipid peroxidation (Niki, 2015 ) The onset of disability prompting levodopa administration Complete—no evidence of clinical benefit Shults et al. ( 2002 ) DPRCT 80 early PD patients who did not require treatment for their disability 16‐month intervention Randomization Coenzyme Q (300 mg/day) Coenzyme Q (600 mg/day) Coenzyme Q (1,200 mg/day) Placebo Coenzyme Q—lipophilic ROS scavenger, reduces lipid peroxidation (Bentinger, Brismar, & Dallner, 2007 ) Safety and tolerability, UPDRS score Complete –safe and tolerable up to 1,200 mg/day, appeared to slow progression of PD motor impairment The Parkinson Study Group DPRCT 600 early PD patients, within 5 years of diagnosis 16‐month intervention Randomization Coenzyme Q (1,200 mg/day) Coenzyme Q (2,400 mg/day) Placebo Each group combined with α‐tocopherol (1,200 IU/day) Coenzyme Q and α‐Tocopherol as above UPDRS score Complete—no evidence of clinical benefit Snow et al. ( 2010 ), Protect Study Group DPRCT 128 early PD patients who do not require treatment for their disability 12‐month intervention Randomization MitoQ (40 mg/day) MitoQ (80 mg/day) Placebo MitoQ—a coenzyme Q mimetic, mitochondria‐targeted ROS scavenging (Tauskela, 2007 ) UPDRS score Complete—no evidence of clinical benefit NET‐PD investigators DPRCT 1,741 early PD patients, within 5 years of diagnosis 5‐ to 8‐year intervention Randomization Creatine (10g/day) Placebo Creatine—induction of antioxidant enzymes (TRx, PRx), effective scavenger of ● OH − and O 2 − , protects against mitochondrial DNA and RNA (Sestili et al., 2011 ) Modified Rankin Scale, Symbol Digit Modalities Test, PDQ−39 Summary Index, Schwab and England Activities of Daily Living scale, and ambulatory capacity Terminated early for futility—no evidence of clinical benefit Mischley, Lau, Shankland, Wilbur, and Padowski ( 2017 ) DPRCT 45 early PD patients (H&Y stage 1–3) 3‐month intervention Randomization GSH (300 mg IN/day) GSH (600 mg IN/day) Placebo GSH—low molecular weight antioxidant; ROS detoxification, redox signaling molecule, substrate for antioxidant enzyme pathways (Forman, Zhang, & Rinna, 2009 ) UPDRS Completed—no evidence of clinical benefit Hauser, Lyons, McClain, Carter, and Perlmutter ( 2009 ) DPRCT 21 PD patients (H&Y stage 2–5), nonresponsive to L‐DOPA 4‐week intervention Randomization GSH (1,400 mg IV, 3×/week) Placebo GSH as above Safety and tolerability, UPDRS Completed—safe and tolerable up to 4,200 mg/week, no preliminary evidence of clinical benefit NINDS Exploratory Trials in Parkinson Disease DPRCT 210 early PD patients, within 5 year of diagnosis (H&Y stage 1–2) 44‐week intervention Randomization Pioglitazone (15 mg/day) Pioglitazone (45 mg/day) Placebo Pioglitazone—induction of peroxisomal and cytosol antioxidant proteins (SOD1, catalase, GPx1; Filograna et al., 2016 ) UPDRS Completed—no evidence of clinical benefit Weill Medical College of Cornell University ( NCT01470027 ) Parallel assignment DPRCT 30 PD patients, less than 15 years postdiagnosis 4‐week intervention Randomization N‐acetyl‐cysteine (1,800 mg/day) N‐acetyl‐cysteine (3,600 mg/day) Placebo N ‐acetyl‐cysteine—ROS scavenger and major substrate for GSH/GPx antioxidant pathway (Firuzi et al., 2011 ) UPDRS, Cerebral GSH levels (measured by Proton Magnetic Resonance Spectroscopy) Completed—results yet to be published NINDs, Michael J. Fox Foundation, The Parkinson Study Group ( NCT02168842 ) Parallel assignment DPRCT 33