NADPH oxidase mediates oxidative stress in the
model of Parkinson’s disease
Du-Chu Wu*, Peter Teismann*, Kim Tieu*, Miquel Vila*, Vernice Jackson-Lewis*, Harry Ischiropoulos†,
and Serge Przedborski*‡§¶
Departments of *Neurology and‡Pathology and§Center for Neurobiology and Behavior, Columbia University, New York, NY 10032; and†Stokes Research
Institute, Department of Pediatrics, Children’s Hospital of Philadelphia, and Department of Biochemistry and Biophysics, University of Pennsylvania
School of Medicine, Philadelphia, PA 19104
Edited by Tomas Ho ¨kfelt, Karolinska Institute, Stockholm, Sweden, and approved March 13, 2003 (received for review November 27, 2002)
Parkinson’s disease (PD) is a neurodegenerative disorder of uncer-
tain pathogenesis characterized by a loss of substantia nigra pars
compacta (SNpc) dopaminergic (DA) neurons, and can be modeled
by the neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine
(MPTP). Both inflammatory processes and oxidative stress may
contribute to MPTP- and PD-related neurodegeneration. However,
whether inflammation may cause oxidative damage in MPTP and
PD is unknown. Here we show that NADPH-oxidase, the main
reactive oxygen species (ROS)-producing enzyme during inflam-
changes coincide with the local production of ROS, microglial
activation, and DA neuronal loss seen after MPTP injections.
Mutant mice defective in NADPH-oxidase exhibit less SNpc DA
neuronal loss and protein oxidation than their WT littermates after
MPTP injections. We show that extracellular ROS are a main
determinant in inflammation-mediated DA neurotoxicity in the
MPTP model of PD. This study supports a critical role for NADPH-
oxidase in the pathogenesis of PD and suggests that targeting this
enzyme or enhancing extracellular antioxidants may provide novel
therapies for PD.
which include tremor, muscle stiffness, paucity of voluntary
movements, and postural instability (1). Its main neuropatho-
logical feature is the loss of the nigrostriatal dopamine (DA)-
containing neurons, whose cell bodies are in the substantia nigra
pars compacta (SNpc) and nerve terminals are in the striatum
defects, PD is a sporadic condition of unknown pathogenesis (1).
Epidemiological studies suggest that inflammation increases
the risk of developing PD (3). Consistent with this view, exper-
imental models of PD show that inflammatory factors may
trigger or modulate SNpc DA neuronal death (4–6). Among
inflammatory mediators capable of promoting neurodegenera-
may deserve particular attention, because oxidative stress is a
leading pathogenic hypothesis of PD (7).
A significant source of ROS during inflammation is NADPH-
oxidase, which is a multimeric enzyme composed of gp91phox,
p22phox, p47phox, p67phox, and p40phoxsubunits (8). In resting
microglia, NADPH-oxidase is inactive because p47phox, p67phox,
and p40phox, which are present in the cytosol as a complex, are
separated from gp91phoxand p22phox, which are transmembrane
proteins. Upon microglial activation, p47phoxbecomes phosphor-
ylated and the entire cytosolic complex translocates to the
membrane, where it assembles with gp91phoxand p22phox, thus
forming a NADPH-oxidase entity now capable of reducing
production of other secondary reactive oxidants (8).
Although NADPH-oxidase is critical to the killing of invading
microorganisms in infections through its abundant and sustained
arkinson’s disease (PD) is a common neurodegenerative
disease characterized by disabling motor abnormalities,
production of O2?(8), its role in noninfectious chronic neuro-
degenerative processes, such as PD, is not known. In the present
study, we show not only that the NADPH-oxidase main subunit
gp91phoxis up-regulated in the SNpc of PD and 1-methyl-4-
phenyl-1,2,3,6-tetrahydropyridine (MPTP) mice, but also that
mitigating inflammation-mediated oxidative attack on SNpc
neurons. These findings indicate that NADPH-oxidase-induced
oxidative stress is instrumental in SNpc DA neurodegeneration
caused by MPTP, and suggest that NADPH-oxidase is a valuable
therapeutic target for the development of neuroprotective strat-
egies for PD.
Materials and Methods
Animals and Treatment. Eight-week-old male C57BL?6 mice
(Charles River Breeding Laboratories), gp91phox-deficient mice
(B6.129S6-Cybbtm1din, The Jackson Laboratory), and their WT
littermates were used. Mice received four i.p. injections of
MPTP-HCl (16 mg?kg free base; Sigma) dissolved in saline at
2-h intervals, and were killed 0–14 days after the last injection.
Control mice received saline only. MPTP handling and safety
measures were in accordance with our published guidelines (9).
Minocycline (2 ? 45 mg?kg per day; Sigma) was given to MPTP
mice as described (5). Bovine erythrocyte superoxide dismutase
1 (SOD1; 20 units?h; Sigma) was infused into the left striatum
with an osmotic minipump (Alzet, Palo Alto, CA) starting 1 day
before and stopping 6 days after the MPTP injections. This
protocol was in accordance with the National Institutes of
Institutional Animal Care and Use Committee of Columbia
University (New York). Striatal 1-methyl-4-phenylpyridinium
levels were determined by HPLC as described (5).
RNA Extraction and RT-PCR. Total RNA was extracted as described
(5). The primer mouse sequences were as follows: gp91phox,
5?-CAGGAGTTCCAAGATGCCTG-3? (forward) and 5?-
GATTGGCCTGAGATTCATCC-3? (reverse); p67phox, 5?-
CAGCCAGCTTCGGAACATG-3? (forward) and 5?-GACAG-
TACCAGGATTACATC-3? (reverse); macrophage antigen
complex 1 (Mac-1), 5?-TTCTCATGGTCACCTCCTGC-3? (for-
ward) and 5?-GGTCTGACCATCTGAACCTG-3? (reverse);
GAPDH, 5?-GTTTCTTACTCCTTGGAGGCCAT-3? (for-
ward) and 5?-TGATGACATCAAGAAGTGGTGAA-3? (re-
verse). PCR was carried out for 29 cycles for gp91phox, 27 cycles
for p67phoxand Mac-1, and 20 cycles for GAPDH. Each cycle
This paper was submitted directly (Track II) to the PNAS office.
Abbreviations: Mac-1, macrophage antigen complex 1; MPTP, 1-methyl-4-phenyl-1,2,3,6-
nigra pars compacta; TH, tyrosine hydroxylase; DA, dopamine/dopaminergic; SOD1, super-
oxide dismutase 1.
¶To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
May 13, 2003 ?
vol. 100 ?
no. 10 ?
consisted of 30 s of denaturation at 94°C, 40 s of annealing at
55°C, and 60 s of extension at 72°C, followed by a final 7-min
extension at 72°C. Products were quantified by a phosphor
and sense RNA probes were prepared by in vitro transcription
from a mouse gp91phoxcDNA fragment (nucleotides 1020–1493;
GenBank accession no. U43384) by using SP6?T7 RNA poly-
merase in the presence of digoxigenin-linked UTP (Roche
Molecular Biochemicals) according to the supplier’s instruc-
tions. Frozen midbrain sections (14 ?m thick) were incubated
with the antisense or sense (for control) digoxigenin-labeled
probes. Hybridization signal was detected by 5-bromo-4-chloro-
indolyl-phosphatase and nitroblue tetrazolium.
Immunohistochemistry and Quantitative Morphology. Mouse brains
were fixed and processed for immunostaining as described (5).
Primary Abs were as follows: for mouse sections, monoclonal
anti-mouse gp91phox(1:1,000; Transduction Laboratories, Lex-
ington, KY), rat anti-MAC-1 (1:200; Serotec), monoclonal anti-
tyrosine hydroxylase (TH; 1:1,000; Chemicon), and polyclonal
anti-TH (1:1,000; Calbiochem, San Diego); for human sections,
monoclonal anti-human gp91phox(gift from Genentech) and
monoclonal anti-human CD68 (DAKO). Immunolabeling was
visualized by using 3,3?-diaminobenzidine (brown), VECTOR
SG (blue?gray), 3-amino-9-ethylcarbazole (red), or fluorescein
and Texas red (all from Vector Laboratories).
Total numbers of TH-positive SNpc neurons were counted by
stereology by using the optical fractionator method described
previously (6). Striatal density of TH immunoreactivity was
determined as described (5).
Western Blots. Total, cytosolic, and plasma membrane proteins
were prepared as described (5). Primary Abs were as follows: for
mouse proteins, monoclonal anti-mouse p67phox(1:1,000; Trans-
duction Laboratories), polyclonal anti-gp91phox(1:5,000, gift
from M. C. Dinauer, Indiana University, Indianapolis), and
polyclonal anti-calnexin (1:5,000; Stressgen Biotechnologies,
Victoria, Canada); for human proteins, monoclonal anti-human
gp91phox(1:500, Genentech). A monoclonal anti-?-actin
(1:5,000; Sigma) was used for both mouse and human pro-
teins. Bound primary Ab was detected by using a horseradish
peroxidase-conjugated secondary Ab against IgG and a chemi-
luminescent substrate (SuperSignal Ultra, Pierce). Films were
quantified by using the NIH IMAGE analysis system.
In Situ Visualization of O2?and O2?-Derived Oxidant Production. In
situ visualization of O2?and O2?-derived oxidant production
was assessed by hydroethidine histochemistry (10). At selected
time points after MPTP administration, mice were injected i.p.
with 200 ?l of PBS containing 1 ?g??l hydroethidine (Molecular
Probes) and 1% DMSO. Brains were harvested 15 min later and
frozen on dry ice. Midbrain sections (14 ?m thick) mounted onto
gelatin-coated glass slides were examined for hydroethidine
oxidation product, ethidium accumulation, by fluorescence mi-
sections were used for Mac-1 immunohistochemistry.
Protein carbonyls were detected after derivatization of brain
homogenates with 2,4-dinitrophenylhydrazine by using a modi-
fication of the method described by Levine et al. (11). The
concentration of protein carbonyls was calculated from the
difference in absorbance at 360 nm between the underivatized
and 2,4-dinitrophenylhydrazine-derivatized samples normalized
to the protein concentration. The extinction coefficient of 21
mM?1?cm?1was applied to calculate the concentration of pro-
Human Samples. Age at death and interval from death to tissue
processing (mean ? SEM) were as follows: for the control group
(n ? 3), 72.2 ? 8.8 y and 13.0 ? 3.5 h, respectively; for the PD
group (n ? 6), 77.2 ? 2.3 y and 10.1 ? 2.4 h, respectively. For
the PD patients, the mean duration of disease was 16.8 ? 2.3 y.
Statistical Analysis. Values represent means ? SEM. Differences
among means were analyzed by using one- or two-way ANOVA
with time, treatment, or genotype as the independent factors.
When ANOVA showed significant differences, pairwise com-
parisons between means were analyzed by Newman–Keuls post
hoc testing. The null hypothesis was consistently rejected at the
NADPH-Oxidase Is Induced in Mouse Ventral Midbrain During MPTP
Neurotoxicity. To define the temporal relationship between
NADPH-oxidase expression and MPTP neurotoxicity, contents
of ventral midbrain (brain region containing SNpc) membrane-
bound subunit gp91phoxand cytosolic subunit p67phoxmRNA
were assessed throughout the time course of MPTP-induced
SNpc DA neurodegeneration (12). In saline-injected mice, ven-
tral midbrain gp91phox, p67phox, and Mac-1 (microglial marker)
mRNAs were low (Fig. 1 A–D). In contrast, in MPTP-injected
mice, ventral midbrain gp91phox, p67phox, and Mac-1 mRNAs
mRNA levels in saline-injected (S) and MPTP-injected mice from 0 to 14 days
after injections. SNpc gp91phoxmRNA labeling is negligible in saline-injected
mice (E), whereas it is copious in MPTP-injected mice at 2 days (F).*, P ? 0.05;
**, P ? 0.001, more than saline-treated mice (n ? 4–6 per time point). (Scale,
(A–D) RT-PCR shows ventral midbrain gp91phox, p67phox, and Mac-1
www.pnas.org?cgi?doi?10.1073?pnas.0937239100 Wu et al.
increased in a time-dependent manner after MPTP injections
(Fig. 1 A–D).
gp91phox, which is one of the main functional subunits of
NADPH-oxidase, revealed no specific labeling in ventral mid-
brain (Fig. 1E), whereas in MPTP-injected mice there was
conspicuous specific labeling over the SNpc at 2 days after
MPTP injections (Fig. 1F). Thus, these results indicate that
NADPH-oxidase is induced after MPTP administration specif-
ically in the area where the demise of DA neurons arises in this
model of PD.
NADPH-Oxidase Is Expressed in Activated Microglia After MPTP Injec-
tion. Consistent with the mRNA data, ventral midbrain gp91phox
protein contents rose in a time-dependent manner after MPTP
injections (Fig. 2 A and B). In cell cultures, NADPH-oxidase has
been identified in different cell types, including neurons (13). In
saline-injected mice, mild gp91phoximmunoreactivity was seen
throughout the substantia nigra (Fig. 2C) without greater
gp91phoximmunolabeling in the SNpc, which hosts the TH-
positive neurons (Fig. 2 E and F). Immunoreactivity of gp91phox
was in small cells with thin ramifications (Fig. 2D) reminiscent
of resting microglia (Fig. 2F). In MPTP-injected mice, robust
2G) in larger cells with thick, shorter ramifications (Fig. 2H)
reminiscent of activated microglia (Fig. 2J). Similar immuno-
histochemical gp91phoxalterations were seen in the striatum,
which contains the nerve terminals of the projecting SNpc DA
neurons, between the saline- and MPTP-injected mice (data not
shown). By confocal microscopy, gp91phoximmunoreactivity
appeared to colocalize with Mac-1 (Fig. 2 K–M). Conversely,
gp91phoximmunoreactivity did not colocalize either with the
astrocytic marker glial fibrillary acidic protein (Fig. 2 N–P) or
with TH (data not shown). Thus, these results demonstrate that,
after MPTP administration, SNpc microglia become activated at
the site of NADPH-oxidase induction.
Expression of gp91phoxIs Increased in PD Midbrain. Consistent with
the finding in the MPTP mice, postmortem SNpc samples from
sporadic PD patients had higher gp91phoxprotein contents than
controls (Fig. 3 A and B). In these autopsy specimens, cellular
gp91phoximmunoreactivity was barely identified in controls (Fig.
3C) but was strong in PD midbrain sections, where it was
identified in microglial cells (Fig. 3D). The similarity of the
In saline-injected mice, gp91phoximmunoreactivity (C and D, brown) is mild and localized in resting microglia, which are not abundant in the SNpc, as shown (E
cells are seen in the SNpc (G and H). These cells resemble activated microglial cells (H vs. J, arrow). At this point there are many fewer TH-positive neurons (I and
J, arrowhead). Confocal microscopy shows that all gp91phox-positive cells are Mac-1-positive, thus confirming their microglial origin (K and L). Conversely, no
gp91phox-positive cells are glial fibrillary acidic protein-positive cells, thus excluding their astrocyctic origin (N–P).*, P ? 0.05, more than saline-treated mice (n ?
6 per time point). [Scale bar, 2.5 mm (C, E, G, and I); 0.25 mm (D, F, H, and J); and 0.2 mm (K–P).]
Western blot shows the time-dependent induction of gp91phoxin mouse ventral midbrain after MPTP injections. ?, Mouse macrophage lysate; s, saline.
Wu et al.
May 13, 2003 ?
vol. 100 ?
no. 10 ?
gp91phoxalterations between the MPTP mice and the PD
postmortem specimens validates the use of the MPTP experi-
mental model to study the role of NADPH-oxidase in the PD
The Lack of gp91phoxAbates MPTP-Associated ROS Production. In
saline-injected mice, ventral midbrain O2?and O2?-derived
oxidant production, evidenced by ethidium fluorescence, was
minimal (Fig. 4A). In contrast, in MPTP-treated mice, ventral
midbrain production of O2?or O2?-derived oxidants shown by
ethidium fluorescence was increased by 12 h (data not shown),
was maximal by 2 days (Fig. 4C), and remained elevated at 7 days
after MPTP (data not shown). SNpc ethidium fluorescence
coincided with the location and the time course of microglial
activation seen after MPTP administration (Fig. 4 C and D).
In mutant mice lacking the gp91phoxsubunit, no translocation
of the cytosolic p67phoxsubunit to the plasma membrane was
seen after MPTP injections (Fig. 4 I and J), which is mandatory
for NADPH-oxidase to become catalytically competent (8).
Unlike WT littermates (Fig. 4 C and D), mutant mice with
defective NADPH-oxidase failed to show any increase in SNpc
ethidium fluorescence (Fig. 4E), despite normal microglial
activation after MPTP administration (Fig. 4F). WT mice
treated with minocycline (i.e., antibiotic that blocks microglial
activation) showed no increase in SNpc ethidium fluorescence
(Fig. 4G) and no microglial activation after MPTP administra-
tion (Fig. 4E). Thus, these findings demonstrate that during the
MPTP neurotoxic process there is an increased production of
ROS in the SNpc that originates from activated microglial cells
and is mediated by NADPH-oxidase.
NADPH-Oxidase Defect Protects Against MPTP Neurodegeneration.In
the ventral midbrain of saline-injected mice, the stereological
counts of SNpc TH-positive neurons did not differ between
gp91phox-deficient mice and their WT littermates (Fig. 5A). In
MPTP-injected mice, the numbers of SNpc TH-positive neurons
midbrain gp91phoxprotein content in two PD and two controls. (B) Bar graph
control ventral midbrain samples. (C and D) Representative gp91phoximmu-
nostaining that shows positive cells in PD samples (arrowhead, gray-blue,
cytosol labeling), but not with neuromelanin (brown pigment).*, P ? 0.05,
higher than controls. (Scale bar, 0.5 mm.)
(A) Representative Western blots illustrating the increase in ventral
minimal in the saline-treated mice. By 2 days after MPTP injections, SNpc
ethidium fluorescence is increased in WT mice (C) and is absent in gp91phox-
deficient mice (E) and minocycline-treated WT mice (G). Microglial activa-
tion is prevented by minocycline (H) but is normal in gp91phox-deficient
mice (E). MPTP stimulates NADPH-oxidase activation, as evidenced by
p67phoxtranslocation from the cytosol to the plasma membrane in WT mice
calnexin is used to normalize the data. Data are means ? SEM for four to
MPTP-injected WT mice, but not different from both saline-injected
Ethidium fluorescence (A) and Mac-1 immunostaining (B) are
www.pnas.org?cgi?doi?10.1073?pnas.0937239100Wu et al.
were reduced in the two groups of animals (Fig. 5A). However,
the loss was smaller in gp91phox-deficient mice compared with
their WT counterparts (Fig. 5A). In the striatum of saline-
injected mice, the density of TH-positive nerve fibers was similar
between gp91phox-deficient mice and their WT littermates (Fig.
5B). Like for the number of SNpc TH-positive neurons, in
MPTP-injected mice the density of striatal TH-positive nerve
fibers was less reduced in the gp91phox-deficient mice than in
their WT counterparts (Fig. 5A). Because MPTP neurotoxic
potency on the nigrostriatal pathway correlates linearly with
1-methyl-4-phenylpyridinium levels in the striatum (14), the
content of this active metabolite of MPTP between the two
genotypes was evaluated. There were no differences in striatal
levels of 1-methyl-4-phenylpyridinium between the gp91phox-
deficient mice (17.8 ? 1.4 ?g?g striatum; n ? 5) and WT
littermates (17.7 ? 2.0 ?g?g striatum; n ? 5; P ? 0.05). These
results show that NADPH-oxidase participates in the MPTP
neurotoxic process affecting DA cell bodies in the SNpc and
nerve fibers in the striatum by a mechanism unrelated to an
alteration in MPTP toxicokinetics.
NADPH-Oxidase Damages Ventral Midbrain Proteins. To assess the
extent of NADPH-oxidase-related oxidative damage, protein
carbonyl levels were determined in ventral midbrain of gp91phox-
deficient and WT mice after saline or MPTP administration. In
saline-injected mice, the levels of ventral midbrain protein
carbonyls were similar between the two groups of animals (Fig.
6A). In MPTP-injected WT mice, levels of ventral midbrain
protein carbonyls were increased (Fig. 6A), but in gp91phox-
deficient mice they were not different from controls (Fig. 6A).
MPTP-Induced Neurotoxicity Is Attenuated by Scavenging Extracellu-
lar Superoxide. To test the noxious role of extracellular ROS, the
membrane-impermeant enzyme SOD1 was infused into the left
striatum. In the MPTP-injected mice, there was a protection of
striatal TH-positive fibers on the infused side compared with the
noninfused side (Fig. 6B). There was also a preservation of SNpc
TH-positive cell bodies ipsilateral to the infused side compared
with the contralateral noninfused side (Fig. 6C). These findings
demonstrate the importance of the oxidative stress that ema-
nates from the extracellular space on the demise of neighboring
This study shows that the microglial activation in MPTP and PD
SNpc specimens is associated with an induction of NADPH-
oxidase. This up-regulation correlates topographically and tem-
porally with the DA neurodegenerative changes seen in MPTP
mouse and human PD brains. It also parallels the production of
ROS seen in the SNpc by 2 days after MPTP injections. The use
of minocycline and mutant mice deficient in gp91phoxdemon-
strates collectively that ROS production originates from acti-
vated microglia and, within these cells, from NADPH-oxidase.
In the MPTP model, ROS can emanate from both cytosol and
mitochondria of DA neurons (15–18). Rise of markers reflecting
oxidative damage in the nigrostriatal DA pathway culminates
during the first 24 h after MPTP injections (19, 20). In contrast,
SNpc NADPH-oxidase-mediated ROS attack becomes signifi-
optical density of striatal DA fibers (B) are higher in gp91phox-deficient mice
4–8 samples per group).*, P ? 0.05, less than saline-injected mice; #, P ? 0.05,
higher than MPTP-injected WT mice.
Stereological counts of TH-positive neurons in the SNpc (A) and
attenuates the striatal (B) and the SNpc lesion on the infused side, but not on
0.05, higher than controls; #, P ? 0.05, less than MPTP-injected WT mice, but
not different from the two saline-injected groups.
(A) Ventral midbrain carbonyl content, used as a marker of protein
Wu et al.
May 13, 2003 ?
vol. 100 ?
no. 10 ?