NOS2A and the Modulating Effect of

Cigarette Smoking in Parkinson’s Disease

Dana B. Hancock, BS,1 Eden R. Martin, PhD,1,2 Kenichiro Fujiwara, BS,1 Mark A. Stacy, MD,1,2
Burton L. Scott, PhD, MD,1,2 Jeffrey M. Stajich, PA-C,1 Rita Jewett, RN,1 Yi-Ju Li, PhD,1,2
Michael A. Hauser, PhD,1,2 Jeffery M. Vance, PhD, MD,1,2 and William K. Scott, PhD1,2

Objective: Inducible nitric oxide synthase, a protein product of NOS2A, generates nitric oxide as a defense mechanism, but
excessive levels threaten cellular survival. NOS2A is a candidate gene for Parkinson’s disease (PD) that potentially interacts with
cigarette smoking. We examined NOS2A for association with PD risk and age at onset (AAO) and for interaction with smoking.
Methods: We genotyped 13 NOS2A single nucleotide polymorphisms (SNPs) in 466 singleton families and in a validation set
of 286 multiplex families. We tested allelic and haplotypic association using the association in the presence of linkage test,
genotypic associations using the genotype pedigree disequilibrium test, AAO effects using the quantitative transmission disequi-
librium test, and interactions using generalized estimating equations.
Results: Among the pooled earliest onset families, rs2255929 and rs1060826 generated significant allelic (p ⫽ 0.000059 and
0.0062, respectively) and genotypic (p ⫽ 0.0039 and 0.0014, respectively) associations with risk and AAO (p ⫽ 0.00070 and
0.0073, respectively); the two-SNP haplotype generated even stronger association with PD (p ⫽ 0.000013). Significant inter-
actions with smoking (p ⫽ 0.0015 for rs 2255929 and p ⬍ 0.0001 for rs 1060826) were detected in a subset of the families;
smoking was inversely associated with PD among risk allele noncarriers, but significance diminished among carriers.
Interpretation: Our findings support NOS2A as a genetic risk factor in PD, potentially by influencing AAO and by modifying
the inverse association between PD and smoking.
Ann Neurol 2006;60:366 –373

Parkinson’s disease (PD) is characterized by degenera-
tion of dopaminergic neurons within the substantia
nigra, which coordinates movement and balance. Af-
fected individuals experience development of resting
tremors, muscular rigidity, bradykinesia, and other
clinical manifestations with severity proportional to
neuronal loss. PD affects more than one million Amer-
icans, but because peak incidence occurs between ages
55 and 66 and the US population is rapidly aging,
prevalence and ensuing societal costs are increasing.1,2
Multiple lines of evidence, including familial aggre-
gation and increased sibling recurrence rate ratios
(ranging from 2–14), suggest that genetic factors confer
susceptibility without excluding the possibility of envi-
ronmental insults.3 Identification of causal genetic vari-
ants in six genes ( parkin, ␣-synuclein, UCH-L1, DJ-1,
PINK1, and LRRK2) confirms the existence of genetic
factors.4,5 A number of environmental exposures have
also been associated with PD, most notably the protec-
tive effect of cigarette smoking.6
Heightened inducible nitric oxide synthase (iNOS)
activity leading to excessive levels of nitric oxide (NO)
and its toxic metabolite, peroxynitrite, have been im-
plicated in infectious, inflammatory, and neurodegen-
erative diseases.7 Three NO synthesizing enzymes,
iNOS, neuronal NOS (nNOS), and endothelial NOS
(eNOS), have been well characterized. The constitu-
tively expressed isoforms (nNOS and eNOS) produce
low NO levels over short time intervals, whereas iNOS
produces high NO levels over long intervals.8 Inflam-
matory mediators regulate iNOS gene (NOS2A) ex-
pression to produce enough NO to combat environ-
mental insults. However, prolonged nitrosative stress
can threaten cellular survival. Significant associations
between PD and a NOS2A polymorphism (rs1060826)
have been reported.9,10 In genomic screens to identify
genes influencing age at onset (AAO) and risk for PD,
we detected evidence for risk and AAO genes on chro-
mosome 17 near NOS2A.11,12 Whether increased
iNOS activity accompanies PD pathogenesis or con-
tributes to disease onset is unclear. However, the toxic
effects of NO suggest that NOS2A is a biologically
plausible candidate gene for PD.13,14
Epidemiologic studies consistently report an inverse
association between PD and cigarette smoking. Meta-
analysis demonstrated that individuals with PD are half

From the 1Center for Human Genetics and 2Department of Med-
icine, Duke University Medical Center, Durham, NC.
Received Jan 19, 2006, and in revised form Apr 17. Accepted for
publication May 9, 2006.
Published online Jul 5, 2006, in Wiley InterScience
(www.interscience.wiley.com). DOI: 10.1002/ana.20915
Address correspondence to Dr Scott, Center for Human Genetics,
Duke University Medical Center, Box 3445, Durham, NC 27710.
E-mail: bill.scott@duke.edu

366 © 2006 American Neurological Association

Published by Wiley-Liss, Inc., through Wiley Subscription Services
as likely to report ever smoking as unaffected individ-
uals.6 We corroborated this inverse association in a
subset of our sample.15 The effect of cigarette smoke
on NOS2A expression is unclear, but one recent study
reported that cigarette smoke condensates inhibited in-
flammatory induction of iNOS and reduced cytotoxic
effects.16 Given the protective effect of cigarette smok-
ing and the biological plausibility of an interaction be-
tween NOS2A and smoking, testing for this interaction
may provide insight into the neurodegenerative process
underlying PD. We therefore set out to investigate the
main effect of NOS2A polymorphisms and NOS2A-
smoking interactions on risk for PD using two family-
based case–control data sets.
Subjects and Methods
Sample Ascertainment
Affected individuals and their family members were ascer-
tained by the Morris K. Udall PD Research Center of Ex-
cellence at the Duke Center for Human Genetics (CHG)
and 13 centers of the PD Genetics Collaboration. The clinic-
based ascertainment enrolled singleton families (only one
sampled affected member or an affected parent–child pair)
and multiplex families (nuclear families with at least two
sampled affected siblings or extended pedigrees with at least
two sampled affected individuals who are not siblings or a
parent–child pair). Any number of unaffected relatives (de-
pending on availability) were collected for both family types.
Only multiplex families are informative for linkage, whereas
both family types are informative for association. Families
with mutations in the parkin, ␣-synuclein, and LRRK2 genes
(20 singleton and 21 multiplex) were removed from the anal-
ysis to minimize heterogeneity. To minimize bias due to
population stratification, we stratified data by self-reported
race. Results for the white families, the only subset with suf-
ficient statistical power, are reported. A total of 752 families
(466 singleton and 286 multiplex families) were studied. A
description of the families is presented in Table 1.
A neurological examination was performed on all partici-
pants with a standard clinical evaluation, including assess-
ment with the Unified PD Rating Scale. Affected individuals
demonstrated at least two cardinal signs of PD (resting trem-
ors, rigidity, and bradykinesia), asymmetry of symptom on-
set, and absence of atypical signs; unaffected individuals pre-
sented no signs of PD; and individuals who showed only one
cardinal sign and/or atypical signs were classified as unclear.
Unclear individuals were excluded from analysis to maximize
phenotypic reliability. AAO was defined as the age at which
Table 1. Description of Pedigree Structures in 466 Singleton Families with 505 Affected Individuals and 286 Multiplex Families
with 637 Affected Individuals
Affected Individuals (n)
Two Parents One Parent
Families Sampled Sampled
Singleton families 87 46
Multiplex families 14 47
affected individuals recalled the development of one of the
cardinal signs. AAO was missing for affected individuals in
10 families. To ensure consistency, a clinical adjudication
board (M.A.S., B.L.S., and J.M.V.) reviewed all clinical ex-
amination data. A blood sample, family history, medical his-
tory, and standard cognitive status test (Blessed Orientation-
Memory-Concentration test, or 3MS) were collected for each
participant. Institutional review boards at the respective cen-
ters approved all study protocols and consent forms, which
were signed by all participants before blood and data collec-
tion.
Environmental Risk Factor Data Collection
Trained interviewers administered a standard telephone ques-
tionnaire to gather environmental risk factor data. These data
were collected only on families enrolled by the Duke Udall
Center; consequently, the interaction analyses utilized data
collected on 243 families (226 singleton and 17 multiplex
families).
Cigarette smoking was assessed by asking: “Have you
smoked at least 100 cigarettes in your lifetime?” and “Did
you ever smoke cigarettes at least once per week?” If appli-
cable, the years of initiating and quitting smoking were re-
corded. Participants who responded “yes” to both questions
and who initiated smoking before the reference age (derived
from the population-averaged time from onset to enroll-
ment) were classified as ever-smokers. Otherwise, participants
were classified as never-smokers.
Single Nucleotide Polymorphism Selection
Genotypes from 13 phase I HapMap single nucleotide poly-
morphism (SNPs) in NOS2A and in the 10kb regions flank-
ing the gene were examined in 30 white European families.17
Data were input into HAPLOVIEW, a program that char-
acterizes linkage disequilibrium (LD) patterns to identify
haplotype blocks and to choose tag SNPs that uniquely iden-
tify haplotypes.18 Eight tag SNPs (one nonsynonymous, one
synonymous, and six intronic) capture all common (minor
allele frequency ⬎ 10%) phase I HapMap variants across
NOS2A with a pairwise correlation coefficient (r2) greater
than or equal to 0.8. One tag SNP located 5⬘ upstream and
one tag SNP located 3⬘ downstream of NOS2A were also
selected for genotyping. Three other validated coding SNPs,
including the previously implicated synonymous SNP
(rs1060826), were selected from the National Center for
Biotechnology Information (www.ncbi.nlm.nih.gov) and En-
sembl (www.ensembl.org) databases.9,10 A total of 13
NOS2A SNPs were selected for genotyping in our singleton
data set of 466 families (Fig). Any SNPs significantly associ-
Informative Relationships (n)
No Parents Affected Discordant Affected
Sampled Sibpairs Sibpairs Relative Pairs
354 0 353 0
265 211 865 126
Hancock et al: NOS2A and Smoking in PD 367
ated with PD ( p ⬍ 0.05) were subsequently genotyped in a
validation data set of 286 multiplex families.
Genotyping
Blood samples were prepared and stored by the Duke CHG
DNA bank. Genomic DNA was extracted from whole blood
via the PUREGENE system (Gentra Systems, Minneapolis,
MN). All SNPs were genotyped using the TaqMan method
with probes and primers obtained through the Applied Bio-
systems (ABI, Foster City, CA) Assay-on-Demand or Assay-
by-Design services. Taqman reactions ran with 5␮l volumes,
and the subsequent polymerase chain reaction (PCR) ampli-
fication, was accomplished with the GeneAmp PCR system
9700 thermocyclers (ABI) for a 50-cycle program. Fluores-
cence resulting from the PCR amplification was detected us-
ing the ABI PRISM 7900HT Sequence Detection Systems
(SDS) and analyzed with SDS software.
Stringent quality-control measures maximized genotyping
accuracy. Each plate contained four controls with known ge-
notype and four nontemplate blanks. Internal controls con-
sisted of 24 duplicated individuals per plate to ensure con-
sistency. Each DNA plate met 100% matching for quality-
control samples and at least 95% efficiency. Genotype data
were then stored and managed by the PEDIGENE informat-
ics system.
Statistical Analysis
Deviations from Hardy–Weinberg equilibrium for all geno-
typed markers were assessed with the Genetic Data Analysis
(GDA) program.19 LD across NOS2A was evaluated for all
pairwise marker combinations using the Graphical Overview
of LD (GOLD) program to generate both r2 and D⬘ val-
ues.20 A total of 457 affected and 299 unaffected individuals
(at most one individual per family) were used in Hardy–
Weinberg equilibrium and LD analyses.
Family-based tests of association were used to examine the
relation between PD and NOS2A SNPs. The association in
the presence of linkage (APL) test captures information from
nuclear families, including triads (parents and offspring), dis-
cordant siblings, and affected sibling pairs or triads, maxi-
mizing genetic information gathered from our collection of
various pedigree structures.21 Parental genotype data are of-
ten missing in PD and other late-onset disease studies. Sib-
lings can be used to infer missing parental genotypes, but
when more than one sibling is affected, the correlations in
transmissions from parents to multiple affected offspring due
Fig. Gene structure of NOS2A and locations of selected single nucleotide polymorphisms (SNPs). Blocks represent the 27 exons, and
lines represent flanking intronic sequences. Shaded blocks denote coding regions, and colorless blocks denote untranslated regions.
Asterisks mark SNPs that show significant allelic association.
368 Annals of Neurology Vol 60 No 3 September 2006
to linkage must be taken into account.21 The APL test ad-
justs for these correlations by incorporating identity-by-
descent parameters into inference of missing parental geno-
types in nuclear families.21 Because APL uses genotypes of
unaffected siblings only to infer missing parental genotypes
to then derive transmitted and untransmitted allele counts to
affected individuals, an age restriction for unaffected individ-
uals is not necessary. The multilocus haplotype APL test was
also applied to combinations of significant SNPs in low to
moderate LD to identify SNP haplotypes associated with
PD. The pedigree disequilibrium test (PDT) also tests for
single-locus allelic association in nuclear pedigrees, but does
not capture information from affected sibling pairs without
both parents typed.22,23 An extension of this test, genotype-
PDT, assesses single-locus genotypic associations; such a test
is not possible in APL.24 SNPs with significant APL results
were analyzed with genotype-PDT to determine whether al-
lelic association was due to an overrepresentation of a partic-
ular genotype in affected individuals.
To examine genetic heterogeneity by age, we stratified by
AAO. Families with at least one affected individual with on-
set before age 40 were classified as early-onset, and families
with a minimum AAO of 40 years or older were classified as
late-onset. Progressively older AAO cutoff points (45, 50, 55,
and 60 years) were applied to detect association by AAO. To
assess the appropriateness of stratification by minimum AAO
when multiple affected individuals exist within families, we
calculated Pearson’s correlation coefficient (r) between indi-
vidual AAO and family reference age (minimum AAO in the
family) for affected individuals in multiplex families. For
SNPs significantly associated with PD in AAO strata, effects
on AAO were explored with the Monks–Kaplan method of
the quantitative transmission disequilibrium test (QTDT)
using AAO as a quantitative trait.25 The previous association
tests use affection status as the outcome, whereas the QTDT
uses AAO as the outcome variable among affected individu-
als.
To test interactions between genetic and environmental
risk factors for PD, we generated models with generalized
estimating equations using PROC GENMOD in SAS ver-
sion 8e (SAS Institute, Cary, NC). Population-averaged gen-
eralized estimating equation models use information from all
sampled family members (concordant and discordant sibling
and relative pairs) and estimate parameters based on marginal
case–control effects. To adjust for correlations between re-
lated individuals, we used an exchangeable correlation matrix
(assuming all pairs of relatives have equal correlation) to cal- Table 2. NOS2A Allelic Association Test Results Using
culate a robust estimate of the variance for each model term. Association in the Presence of Linkage in 466 Singleton
Although this correlation matrix structure might be incor- Families
rect, generalized estimating equation models are quite robust
to misspecification of the correlation matrix.26 GEE models APL p
assessed interactions between smoking and the risk allele of Overall Early-Onset Late-Onset
each significantly associated SNP by including the risk allele SNP (N ⫽ 466) (n ⫽ 51) (n ⫽ 409)
carrier status (carrier vs noncarrier), cigarette smoking history
(ever vs never), an interaction term between carrier status rs2531860 0.94 0.47 0.70
and smoking, age at examination (AAE), and sex. rs3730014 0.68 0.87 0.53
For SNPs that significantly interacted with smoking, a rs3794766 0.65 0.97 0.80
stratified analysis was performed to assess the differential ef- rs2072324 0.56 0.45 0.72
fects of smoking by risk allele carrier status. GEE models rs8072199 0.46 0.44 0.38
with smoking, AAE, and sex were constructed in allele car- rs16966563 0.88 0.58 0.97
riers and noncarriers separately. Due to an excess of familial rs3794764 0.64 0.68 0.52
rs1137933 0.86 0.96 0.85
clusters containing only one individual in stratified analyses, rs2248814 0.93 0.036 0.55
exchangeable correlation matrices could not be estimated. rs2297518 0.97 0.22 0.78
Thus, an independent correlation matrix was used in strati- rs1060826 0.55 0.040 0.98
fied models. Given the robustness of GEE models to mis- rs2255929 0.37 0.0018 0.87
specification of the correlation matrix and the excess of un- rs4796017 0.91 0.11 0.79
related cases and control subjects in stratified analyses (the
source of the exchangeable correlation matrix estimation er- The affected member of 51 singleton families presented Parkinson’s
disease symptoms before age 40, and the affected member of 409
ror), this adjustment is not likely to significantly impact the singleton families had onset of symptoms at or beyond age 40.
validity of the stratified analyses.26,27 Empirical p values were APL ⫽ association in the presence of linkage; SNP ⫽ single nucle-
generated to evaluate significance for initial GEE models, otide polymorphism.
and odds ratios and 95% confidence intervals were generated
to assess the direction and magnitude of association for strat-
ified analyses. age for affected individuals in multiplex families, sug-
gesting that stratification by minimum AAO is appro-
Results priate. Significant allelic association was replicated in
The 13 NOS2A polymorphisms were genotyped and the 22 multiplex families with minimum AAO less
analyzed in 466 nuclear singleton families. No signifi- than 40 years for both rs2255929 ( p ⫽ 0.018) and
cant deviations from Hardy–Weinberg equilibrium rs1060826 ( p ⫽ 0.027). No significant allelic associa-
were detected in affected and unaffected individuals. As tion of either polymorphism was detected in the overall
shown in Table 2, the APL test detected no significant multiplex data set (286 families) or in the 260 multi-
allelic associations in the overall data set of 466 single- plex families with a minimum AAO of 40 years or
ton families or in the 409 late-onset (AAO ⱖ 40 years) older.
singleton families. However, APL detected significant Considering the significant allelic association of
association ( p ⬍ 0.05) between PD and three SNPs, rs2255929 and rs1060826 in singleton and multiplex
rs2255929, rs1060826, and rs2248814, in the 51 families, these data sets were combined for further
early-onset families. The positions of these significantly analysis. No significant allelic association was observed
associated SNPs in NOS2A are noted with asterisks in for either rs2255929 or rs1060826 in the 752 com-
the Figure. SNP rs2255929 provided the most signifi- bined families. Because previous analyses showed sig-
cant evidence for association (see Table 2), with mod- nificant associations only in earlier onset families, we
erate LD existing between this polymorphism and both stratified the 742 families with AAO data available by
rs1060826 and rs2248814 among affected individuals minimum AAO using progressively increasing cutoff
(Table 3). SNPs rs2248814 and rs1060826 were sig- points of 40, 45, 50, 55, and 60 years. As shown in
nificantly associated with PD (see Table 2), with high Table 4, the most significant association between the
LD existing between these SNPs (see Table 3). Similar more frequent T allele of rs2255929 and PD was
levels of LD between these three markers were shown achieved in the 73 families with at least one member
in the unaffected individuals (data not shown). Con- affected before age 40; however, association between
sidering the LD pattern among these significantly as- PD and rs2255929(T) remained significant across all
sociated markers, we selected rs2255929 and early-onset strata. The most significant association be-
rs1060826 for analysis in a second data set. tween PD and the less frequent A allele of rs1060826
We genotyped rs2255929 and rs1060826 in a vali- was also achieved in the 73 earliest onset families; the
dation data set consisting of 286 nuclear and extended association remained significant only in the families
multiplex families. A strong correlation existed (r ⫽ with a minimum AAO of younger than 45 and 50
0.86) between individual AAO and family reference years (see Table 4). Among the 73 earliest onset fami-

Hancock et al: NOS2A and Smoking in PD 369

Table 3. Linkage Disequilibrium between NOS2A Polymorphisms in 457 Affected and Unrelated Individuals as Measured by r2
(Above Diagonal) and D⬘ (Below Diagonal)

rs2531860 rs3730014 rs3794766 rs2072324 rs8072199 rs16966563

rs2531860 — 0.003 0.042 0.127 0.096 0.085

rs3730014 1a — 0.001 0.005 0.004 0
rs3794766 0.858 0.110 — 0.077 0.094 0.008
rs2072324 0.421 1a 0.953a — 0.181 0.006
rs8072199 0.845 0.562 0.470 0.961a — 0.019
rs16966563 0.794 1a 1a 0.993a 1a —
rs3794764 0.403 0.999a 0.906 0.985a 0.981a 0.996a
rs1137933 0.774 0.021 0.965a 1a 0.535 1a
rs2248814 0.434 0.355 0.641 0.937 0.246 1a
rs2297518 0.722 0.016 0.824 0.696 0.491 0.999a
rs1060826 0.435 0.322 0.652 0.956a 0.249 1a
rs2255929 0.353 0.508 0.679 0.311 0.050 1a
rs4796017 0.642 0.364 0.605 0.181 0.070 1a
Similar levels of linkage disequilibrium were observed in 299 unaffected and unrelated individuals.
a 2
r values greater than 0.8 and D⬘ values greater than 0.95.

lies, the global test for association of all haplotypes be-
tween the two polymorphisms (rs2255929 and
rs1060826) and PD risk provided stronger evidence for
association than either of the individual polymor-
phisms ( p ⫽ 0.000013). Furthermore, among the 73
earliest onset families, the global genotype-PDT test
showed a significant association for the TT genotype at
rs2255929 ( p ⫽ 0.0039) and for the AA genotype at
rs1060826 ( p ⫽ 0.0014). Given the AAO trend in as-
sociations of these SNPs and PD, we applied the
QTDT using AAO as a quantitative trait in the com-
bined data set. QTDT analysis demonstrated a signif-
icant association between earlier AAO of PD and the T
allele at rs2255929 ( p ⫽ 0.00070) and the A allele at
rs1060826 ( p ⫽ 0.0073).
GEE modeling showed significant interactions be-
tween smoking and NOS2A SNPs. In a model contain-
ing rs2255929, smoking, AAE, and sex, the interactive
term between rs2255929 T-allele carrier status and
smoking was statistically significant ( p ⫽ 0.0015).
When stratifying by rs2255929 genotype, the consis-
tently documented significant inverse association be-
tween ever-smoking and PD was present in 90 carriers
of the AA genotype; in 366 carriers of the risk allele
[rs2255929(TA/TT)], this significant inverse associa-
tion diminished (Table 5). The model containing
rs1060826, smoking, AAE, and sex also showed a sig-
nificant interaction between rs1060826 and smoking
( p ⬍ 0.0001). When stratifying by rs1060826 geno-
type, the inverse association between ever-smoking and
PD was observed in 167 carriers of the GG genotype,
but among 289 carriers of the risk allele
[rs1060826(AG/AA)]; the inverse association again di-
minished (see Table 5).
Discussion
We identified two NOS2A polymorphisms and their
haplotype to be significantly associated with PD among
earlier onset singleton and multiplex families. SNP
rs2255929 is an intronic variation between exons 23
and 24, and rs1060826 is a synonymous variation in
exon 22. The risk alleles, rs2255929(T) and
rs1060826(A), were also shown to be significantly as-
sociated with earlier AAO. We have presented a large
number of statistical tests without adjustment for mul-
tiple testing. After applying the conservative Bonferroni
correction for 76 tests, the significant associations of
rs2255929(T) and the rs2255929-rs1060826 haplotype
with early-onset PD and the significant interaction be-
tween rs1060826(A) and smoking remained statistically
significant. Our findings strongly suggest that NOS2A
influences the development of PD.
An alteration in NOS2A regulation and/or constant
environmental induction could result in prolonged
iNOS activity and contribute to the depletion of do-
paminergic neurons leading to PD. Mice with an
ablated NOS2A resist the neurotoxic effects of
1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP),
a parkinsonism-inducing agent. The mice have a simi-
lar immunological response, yet a statistically signifi-
cant increase in neuronal survival rate compared with
wild-type littermates.28 –30 In rats, selective pharmaco-
logical inhibition of NOS2A attenuates inflammatory-
induced dopaminergic neuronal loss.8,31 In postmor-
tem human brains, a substantially increased iNOS
concentration has been detected in the substantia nigra
of PD patients compared with matched control sub-
jects.32 These biological data substantiate NOS2A as a
strong candidate for PD susceptibility.
Two previous case–control studies examined only

370 Annals of Neurology Vol 60 No 3 September 2006

Table 3. Continued

rs3794764 rs1137933 rs2248814 rs2297518 rs1060826 rs2255929 rs4796017

0.107 0.030 0.023 0.023 0.023 0.016 0.053

0.005 0 0.001 0 0.001 0.003 0.002
0.073 0.825a 0.089 0.535 0.094 0.206 0.156
0.918a 0.073 0.156 0.031 0.158 0.035 0.011
0.195 0.109 0.055 0.080 0.055 0.002 0.005
0.007 0.007 0.016 0.006 0.016 0.018 0.018
— 0.077 0.166 0.035 0.169 0.035 0.014
1a — 0.090 0.640 0.093 0.238 0.193
0.939 0.680 — 0.174 0.953a 0.464 0.369
0.718 0.850 1a — 0.174 0.358 0.304
0.958a 0.687 0.976a 1a — 0.490 0.371
0.301 0.766 0.975a 1a 1a — 0.675
0.197 0.707 0.850 0.948 0.858 0.837 —
Similar levels of linkage disequilibrium were observed in 299 unaffected and unrelated individuals.
a 2
r values greater than 0.8 and D⬘ values greater than 0.95.

one polymorphism in this candidate gene and reported
significant inverse associations between the rs1060826
AA genotype and PD.9,10 Our family-based case–con-
trol data showed a significant positive association be-
tween the A allele and AA genotype of rs1060826 and
earlier onset PD. In 752 families, we examined multi-
ple polymorphisms across NOS2A to provide compre-
hensive coverage of common variation and detected
significant associations between PD and two SNPs:
rs2255929 and rs1060826. To limit confounding due
to population stratification, we restricted analyses to
white families only. This is likely sufficient control for
such confounding, because self-reported race has nearly
perfect correlation with genetic determination of ethnic
background among the major US racial groups.33
However, spurious association is possible if substantial
heterogeneity exists within major racial groups.33
Numerous differences exist between our study and
previous studies that may explain the opposing direc-
tions of association. First, we detected significant asso-
ciation between rs1060826 and PD in earlier onset
families, whereas Levecque and colleagues9 examined
the association in a sample consisting mostly of later
onset cases and older unrelated control subjects. Al-
though a protective effect of rs1060826 was not de-
tected in our late-onset families, the opposing associa-
tions may be reflective of differing effects of NOS2A
SNPs by AAO. Second, our family-based case–control
sample consists of white individuals ascertained in the
United States, whereas the two previous studies ascer-

Table 4. Allelic Association Test Results for rs1060826 and rs2255929 Using Association in the Presence of Linkage in 742
Singleton and Multiplex Parkinson’s Disease Families at Progressively Older Age-at-Onset Cutoff Points

Early-Onset Families Late-Onset Families

Minimum Observed Expected Observed Expected

SNP Allele AAO (yr) na Transmissionsb Transmissionsc APL p na Transmissionsb Transmissionsc APL p

rs1060826(A) 40 73 74 63.5 0.0062 669 668 666.9 0.92

45 145 150 135.1 0.0063 597 592 594.3 0.82
50 224 236 216.2 0.0026 518 506 512.3 0.50
55 360 366 352.3 0.094 382 376 377.7 0.84
60 443 439 428.1 0.23 299 303 302.1 0.90
rs2255929(T) 40 73 96 81.8 0.000059 669 935 936.7 0.87
45 145 203 183.2 0.00046 597 828 834.3 0.54
50 224 315 296.6 0.0059 518 716 719.3 0.72
55 360 513 490.2 0.0062 382 518 526.1 0.34
60 443 613 593.0 0.024 299 417 424.5 0.33
a
Number of families categorized into early- and late-onset strata at each minimum AAO cutoff point.
b
Observed number of alleles transmitted from parents to affected offspring.
c
Expected number of alleles transmitted from parents to affected offspring.
SNP ⫽ single nucleotide polymorphism; AAO ⫽ age at onset; APL ⫽ association in the presence of linkage.

Hancock et al: NOS2A and Smoking in PD 371

Table 5. Effect of Cigarette Smoking on Parkinson’s Disease
Risk as Stratified by Genotypes at rs1060826 and rs2255929
Cases/Control Lower Upper
SNP Genotype Subjects (n) OR CI CI
rs1060826(GG) 96/71 0.29 0.14 0.59
rs1060826(AG/AA) 155/134 0.91 0.57 1.44
rs2255929(AA) 52/38 0.26 0.11 0.66
rs2255929(TA/TT) 199/167 0.77 0.51 1.19
SNP ⫽ single nucleotide polymorphism; OR ⫽ odds ratio; CI ⫽
confidence interval.
tained cases and unrelated control subjects from Euro-
pean populations.9,10 Haplotypic differences between
these populations could result in differing associations
with a polymorphism that is not considered to be a
causal variant, but potentially resides in differing de-
grees of LD with the true variant. We detected a highly
significant association between PD and haplotypes con-
taining rs2255929 and rs1060826, but the previous
studies considered only rs1060826. This illustrates the
need to evaluate multiple polymorphisms across candi-
date genes. Finally, differences in the frequency of
smoking exist between our study (46.2% of cases and
52.3% of related control subjects reported ever having
smoked) and the Levecque and colleagues’ study,9
which reported smoking characteristics (33% of cases
and 25% of unrelated control subjects reported ever
having smoked), but did not report testing for an in-
teractive effect.
A limitation of our study is the reduced number of
families in which gene–environment interactions can
be examined. Even though this portion of our study
has reduced power for testing marginal genetic and en-
vironmental effects, our sample size is sufficient to de-
tect large interactions. Using parameters similar to
those in the NOS2A-smoking interaction models
(dominant mode of inheritance, minor allele frequency
of 40%, marginal genetic effect of OR ⫽ 1.15, envi-
ronmental prevalence of 50%, marginal environmental
effect of OR ⫽ 0.69, and interactive effect of OR ⫽
8.00), our sample size achieves a power of 99.68% to
detect a gene–environment interaction of this magni-
tude as calculated by QUANTO.34 A larger sample
size is required for replication if this interactive effect is
overestimated. Nevertheless, in the presence of such in-
teraction, examination of NOS2A polymorphism ef-
fects without considering the effect of smoking could
lead to differing results across studies. More complete
evaluation of NOS2A variants and molecular investiga-
tion into the main effect of NOS2A and its interactive
effect with smoking is necessary to resolve these dis-
crepancies.
In synthesizing NO from L-arginine, the oxygenase
domains of two iNOS molecules dimerize and align
with the reductase domain of one subunit, which en-
372 Annals of Neurology Vol 60 No 3 September 2006
compasses a flavin adenine dinucleotide (FAD) binding
domain.35 The FAD binding domain spans exons 19
through 23 of NOS2A (Ensembl). SNP rs1060826 re-
sides in exon 22, and rs2255929 resides in the intronic
region flanked by exons 23 and 24. The significantly
associated markers potentially reside in strong LD with
a variant that disrupts the FAD binding domain of
NOS2A and alters NO synthesis. More extensive char-
acterization of the FAD binding domain variation, its
LD block, and its regulatory elements is necessary to
identify this causal variant.
Examination of a statistical interaction between
NOS2A and smoking in relation to PD risk was previ-
ously unexplored, and limited biological evidence is
available to decipher the mechanism of an interaction.
One previous study reported that cigarette smoke con-
densates attenuate iNOS induction in vitro, ultimately
reducing cytotoxic effects.16 Among noncarriers of the
reported risk alleles, our results are consistent with this
observed protective effect of cigarette smoking. How-
ever, a genetic alteration that disrupts the interactive
mechanism could potentially hinder the inhibitory ef-
fect of cigarette smoke on iNOS induction as demon-
strated by the diminished protective effect of smoking
among carriers of the risk alleles. Given the strong ep-
idemiologic evidence supporting an environmental ef-
fect by smoking and our statistical suggestion of an in-
teraction, further investigation is warranted to establish
the biological mechanism of an interaction. Our find-
ings strongly suggest that NOS2A is a genetic risk fac-
tor for PD, and its interaction with cigarette smoking
merits consideration in future studies of NOS2A and
PD.
This research was supported by the NIH (National Institute on
Neurological Disorders and Stroke, P50 NS39764-03, J.M.V.) and
the Duke University Graduate School (D.B.H.).
We are grateful to the PD families for their participation in our
study. We thank the PD Genetics Collaboration members who con-
tributed families: M. Nance, R. Watts, J. Hubble, W. Koller, K.
Lyons, R. Pahwa, M. Stern, A. Colcher, B. Hiner, J. Jankovic, W.
Ondo, F. Allen Jr, C. Goetz, G. Small, D. Masterman, F. Mastaglia,
and J. Haines. We also thank B. Wheeler, E. Tegnell, and the CHG
and Duke University Medical Center personnel for their exceptional
clinical, technical, and administrative contributions.
References
1. Olanow CW, Koller WC. An algorithm (decision tree) for the
management of Parkinson disease: treatment guidelines. Neu-
rology 1998;50:S1–S57.
2. McDonald WM, Richard IH, DeLong MR. Prevalence, etiol-
ogy, and treatment of depression in Parkinson’s disease. Biol
Psychiatry 2003;54:363–375.
3. Gasser T. Genetics of Parkinson’s disease. Ann Neurol 1998;
44:S53–S57.
4. Vila M, Przedborski S. Genetic clues to the pathogenesis of
Parkinson’s disease. Nat Med 2004;(suppl 10):S58 –S62.
 5. Foroud T. LRRK2: both a cause and a risk factor for Parkinson
disease? Neurology 2005;65:664 – 665.
6. Hernan MA, Takkouche B, Caamano-Isorna F, Gestal-OteroJJ.
A meta-analysis of coffee drinking, cigarette smoking, and the
risk of Parkinson’s disease. Ann Neurol 2002;52:276 –284.
7. Kroncke KD, Kehsel K, Kolb-Bachofen V. Inducible nitric ox-
ide synthase in human diseases. Clin Exp Immunol 1998;113:
147–156.
8. Arimoto T, Bing G. Up-regulation of inducible nitric oxide
synthase in the substantia nigra by lipopolysaccharide causes
microglial activation and neurodegeneration. Neurobiol Dis
2003;12:35– 45.
9. Levecque C, Elbaz A, Clavel J, et al. Association between Par-
kinson’s disease and polymorphisms in the nNOS and iNOS
genes in a community-based case-control study. Hum Mol
Genet 2003;12:79 – 86.
10. Hague S, Peuralinna T, Eerola J, et al. Confirmation of the
protective effect of iNOS in an independent cohort of Parkin-
son disease. Neurology 2004;62:635– 636.
11. Scott WK, Nance MA, Watts RL, et al. Complete genomic
screen in Parkinson disease: evidence for multiple genes. JAMA
2001;286:2239 –2244.
12. Li YJ, Scott WK, Hedges DJ, et al. Age at onset in two com-
mon neurodegenerative diseases is genetically controlled. Am J
Hum Genet 2002;70:985–993.
13. Kroncke KD, Suschek CV, Kolb-Bachofen V. Implications of
inducible nitric oxide synthase expression and enzyme activity.
Antioxid Redox Signal 2000;2:585– 605.
14. Ebadi M, Sharma SK. Peroxynitrite and mitochondrial dysfunc-
tion in the pathogenesis of Parkinson’s disease. Antioxid Redox
Signal 2003;5:319 –335.
15. Scott WK, Zhang F, Stajich JM, et al. Family-based case-
control study of cigarette smoking and Parkinson disease. Neu-
rology 2005;64:442– 447.
16. Mazzio EA, Kolta MG, Reams RR, Soliman KFA. Inhibitory
effects of cigarette smoke on glial inducible nitric oxide syn-
thase and lack of protective properties against oxidative neuro-
toxins in vitro. Neurotoxicology 2005;26:49 – 62.
17. International HapMap Consortium. The International Hap-
Map Project. Nature 2003;426:789 –795.
18. Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and
visualization of LD and haplotype maps. Bioinformatics 2005;
21:263–265.
19. Lewis PO, Zaykin D. Genetic data analysis: computer program
for analysis of allelic data Version 1.0(d15). 2000, http://
hydrodictyon.eeb.uconn.edu/people/plewis/software.php
20. Abecasis GR, Cookson WOC. GOLD—graphical overview of
linkage disequilibrium. Bioinformatics 2000;16:182–183.
21. Martin ER, Bass MP, Hauser ER, Kaplan NL. Accounting for
linkage in family-based tests of association with missing paren-
tal genotypes. Am J Hum Genet 2003;73:1016 –1026.
22. Martin ER, Monks SA, Warren LL, Kaplan NL. A test for link-
age and association in general pedigrees: the Pedigree Disequi-
librium Test. Am J Hum Genet 2000;67:146 –154.
23. Martin ER, Bass MP, Kaplan NL. Correcting for a potential
bias in the Pedigree Disequilibrium Test. Am J Hum Genet
2001;68:1065–1068.
24. Martin ER, Bass MP, Gilbert JR, et al. Genotype-based associ-
ation tests for general pedigrees; the genotype-PDT. Genet Epi-
demiol 2003;23:203–213.
25. Monks SA, Kaplan NL. Removing the sampling restrictions
from family-based tests of association for a quantitative-trait lo-
cus. Am J Hum Genet 2000;66:576 –592.
26. Zeger SL, Liang KY. Longitudinal data analysis for discrete and
continuous outcomes. Biometrics 1986;42:121–130.
27. Hardin JW, Hilbe JM. Generalized Estimating Equations. Boca
Raton, FL: Chapman & Hall/CRC, 2003.
28. Liberatore GT, Jackson-Lewis V, Vukosavic S, et al. Inducible
nitric oxide synthase stimulates dopaminergic neurodegenera-
tion in the MPTP model of Parkinson disease. Nat Med 1999;
5:1403–1409.
29. Grunewald T, Beal MF. NOS knockouts and neuroprotection.
Nat Med 1999;5:1354 –1355.
30. Dehmer T, Lindenau J, Haid S, et al. Deficiency of inducible
nitric oxide synthase protects against MPTP toxicity in vivo.
J Neurochem 2000;74:2213–2216.
31. Herrera AJ, Castano A, Venero JL, et al. The single intranigral
injection of LPS as a new model for studying the selective ef-
fects of inflammatory reactions on dopaminergic system. Neu-
robiol Dis 2000;7:429 – 447.
32. Hunot S, Boissiere F, Faucheux B, et al. Nitric oxide synthase
and neuronal vulnerability in Parkinson’s disease. Neuroscience
1996;72:355–363.
33. Tang H, Quertermous T, Rodrigues B. Genetic structure, self-
identified race/ethnicity, and confounding in case-control asso-
ciation studies. Am J Hum Genet 2005;76:268 –275.
34. Gauderman JW. Sample size requirements for matched case-
control studies of gene-environment interaction. Stat Med
2002;21:35–50.
35. Siddhanta U, Presta A, Fan B, et al. Domain swapping in in-
ducible nitric-oxide synthase. J Biol Chem 1998;273:
18950 –18958.
Hancock et al: NOS2A and Smoking in PD 373