␣-Synuclein, pesticides, and Parkinson disease A case– control study
L. Brighina, MD R. Frigerio, MD N.K. Schneider, BA T.G. Lesnick, MS M. de Andrade, PhD J.M. Cunningham, PhD M.J. Farrer, PhD S.J. Lincoln, BS H. Checkoway, PhD W.A. Rocca, MD, MPH D.M. Maraganore, MD
ABSTRACT
Background: Aggregation and fibrillization of the ␣-synuclein protein (encoded by the SNCA gene) may represent key events in the pathogenesis of Parkinson disease (PD). Variability in the length of a dinucleotide repeat sequence (REP1) within the SNCA promoter confers susceptibility to sporadic PD. Pesticide exposures may also confer susceptibility to PD. Our objective was to test possible joint effects of SNCA REP1 genotypes and pesticide exposures on the risk of PD.
Methods: This was a case– control study. Cases were recruited prospectively from the Department of Neurology of the Mayo Clinic, Rochester, MN, after June 1, 1996. The control subjects included unaffected siblings of cases and unrelated population control subjects. We assessed pesticide exposures by telephone interview and genotyped SNCA REP1. Odds ratios (ORs) and 95% CIs were determined using conditional logistic regression models. Results: There were 833 case– control pairs. We observed an increased risk of PD with increasing SNCA REP1 bp length (OR, 1.18 for each score unit; 95% CI, 1.02–1.37; p ⫽ 0.03). Pesticide exposures were associated with PD in younger subjects only (lowest quartile of age at study, ⱕ59.8 years; OR, 1.80; 95% CI, 1.12–2.87; p ⫽ 0.01 for all pesticides; OR, 2.46; 95% CI, 1.34 – 4.52; p ⫽ 0.004 for herbicides). In multivariate analyses, both SNCA REP1 score and pesticide exposures were significantly associated with PD in younger subjects, but there were no pairwise interactions. Conclusions: Our findings suggest that SNCA REP1 genotype and herbicides have independent ef-
Address correspondence and reprint requests to Dr. D.M. Maraganore, Department of Neurology, Mayo Clinic College of Medicine, 200 First Street SW, Rochester, MN 55905
[email protected]
fects on risk of Parkinson disease, primarily in younger subjects. Neurology® 2008;70:1461–1469
Accumulating evidence supports the involvement of both genetic and environmental factors in the etiology of Parkinson disease (PD). This evidence largely builds on two landmark discoveries regarding the causes of PD. The first discovery was the occurrence of parkinsonism secondary to the injection of meperidine analogs, and more specifically 1-methyl-4-phenyl-4-propionpiperidine.1 The 1-methyl-4-phenyl-4-propionpiperidine discovery led to the hypothesis that some pesticide exposures might confer susceptibility to PD because 1-methyl-4-phenyl-4-propionpiperidine chemically resembles the herbicide paraquat.2 Since then, a number of toxicologic and epidemiologic studies have provided evidence supporting this hypothesis.3-5 The second landmark discovery was that mutations in the ␣-synuclein gene (SNCA) may cause PD in some families.6 The SNCA discovery led to the hypothesis that aggregation and e-Pub ahead of print on March 5, 2008, at www.neurology.org. From the Department of Neurology (L.B., R.F., W.A.R., D.M.M.), the Division of Epidemiology (W.A.R.) and the Division of Biostatistics (N.K.S., T.G.L., M.A.), Department of Health Sciences Research, and the Department of Laboratory Medicine (J.M.C.), Mayo Clinic College of Medicine, Rochester, MN; the Department of Neuroscience (S.J.L., M.J.F.), Mayo Clinic College of Medicine, Jacksonville, FL; and the Department of Environmental and Occupational Sciences (H.C.), University of Washington, Seattle, WA. Dr. Brighina was on leave from the Department of Neuroscience and Biomedical Technologies, University of Milano-Bicocca, Monza, Italy, at the time of the study. Supported by grants NS 33978 and ES 10751 (National Institutes of Health). Disclosure: Dr. Maraganore is the co-inventor of a US provisional patent application entitled “Method of treating neurodegenerative disease.” It has been licensed to Alnylam Pharmaceuticals, Inc. and he has received less than $10,000 in royalty payments. Dr. Maraganore has consulted for Alnylam Pharmaceutical, Inc., Pfizer, Inc., and Merck and Co., Inc. during the last 5 years but received no personal compensation. Dr. Farrer is the co-inventor of a US provisional patent application entitled “Method of treating neurodegenerative disease.” It has been licensed to Alnylam Pharmaceuticals, Inc. and he has received royalty payments in excess of $10,000. Dr. Farrer has a consulting relationship with Amgen, Inc. for which no payments have been made to date. Mayo Clinic has received royalty payments in excess of $10,000 from Alnylam Pharmaceuticals, Inc., in relation to the US provisional patent application entitled “Method of treating neurodegenerative disease.” Copyright © 2008 by AAN Enterprises, Inc.
1461
fibrillization of the ␣-synuclein protein is a common cause of PD.7 ␣-Synuclein is a major component of Lewy bodies, the pathologic hallmark of PD,8 and overexpression of the SNCA gene is sufficient to cause PD in some families.9 Although SNCA causal mutations are rare, variability in the length of a dinucleotide repeat sequence (REP1) within the SNCA promoter has been shown to confer risk of PD across populations worldwide, presumably by the same mechanisms of gene overexpression and protein aggregation.10 Experimental studies have provided a biologic link of those two discoveries. Pesticides directly accelerate the rate of ␣-synuclein fibril formation in vitro.11 In addition, paraquat causes upregulation and aggregation of ␣-synuclein in wildtype mice substantia nigra,12 and pesticides exacerbate the pathology associated with causal SNCA mutations in transgenic mice.13 In light of these findings, the objective of this study was to test possible joint effects of SNCA REP1 genotypes and pesticide exposures on the risk of PD. METHODS Subjects. Case subjects were patients with PD sequentially referred to the Department of Neurology of the Mayo Clinic, Rochester, MN, after June 1996, who resided in Minnesota or in one of the surroundings four states (Wisconsin, Iowa, South Dakota, or North Dakota). The diagnosis of PD was made by a movement disorder specialist using previously reported criteria.14 Control subjects included unaffected siblings of cases or unrelated population control subjects when there were no available siblings. The unrelated control subjects were identified by random digit dialing (younger than age 65 years) or from the Centers for Medicare and Medicaid Services lists (ages 65 years and older) and resided in the same five-state region. Potential control subjects were screened for parkinsonism by a validated telephone instrument.15 Only potential control subjects who screened negative for PD or who were confirmed not to have PD by clinical assessment (despite having screened positive by telephone interview) were included in the study. The Mayo Clinic Investigational Review Board approved all study methods.
Genotyping. Venous blood samples were obtained from all cases and control subjects with a written informed consent. Specimens were collected directly from the subjects examined (cases or control subjects screening positive for PD at telephone interview) or by mail-in blood kits (control subjects who screened negative for PD at telephone interview). Genomic DNA was extracted from leukocytes with the Puregene procedure (Gentra Systems, Minneapolis, MN). SNCA REP1 allele lengths were measured using previously de1462
Neurology 70
April 15, 2008 (Part 2 of 2)
scribed genotyping methods.10 Genotyping was performed on an ABI 3730XL platform and allelic sizes assessed using GeneMapper version 4.0 software (Applied Biosytems, Inc., Foster City, CA).
Ascertainment of exposures. Exposures data were obtained by telephone by direct or proxy (for subjects who had died or incapacitated subjects) interviews using a structured risk factors questionnaire administered by specifically trained study assistants. We first collected information for all occupations held for more than 1 year and before the year of onset of PD (or the index year). All subjects who farmed for 5 or more years were asked additional detailed questions about the total duration of farming in years; main types of crops; the indication for use of pesticides (including herbicides, insecticides, or fungicides); and the names of the specific products used. Details about the questionnaire are reported elsewhere.4 In addition, we inquired about gardening as a hobby for more than 1 year. Subjects who gardened were asked about duration in years; type of gardening (flowers, vegetables, or fruits); and the names of specific products used. Reported pesticide exposures were post-coded according to indication for use (herbicides, insecticides, or fungicides) and according to the chemical class of the active ingredients (e.g., organophosphates) using the Pesticide Action Network database (www.pesticideinfo.org). All postcoding was done by two physicians (L.B. and R.F.) kept unaware of the case or control status of subjects to avoid bias. We conducted a reliability study of the risk factors questionnaire including the questions regarding pesticide exposures (overall and by indication for use). We included 55 subjects (20 with PD, 35 control subjects) in the reliability study. The initial telephone interview and the repeated telephone interview were performed by two different interviewers (by design). We calculated kappa statistics for categorical variables16,17 and the intraclass correlation coefficients for continuous variables.18 Statistical analysis. To study the association of SNCA REP1 variants and of pesticides with PD, we first matched cases to a single unaffected sibling of the same sex (when possible) and then of closest age at study. For cases without an available unaffected sibling, we matched an unrelated population control subject by sex, age (⫾ 2 years), and geographic region. Hardy-Weinberg equilibrium was estimated for the SNCA REP1 genotypes among control subjects. We calculated a SNCA REP1 genotype score for each subject as previously described.10,19 Higher SNCA REP1 genotype scores are correlated with higher expression levels of the SNCA gene.19 We assigned an allele score of “0” to the 259 bp allele, a score of “1” to the 261 bp allele, and a score of “2” to the 263 bp allele. We calculated the SNCA REP1 genotype score as the sum of the score for the two alleles. SNCA REP 1 genotype scores ranged from 0 to 4: 0 (259/259), 1 (259/261), 2 (261/261 or 259/263), 3 (261/263), or 4 (263/263). We then studied the association of SNCA REP1 genotype score with PD susceptibility by conditional logistic regression analysis. We considered exposure to pesticides overall (occupational, including farming; residential, including gardening; or both; ever/never) and categorized by indication for use (herbicides, insecticides, or fungicides) and by chemical class of the active ingredients. We then studied the association of
Table 1
Characteristics of Parkinson disease cases, unaffected sibling control subjects, and unrelated control subjects Cases-unaffected siblings
Cases-unrelated control subjects
All cases and control subjects
General characteristics
Cases
Control subjects
Cases
Control subjects
Cases
Total sample, n (%)
472 (56.7)
472 (56.7)
361 (43.3)
361 (43.3)
833 (100)
833 (100)
Men
294 (62.3)
239 (50.6)
236 (65.4)
236 (65.4)
530 (63.6)
475 (57.0)
Women
178 (37.7)
233 (49.4)
125 (34.6)
125 (34.6)
303 (36.4)
358 (43.0)
Age at onset, median (range)
59.5 (30.6–86.9)
Age at study, median (range)*
65.3 (32.8–91.4)
64.4 (32.0–86.8)
70.2 (42.3–90.4)
70.9 (44.9–92.8)
67.7 (32.8–91.4)
66.8 (32.0–92.8)
409 (86.7)
395 (83.7)
286 (79.2)
312 (86.4)
695 (83.4)
707 (84.9)
108 (26.4)
100 (25.3)
89 (31.1)
103 (33.0)
197 (28.3)
203 (28.7)
163 (39.9)
152 (38.5)
98 (34.3)
98 (31.4)
261 (37.6)
250 (35.4)
2 (0.7)
3 (1.0)
—
65.1 (23.3–88.0)
—
Control subjects
61.9 (23.3–88.0)
—
Region of origin of parents† Both parents of European origin Both parents Northern European‡ Both parents Central European
§
Both parents Southern European Both parents European, mixed region
3 (0.7)
3 (0.8)
5 (0.7)
6 (0.8)
135 (33.0)
140 (35.4)
97 (33.9)
108 (34.6)
232 (25.9)
Only one parent of European origin¶
38 (8.1)
46 (9.7)
53 (14.7)
32 (10.3)
91 (10.9)
248 (35.1) 78 (9.4)
One parent declared “American”**
17 (3.6)
16 (3.4)
17 (4.7)
9 (2.5)
34 (4.1)
25 (3.0)
Both parents declared “American”**
17 (3.6)
17 (3.6)
11 (3.0)
7 (1.9)
28 (3.7)
24 (2.9)
Both parents Asian
1 (0.2)
2 (0.5)
3 (0.8)
0 (0.0)
4 (0.5)
2 (0.2)
Both parents Mexican
1 (0.2)
1 (0.2)
0 (0.0)
1 (0.3)
1 (0.1)
2 (0.2)
Both parents African
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
0 (0.0)
Unknown
4 (0.9)
10 (2.1)
7 (1.9)
4 (1.1)
11 (1.3)
14 (1.7)
*Age at examination (cases) or blood draw (control subjects). †Region of origin of parents was self-reported by subjects. Note that Parkinson disease cases and their siblings were not always in agreement. ‡“Northern European” includes Scandinavian, Swedish, Norwegian, Finnish, Danish, Irish, or British origins. § “Central European” includes French, Belgian, Dutch, Swiss, Luxemburgian, German, Austrian, Hungarian, Polish, Czechoslovakian, or Russian origins. ”Southern European” includes Italian, Spanish, Portuguese, Greek, or Yugoslavian origins. Includes subjects for whom origin of one parent is unknown. ¶ Includes subjects for whom origin of one parent is unknown. **These subjects were all white and not Native Americans.
pesticide exposures with PD susceptibility by conditional logistic regression analyses. All analyses of main effects were adjusted for age and sex as appropriate. Analyses were performed for subjects overall and stratified by family history of PD (defined as at least one first-degree relative with PD), age at study (divided in quartiles), and sex. All stratified analyses were preplanned and were supported by a biologic rationale. We calculated odds ratios (ORs), 95% CIs, and p values (two-tailed tests, ␣ ⫽ 0.05). We also performed analyses adjusted by education (four quartiles), smoking (ever smoked more than 100 cigarettes vs never), type of respondent (direct vs proxy), and type of control (siblings vs unrelated control subjects). We also performed analyses restricted to case-unaffected sibling or case-unrelated control pairs or to occupational pesticide or residential pesticide exposures. We explored the joint effects of the SNCA REP1 genotype score and pesticide exposures (overall or by indication for use) on PD susceptibility using conditional logistic regression models. We included age at study and sex in all of the models (baseline). We compared the goodness-of-fit of the models using the Akaike Information Criterion20 and assessed their significance using likelihood ratio tests. For each interaction term in the models, we calculated ORs, 95% CIs, and p values. We similarly explored the joint effects of the SNCA REP1 genotype score and pesticide exposures (overall or by indication for use) on age at onset of PD using Cox
proportional hazard models and calculated hazard ratios, 95% CIs, and p values. We included sex in all of the models (baseline). All statistical analyses were performed with SAS version 9.1 (SAS Institute Inc., Cary, NC) or S-Plus version 7 (Insightful Corp., Seattle, WA). RESULTS
Exposure interviews and blood samples were collected from 833 case– control pairs (472 case-unaffected sibling pairs and 361 caseunrelated control pairs). A summary of the clinical and demographic characteristics of the cases and control subjects is provided (table 1). Overall, the case group included more men than the control group (63.6% vs 57.0%). However, the cases and control subjects had a similar median age at study (67.7 vs 66.8 years). Overall, the median age at onset of PD in the cases was 61.9 years. Cases and control subjects were primarily white and of European descent. The frequencies of SNCA REP1 genotypes did not deviate significantly from Hardy-Weinberg equilibrium in control subjects. We noted a trend toward increased risk of PD with increasing
Neurology 70
April 15, 2008 (Part 2 of 2)
1463
Table 2
Results of case– control analyses for SNCA REP1 genotype score and risk of Parkinson disease Genotype score,* n (%)
Trend model†
Sample or stratum
No. of subjects
0
1
2
3
4
OR (95% CI)
p
Cases, all
833
49 (5.9)
261 (31.3)
426 (51.1)
93 (11.2)
4 (0.5)
1.18 (1.02–1.37)
0.03
Control subjects, all
833
58 (7.0)
294 (35.3)
395 (47.4)
83 (10.0)
3 (0.4)
1.00 (reference)
Cases, family history‡
141
7 (5.0)
51 (36.2)
73 (51.8)
9 (6.4)
1 (0.7)
0.91 (0.63–1.31)
Control subjects, family history‡
141
8 (5.7)
51 (36.2)
65 (46.1)
16 (11.3)
1 (0.7)
1.00 (reference)
Cases, no family history‡
689
41 (6.0)
209 (30.3)
353 (51.2)
83 (12.0)
3 (0.4)
1.26 (1.07–1.49)
Control subjects, no family history‡
689
50 (7.3)
242 (35.1)
329 (47.8)
66 (9.6)
2 (0.3)
1.00 (reference)
Cases, age ⱕ59.8 (Q1)‡
208
10 (4.8)
62 (29.8)
108 (51.9)
27 (13.0)
1 (0.5)
1.67 (1.18–2.37)
Control subjects, age ⱕ59.8 (Q1)‡
208
15 (7.2)
77 (37.0)
99 (47.6)
17 (8.2)
0 (0.0)
1.00 (reference)
Cases, age ⱕ67.7 (Q2)‡
417
21 (5.0)
132 (31.7)
211 (50.6)
50 (12.0)
3 (0.7)
1.30 (1.04–1.61)
Control subjects, age ⱕ67.7 (Q2)‡
417
27 (6.5)
153 (36.7)
193 (46.3)
41 (9.8)
3 (0.7)
1.00 (reference)
Cases, age ⬎67.7 (Q2)‡
416
28 (6.7)
129 (31.0)
215 (51.7)
43 (10.3)
1 (0.2)
1.10 (0.89–1.35)
Control subjects, age ⬎67.7 (Q2)‡
416
31 (7.5)
141 (33.9)
202 (48.6)
42 (10.1)
0 (0.0)
1.00 (reference)
Cases, female
250
10 (4.0)
74 (29.6)
134 (53.6)
29 (11.6)
3 (1.2)
1.24 (0.95–1.61)
Control subjects, female
250
15 (6.0)
85 (34.0)
121 (48.4)
27 (10.8)
2 (0.8)
1.00 (reference)
Cases, male
422
29 (6.9)
135 (32.0)
211 (50.0)
46 (10.9)
1 (0.2)
1.21 (0.99–1.48)
Control subjects, male
422
33 (7.8)
159 (37.7)
189 (44.8)
41 (9.7)
0 (0.0)
1.00 (reference)
— 0.61 — 0.005 — 0.003 — 0.02 — 0.37 — 0.11 — 0.06 —
*Genotype score was calculated by assigning a score of “0” to 259 bp alleles, “1” to 261 bp alleles, and “2” to 263 bp alleles and then summing scores for the two alleles. †Adjusted for sex and age at study, as appropriate. ‡Based on characteristics of Parkinson disease cases. OR ⫽ odds ratio; Q1 ⫽ sample cut by lowest quartile of age at study; Q2 ⫽ sample cut by median age at study.
SNCA REP1 genotype score (OR, 1.18 for each score unit; 95% CI, 1.02–1.37; p ⫽ 0.03). This effect was observed in the sample overall but was stronger for sporadic PD and younger subjects (lowest quartile of age at study, ⱕ59.8 years) (table 2). The percent agreement between the initial and the repeated interview for pesticide exposures overall was 85.5% (kappa ⫽ 0.71; 95% CI, 0.54 – 0.89). A similar agreement was achieved for exposures to pesticides by indication for use (data not shown). Duration of exposure to pesticides overall and by indication for use was recalled with fair to moderate reliability (intraclass correlation range, 0.78–0.83). Occupational exposures were recalled with slightly greater reliability than residential exposures in agreement with results from another recent study.21 More importantly, there was no significant difference in the agreements between cases and control subjects (data not shown). For control subjects, the pesticide exposures frequencies (ever/never) were 6.6%, occupational; 22.2%, residential; and 33.4%, either occupational or residential (accounting for 4.6% of overlapping exposures). Pesticides use (ever/never) was not associated with an increased risk of PD in the sample overall (OR, 1.11; 95% CI, 0.89–1.38; p ⫽ 0.37), but 1464
Neurology 70
April 15, 2008 (Part 2 of 2)
was associated in the younger subjects (lowest quartile of age at study, ⱕ59.8 years; OR, 1.80; 95% CI, 1.12–2.87; p ⫽ 0.01). Similarly, for the subgroups of pesticides categorized by indication for use (herbicides, insecticides, or fungicides), only herbicides were associated with PD and in younger subjects (OR, 2.46; 95% CI, 1.34– 4.52; p ⫽ 0.004; table 3). The frequencies of pesticide exposures and of herbicide exposures considered separately were significantly greater in younger subjects than older subjects (significant trend across quartiles of age, data not shown). Patients with PD were more likely than control subjects to have used pesticides belonging to the chlorophenoxy acid or esters chemical class (OR, 1.52; 95% CI, 1.04 –2.22; p ⫽ 0.004); these chemicals are used as herbicides. In total, our subjects reported exposures to 44 different chemical subclasses of pesticides, but no other chemical subclass of pesticides was significantly associated with PD. The number of chemical subclasses reported was significantly greater in younger subjects than older subjects (test for trend between age at study in quartiles and number of chemical classes reported, p ⫽ 0.0046). Analyses adjusted for education level, smoking, and interview type yielded similar results for
Table 3
Results of case– control analyses for pesticide exposures and risk of Parkinson disease Pesticide use n (%)
Pesticides (ever/never)*
Herbicides (ever/never)*
Neurology 70
Sample or stratum
Subjects n
Ever
Never
OR (95% CI)
p
OR (95% CI)
Cases, all
833
303 (36.4)
530 (63.6)
1.11 (0.89–1.38)
0.37
1.25 (0.94–1.66)
Control subjects, all
833
278 (33.4)
555 (66.6)
1.00 (reference)
Cases, family history†
141
46 (32.6)
95 (67.4)
Control subjects, family history†
141
42 (29.8)
99 (70.2)
Cases, no family history†
689
256 (37.2)
433 (62.8)
Control subjects, no family history†
689
235 (34.1)
454 (65.9)
1.00 (reference)
Cases, age ⱕ59.8 (Q1)†
208
87 (41.8)
121 (58.2)
1.80 (1.12–2.87)
Controls, age ⱕ59.8 (Q1)†
208
62 (29.8)
146 (70.2)
1.00 (reference)
Cases, age ⱕ67.7 (Q2)†
417
157 (37.6)
260 (62.4)
1.03 (0.76–1.40)
1.09 (0.65–1.83) 1.00 (reference) 1.11 (0.87–1.41)
April 15, 2008 (Part 2 of 2)
Control subjects, age ⱕ67.7 (Q2)†
417
148 (35.5)
269 (64.5)
1.00 (reference)
Cases, age ⬎67.7 (Q2)†
416
146 (35.1)
270 (64.9)
1.19 (0.87–1.64)
Control subjects, age ⬎67.7 (Q2)†
416
130 (31.3)
286 (68.8)
1.00 (reference)
Cases, women
250
72 (28.8)
178 (71.2)
1.06 (0.71–1.59)
Control subjects, women
250
69 (27.6)
181 (72.4)
1.00 (reference)
Cases, men
422
168 (39.8)
254 (60.2)
1.01 (0.75–1.37)
Control subjects, men
422
167 (39.6)
255 (60.4)
1.00 (reference)
— 0.74 — 0.40 — 0.01 — 0.84 — 0.28 — 0.76 — 0.94 —
1.00 (reference) 1.18 (0.57–2.44) 1.00 (reference) 1.25 (0.91–1.70) 1.00 (reference) 2.46 (1.34–4.52) 1.00 (reference) 1.54 (1.02–2.32) 1.00 (reference) 1.03 (0.69–1.55) 1.00 (reference) 1.74 (0.86–3.54) 1.00 (reference) 1.04 (0.73–1.49) 1.00 (reference)
Insecticides (ever/never)*
p 0.12 — 0.65 — 0.16 — 0.004 — 0.04 — 0.87 — 0.13 — 0.82 —
*Conditional logistic regression models, adjusted for sex and age at study as appropriate; results for multivariate models were similar (not shown). †Based on characteristics of Parkinson disease cases. OR ⫽ odds ratio; Q1 ⫽ sample cut by lowest quartile of age at study; Q2 ⫽ sample cut by median age at study.
Fungicides (ever/never)*
OR (95% CI)
p
OR (95% CI)
p
0.95 (0.74–1.22)
0.69
0.83 (0.44–1.59)
0.58
1.00 (reference) 1.53 (0.83–2.83) 1.00 (reference) 0.87 (0.66–1.14) 1.00 (reference) 1.04 (0.64–1.70) 1.00 (reference) 0.87 (0.62–1.22) 1.00 (reference) 1.04 (0.72–1.52) 1.00 (reference) 0.89 (0.56–1.43) 1.00 (reference) 0.79 (0.56–1.10) 1.00 (reference)
— 0.18 — 0.32 — 0.87 — 0.43 — 0.83 — 0.64 — 0.16 —
1.00 (reference) 0.68 (0.19–2.37) 1.00 (reference) 0.97 (0.45–2.08) 1.00 (reference) 0.98 (0.30–3.16) 1.00 (reference) 0.62 (0.27–1.43) 1.00 (reference) 1.12 (0.39–3.25) 1.00 (reference) 0.63 (0.21–1.94) 1.00 (reference) 0.76 (0.30–1.90) 1.00 (reference)
— 0.54 — 0.93 — 0.97 — 0.26 — 0.83 — 0.42 — 0.56 —
1465
Table 4
Multivariate analyses for SNCA REP1 genotype score, pesticide exposures, and risk of Parkinson disease Susceptibility*
Model‡
Terms
Age at onset† p
HR, 95% CI
p
.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.55
OR, 95% CI
Univariate analyses 1
Rep1 score
2
Pesticides
1.11 (0.89–1.38)
0.37
1.22 (1.05–1.41)
0.008
3
Herbicides
1.25 (0.94–1.66)
0.12
1.29 (1.07–1.55)
0.007
4
Insecticides
0.95 (0.74–1.22)
0.69
1.22 (1.03–1.46)
0.03
5
Fungicides
0.83 (0.44–1.59)
0.58
0.96 (0.59–1.55)
0.86
Rep1 score
1.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.55
Pesticides
1.10 (0.88–1.37)
0.39
1.22 (1.05–1.41)
0.008
Rep1 score
1.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.51
Herbicides
1.24 (0.93–1.65)
0.14
1.29 (1.08–1.56)
0.006
Rep1 score
1.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.59
Insecticides
0.95 (0.74–1.22)
0.67
1.22 (1.02–1.45)
0.03
Rep1 score
1.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.56
Fungicides
0.87 (0.45–1.66)
0.66
0.97 (0.60–1.56)
0.89
Rep1 score
1.18 (1.02–1.37)
0.03
1.03 (0.94–1.12)
0.59
DISCUSSION
Multivariate analyses 6
7
8
9
10
Herbicides
1.28 (0.95–1.72)
0.10
1.25 (1.03–1.51)
0.02
Insecticides
0.91 (0.69–1.19)
0.49
1.19 (0.98–1.44)
0.07
Fungicides
0.92 (0.46–1.84)
0.81
0.83 (0.50–1.37)
0.46
*Adjusted for sex and age at study, as appropriate. †In cases only, adjusted for sex, as appropriate. There were no significant associations of pesticides, herbicides, insecticides, and fungicides (ever/never) with age at study in control subjects (data not shown). ‡Models including pairwise interaction terms were all nonsignificant (not shown). OR ⫽ odds ratio; HR ⫽ hazard ratio.
pesticide exposures overall (data not shown). Similarly, sensitivity analyses restricted to casesibling or to case-unrelated control pairs or to occupational or residential exposures yielded similar results for pesticide exposures overall (data not shown). We constructed several multivariate models, including SNCA REP1 genotype score and ever/ never exposures to pesticides (overall and classified by indication for use). For susceptibility, only SNCA REP1 genotype score was significant in the models (OR, 1.18; 95% CI, 1.02–1.37; p ⫽ 0.03; table 4). For age at onset, pesticides, herbicides, and insecticides (but not SNCA REP1 genotype score) were significant in the models (table 4). The model that was most parsimonious (i.e., all terms in the model significant) and best fitting (i.e., smallest Akaike Information Criterion value) included herbicides only (hazard ratio, 1.29; 95% CI, 1.07–1.55; p ⫽ 0.007; model 3), although the observed effect size was small. Cases who had been exposed to herbicides had a median age at onset of 58.8 years (range, 31.5– 81.9 years), 1466
whereas unexposed cases had a median age at onset of 62.4 years (range, 23.3– 88.0 years). After restricting the susceptibility analyses to the youngest quartile of subjects, several models became significant (table 5). The best-fitting model (model 7) included SNCA REP1 genotype score and herbicides; both SNCA REP1 genotype score (OR, 1.65; 95% CI, 1.16 –2.35) and herbicides (OR, 2.39; 95% CI, 1.29 – 4.41) were significant. However, there was no evidence for a significant pairwise interaction in that model.
Neurology 70
April 15, 2008 (Part 2 of 2)
In younger case– control pairs, we observed an association of pesticide exposures with PD and an association of SNCA REP1 genotype score with PD. The age-dependent findings for pesticides are consistent with secular trends in the number of pesticides in use. In 1939, there were 32 pesticide products registered with the US Department of Agriculture. By 1989, there were 729 active-ingredient pesticide chemicals mixed with other ingredients and formulated into 22,000 commercial products.22,23 In our study, younger subjects were exposed to a greater number of chemical classes of pesticides than older subjects. The age-dependent findings for SNCA REP1 are consistent with previous studies of SNCA promoter variability and PD10 and consistent with the greater heritability of PD in younger subjects.24,25 Although our multivariate analyses of SNCA REP1 score, pesticide exposures, and risk of PD were significant in younger subjects, there were no pairwise interactions. The lack of association of SNCA REP1 score with familial PD was unexpected. However, it is possible that other genetic loci confer susceptibility to PD in families.26 There has been only limited evidence for the interaction of susceptibility genes and pesticide exposures in PD. The first reported interaction was for a polymorphism in the glutathione transferase gene (GSTP1) and PD.27 More recently, an interaction of herbicide exposures and a GSTP1 haplotype on age at onset of PD was reported.28 However, neither of those observations has been independently replicated. Interactions of paraoxonase and dopamine transporter genes and pesticides and PD also await replication.29,30 To date, only an interaction of CYP2D6 gene polymorphisms and pesticides on the risk of PD has been independently replicated.31,32 Information regarding specific subtypes of pesticides (e.g., by indication for use or by chemical class) and their association with PD is limited. We observed an association that was specific for herbicides and for chlorophenoxy acids or esters in par-
Table 5
Multivariate analyses for SNCA REP1 genotype score, pesticide exposures, and risk of Parkinson disease in younger subjects (lowest quartile of age at study ⱕ59.8 years) Susceptibility
Model*
Terms
OR, 95% CI
p†
Univariate analyses 1
Rep1 scored
1.67 (1.18–2.37)
0.003
2
Pesticides
1.80 (1.12–2.87)
0.01
3
Herbicides
2.46 (1.34–4.52)
0.004
4
Insecticides
1.04 (0.64–1.70)
0.87
5
Fungicides
0.98 (0.30–3.16)
0.97
Multivariate analyses 6
7
8
9
10
Rep1 scored
1.73 (1.21–2.48)
0.003
Pesticides
1.90 (1.17–3.06)
0.009
Rep1 scored
1.65 (1.16–2.35)
0.006
Herbicides
2.39 (1.29–4.41)
0.006
Rep1 scored
1.68 (1.18–2.37)
0.004
Insecticides
1.07 (0.65–1.75)
0.80
Rep1 scored
1.67 (1.18–2.37)
0.004
Fungicides
1.09 (0.32–3.70)
0.90
Rep1 scored
1.65 (1.16–2.36)
0.006
Herbicides
2.43 (1.30–4.57)
0.006
Insecticides
0.91 (0.51–1.61)
0.74
Fungicides
1.12 (0.29–4.32)
0.87
*Models including pairwise interaction terms were all nonsignificant (not shown). †Adjusted for sex. OR ⫽ odds ratio.
ticular. The chlorophenoxy herbicide most commonly reported by our subjects was 2,4-D. This agent is contained in several broad leaf herbicide products and was a component of Agent Orange.33 Laboratory experiments have shown that 2,4-D may induce ␣-synuclein fibrillation in vitro34 and may delay ontogeny of dopamine levels in utero in animals.35 Intracerebral administration of 2,4-D in the basal ganglia induces behavioral and neurochemical alterations consistent with other adult animal models of parkinsonism.36 Our findings warrant further studies more specifically exploring the role of 2,4-D exposure in the pathogenesis of PD (particularly in animals or human subjects genetically overexpressing SNCA). Our case–control study has some methodological strengths. The large sample size provided sufficient statistical power to detect the main and joint effects considered. We adjusted our analyses for possible confounders, including sex and age at study, but also smoking, education, and type of control (unaffected sibling or unrelated control subject).
Our study also has some limitations. Pesticide exposures history was recalled by telephone interviews. Nevertheless, consistent with previous studies, we found that our subjects were able to report pesticide exposures reliably particularly at the ever or never level.37 To minimize recall bias, the subjects were kept unaware of the study hypothesis and the interviewers were not aware of the subjects’ case– control status. It is possible that older subjects underreported pesticide exposures as compared with younger subjects, because older subjects, and in particular patients with PD, have a greater frequency of cognitive impairment. However, we screened all subjects for cognitive impairment and used a proxy informant when subjects were incapacitated. Indeed, proxy interviews were conducted more frequently for cases (12.8%) than for control subjects (2.6%). This might have led to a further underestimation of pesticide exposures by cases, in turn concealing a disease association. However, when we adjusted our analyses of pesticide exposures for type of respondent, we obtained similar results. Furthermore, previous studies reported good agreement between PD cases and proxies in reporting pesticide exposures.38 Although exposures to herbicides, and specifically to chlorophenoxy acid or ester herbicides, were reported more frequently by PD cases than control subjects in our study, other types of pesticides (by indication for use or by chemical class) were not reported more frequently by PD cases than control subjects. This observation argues against a sizeable recall bias. It is also reassuring that at least three cohort studies collected pesticide exposure data before the onset of PD and observed an association of pesticides with PD. These studies provide some external support to our findings.5,39,40 Also, the risk factors questionnaire that we used did not allow us to measure the intensity of specific pesticide exposures (neither doses used nor years exposed). We investigated years of pesticides use overall, but we observed no association with PD. This is by contrast to a recent study that showed an association of cumulative lifetime days of pesticide exposures and incident PD within an agricultural cohort.5 Finally, although most of the analyses that we performed were preplanned, our multivariate analyses in younger subjects were performed only in light of our findings. Because of multiple comparisons (and the other limitations noted), our findings are preliminary and await independent replication. Neurology 70
April 15, 2008 (Part 2 of 2)
1467
ACKNOWLEDGMENT
18.
The authors thank Patricia J. Schultz for her assistance in preparing the manuscript for submission.
19.
Received June 23, 2007. Accepted in final form November 1, 2007. REFERENCES 1. Langston JW, Ballard P, Tetrud JW, Irwin I. Chronic parkinsonism in humans due to a product of meperidine-analog synthesis. Science 1983;219:979– 980. 2. Snyder SH, D’Amato RJ. MPTP a neurotoxin relevant to the pathophysiology of Parkinson’s disease. The 1985 George C. Cotzias lecture. Neurology 1986;36: 250–258. 3. Brown TP, Rumsby PC, Capleton AC, Rushton L, Levy LS. Pesticides and Parkinson’s disease—is there a link? Environ Health Perspect 2006;114:156–164. 4. Frigerio R, Sanft KR, Grossardt BR, et al. Chemical exposures and Parkinson’s disease: a population-based case– control study. Mov Disord 2006;21:1688–1692. 5. Kamel F, Tanner C, Umbach D, et al. Pesticide exposure and self-reported Parkinson’s disease in the agricultural health study. Am J Epidemiol 2007;165:364–374. 6. Polymeropoulos MH, Lavedan C, Leroy E, et al. Mutation in the alpha-synuclein gene identified in families with Parkinson’s disease. Science 1997;276:2045–2047. 7. Volles MJ, Lansbury PT Jr . Zeroing in on the pathogenic form of alpha-synuclein and its mechanism of neurotoxicity in Parkinson’s disease. Biochemistry 2003;42:7871–7878. 8. Spillantini MG, Schmidt ML, Lee VM, Trojanowski JQ, Jakes R, Goedert M. Alpha-synuclein in Lewy bodies. Nature 1997;388:839–840. 9. Singleton AB, Farrer M, Johnson J, et al. Alphasynuclein locus triplication causes Parkinson’s disease. Science 2003;302:841. 10. Maraganore DM, de Andrade M, Elbaz A, et al. Collaborative analysis of ␣-synuclein gene promoter variability and Parkinson disease. JAMA 2006;296:661–670. 11. Uversky VN, Li J, Fink AL. Pesticides directly accelerate the rate of ␣-synuclein fibril formation: a possible factor in Parkinson’s disease. FEBS Lett 2001;500:105–108. 12. Manning-Bog AB, McCormack AL, Li J, Uversky VN, Fink AL, Di Monte DA. The herbicide paraquat causes up-regulation and aggregation of ␣-synuclein in mice: paraquat and ␣-synuclein. J Biol Chem 2002;277:1641– 1644. 13. Norris EH, Uryu K, Leight S, Giasson BI, Trojanowski JQ, Lee VM. Pesticide exposure exacerbates alphasynucleinopathy in an A53T transgenic mouse model. Am J Pathol 2007;170:658–666. 14. Bower JH, Maraganore DM, McDonnell SK, Rocca WA. Incidence and distribution of parkinsonism in Olmsted County, Minnesota, 1976 –1990. Neurology 1999;52:1214–1220. 15. Rocca WA, Maraganore DM, McDonnell SK, Schaid DJ. Validation of a telephone questionnaire for Parkinson’s disease. J Clin Epidemiol 1998;51:517–523. 16. Cohen J. A coefficient of agreement for nominal scales. Educ Psychol Measure 1960;20:37–46. 17. Fleiss JL. Statistical Methods for Rates and Proportions. 2nd ed. New York: Wiley, 1981. 1468
Neurology 70
April 15, 2008 (Part 2 of 2)
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull 1979;86:420–428. Bo¨nsch D, Lederer T, Reulbach U, Hothorn T, Kornhuber J, Bleich S. Joint analysis of the NACP-REP1 marker within the ␣-synuclein gene concludes association with alcohol dependence. Hum Mol Genet 2005; 14:967–971. Akaike HA. A new look at the statistical model identification. IEEE Transactions on Automatic Control 1974;19:716 –723. Gartner CE, Battistutta D, Dunne MP, Silburn PA, Mellick GD. Test–retest repeatability of self-reported environmental exposures in Parkinson’s disease cases and healthy controls. Parkinsonism Relat Disord 2005; 11:287–295. Moses M, Johnson ES, Anger WK, et al. Environmental equity and pesticide exposure. Toxicol Ind Health 1993;9:913–959. Maroni M, Fait A. Health effects in man from longterm exposure to pesticides. A review of the 1975–1991 literature. Toxicology 1993;78:1–180. Tanner CM, Ottman R, Goldman SM, et al. Parkinson disease in twins: an etiologic study. JAMA 1999;281: 341–346. McDonnell SK, Schaid DJ, Elbaz A, et al. Complex segregation analysis of Parkinson’s disease: the Mayo Clinic Family Study. Ann Neurol 2006;59:788–795. Farrer MJ. Genetics of Parkinson’s disease: paradigm shifts and future prospects. Nat Rev Genet 2006;7:306– 318. Menegon A, Board PG, Blackburn AC, Mellick GD, Le Couteur DG. Parkinson’s disease, pesticides, and glutathione transferase polymorphisms. Lancet 1998;352: 1344–1346. Wilk JB, Tobin JE, Suchowersky O, et al. Herbicide exposure modifies GSTP1 haplotype association to Parkinson onset age: the GenePD Study. Neurology 2006;67:2206–2210. BenMoyal-Segal L, Soreq H. Gene– environment interactions in sporadic Parkinson’s disease. J Neurochem 2006;97:1740–1755. Kelada SN, Checkoway H, Kardia SL, et al. 5= and 3= region variability in the dopamine transporter gene (SLC6A3), pesticide exposure and Parkinson’s disease risk: a hypothesis-generating study. Hum Mol Genet 2006;15:3055–3062. Elbaz A, Levecque C, Clavel J, et al. CYP2D6 polymorphism, pesticide exposure, and Parkinson’s disease. Ann Neurol 2004;55:430–434. Deng Y, Newman B, Dunne MP, Silburn PA, Mellick GD. Further evidence that interactions between CYP2D6 and pesticide exposure increase risk for Parkinson’s disease. Ann Neurol 2004;55:897. Stellman JM, Stellman SD, Christian R, Weber T, Tomasallo C. The extent and patterns of usage of Agent Orange and other herbicides in Vietnam. Nature 2003; 422:681–687. Uversky VN, Li J, Bower K, Fink AL. Synergistic effects of pesticides and metals on the fibrillation of alpha-synuclein: implications for Parkinson’s disease. Neurotoxicology 2002;23:527–536. Mohammad FK, St. Omer VE. Developing rat brain monoamine levels following in utero exposure to a mix-
36.
37.
ture of 2,4-dichlorophenoxyacetic and 2,4,5 trichlorophenoxyacetic acids. Toxicol Lett 1985;29:215–223. Bortolozzi A, Evangelista de Duffard AM, Dajas F, Duffard R, Silveira R. Intracerebral administration of 2,4-diclorophenoxyacetic acid induces behavioral and neurochemical alterations in the rat brain. Neurotoxicology 2001;22:221–232. Blair A, Tarone R, Sandler D, et al. Reliability of reporting on life-style and agricultural factors by a sample of participants in the Agricultural Health Study from Iowa. Epidemiology 2002;13:94–99.
38.
39.
40.
Wang FL, Semchuk KM, Love EJ. Reliability of environmental and occupational exposure data provided by surrogate respondents in a case– control study of Parkinson’s disease. J Clin Epidemiol 1994;47:797–807. Petrovitch H, Ross GW, Abbott RD, et al. Plantation work and risk of Parkinson’s disease in a populationbased longitudinal study. Arch Neurol 2002;59:1787– 1792. Ascherio A, Chen H, Weisshopf MG, et al. Pesticide exposure and risk for Parkinson’s disease. Ann Neurol 2006;60:197–203.
Help Us Complete the Neurology® Archive! The Editorial Office is seeking the following unbound issues. These are the only missing issues from the comprehensive archive currently posted online at www.neurology.org. We would be unable to return the issues to you because they would have to be de-spined for scanning. If you can help, please contact Kathy Pieper, Deputy Managing Editor, at
[email protected]. January 1, 1955 (Volume 5, Number 1) November 1, 1964 (Volume 14, Number 11) January 1, 1968 (Volume 18, Number 1) October 1, 1979 (Volume 29, Number 10) December 1, 1979 (Volume 29, Number 12) May 1, 1991 (Volume 41, Number 5)
Neurology 70
April 15, 2008 (Part 2 of 2)
1469
