Our population-based case–control study of PD conducted in a California population heavily exposed to pesticides replicates and extends evidence for an association between DAT
variants and PD (Kelada et al. 2006
) and highlights possible interactions of disease-associated DAT
susceptibility alleles and pesticide exposure (–).
All eight major DAT
5′ region haplotypes are part of two evolutionary clades (clades A and B) (Kelada et al. 2006
). In vitro
study of the six common haplotypes indicated that the two most prevalent haplotypes of clade A have 40–50% higher luciferase activity compared with the two most prevalent haplotypes of clade B (Kelada 2005
). A caveat is that in vitro
gene reporter assays are often dependent on cell type, transcription factor, and chromatin context. In contrast, direct in vivo
human imaging of DAT binding and gene expression studies in postmortem brains suggest that B clade haplotypes are associated with higher DAT levels (Drgon et al. 2006
). Conflicting functional results are difficult to reconcile, but, in agreement with Kelada et al. (2006)
, we subscribe to the view that in vivo
measures are more likely to be representative of the true physiologic picture. Thus, our data suggest that lower DAT function/levels due to clade A may increase susceptibility to PD and, by extension, that DAT levels affect PD risk only in those subjects who are pesticide exposed.
Thus, the combination of these genetic associations and previous in vivo
functional observations appears to contradict the long-held belief that DAT provides a gateway for MPP+
, paraquat, and maneb and thus potentiates their neurotoxic effects in dopaminergic neurons (Edwards 1993
). To date, there remains no conclusive evidence that pesticides enter dopaminergic neurons via DAT; in fact, a recent study showed DAT to be an unlikely transporter for paraquat (Richardson et al. 2005
). Indeed, the selective vulnerability underlying loss of dopaminergic neurons remains enigmatic. Placing the observed and now confirmed gene–pesticide interaction association in biologic context likely requires further understanding of the roles of DAT, paraquat and maneb, and the toxic mechanisms they exert on neurons. Paraquat’s toxic action is often attributed to reduction–oxidation cycling that generates reactive oxygen species (Przedborski and Ischiropoulos 2005
). For maneb, the neurotoxic mechanism may be mediated by ubiquitin-proteasome system inhibition (Wang et al. 2006
; Zhou et al. 2004
Most important in our study, risk of PD seems to depend on whether subjects are exposed to pesticides. We observed little indication that DAT
susceptibility allele(s) affect risk in those unexposed to agriculturally applied maneb and paraquat or occupationally (albeit self-reported) exposure to any type of pesticide. For occupationally exposed males, we estimated an almost 3-fold increase in risk for those carrying two or more susceptibility alleles and a 2-fold increase in risk for those with only one allele, compared with those not carrying DAT
susceptibility alleles. Our results thus replicate a strong gene–pesticide interaction (> 5-fold risk increase; ) previously reported for occupationally pesticide-exposed males (Kelada et al. 2006
). Moreover, we employed our GIS-derived, record-based residential pesticide exposure estimates for maneb and paraquat and found that highly exposed subjects with one DAT
susceptibility allele have an estimated 3-fold increase, and subjects with two and more alleles a 4.5-fold increase, in risk of PD compared with those with no DAT
susceptibility alleles. There was little or no indication of a DAT
susceptibility allele association in subjects with no or low residential pesticide exposure as estimated by our GIS model ().
A limitation of our study is the relatively small sample size for some strata of our gene–environment interaction analysis. This may affect the informativeness of the data and the interpretability of results. In research of rare diseases, such as PD, sample size is always an issue for gene–environment interaction studies. However, collaborations and data pooling efforts to increase sample size are usually limited by the need to arrive at comparable and valid measures of exposures, in addition to identifying a large enough exposed population in each study.
A primary strength of our study is the estimation of residential pesticide exposure using a GIS-based computer model and not subjects’ self-reports; therefore, our residential estimates are unlikely to be biased by differential recall. Some nondifferential exposure misclassification possibly attenuates effect estimates, and we encountered some missing or incomplete address information and geocoding problems in our GIS approach. Residential pesticide exposure will also depend on wind patterns at the time of application, open windows, and the likelihood of tracking dust and pesticide residues into homes. There is no obvious reason why subjects would have participated in our study based on a history of living near agricultural plots, and most rural residents might not know what was applied on fields near their homes in the past decades. It is even less likely that subjects would be able to self-select themselves according to both genotype and pesticide exposure. A strength of our study is that, in contrast to most previous occupational and environmental epidemiologic studies of PD, all of our diagnoses were clinically confirmed by one or more examinations by UCLA movement disorder specialists, so disease misclassification is likely to be minimal.
Our residential pesticide estimates are unique in the field of epidemiologic exposure assessment, and our rationale for their use in this study is strong. In this region of California, our two pesticides of particular interest, maneb and paraquat, are both applied on common crops such as potatoes, dry beans, and tomatoes, and both survive in the soil for > 30 days (Oregon State University 1996a, 1996b
; U.S. EPA. 2005
). Pesticide drift can expose rural residents to pesticides without direct occupational contact. For example, measurable concentrations have been detected in the air, in plants, and in animals away from application sites (Chester and Ward 1984
; Currier et al. 1982
; MacCollom et al. 1986
). Outdoor and indoor air concentrations for agriculturally applied pesticides correlate with each other and also correlate with distance to the application sites (Kawahara et al. 2005
). Our GIS-derived method of residential exposure estimation has been validated using values determined for organochlorines and measurements of serum biomarkers for dichlorodiphenyldichloroethylene, a metabolite of dichlorodiphenyltrichloroethane (Ritz and Costello 2006
Direct evidence of any particular pesticide compound contributing to PD in humans is lacking (Brown et al. 2006
). This is partially due to a dearth of exposure assessment tools that accurately document past and long-term pesticide exposures in humans. The relatively small effect size any single environmental toxin may exhibit, and the necessity for large sample sizes that allow for an efficient investigation of gene–environment interactions among vulnerable subgroups, may have further hindered progress in this area. The availability of historical pesticide application data for California allowed us to develop a GIS-based method of assessing exposures to pesticides for residents of the highly agricultural California Central Valley. Our study is unique in that a record-based rather than recall-based assessment of historical residential pesticide exposures was possible.