Lifetime Pesticide Use and Telomere Shortening among Male Pesticide Applicators in the Agricultural Health Study

Background: Telomere length (TL) in surrogate tissues may be influenced by environmental exposures. Objective: We aimed to determine whether lifetime pesticides use is associated with buccal cell TL. Methods: We examined buccal cell TL in relation to lifetime use of 48 pesticides for 1,234 cancer-free white male pesticide applicators in the Agricultural Health Study (AHS), a prospective cohort study of 57,310 licensed pesticide applicators. Participants provided detailed information on lifetime use of 50 pesticides at enrollment (1993–1997). Buccal cells were collected from 1999 to 2006. Relative telomere length (RTL) was measured using quantitative real-time polymerase chain reaction. We used linear regression modeling to evaluate the associations between specific pesticides and the logarithm of RTL, adjusting for age at buccal cell collection, state of residence, applicator license type, chewing tobacco use, and total lifetime days of all pesticide use. Results: The mean RTL for participants decreased significantly in association with increased lifetime days of pesticide use for alachlor (p = 0.002), 2,4-dichlorophenoxyacetic acid (2,4-D; p = 0.004), metolachlor (p = 0.01), trifluralin (p = 0.05), permethrin (for animal application) (p = 0.02), and toxaphene (p = 0.04). A similar pattern of RTL shortening was observed with the metric lifetime intensity-weighted days of pesticide use. For dichlorodiphenyltrichloroethane (DDT), we observed significant RTL shortening for lifetime intensity-weighted days (p = 0.04), but not for lifetime days of DDT use (p = 0.08). No significant RTL lengthening was observed for any pesticide. Conclusion: Seven pesticides previously associated with cancer risk in the epidemiologic literature were inversely associated with RTL in buccal cell DNA among cancer-free pesticide applicators. Replication of these findings is needed because we cannot rule out chance or fully rule out bias.


Introduction
Pesticides are widely used in the United States and worldwide and are pervasive in our environment. Several pesti cides have been associated with various cancers in experi mental studies [summarized by the U.S. Environmental Protection Agency (U.S. EPA 2012)] and in epidemiologic studies of farmers (Blair and Freeman 2009) and pesti cide manufacturing workers Fryzek et al. 1997;Kogevinas et al. 1997;Leet et al. 1996) and among register pesti cide applicators, including the Agricultural Health Study (AHS) cohort, one of the largest prospective studies of pesti cide applicators (Alavanja et al. 1996).
The mechanisms by which pesti cides may be linked to cancers in humans are unclear. Potential mechanisms include oxidative stress, DNA damage, chromosome aberra tion, immune response abnormality, and chronic inflammation (Casale et al. 1998;Hooghe et al. 2000;StillerWinkler et al. 1999;Undeger and Basaran 2005). These biological processes are also involved in telo mere shortening (von Zglinicki 2002). The genetic integrity of the genome is main tained, in part, by the architecture of telo meres (Artandi et al. 2000;Meeker et al. 2004). Telomeres typically shorten with each cell division. When telomeres reach a criti cal length, cellular apoptosis or senescence is triggered. Cancer cells bypass these pathways and continue to divide despite the presence of chromosomal instability (Blasco 2005). Human epidemiologic investigations, mostly from case-control studies, have suggested that telomere length (TL) in surrogate tissues (i.e., blood or buccal cells) is associated with some but not all cancers (Broberg et al. 2005;Hou et al. 2009;Jang et al. 2008;McGrath et al. 2007;Risques et al. 2007;Shao et al. 2007;Widmann et al. 2007). While most studies have reported that shorter telomeres in surrogate tissue are positively associated with cancer [as reviewed by Ma et al. (2011) and Wentzensen et al. (2011)], longer telo meres have been associated with cancer in some studies and reviews (Gramatges et al. 2010;Han et al. 2009;Lan et al. 2009;Shen et al. 2011;Svenson et al. 2009).
The potential effects of pesti cide exposure on TL in surrogate tissues have not been well characterized, although associations with telo mere shortening have been reported for occu pational exposures (Eshkoor et al. 2011) and persistent organic pollutants (Shin et al. 2010). In the present study, we examined whether lifetime use of 48 pesti cides is associated with telomere shortening in buccal cell DNA from 1,234 white male cancerfree licensed pesti cide applicators participating in the AHS.

Study population.
A detailed description of the AHS has been published (Alavanja et al. 1996). Briefly, 57,310, or 82%, of pesti cide applicators seeking pesti cide licensing in the U.S. states of Iowa and North Carolina were enrolled between 13 December 1993 and 31 December 1997. All data used in these analyses were based on AHS data release P1REL0506.01. Enrolled participants were licensed private pesti cide applicators (mostly farmers) residing in Iowa and North Carolina volume 121 | number 8 | August 2013 • Environmental Health Perspectives and commercial applicators residing in Iowa. All pesti cide applicators completed an enroll ment questionnaire that inquired about ever/ never use of 50 pesti cides (National Institutes of Health 2013) as well as the duration (years) and frequency (average days per year) of use for 22 of these pesti cides. In addi tion, 44% of the participants completed a second takehome questionnaire at enroll ment that inquired about the duration and frequency of use of the remaining 28 pesti cides. Approximately 5 years after enrollment, participants completed a followup phone interview to collect additional pesti cide use and medical history information and were asked to provide a mouthwash rinse sample for extraction of DNA from buccal cells. Participants who agreed to provide buccal cells were sent a buccal cell collection kit and a postagepaid, padded envelope. Informed consent for buccal cell collection and buccal cell analyses associated with potential carcino genic risk assessment was obtained at the time of collection, and the study protocol and informed consent was reviewed by all relevant institutional review boards. A total of 36,342 (63%) participants completed the followup interview, and 20,421 (56%) returned buc cal cells. No meaningful difference in demo graphics was observed between those who donated buccal cells and those who did not .
As part of a nested case-control study of prostate cancer, 1,372 participants who had no history of prostate cancer (controls) and were > 40 years of age at the time of buccal cell collection were initially selected for TL measurement. Of these, 1,234 met inclu sion criteria for the TL analysis. Participants excluded from the analysis were 115 men who had been diagnosed with any cancer prior to or within 3 years of buccal cell col lection, 14 men who donated a buccal cell sample but did not provide a completed written consent at the time of the present analysis, and 9 men who were nonwhite. The present analysis was limited to white applicators because of the small numbers of nonwhite applicators in the AHS cohort (Alavanja et al. 1996).
Pesticide exposure. Two pesti cide expo sure metrics were used. Pesticide use was evalu ated as both lifetime days of pesti cide use (years of use × days per year) and lifetime intensity weighted days of pesti cide use (lifetime expo sure days × intensity score). The intensity score was computed from an algorithm that took into account exposure modifying factors such as application method and protective equip ment use (Coble et al. 2011).
Buccal cell collection. Buccal cells were collected from 1999 through 2006 using a mouthwash "swish and spit" collection tech nique (GarcíaClosas et al. 2001). Buccal cells collection vials were returned to the National Cancer Institute (NCI) repository via a postagepaid, padded envelope. All samples were stored at the NCI repository at -80°C. DNA from buccal cells was extracted using the Wizard® Genomic DNA Purification Kit (Promega Corp., Madison, WI, USA).
TL measurements. Relative telomere length (RTL) in buccal DNA was mea sured at the Laboratory of Environmental Epigenetics, Center of Molecular and Genetic Epidemiology, Milan University, Italy, using quantitative realtime polymerase chain reac tion (RTqPCR) as described previously (Cawthon 2002). Briefly, this method mea sures RTL in genomic DNA by determining the ratio of telomere repeat copy number (T) to single copy gene (S) (36B4 gene located on chromosome 12, which encodes acidic ribosomal phosphoprotein P0) copy num ber (T/S ratio) in individual samples relative to a reference pooled DNA (Boulay et al. 1999). The reference pooled DNA was cre ated using samples from 60 participants ran domly selected from the population sample selected for this study and was used to gen erate a fresh standard curve, ranging from 0.25 to 8 ng/µL in every T and S RTqPCR run [see Supplemental Material, Figure S1 (http://dx.doi.org/10.1289/ehp.1206432)]. All samples were successfully run in dupli cate with a 100% completion rate. The inter batch variability [coefficient of variation (CV)] in this study was 8.1%. The primer sequences and concentrations were GGTT TTTG AGGG TGAG GGTG AGGG TGAG GGT (270 nM) and TCCC GACT ATCC CTAT CCCT ATCC CTAT CCCT ATCC CTA (900 nM) for telomere; and CAGC AAGT GGGA AGGT GTAA TCC (300 nM) and CCCA TTCT ATCA TCAA CGGG TACA A (500 nM) for human betaglobin.
The T (telomere) RTqPCR mix was iQ SYBR Green Supermix (BioRad, Hercules, California, USA) 1×, tel1b 100 nM, tel2b 900 nM, DMSO 1%, EDTA 1×. The S (human betaglobin) RTqPCR mix was iQ SYBR Green Supermix (BioRad) 1×, hbg1 300 nM, hbg2 700 nM, DMSO 1%, DTT 2,5 mM, EDTA 1×. We used the RTqPCR primer sets previously described by McGrath et al. (2007). We used pooled DNA from 20 referents (500 ng for each sample), randomly selected from samples of this same study, to create a fresh standard curve, ranging from 8 ng/µL to 0.5 ng/µL, at every T and S RTqPCR run. All samples contained Eschericheria coli DNA heated at 96°C × 10 min and cooled at room tem perature. 15 ng of DNA samples was added to each reaction (final volume, 20 µL). All RTqPCRs were performed on a DNA Engine thermal cycler Chromo4 (BioRad). The ther mal cycling profile for both amplicons started with a 95°C incubation for 3 min to activate the hotstart iTaq DNA polymerase. The T RTqPCR continued with 25 cycles at 95°C for 15 sec, and anneal/extend at 54°C for 49 sec. The S RTqPCR continued with 35 cycles at 95°C for 15 sec, anneal at 58°C for 1s, extend at 72°C for 15 sec. At the end of each reaction, a melting curve was used for both T and S RTqPCRs. All samples were run in triplicates.
Statistical analysis. The means of all three RTL measurements were used in the statisti cal analyses. Mean RTL values reported in Table 1 are the arithmetic mean stratum speci fic estimates of RTL ± SD. Because RTL and both pesti cide exposure metrics (i.e., lifetimedays and lifetime intensityweighted days) had rightskewed frequency distribu tions, a natural logarithm transformation was applied to RTL and to both exposure metrics. Linear regression models were used to estimate the change in RTL with increas ing pesti cide use on a continuous scale (i.e., lifetimedays and lifetime intensityweighted days). For presenta tion purposes, we also cal culated the mean RTL for the reference group (i.e., no use of the pesti cide) and each tertile of lifetime intensityweighted days of pesti cide use; however, the continuous measure was used in the analysis for testing statisti cal significance. Two pesti cides (trichlorofon and ziram) had small numbers of exposed participants (n < 20) and were dropped from the analyses. pValues for the linear regres sion coefficient between each pesti cide and RTL, showing the continuous change in RTL with increasing days of use, were adjusted for age at buccal cell sample collection (as a continuous variable), state of residence (Iowa vs. North Carolina), applicator license type (private, commercial), use of chewing tobacco regularly for 6 months or longer (yes vs. no) and total lifetimedays of all pesti cide use (continuous). Further adjusting the linear regression model for body mass index (BMI) (continuous), alcohol consumption (none vs. any; or none, < 3, or ≥ 3 drinks per week), smoking (packyears, current vs. never, ever vs. never), and selfreported cardio vascular disease, diabetes, and high blood pres sure produced comparable results (data not shown), and these factors were not included in our final models. All tests were twosided and p ≤ 0.05 was considered significant. All statistical analyses were conducted using AHS Data Release, version P1REL0506.01, and SAS, version 9.2 (SAS Institute Inc., Cary, NC, USA).

Results
Regression analysis pvalues for model coef ficients of lntransformed pesti cide exposures adjusted for age at buccal cell collection are shown according to selected characteristics in Table 1. RTL was negatively associated with age at buccal cell sample collection (p = 0.003). Commercial pesti cide applica tors had shorter RTL (mean = 1.08) than private applicators (mean = 1.21, p = 0.01). The mean RTL in Iowa applicators (1.19) was significantly shorter than the mean value in North Carolina applicators (mean = 1.24, p = 0.03). The mean RTL was significantly longer for men who used chewing tobacco for ≥ 6 months (mean = 1.27) than men who did not (mean = 1.19, p = 0.01). Selfreported education, BMI, smoking status, packyears smoked, alcohol consumption amount, fam ily history of any cancers, and cardiovascu lar disease, diabetes, and high blood pressure were not significantly associated with RTL. It is worth noting that although RTL among current smokers is not statistically significant (overall p > 0.05), both current smokers and those chewing tobacco had longer telomeres. All study participants applied some pesti cides, but a comparison of those with more than the median number of applications days of total pesti cide use versus those with less than the median number did not show any difference in RTL.
Comparable results were observed with lifetime intensityweighted days of pesti cide use (Table 2) for alachlor (p = 0.005), metola chlor (p = 0.01), permethrin (p = 0.02), 2,4D (p = 0.02), and toxaphene (p = 0.05). RTL also decreased with increasing use of tri fluralin (p = 0.06), but the association was not statistically significant. A statistically signifi cant negative association with RTL was esti mated for intensityweighted lifetimedays of dichlorodiphenyltrichloroethane (DDT) use (p = 0.03), but not for lifetimedays of DDT use (p = 0.08). Adjustment for total days of pesti cide use and for use of chewing tobacco did not affect the associations with pesti cides in any meaningful way.

Discussion
Seven pesti cides were negatively associated with RTL in buccal cell DNA among cancer free pesti cide applicators > 40 years of age. Increasing lifetimedays of use of six pesti cides used in agriculture (alachlor, metolachlor, tri fluralin, 2,4D, permethrin, and toxaphene) were associated with significantly shorter telomeres after controlling for age at buccal cell collection, state of residence, license type, use of chewing tobacco, and total pesti cide use days. Associations were similar between lifetimedays and intensityweighted days of use of alachlor, metolachlor, 2,4D, perme thrin, and toxaphene. RTL also decreased with increasing lifetime intensityweighted days of use of trifluralin, but the association was not statistically significant. For DDT, RTL shortening was significant in association with lifetime intensityweighted days, but not lifetimedays of use.
The only study reporting data on pesti cides and TL found that high levels of expo sure to persistent organic pollutants (POPs), including organochlorine (OC) pesti cides, polychlorinated biphenyls (PCBs), and poly brominated diphenylethers (PBDEs) were associated with decreased RTL in peripheral blood leukocyte DNA, yet low levels of expo sure were associated with increased RTL in an apparently healthy Korean population (Shin et al. 2010). However, because of the limited sample size, and the very small numbers of observations at higher levels of exposure, more studies are needed to confirm the observation.
We observed that use of seven pesti cides, as estimated by one or more exposure metrics, was associated with significantly decreasing RTL. Pesticide use has been noted to cause oxidative stress in humans (Honda et al. 2001;Houben et al. 2008;Meeker et al. 2004;von Zglinicki 2002), and telomere shortening has been associated with cumula tive oxida tive stress (Houben et al. 2008;von Zglinicki 2002). Telomeres are remarkably sensitive to damage by oxidative stress because of the high guanine content in specific telomere sequences and the deficiency in the repair of single strand breaks (Honda et al. 2001;Meeker et al. 2004;von Zglinicki 2002). Pesticide exposure may also lead to telomere shortening by causing inflammation. Figgs et al. (2000) reported that urinary 2,4D concentration was associated with an increased lymphocyte replicative index, a cell proliferation bio marker. The replicative index for lympho cytes was higher among applicators than nonapplicators and higher among applicators after spraying than before spraying. Telomeric DNA is dynamic, and TL typically shortens with each cell division (Hou et al. 2012). Therefore, the increased lymphocyte replica tive index associated with the use of 2,4D could be associated with inflammation because extensive cell proliferation and clonal expan sion is an essential part of an inflammatory response (Hodes et al. 2002).
Of the seven pesti cides associated with telomere shortening, four were herbicides and three were insecticides. The use of these chem icals was not strongly correlated with each other in our sample (the range of correlation coefficients between the seven pesti cides varied from 0.1 to 0.24). These pesti cides belong to different chemical classes and there were no chemical or functional classes where all pesti cides were linked to RTL shortening.
Alachlor has been shown to induce chromo somal aberrations in mouse bone mar row cells (Meisner et al. 1992) and to cause chromosomal damage in in vitro experi mental studies using Chinese hamster ovary cells (Lin et al. 1987), which may be related to TL shortening. In the AHS, a positive association between alachlor and the incidence of lym phohematopoietic cancers was found among applicators (Lee et al. 2004), but no excess cancer risk was observed in a study of alachlor manufacturing workers (Acquavella et al. 2004).
Metolachlor has been associated with lung cancer in the AHS (Alavanja et al. 2004), and the U.S. EPA (1995) classifies metolachlor as a possible human carcinogen (U.S. EPA group C). Trifluralin has been associated with colon cancer in the AHS (Kang et al. 2008), and the U.S. EPA (1996) classifies trifluralin as a possible human carcinogen (U.S. EPA group C). 2,4D has been associated with prostate cancer in a large case-control study in British Columbia (Band et al. 2011), but not in the AHS cohort (Koutros et al. 2013), and with nonHodgkin lymphoma (NHL) in a number of studies (Burns et al. 2011;Hoar et al. 1986;McDuffie et al. 2001;Miligi et al. 2006). The U.S. EPA (2005) classifies 2,4D as not classifiable (U.S. EPA group D) with regard to human carcino genicity, and the International Agency for Research on Cancer (IARC 2001) classifies the chlorphe noxy herbicide group as possibly carcinogenic to humans (IARC group 2B). Permethrin has been associated with multiple myeloma in the AHS (Rusiecki et al. 2009). In laboratory studies, treatment with a high dose of perme thrin induced significant lymphocyte DNA damage in a rat model (Gabbianelli et al. 2004). The IARC (2001) considers DDT and toxaphene possible human carcinogens (IARC group 2B). In a recent Danish study conducted within a prospective cohort, pre diagnostic adipose concentrations of DDT demonstrated a significant positive monotonic dose-response trend with NHL incidence (Brauner et al. 2012).
The present study is unique in that we have a relatively large population of licensed pesti cide applicators who provided reliable information regarding their pesti cide appli cation history (Blair et al. 2002;Coble et al. 2011). We examined the relationship between cumulative lifetime use of specific pesti cides and RTL in a cancerfree popula tion (i.e., the study participants had no can cer diagnoses). Furthermore, we were able to adjust for potential confounding factors related to RTL. In the AHS, a priori-derived algorithm scores that incorporated several exposure determinants were used to predict pesti cide exposure intensity. These algorithm scores have been shown to predict urinary pesti cides levels (Coble et al. 2011;Thomas et al. 2010). The significant decrease in RTL with age that we observed was expected  (Blasco 2005) based on the literature, add ing confidence to our pesti ciderelated find ings. Most pesti cides used by applicators are mixtures of one active ingredient and sev eral "inerts" that are not pesti cides, and there are some products that contain more than one active ingredient. The cooccurrence of active ingredients is an intractable problem for an epidemiologic study assessing a lifelong occupational history of pesti cide use. This problem, however, is less common than the concern we were able to address, which is that over a lifetime applicators may use more than one product for the same or different crops. In this study we controlled for the total pesti cideexposure days and found no mean ingful change in the associations, although confounding by specific pesti cides cannot be ruled out. Although potential confounding from other occupational exposures is possible, the magnitude of bias due to this confound ing was reported to be minimal in the AHS and not likely to be associated with pesti cide exposures (Coble et al. 2002). The fact that other lifestyle factors (e.g., smoking and alco hol drinking) and chronic health conditions (e.g., cardiovascular disease, high blood pres sure) were not significantly associated with TL in our population may be a result of our population being more physically active and older than in some previous studies. Physical activity has been shown to increase TL (Du et al. 2012) and TL seems to be less influ enced by lifestyle factors in older populations (Tiainen et al. 2012). Future studies with increased statistical power will be needed to assess this issue more completely.
In this study, we used buccal cell DNA for TL measurement collected at a single point in time after exposure assessment was made. Buccal cells can be easily and inexpen sively collected and are therefore a con venient resource for evaluating TLrelated disease risks. Future studies may benefit from collecting buccal cells at multiple points in time to assess the rate at which telomeres are shortened. Our results should, however, be interpreted with caution because TL in buccal DNA may be different from that in diseased tissue or blood DNA Wong et al. 2011). We observed that cumula tive lifetime use of some pesti cides is associ ated with telomere shortening in buccal cell DNA, although we cannot rule out the pos sibility of bias due to uncontrolled confound ing, or associations due to chance because of the multiple comparisons that were made. Our findings suggest that specific pesti cides may contribute to telomere shortening and may serve as a mechanism for development of some diseases. Given that this is one of the first studies of pesti cide exposure and TL, we view this work as hypothesis generating rather than hypothesis testing.