Agricultural exposures and DNA damage in PBMC of female farmers measured using the alkaline comet assay

We visited 410 farms in Calvados (France) over the period 1997–2000, where all willing persons (319 females and 443 males gave informed consent) were included in a cohort by completing a questionnaire and donating blood samples. This cohort was described by Roulland et al. (2004). Ethical approval was obtained from the ethical committee (Comité Consultatif Pour les Personnes Se Prêtant à la Recherche Biomédicale #99-07, Caen, France). The questionnaire contained information regarding lifetime occupational exposure (use of pesticides, tasks undertaken on the farm) as well as socio-demographic information and individual habits. A second survey detailing farm characteristics and agricultural activities was also filled out for each farm (Utilized Agricultural Land, crops and livestock found on the premises, equipment). Information on all female participants is shown in a flow chart (Fig. 1).

From heparinized blood samples, peripheral blood mononuclear cells (PBMCs) were separated by Ficoll Hypaque gradient centrifugation (GE Healthcare) following manufacturer’s instructions. The PBMCs obtained were then suspended in a solution of RPMI1640 (Gibco) with 20% FBS (Fetal Bovine Serum, heat-inactivated in all procedures) (Gibco) and 10% DMSO (Dimethyl sulfoxide) (Euromedex) to ensure long-term viability of cells in liquid nitrogen.

For each experiment, a historical control was integrated to assess the potential intra- and inter-experimental variability. This control sample was constituted of PBMCs from blood drawn from a healthy volunteer unexposed to agriculture. One aliquot of this control was used per experiment to avoid freeze–thaw cycles.

Cells were quickly thawed at 37 °C in a water bath. The aliquot was re-suspended in 6 mL of RPMI1640 with 10% FBS and centrifuged at 200 RCF for 10 min. The pellet was re-suspended in 1 mL of RPMI1640 with 10% FBS. The viability of the cells was tested with Trypan blue. Samples with <90% viability were excluded (3.5% of the 254 samples available, Fig. 1). Cells were brought to a concentration of 120,000 cells/mL in RPMI1640 with 10% FBS and centrifuged at 200 RCF for 10 min.

We then followed the protocol described by Perdry et al. (2018), although with a reduced throughput: in each experiment, we used a Gelbond® (GE Healthcare) sheet holding 10 samples in duplicate and a duplicate of the historical control.

The methodology we used is in accordance with the Minimum Information for Reporting on the Comet Assay (MIRCA) recommendations (Møller et al. 2020).

We acquired a virtual image of each slide using a fluorescence scanner (VS120, Virtual Slide Microscope, Olympus Life Science).

For each gel deposit, 100 randomly selected cells were visually analyzed (200 cells per sample). They were graded into four categories according to a visual scoring system: undamaged, lightly damaged, moderately damaged and highly damaged cells. We assigned a rank number ranging from 0 to 3 to each category.

Some studies state that highly damaged cells (HDCs) are not all apoptotic cells (Lorenzo et al. 2013), while others indicate that HDCs are rather dying (Collins et al. 2008). Therefore, at least a fraction of HDCs would be apoptotic cells, which are not segregated by the comet assay. We thus ran our analyses first without HDCs to exclude DNA damage resulting from apoptosis in dying cells.

The DNA damage score per deposit was expressed as Eq. 1. Both deposit scores were then added together and normalized from the standardization control of their experiment.

In the sensitivity analysis, the comet damage score included HDCs as per Eq. 2.

All scoring was blinded and carried out by a trained researcher.

By analyzing the data from the 41 experiments ran, the mean damage score was of 8.77 with a SD of 4.95 (range 2.0–21.64) (Supplementary Table S1). The variability of damage scores does not show a trend over time (Fig. 1), and with the exception of 3 experiments falls within 2 SD, which is acceptable for comet assay (De Boeck et al. 2000). The removal of the 3 experiments does not change our results (data not shown). Two people performed the experiments, and no deviation in scores appear between these (Fig. 1). We normalized the comet scores of the samples by dividing the comet score by the score of its nested control.

The study population was described by mean and standard deviation when the data was normally distributed. For other continuous variables, median and IQR were expressed. Categorical variables were described by proportion. For categorical variables, χ² or Fisher exact tests were used when appropriate. For continuous variables, Student t test or Wilcoxon non-parametric test was used when appropriate. Explanatory variables were selected based on the number of observations (excluded if n < 5). Complete case analysis was performed, all selected variables had <5% missing data (bar one at 6%). Comet DNA damage score was log-transformed, which allowed a normal distribution. To test associations between explanatory variables and the log-transformed comet score, univariable linear regression analyses were performed. For occupational and environmental farmland exposures, individuals unexposed to the considered exposure were taken as a reference. For all tested associations between a variable and DNA damage, we considered that a p value below 0.2 was worth being reported. Such associations are termed as “tendencies” and referred to accordingly in the manuscript. Associations with a p value below 0.05 are reported as “associations” and are referred to accordingly in the manuscript. Our approach to analyze our findings beyond the classical view of statistical significance is in accordance with recommendations (Amrhein et al. 2019). For multivariable analysis, all variables associated with DNA damage with a p value below 0.20 were included in a first multivariable model. A backwards selection was performed to obtain a final multivariable model. Sensitivity analyses were run by introducing the HDCs. The statistical analysis was generated using SAS software, Version 9.4 of the SAS System for Windows.

Our sample selection is described with the flow chart in Fig. 2. Of the 319 females included, 261 gave blood samples, of which 254 could be used for the comet assay. Nine samples were excluded due to insufficient cell viability. The results presented in this study rely on 245 female samples.

Mean age of the 245 females analyzed was 46 years old at enrollment (ranging from 19 to 73 years), 58% of them were of normal BMI (18.5–25 kg/m²). The majority of the females were never smokers (80%) and 45% consumed alcohol occasionally (Table 1).

There was no variation in DNA damage levels with age, BMI, or alcohol consumption. Smoking at the time of sampling was associated with a decrease in DNA damage compared with no smoking history (p = 0.01). Former smoking revealed a tendency of decreased DNA damage compared with no smoking history (p = 0.09). We observed a trend of decreased DNA damage in association with smoking and former smoking (p = 0.01). A tendency of decreased DNA damage was found with an increased quantity of cigarettes smoked per day (p = 0.09) (Table 1), only when current smokers are compared to ex- and never-smokers together.

Participants undertaking tasks associated with crops were lowly represented, where the most implication was for harvest work with 34 women (14%). The use of herbicides on meadows was not associated with differences in DNA damage levels (p = 0.68). However, a longer duration of use of herbicides on meadows for people undertaking this task show a tendency of higher DNA damage (n = 5, p = 0.05). Cleaning and upkeep of agricultural equipment was not associated with modified DNA damage levels (p = 0.65), but a longer duration of practicing this task showed a tendency to have higher DNA damage levels (n = 18, p = 0.06). Regarding tasks related to livestock, people undertaking milking tended to have higher DNA damage levels (p = 0.16) (Table 2).

Farm characteristics were assessed by a separate questionnaire which was filled in by the farm owner, describing what is present on the farm regardless of individual activities, therefore assessing the general working and living environment participants are exposed to.

Regarding specific crops, no modification of DNA damage levels was found in people working on farms where rape, meadows or wheat crops are present (rape p = 0.64, meadows p = 0.93 and wheat p = 0.56 respectively). However, an increased area of rape and meadows were associated with reduced levels of DNA damage (p = 0.04 and p = 0.03 respectively), there was also a tendency of reduced DNA damage levels with increased area of wheat crops (p = 0.07).

No association was visible with the presence of orchards (p = 0.98), but DNA damage levels tended to be higher with increasing orchard size (p = 0.19).

The presence of flax and pea crops on the farm led to a tendency to lower DNA damage (p = 0.18 and p = 0.15). However, size of these crops was not associated with modified DNA damage levels (p = 0.34 for flax and p = 0.33 for pea).

Concerning livestock, the presence of sheep, cattle, and goats was not associated with modification of DNA damage levels (sheep p = 0.62, cattle p = 0.82, goats p = 0.53). However, increasing number of sheep on the farm was associated with reduced DNA damage levels (p = 0.04), and there was also a tendency to reduced DNA damage levels with the number of cattle and goats (p = 0.17 and p = 0.06, respectively) (Table 3).

Having pigs present on farms was associated with a higher amount of DNA damage (p = 0.04), though the number of pigs was not associated with modified DNA damage levels (p = 0.95).

There was a tendency to lower DNA damage when poultry were present on the farm (p = 0.15), but without link with the number of animals (p = 0.56).

We ran a multivariate analysis with the parameters with which an association or tendency with modified DNA damage levels, either increase of decrease, was detected. Current and former smoking remained associated with lower DNA damage levels (p = 0.0008 with β = −0.80 and p = 0.03 with β = −0.48 respectively), as well as the trend associating smoking with decreased DNA damage (p = 0.0009).

No direct occupational exposure remained associated with DNA damage in the multivariate analysis.

Regarding exposures to the farming environment, pig farming was associated with higher DNA damage, (p = 0.01) though the presence of poultry farming and the bigger the surface of meadows were associated with lower DNA damage (p = 0.003 and p = 0.04, respectively) (Table 4).

We replicated our analysis with HDCs included in the comet score.

Concerning occupational exposure, all associations/tendencies remained, although the p values were numerically changed, as could be expected (supplementary table S3).

Concerning farm characteristics, the tendency of lower DNA damage with the presence of flax crops on the farm remained (p = 0.16), while it was lost with peas crops. The utilized agricultural area was no longer associated with decreased DNA damage. The tendency of decreased DNA damage with increasing wheat surface disappeared. Increasing size of rape crops and meadows continued to be associated with lower DNA damage, though only as a tendency for meadows (p = 0.04 and p = 0.14 respectively). The tendency of higher DNA damage with the presence of orchards remained (p = 0.19), as it was with the presence of pigs (p = 0.12), although as a tendency. The presence of poultry however lost its association with DNA damage. Regarding the number of sheep and goats, only the number of sheep continued to be associated with lower DNA damage (p = 0.04) (supplementary data table S4).

Regarding occupational pesticide exposure, farm workers have an overall increased level of DNA damage compared to unexposed populations (Nascimento et al. 2022) although it depends on the molecules they are exposed to (Lebailly et al. 1998a, 2003).

Studies regarding female agricultural professional exposure to pesticide are scarce likely because women are rarely direct users of pesticides on crops (Dahiri et al. 2021). In our cohort, the low number of women involved in pesticide spraying prevented us to assess its consequences. The repertoire of tasks reported in our study is more detailed than usually reported, and we could report the absence of effect on DNA damage of seed treatment, and the tendency of having increased DNA damage the longer the duration of use of herbicide on meadows, although this was lost in multivariate analysis. However, with n = 5, our findings would need to replication in larger cohorts.

There are however several tasks involving the use of potentially harmful chemicals to which more women were exposed, with no association with DNA damage: use of antiparasitic or insecticide on livestock (n = 138) or of rodent poison (n = 92); use of herbicide on courtyards (n = 113) or on embankments (n = 32); disinfection of milking equipment (n = 117), of premises and facilities (n = 94) or the cleaning and upkeep of agricultural equipment (n = 18). For this latter task however, there is a tendency to increased DNA damage for a longer duration of practice, which would need further attention in additional studies, although this association is lost in multivariate analysis. Some of these tasks have been associated to an increased risk of cancer in other studies, such as the use of pesticide on livestock with multiple myeloma (Tual et al. 2019), or the disinfection of barns with multiple myeloma or sarcoma (Renier et al. 2022), meaning that either women perform these tasks in a way is less harmful than men, or that the associated cancer risk does not rely on an early increase in DNA damage in PBMC.

Regarding interaction with animals, women involved in milking did present a tendency to increased DNA damage (p = 0.16). Although this disappeared in multivariate analysis, it would be worth exploring further this relationship since milking has been associated with an increased risk of sarcoma (Renier et al. 2022), and because this task is involving a large number of women.

Other associations between DNA damage and exposure to animals were present, although in an indirect fashion. The presence of swine on the farm was associated with increased DNA damage, though the presence of poultry was associated with less. To our knowledge, literature shows no previous studies regarding associations between exposure to these animals and DNA damage. However, poultry (Beane Freeman et al. 2012) and pig (Hofmann et al. 2018) farming were associated with a decreased risk of lung cancer in the AHS cohort, whereas such associations were not present in the AGRICAN cohort (Tual et al. 2017). Additionally, raising poultry was associated with a greater risk of colorectal cancer (CRC) in AHS (Beane Freeman et al. 2012) and AGRICAN (Talibov et al. 2022), and pig farming with an increase in meningioma (Piel et al. 2017) and CRC risk (Talibov et al. 2022) in AGRICAN, although this last association does not last when only women are considered.

A larger surface of meadows on the farm was associated with less DNA damage, our study is the first to report such association. A study reported an increased risk of glioma for workers performing tasks associated with meadows (Piel et al. 2017), but the association we report is associated the presence of meadows, not performing tasks on this crop.

At this stage, it would be arbitrary to infer any causal relationship between the alterations in DNA damage we observe and a specific cancer risk. Moreover, our study, unlike most others, is reporting results specific for women, who can experience cancer risks different from men for similarly reported exposures, such as for CRC and pig farming (Talibov et al. 2022), or for bladder cancer where the risk reported in women is higher than in men for field-grown vegetables workers (Boulanger et al. 2017). If anything, observed modifications in the level of DNA damage associated with exposure to meadows, poultry and pig farming, underline that these exposures do impact women biology at the level of DNA damage in circulating PBMC, and hence alter genome stability. This warrant further studies on female cohorts to better characterize their exposure, by questionnaire and/or biomarkers of exposure. The study of additional markers linked with genome stability and DNA damage, such as micronuclei or levels of DNA methylation, could help better characterize the biological consequences of these exposures. Eventually, the ability to assess cancer risk in women farmers specifically would also be desirable.

Smokers were found to have less DNA damage than non-smokers in our analysis. While smoking has been associated with increased DNA damage in some comet-based studies (Hoffmann et al. 2005), some others did not, including in a study analyzing data from more than 19,000 individuals (Milić et al. 2021). It was however underlined that most studies did not report precisely enough the smoking status or intensity of participants. Interestingly, another study reported a decrease in DNA damage, measured by micronuclei, for smokers and former smokers (Bonassi et al. 2003).

A previous study did mention an increase in DNA damage in male smoking farmers (Lebailly et al. 1998b). It is thus possible that either different tasks performed on farms and/or sex are influencing the DNA damage to PBMC induced by smoking.

Some studies suggested that females repair DNA damage less efficiently than males (Wei et al. 2000), while others did find no difference (Soares et al. 2014); thus preventing us to attribute an effect of sex on our association.

However, our observation of a lesser extent of DNA damage decrease in former smokers compared to smokers echoes with a recent study exploring the changes in DNA methylation induced by smoking in blood samples from 745 women, where smoking-associated changes were slowly returning to normal after smoking cessation over years (Guida et al. 2015). Because PBMCs are short-lived, former smoking cannot affect them directly. It is however possible that smoking may have affected the repair capacities of progenitor cells for PBMC at the epigenetic level, which would therefore be passed on to PBMC daughter cells, even after smoking cessation.

Further studies exploring DNA repair capacities and epigenetic status in women with regard to smoking would be warranted to explore these aspects.