Despite the small sample size, our study found indications of a gene-environment interaction between prenatal pesticide exposure and the PON1 Q192R genotype that might affect the risk of obesity and related diseases later in life. Prenatally pesticide-exposed children carrying the R-allele had higher abdominal circumference, body fat content, BMI Z-scores, systolic and diastolic blood pressure, and serum concentrations of leptin and IGF-I at school age than did unexposed children. For QQ-homozygous children, none of these variables were significantly affected by prenatal pesticide exposure. The results indicate that R-carriers may be especially susceptible towards developmental disturbances after prenatal pesticide exposure.
An important strength of this study is the longitudinal design that minimizes exposure misclassification and bias. The classification of the mothers as high, medium or not exposed was done independently by two toxicologists with special expertise in working conditions in greenhouse horticultures at enrolment early in pregnancy 
and hence completely independent of subsequent examination outcomes for the children.
The unexposed group was extended at the follow-up examination to provide better comparison data. None of the mothers of these children had worked in greenhouses or other occupations where pesticides were used during pregnancy. They were distributed evenly between the group of children with the R-allele and the QQ homozygotes, and, therefore, did not affect the differences observed in response to prenatal pesticide exposure between the two genotypes. In addition, all examinations of the children were done blinded to the exposure information and by the same paediatrician using standardized procedures. Genotyping of PON1 and analyses of serum samples were performed blinded to both exposure information and examination outcomes.
The small number of exposed and unexposed children within each of the two PON1 genotypes is a limitation of the study, because it decreases the power of the study and diminishes the possibility of detecting exposure related effects. Nonetheless, if detectable in a small study like ours, the associations may be important. An additional limitation is that we conducted many statistical analyses and chose not to adjust for multiple comparisons, although they may increase the likelihood of spurious associations. However, several of the outcomes are interrelated and the observed associations between prenatal pesticide exposure and effects in the group of children with the R-allele were very robust being significant both before and after adjusting for covariates. In addition, the magnitudes of the effects were related to the exposure level which again supports a causal association. Finally, the distribution of known risk factors such as social class and maternal smoking in pregnancy did not differ between the children with the R-allele and the QQ-homozygotes.
Another possible limitation of the study is that the PON1 genotypes of the mothers were not determined. The maternal genotype and related ability to metabolize pesticides might be important for protection of the foetus against exposure. However, only 11 children (6 unexposed and 5 exposed) were homozygous for the R-allele, and for the heterozygous children approximately half of the mothers would be assumed to be QQ homozygotes. Therefore, it is unlikely that the observed associations would be explained by the maternal genotype.
In a previous study, women who worked in floricultures in Mexico and had the RR genotype had higher risk of having children with low birth weight than women with the QQ or QR genotype 
. In another study, maternal organophosphate exposure during pregnancy was associated with reduced head circumference and head size of the children at birth if the mothers had a low PON1 concentration but no association was seen between neither maternal nor child PON1
genotype and head circumference or birth weight or birth length 
. However, at 12 months of age, negative associations between prenatal organophosphate exposure and cognitive development were seen if the mothers had the QR/RR genotype, although later in childhood children of mothers with the QQ-genotype appeared more affected 
. In a recent study, shorter gestation and smaller head circumference at birth were associated to low infant, but not maternal, PON1 enzyme activity 
. This association was independent of maternal organophosphate exposure in pregnancy but for infants with low activity (and PON1
-108TT and PON1
192QQ genotype) an association between maternal organophosphate exposure in pregnancy and shorter gestational age was seen. Although, the R-allele seems to provide better protection than the Q-allele toward toxic effects of certain organophosphates such as chlorpyrifos 
, the capacity of PON1
allozymes to protect LDL from oxidation might be different, and even the reverse, of the paraoxonase activity, with the lowest capacity for the R-allozyme 
A weak association between the R-allele and increased CVD risk has been confirmed in most case-control studies, and is supported by meta-analyses 
. Results from studies on associations between PON1
R192Q genotype and risk markers for CVD in healthy populations are inconsistent, as some found no association 
or a higher frequency of the R-allele among obese (BMI ≥30 kg/m2
) compared to normal-weight pre-menopausal women 
, while increased blood pressure was associated with the R-allele in women above 60 years 
, and in a rural population in Greece 
. In general, these studies do not include information about environmental exposures, and discrepancies might be due to unidentified gene-environment or gene-gene interactions. In a recent study the combination of low serum HDL concentration and the RR genotype markedly increased the risk of CVD 
. This interaction is also supported by the results from our study where no marked difference in the risk profile between unexposed R-carriers and QQ-homozygotes was seen. In fact, the R-carriers were leaner than the QQ homozygotes at school age. However, the combination of the R-allele and prenatal pesticide exposure had a pronounced effect on the risk profiles.
In our study, pesticide-exposed children with the R-allele had a higher blood pressure and a lower CVRR
, although only the former reached statistical significance. Increased blood pressure was also reported for children, whose mothers were occupationally exposed to pesticides during pregnancy in Ecuador 
. The PON1
genotype was not determined in these studies, but the R-allele frequency is likely to be higher than in our study, since an R-allele frequency of 40 to 60% has been reported for many populations in Latin and South America 
. Due to maturation of the cardiac autonomic activity, CVRR
increases during gestation and early postnatal life followed by a decline 
. Prenatal exposure to the neurotoxicant methyl mercury caused decreased CVRR
in children at 7 and 14 years of age 
Children prenatally exposed to pesticides had lower birth weight than unexposed children as previously reported for the children in this study 
, and also reported in several other studies 
. For children with the R-allele, this was followed by accelerated body fat accumulation until school age. Low birth weight followed by catch-up growth and body fat accumulation during childhood is associated with increased risk of obesity, insulin and leptin resistance and CVD later in life 
. There is increasing evidence that prenatal and early postnatal environmental factors might influence the risk 
. In addition, prenatal exposure to low doses of some endocrine disrupting chemicals has been demonstrated to cause obesity in rodents 
, and epidemiological studies also have linked exposure to some of these chemicals to obesity and metabolic syndrome in humans 
. The mechanisms behind these associations are not fully understood, but might include epigenetic changes that affect gene expression 
and fetal programming of energy balance 
, perhaps including disturbance of the hypothalamic-pituitary-adrenal axis 
. In rats, low doses of chlorpyriphos during development caused excess weight gain and leptin and insulin dysregulation later in life 
and neonatal parathion exposure disrupted the production of adipokines, including leptin, that regulate appetite by communicating metabolic status between adipose tissues and the brain 
. Hence, these effects resemble those seen in the exposed children expressing the R-allele.
In the present study, the mothers were exposed early in their pregnancies to a variety of different pesticides including organophosphates and several fungicides with known endocrine disrupting properties 
. All of the substances had been approved by the national regulatory agency in accordance with European Union legislation. Due to the complex exposure setting, it was not possible to identify specific pesticides as responsible for the effects. They may be due to the combined exposure to several pesticides and maybe also to other ingredients added to the pesticide formulations. For most of the exposed mothers, organophosphates had been used occasionally in the working areas, but only very few mothers had been involved directly in their application. Whether PON1 can detoxify other types of pesticides than organophosphates has, to our knowledge, not been investigated.
In conclusion, this study indicates a gene-environment interaction between prenatal pesticide exposure and PON1
gene heterogeneities that affects cardiovascular risk markers already thought to be associated with the R-allele, as such 
. Individuals with the R allele in the 192 position seem particularly susceptible, and the combination of this genotype and prenatal pesticide exposure significantly stimulated fat accumulation from birth to school age, caused higher blood pressure and enhanced serum concentration of leptin and IGF-I. These effects are all related to increased risk for development of metabolic syndrome and CVD later in life. The results also illustrate that a susceptible subgroup of a population could be affected, although effects may not be evident in the entire population.