We first identified pesticides that have caused neurotoxic effects in humans from the Hazardous Substances Data Bank (HSDB) of the U.S. National Library of Medicine 16. We searched for the terms "pesticide" and "neuro*". From the list of substances obtained in this way, we identified the pesticides, for which neurotoxic effects in humans had been reported. In addition, we searched the U.S. National Institute of Occupational Safety and Health (NIOSH) – Pocket Guide to Chemical Hazards http://www.cdc.gov/Niosh/npg/npgsyn-p.html, using the search terms "pesticide", "insecticide", "herbicide", "fungicide", "fumigant", and "rodenticide" in combination with "central nervous system". The list of neurotoxic pesticides identified in this way was then compared to the current Annex 1 list (as of August, 2008) of pesticides authorised in the European Union according to Plant Protection Directive 91-414-EEC (an Excel data sheet with the status of active substances under EU review can be downloaded from http://ec.europa.eu/food/plant/protection/pesticides/index_en.print.htm). We chose pesticides with an Annex 1 status "in" or "pending" for consideration (Table 1).

For each neurotoxic pesticide in current use, we searched PubMed to identify published data on developmental neurotoxicity. We used pesticide synonyms, commercial names and the CAS number, in combination with each of the terms "neurotoxic", "neurotoxicity", "neurologic", "neurological" and "nervous system", and additional searches included the terms "prenatal", "pregnancy", "fetus", "fetal", "maternal", "developmental" and "child".

Carbamate insecticides, like the OP insecticides, inhibit AChE and elicit cholinergic hyperstimulation. However, carbamates cause only reversible inhibition of AChE 67. Thus, AChE inhibition by carbamates lasts only minutes or hours, whereas the effects of OPs with respect to AChE can last for 3–4 months (reviewed in 32). Because of the transient inhibition of AChE, acute intoxication by carbamates generally resolves within a few hours 67.

When comparing the clinical course of carbamate poisoning (by aldicarb or methomyl) in young children (1–8 years old) and adults (17–41 years old), it was found that the predominant symptoms in children were CNS depression and hypotonia, and the most common muscarinic effect was diarrhoea. In adults the main symptoms were miosis and fasciculations, whereas CNS depression, hypotonia, and diarrhoea were uncommon 68. Symptoms in children poisoned with OPs were found to be similar to symptoms for carbamate poisoning 69. Thus symptoms of carbamate poisoning do not differ markedly from symptoms of OP poisoning in children, but rather the symptoms in children, differ from symptoms in adults.

It is possible that some carbamates may also be involved in oxidative stress 7071. The carbamate, carbofuran, has been observed to accentuate oxidative stress in rat brain by inducing lipid peroxidation and diminishing the antioxidant defence 70.

As for the OPs, it is likely that poisoning with carbamates may result in long term neurological effects 72. Two patients showed cognitive deficit in attention, memory, perceptual, and motor domains 12 months after a poisoning incident 72. With respect to long term, low level exposures to carbamates, reports concerning chronic toxicity are almost non-existent.

No epidemiological studies of developmental neurotoxicity of carbamates in humans could be found, and data from animal experiments are very sparse as well.

Assuming that some of the neurotoxic effects observed in association with prenatal exposure to OPs, such as chlorpyrifos, are due to inhibition of AChE, it is possible that carbamates may have similar developmental effects, even though the inhibition of AChE by carbamates is only transient. Induction of oxidative stress by some carbamates might also cause developmental neurotoxicity. It should also be noted that the carbamate physostigmine inhibits DNA synthesis in undifferentiated neuronotypic PC12 cells (a standard in vitro model for neuronal development). When differentiation was induced, adverse effects on DNA synthesis were intensified, and effects on cell number after prolonged exposure were also worsened by differentiation. Furthermore, differentiating cells displayed signs of oxidative stress, as measured by lipid peroxidation. Finally, the transmitter fate of the cells was shifted away from cholinergic phaenotype toward the catecholaminergic phaenotype. Similar findings were made when incubating the cells with the OPs chlorpyrifos, diazinon and parathion 73.

The pyrethroids are a class of insecticides derived from naturally occurring pyrethrins from the Chrysanthemum genus of plants 74. Pyrethroids contain several common features: an acid moiety, a central ester, and an alcohol moiety. Several stereoisomers exist of each pyrethroid compound, and their effects are stereospecific, indicating presence of specific binding sites (reviewed in 75).

The acute toxicity of pyrethroids is mainly mediated by prolongation of the kinetics of voltage-gated sodium channels, which are responsible for generation of the inward sodium current that produces the action potential in excitable cells. Specific interaction of pyrethroids with the sodium channel slows down both the activation and inactivation properties of the channel, leading to a hyperexcitable state. Although activation is slowed at the single channel level, the density of sodium channels in excitable cells is so high that there are always sufficient unmodified channels to ensure that the activation phase of the action potential is not delayed. However, in the falling phase of the action potential, even a low proportion of modified channels can generate enough extra current to delay inactivation. This delay causes prolonged depolarisation, which, if the current is large enough and lasts long enough for neighbouring unmodified channels to recover excitability, can trigger a second action potential (reviewed in 76).

Two types of pyrethroid structures exist. The type II pyrethroids contain a cyano-group in the α-position, whereas type I pyrethroids do not contain a cyano-group 77. The two types differ with respect to the toxic signs they produce in rats, and with respect to the prolongation time of the sodium current they induce. Type I compounds prolong channel opening just long enough to induce repetitive firing of action potentials (time constants less than 10 msec), whereas type II compounds (time constants of more than 10 msec) hold the channels open for so long that the membrane potential ultimately becomes depolarised to the point at which generation of action potentials is no longer possible (reviewed in 75).

Human pyrethroid poisoning is rare, and almost entirely involves type II pyrethroids. The main adverse effect of dermal exposure to type II pyrethroids is paresthesias, presumably due to hyperactivity of cutaneous sensory nerve fibres. Dizziness, headache and fatigue are common symptoms following ingestion of type II pyrethroids. In severe cases coma and convulsions are the principal life-threatening features 77.

The effects of pyrethroids on the CNS are complex and may also involve antagonism of γ-aminobutyric acid (GABA), modulation of nicotinic cholinergic transmission, enhancement of noradrenalin release, and direct actions on calcium or chloride ion channels. Still, because neurotransmitter-specific pharmacological agents do not protect very well against pyrethroid poisoning, it is unlikely that any one of these effects represents a primary toxic mechanism of action of pyrethroids. More likely, they are secondary to the effects on sodium channels, since most neurotransmitters are released secondary to increased sodium entry (reviewed in 76).

In the few existing accounts of poisonings of adults with pyrethroids, successful recovery after the acute phase of poisoning has been described 7879. However, no detailed neuropsychological testing was applied to these patients, and also no post mortem examinations have been reported, and therefore it is unknown if such poisonings may have lasting effects. Likewise, no information is available on long term effects of low level chronic exposure in humans.

Neonatal rats are 4–17 times more vulnerable to the acute toxicity of pyrethroids (including permethrin (type I), deltamethrin (type II), cypermethrin (type II)) than adult rats 8081. The higher toxicity in neonates is affected by the lower capacity for metabolic detoxification, since neonates and adults have similar brain concentrations at different, but equitoxic, doses 80. However, another study did not observe any age-dependency of the toxicity of the two type I pyrethroids, cismethrin and permethrin 82. It has therefore been argued that age-dependent sensitivity to pyrethroids is only apparent at high acute doses, not at doses below those causing overt toxicity 82.

In addition to the possibility that young animals are more vulnerable to pyrethroids due to lower metabolic detoxification, there is also a possibility that increased vulnerability in young animals may be due to more specific effects of early life exposures. For example, several studies have found that embryonically expressed forms of voltage-gated sodium channels are replaced by adult forms as neurodevelopment proceeds (reviewed in75), and this difference in expression profile may affect the sensitivity towards pyrethroids. In mutation and knockout models of the voltage-gated sodium channels, perturbation of channel function during development impairs nervous system structure and function, underlining the importance of these channels in neurodevelopment. (reviewed in 75).

Also in humans, perturbations of nervous system development have been associated with altered structure and function of voltage-gated sodium channels. Mutations in genes encoding sodium channel subunits have been identified, which result in neuronal hyperexcitability due to subtle changes in channel gating and inactivation 83. Since pyrethroids also alter the activation and inactivation of sodium channels, and thereby the neuronal excitability, it is possible that these may have effects similar to mutations in the sodium channels. However, the mechanisms and magnitude of mutational versus pyrethroid effects are different, and also the duration of effect will differ (pyrethroids have a relatively short half-life, whereas mutations are permanent) 75.

Another possible indication that pyrethroid effects on sodium channels may be relevant to neurodevelopment is the observation that developmental exposure to phenytoin, an anticonvulsant that blocks sodium channels and other ion channels, disrupts nervous system structure and function 84. The use of anticonvulsants during pregnancy has been associated with adverse effects, including microcephaly and intellectual impairment (reviewed in 75). Although differences in doses and in pathogenesis may occur, this evidence would support a concern about the effect of pyrethroids on ion channels.

All existing studies of developmental neurotoxicity of pyrethroids were conducted with rodents as test animals, and although several of them have reported persistent changes in behaviour and/or neurochemistry in the animals, results appear somewhat inconsistent (reviewed in 75). Several studies performed by Eriksson's group 858687 have shown that mice exposed to pyrethroids on postnatal day 10–16 exhibit increased motor activity and a lack of habituation. These mice exhibit changes in density of muscarinic acetylcholine receptor (mAChR) binding for as long as 5 months after cessation of exposure 88. Others have reported persistent changes in behaviour and/or biochemistry, including learning 89, motor activity 90, sexual behaviour 91, mAChR expression 9293, and blood-brain barrier permeability 94. A recent study in rats showed that neonatal exposure to permethrin and cypermethrin caused lasting behavioural effects, changes in monoamine concentrations in the striatum as well as increased oxidative stress 95. In one study, both male and female mice were exposed to the type I pyrethroid, permethrin, before mating, and the following functions were affected in the offspring (with parental exposure to 9.8 mg/kg/day or more for 4 weeks before mating): development of reflexes, swimming ability and open field activity 96.

The potential developmental neurotoxicity of pyrethroids has also been investigated in vitro using cell lines. For example non-toxic concentrations (10^-6 M) of bifenthrin inhibited neurite outgrowth in PC12 cells, indicating that bifenthrin may have deleterious effects on the developing nervous system at concentrations lower than those capable of causing toxicity in the adult brain 97.

Existing data indicate that human exposures to pyrethroids occur and result in detectable concentrations in body fluids 9899100, but there is insufficient information available to adequately evaluate the range of internal doses in humans, and the consequences of these exposures are so far unknown.

Dithiocarbamates are non-cholinesterase inhibiting, sulfur-containing carbamates, which are primarily used as fungicides and herbicides. Four major classes of dithiocarbamates exist; the methyldithiocarbamates, the dimethyldithiocarbamates, the diethyldithiocarbamates (DEDC), and the ethylenebisdithiocarbamates (EBDCs) (reviewed in 101). The dithiocarbamates used as fungicides include metam sodium (methyldithiocarbamate), thiram (dimethyldithiocarbamate/tetramethyldithiocarbamate), and several EBDCs (mancozeb, maneb, metiram, zineb and nabam).

Dithiocarbamates form lipophilic complexes with di- and trivalent metallic cations, bonding through the sulfur atoms 102. They are non-specific in action, and it is difficult to identify a single mechanism for their neurotoxic effects. Because of their metal-chelating capacity and their affinity for sulfhydryl groups, they are biologically highly active 103104. DEDCs are particularly known to modify the cellular redox state by inducing a copper-dependent oxidative stress 105106, and inhibition of cytosolic Cu/Zn superoxidedismutase (SOD1), a key enzyme in the antioxidant response, has been observed in mice treated with DEDC 107. The EBDCs can uncouple the mitochondrial electron transport chain 108109. Mitochondrial dysfunction is often associated with generation of reactive oxygen species (ROS), and ROS production was also found to play a role in mancozeb induced neuronal toxicity in mesencephalic cells, likely via redox cycling with extracellular and intracellular oxidases 110. Further, ethylenethiourea (ETU), which is a degradation product of EBDCs, has been shown to inhibit thyroid peroxidase (TPX), the enzyme that catalyses synthesis of the thyroid hormones 111112. In addition, interference of dithiocarbamates with the vesicular transport of glutamate may play a role in their neurotoxicity 113. Due to the differences in biochemical effects, these compounds seem to exhibit a range of different potencies in regard to developmental neurotoxicity.

Dithiocarbamates are reported to display low acute toxicity in humans and experimental animals 114. Both in humans and laboratory animals, prolonged exposure to dithiocarbamates may cause neurotoxicity. Notably, peripheral neuropathy and extrapyramidal symptoms resembling parkinsonism have been associated with chronic exposure to dithiocarbamate pesticides 115.

As mentioned, chronic exposure of humans to EBDCs has been associated with neurocognitive impairment and parkinsonism 116. In particular, exposure to maneb, which contains manganese, has been linked to development of parkinsonian-like symptoms in agricultural workers 117118. This finding may be related to the inhibition of complex III of the mitochondrial electron transport chain 108, disruption of the glutathione antioxidant system in dopaminergic cells 119, inhibition of proteasomal function and induction of α-synuclein aggregates in dopaminergic cells 120, induction of catechol autooxidation 121, and potentiation of the neurotoxicity of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) in mice 122123124. All of these observations support the notion that maneb may cause parkinsonian-like symptoms. DEDCs, though not methyldithiocarbamate, can also enhance MPTP- induced striatal dopamine depletion in mice 124.

Both thiram and ziram (dimethyldithiocarbamates) can induce apoptotic cell death in PC12 cells, in a dose- and time-dependent manner 125. Both compounds induced rapid and sustained increases of intracellular Ca²⁺ in the cells, which were almost completely blocked by flufenamic acid, an inhibitor of non-selective cation channels. Also, BAPTA-AM, which is an intracellular Ca²⁺ chelator, inhibited the thiram and ziram induced apoptotic cell death, indicating that thiram and ziram induce apoptotic neuronal cell death by Ca²⁺ influx through non-selective cation channels 125.

The EBDCs maneb, mancozeb and metiram can induce malformations in rat foetuses, apparently mediated through formation of the ETU metabolite. The malformations predominantly affect the nervous system and the head, and they correspond to those expected as the result of thyroid insufficiency. They occur only at doses in excess of those that produce significant thyroid inhibition in adult rats, and they have been prevented, at least in part, by co-administration of thyroxin (reviewed in 126). A key concern with thyroid inhibitors is that impaired thyroid function may alter hormone-mediated events during development, thereby possibly leading to permanent alterations in brain morphology and function 127128. Functional deficits are likely to occur during brain development even at mild degrees of hypothyroidism 129. Even withing the normal range, a relatively slight reduction of the concentration of maternal thyroid hormones during pregnancy can lead to intelligence deficits in the children 130. In addition to EBDCs/ETU, many other environmental contaminants have been found to interfere with thyroid function, for example the chlorophenoxy herbicide, 2,4-D (se below). Some of the mechanisms of action with respect to thyroid inhibition are shared by mancozeb/ETU and 2,4-D (including interference with uptake of iodide by the thyroid gland and interference with serum protein-bound iodide level) 131, and exposure to both EBDCs and chlorophenoxy herbicides may therefore result in additive effects.

Evidence that developmental exposure to maneb may be involved in development of Parkinson's disease (PD) later in life includes the finding that postnatal exposure of mice to maneb in combination with paraquat (a classic bipyridyl herbicide, which is no longer authorised in EU) led to a permanent and selective loss of dopaminergic neurons in the substantia nigra pars compacta 132. The postnatal exposure to these pesticides enhanced the effect of the same pesticides administered during adulthood, relative to exposures during development only or adulthood only. Furthermore, exposure to maneb alone during gestation resulted in a dramatic response to paraquat in adulthood, including notable reductions in levels of dopamine and a loss of nigral dopamine neurons. Thus, these results support the notion that a silent neurotoxicity produced by developmental insults can be unmasked by insults later in life 132.

For specific dithiocarbamates, especially the EBDCs maneb and mancozeb, substantial evidence supports the possibility of developmental neurotoxicity. In addition, the likely mechanisms of toxicity for thiram and ziram indicate that these compounds too may be capable of causing developmental neurotoxicity in small doses.

The chlorophenoxy herbicides are widely used for the control of broad-leaved weeds. Structurally, they consist of a simple aliphatic carboxylic acid moiety, which is attached to a chlorine- (or methyl-) substituted aromatic ring by an ether bond. In vivo the salts and esters are rapidly dissociated or hydrolysed, and therefore the toxicity of each chlorophenoxy compound depends principally on the acid form of the pesticide 133. The chlorophenoxy herbicides bind strongly to albumin 134, and binding is favoured by longer acid chains and by more greatly substituted aromatic rings. Therefore the bioavailability and toxicity of the herbicides vary for different herbicides 135. The mechanisms of neurotoxicity of the chlorophenoxy herbicides are incompletely known, but they seem to primarily involve cell membrane damage (reviewed in 135).

2,4-Dichlorophenoxyacetic acid (2,4-D) is the most widely used chlorophenoxy herbicide and also the most widely studied. With respect to membrane damage, it does not cause significant penetration of lipid monolayers in vitro at concentrations below 0,1 μM 134, but at higher concentrations (10–100 μM) it increases bilayer width and causes deep structural perturbations of the hydrophobic region of model membrane systems. At the higher concentrations it also damages human erythrocyte cell membranes 136. This dose- dependent effect on plasma membranes may in part explain the dose-dependent CNS toxicity caused by chlorophenoxy herbicides. In experimental animals (rats, mice and rabbits), only small amounts of herbicide were found in the brain following administration of 100 mg/kg or less 137138139, likely because low concentrations of herbicide have little effect on the plasma membranes comprising the blood-brain barrier. When exposing rats to high doses (250–500 mg/kg) of the herbicide, a reversible selective damage to the blood-brain barrier occurred, and as a result serum albumin and IgG could be detected in the brain along with the herbicide itself 140. The severity of the herbicide-induced cerebrovascular damage in rats has been reported to increase in the order 2,4,5-T (2,4,5-trichlorophenoxyacetic acid) < MCPA (4-chloro-2-methylphenoxyacetic acid) < 2,4-D 141.

Chlorophenoxy herbicides can also disrupt cell membrane transport mechanisms. They competitively inhibit and ultimately saturate the organic anion transport system in the choroid plexus, which facilitates the removal of potentially toxic anions (including endogenous neurotransmitter metabolites and exogenous organic acids) 139142143. Homovanillic acid and 5-hydroxy-3-indoleacetic acid, i.e. metabolites of dopamine and serotonin, respectively, accumulate in the CNS of rats following 2,4-D administration 144.

It has also been reported that 2,4-D induced neurotoxicity may be partly due to generation of free radicals. When incubating rat cerebellar granule cells with 2,4-D in vitro, glutathione (GSH) levels and catalase activity were significantly reduced, whereas generation of reactive oxygen species (ROS) and activity of selenium-glutathione peroxidase (Se-GPx) were augmented 145.

Furthermore, chlorophenoxy acids are structurally related to acetic acid and are able to form analogues of acetyl-CoA (e.g. 2,4-D-CoA) in vitro. Formation of such analogues has the potential of disrupting several pathways involving acetyl-CoA, including the synthesis of acetylcholine. Possible formation of choline esters (e.g. 2,4-D-Ach) may act as false cholinergic messengers (reviewed in 135).

In cerebellar granule cells, 2,4-D produced a striking and dose-dependent inhibition of neurite extension, and in vitro 2,4-D inhibited polymerisation of purified tubulin. Thus, it was suggested that at least one mechanism of 2,4-D neurotoxicity involves inhibition of microtubule assembly 146. Yet another study with cerebellar granule cells showed that 2,4-D induced apoptosis when cells were exposed to millimolar concentrations of the compound 147.

Chlorophenoxy herbicide poisoning in humans is uncommon, but it may produce severe sequelae. In a review of 66 cases of chlorophenoxy herbicide poisoning 135, the majority of cases involved ingestion of 2,4-D, either alone or in combination with other chlorophenoxy herbicides. Neurotoxic effects included coma, hypertonia, hyperreflexia, ataxia, nystagmus, miosis, hallucinations, convulsions, fasciculations, and paralysis. Some degree of peripheral neuromuscular involvement occurred in approximately one third of the cases reviewed. Still, other constituents, such as surfactants or solvents, in the formulations of the herbicides could possibly have contributed to some of the effects observed 135.

The information with respect to possible neurological effects of chronic exposures to low doses of chlorophenoxy herbicides is sparse, and in a review from 2002, it was concluded that it is unlikely that 2,4-D has any neurotoxic potential at doses below those required to induce systemic toxicity 148. However, a cohort study suggested an increased risk of amyotrophic lateral sclerosis (ALS) among workers chronically exposed to 2,4-D, compared to non-exposed employees at the same company, although this conclusion was based on only three deaths 149.

Although neurotoxicity in adults from low, chronic exposures to chlorophenoxy herbicides has not been reported, developmental exposure to low levels of these herbicides may still pose a threat. One case of cephalic malformations and severe mental retardation has been observed in an infant whose parents received prolonged exposure to 2,4-D via the dermal route from forest spraying 150.

Evidence of developmental neurotoxicity of chlorophenoxy herbicides, in particular 2,4-D, has also been obtained from experimental animals. For example, external treatment of fertilised hens' eggs with 2,4-dichlorophenoxyacetic butyl ester produced hypomyelination in the chicks, and reductions in "myelin markers" (including sulfatides, cerebrosides and 2'3'-cyclic nucleotide 3'-phosphohydrolase activity) were seen in chick embryos even before the period of active myelination 151. A deficit in myelin lipid deposition was also detected in neonatal rats exposed to 2,4-D through lactation 152. Other findings in response to neonatal exposure of rats to 2,4-D through lactation include a delay in CNS development 153, an increase in size and density of serotonin immunoreactive neuronal somata as well as an increase in fibre length in the dorsal and medial raphe nuclei 154; and oxidative stress in specific brain areas, including midbrain, striatum, and prefrontal cortex 155.

Behavioural effects in the offspring have also been reported following prenatal and continued exposure to 2,4-D 156. Also following prenatal and continued exposure of rats to 2,4-D, even beyond lactation, the dopamine D₂-type receptor was increased about 40% in the striatum. Increased levels of the receptor were also found in the prefrontal cortex and cerebellum. However, when discontinuing exposure after weaning, no differences in dopamine D₂-type receptors could be detected compared to control rats, suggesting that the effects of 2,4-D on these receptors may be reversible 157.

Thus, even though the evidence is sparse, some chlorophenoxy herbicides, in particular 2,4-D, have neurotoxic potentials and may cause developmental neurotoxicity.

The bipyridyl herbicides share common toxic mechanisms 158159; paraquat has been used as a model substance, but is no longer allowed in the EU. Intracellularly, both paraquat and diquat undergo redox cycling, leading to the generation of superoxide anions. These anions may react to form hydrogen peroxide and subsequently the highly reactive hydroxyl radical, which may then cause lipid peroxidation and cell death 159160. Another factor contributing to toxicity is the depletion of nicotinamide adenine dinucleotide phosphate with a bound hydrogen ion (NADPH), as both herbicide redox cycling and hydrogen peroxide detoxification via glutathione are NADPH dependent 159161. In addition to redox-cycling, there is some evidence that paraquat may be able to interact with enzymatic targets in the CNS, such as AChE and butylcholinesterase 162.

The initial phase of moderate to severe intoxication with paraquat and diquat is characterised by renal and liver failure, but the subsequent clinical course differs between the two, with intestinal paralysis and fluid loss as prominent features of diquat intoxication 160163164165. In severe and usually fatal cases of diquat poisoning, coma has also been reported 160. Severe neurological and neuropsychiatric complications due to brain stem infarction and/or intracranial haemorrhage have also been described 161163166.

In regard to long-term consequences of exposure to bipyridyl herbicides, paraquat is a prime suspect with respect to induction of PD. It causes selective degeneration of tyrosine hydroxylase immunopositive (TH⁺) neurons in the substantia nigra pars compacta, and long-term exposure has been found to increase the risk of PD in a Taiwan population that sprays paraquat on rice fields 167168169. A case report has described PD following diquat exposure 170, but because of a long induction period and the difficulties in retrospective exposure assessment, the hypothesis of delayed appearance of degenerative nervous system disease is difficult to verify. Since both paraquat and diquat can generate the formation of ROS, these compounds may well be involved in neurodegenerative diseases other than PD, such as Alzheimer's disease, but little evidence is available to evaluate this potential.

Even though it is rather clear that the cytotoxicity of paraquat involves oxidative stress 171, the sensitivity of dopaminergic neurons is difficult to explain 172. Possibly, the dopaminergic neurons may be particularly sensitive to the reactive oxygen species (ROS) from paraquat, since dopamine metabolism also creates ROS 173. In mice treated with paraquat once a week for 3 weeks, the effect on catecholaminergic neurons was reminiscent of that in PD, with a preferential loss of dopaminergic neurons in the substantia nigra pars compacta. This is consistent with the results from several similar studies 168169171.

PD has also been explored as a relevant outcome with respect to developmental neurotoxicity. When neonatal mice were exposed to paraquat, a marked hypoactive condition was apparent at 60 days of age and became even more pronounced at 120 days of age 174. Furthermore, paraquat reduced the striatal content of dopamine and metabolites without affecting serotonin 174. As already mentioned above under dithiocarbamates, other evidence suggests that maneb and paraquat may jointly and individually induce loss of dopaminergic neurons in mice. Administration of these pesticides postnatally enhanced the effect of the same pesticides administered during adulthood. Furthermore, exposure to maneb alone during gestation resulted in a dramatically increased response to paraquat in adulthood, including notable reductions in levels of dopamine and a loss of nigral dopamine neurons 132. Similarly, the greatest effect on locomotor activity in mice occurred in males after exposure to maneb prenatally and to paraquat in adulthood 175. This finding was supported by decreased levels of striatal dopamine, increased striatal dopamine turnover, and selective reduction in tyrosine hydroxylase-immunoreactive neurons of the substantia nigra pars compacta.

These observations are in agreement with the notion that an initially silent toxicity was later unmasked, and was affected by the specific order-of-presentation of the pesticides in regard to the developmental stage (not just an effect of the combination of pesticides). Thus, it seems that prenatal exposure to maneb, rather than paraquat, may sensitise/predispose mice to development of PD (or lead to a state of increased vulnerability), whereas paraquat exposure later in life may unmask the silent toxic effect of the earlier maneb exposure and then lead to clinical symptoms of the disease. Therefore it is possible that in the case of PD, developmental exposure to paraquat may not be as damaging as later exposure, particularly if this later exposure follows developmental exposure to maneb.

The mechanisms of toxicity employed by various types of fumigants are poorly known. A common mechanism of action is not expected, and the fumigants are therefore reviewed one by one.

Among metal phosphide fumigants, aluminium phosphide is one of the most extensively used. The phosphides are very toxic, because of their ability to liberate phosphine under moist conditions (reviewed in 176). Phosphine is a reductant and predictably reacts with metal ions such as the iron in haem and the divalent metals of metal dependent enzymes 177. Cytochrome c oxidase, of the mitochondrial electron transport chain, has been suggested as the primary site of action for phosphine 176178179. A 50% inhibition of this enzyme was found to be sufficient for generation of superoxide anions, and it was suggested that the toxicity of phosphine was due to damage by free radicals 178. In agreement with this hypothesis, aluminium phosphide has been found to increase lipid peroxidation in rat brain 180.

Further, in 45 phosphine poisoning patients, increased levels of superoxide dismutase (SOD) and malondialdehyde (MDA) were detected in non-survivors, while catalase was inhibited 181. Oxidation of phosphine can lead to formation of reactive phosphorylating species 182, thus suggesting that effects on cholinesterase may be possible 183. Studies of grain fumigant applicators 184 and in vitro studies of human red blood cells 185 have shown that significant phosphine-induced inhibition of red blood cell cholinesterase occurs at concentrations of phosphine exceeding 10 μg/ml.

Neurological changes like ataxia, stupor, tremors and convulsions have been observed following aluminium phosphide poisoning. Acute hypoxic encephalopathy has also been observed following aluminium phosphide poisoning, which may lead to death as a result of complete depression of the central nervous system and paralysis of the respiratory centres of the brain (reviewed in 176).

With respect to consequences of chronic phosphide exposure knowledge is sparse, but one descriptive study reported that most of a group of workers exposed to zinc phosphide had one or more neuropsychiatric symptoms including anxiety, impotence and easy fatigue. About half of the workers showed hyperreflexia, polyneuropathy, lumbar radiculopathy, and cervical myelopathy, as well as anxious mood, impaired attention, and psychomotor stimulation. EEG recordings showed abnormal findings in 17.4% of the subjects, mainly those with longer exposure 186. These preliminary findings should invite further studies in this area.

For the fumigant sulfuryl fluoride, very little is known concerning the mechanism of toxicity. The fluoride ion may play a role, since many of the observations in rodents overexposed to sulfuryl fluoride are typical of acute fluoride poisoning 187. In humans, short-term inhalation exposure to high concentrations of sulfuryl fluoride has been reported to cause central nervous system effects 188. A case report describes an elderly couple, who returned to their home 5–8 hours after fumigation with sulfuryl fluoride. The wife experienced weakness, nausea, and repeated vomiting, while the husband complained of dyspnea and restlessness. Within 48 hours the husband had a generalised seizure followed by cardiopulmonary arrest. The wife died within 7 days due to ventricular fibrillation. The serum fluoride concentration of the wife six days after the fumigation was reported to be as high as 0.5 mg/L 189.

Workers with a chronic, low level exposure to sulfuryl fluoride showed non-significantly reduced performance on all applied neurobehavioural tests compared to the control group in one study 190. Education levels, ethnicity and drug use differed between the workers and the control group in this study. In a later study of structural fumigation workers 191, sulfuryl fluoride exposure during the year preceding the examination was associated with significantly reduced performance on the Pattern Memory Test (a test of cognitive and visual memory) and an olfactory test. No pattern of cognitive deficits was detected.

None of these fumigants has been examined in detail for possible developmental neurotoxicity. Pregnant rats and rabbits exposed to sulfuryl fluoride were reported to show no evidence of embryotoxicity, foetotoxicity, or teratogenicity at concentrations of sulfuryl fluoride as high as 225 ppm, although body weights of rabbit foetuses as well as the dams at the highest exposure were lower than in the control group 192.

In regard to phosphine, a large epidemiological study found that adverse neurological and neurobehavioural developmental effects clustered among children fathered by applicators of phosphine (odds ratio = 2,48; 95% confidence interval: 1.2, 5.1) 193. Other than this study, no information regarding developmental neurotoxicity of phosphine was identified.

The present review on neurotoxicity has focused on a small number of substances out of the total number approved for use as pesticides in the EU. Quite likely, much evidence exists on neurotoxicity, but has not been published in biomedical journals. Nicotine, warfarin and ethanol are additional well documented neurotoxicants, but their primary use is not as pesticides. The same applies to other substances listed, such as sodium hypochlorite and aluminium sulfate, which may potentially add to neurotoxic hazards.

Some of the substances belonging to the groups of pesticides reviewed here (including OPs, carbamates, pyrethroids, ethylenebisdithiocarbamates, chlorophenoxy herbicides, and bipyridyl herbicides) appear to share common mechanisms of action with respect to induction of neurotoxicity. Thus, members of these chemical groups of pesticides other than those identified as neurotoxic in the present review, would then be highly likely also to cause neurotoxicity. For other groups of pesticides without a plausible common mechanism of action (e.g. the fumigants), it is not possible to predict whether group members might share neurotoxicity potentials.

Further refinement of this prediction is difficult. As anticipated, the literature on developmental neurotoxicity is sparse for most of the pesticides. However, some evidence does exist to suggest that several of the neurotoxic pesticides in current use in the EU may cause developmental neurotoxicity in small doses. Table 2 summarises the existing evidence of developmental neurotoxicity for groups of pesticides with common mechanisms of action.

Most evidence is available for the OPs, especially chlorpyrifos. The evidence strongly supports the notion that developmental neurotoxicity may be induced by very low exposure levels, i.e. much below those causing any neurotoxicity in adults. Most evidence still comes from studies in laboratory animals, but some epidemiological data are highly suggestive of neurotoxic effects caused by developmental exposure of humans to OPs (including chlorpyrifos). In the case of OPs, which share inhibition of AChE as a common mechanism of action in high doses, chlorpyrifos may employ other mechanisms of action at lower doses associated with developmental neurotoxicity. In fact developmental neurotoxicity in mice and rats can be induced at doses, which cause no detectable inhibition AChE 4344454651. Thus, even though a group of pesticides shares a common mechanism of action at larger doses, it cannot be excluded that compound specific mechanisms may also exist at lower doses. This fact unfortunately complicates the extrapolation of developmental neurotoxicity from one member of a group of pesticides to another. However, the combined human evidence on developmental neurotoxicity associated with OP exposure cannot be ascribed to chlorpyrifos alone.

Other than for OPs, the evidence of developmental neurotoxicity in humans is sparse, but evidence on developmental neurotoxicity in laboratory animals exists for pyrethroids, ethylenebisdithiocarbamates, and chlorophenoxy herbicides (mainly 2,4-D).

In the case of dithiocarbamates, evidence from laboratory animals suggests that developmental exposure to, e.g. maneb, may predispose the individual to development of PD later in life in response to another exposure, in particular paraquat. Other experimental studies suggest that prenatal exposure to paraquat can also predispose to development of PD later in life. It seems that the greatest effect of paraquat with respect to induction of PD is obtained from exposure later in life, following early priming exposure to maneb 175. Although PD is a degenerative disease associated with aging, these data suggest that developmental exposure to pesticides (e.g. maneb) may constitute an aetiological factor that sensitises the individual to later insults (e.g. subsequent pesticide exposure, and aging).

For the remaining pesticides that belong to groups without a common mechanism of toxicity, the lack of research on developmental neurotoxicity complicates the evaluation of their safety. In a few cases (e.g. the fumigant sulfuryl fluoride), the existing evidence from animal experiments indicates that developmental neurotoxicity may be unlikely to occur at doses below those causing maternal toxicity 192194. Still, in these experiments, possible later emerging effects or sensitisation caused by developmental exposure has not been studied, so any conclusion in this regard would be tentative.

On the other hand, with respect to the metal phosphide fumigants, which release phosphine under moist conditions, some evidence of developmental neurotoxicity does exist. An epidemiologic study has found adverse neurological and neurodevelopmental effects among children fathered by applicators of phosphine 193. For the remaining pesticides reviewed, no data from either human or animal studies could be located by our search.

This review has focused on those pesticides, for which human neurotoxicity has been reported in relation to specific exposures to the particular pesticide. This means that we have excluded poisoning cases involving more than one compound, where the contribution by each substance may be unknown. Thus, our list of neurotoxic pesticides is likely a substantial underestimate of the true number of neurotoxic pesticides. The fact that no poisoning incident with neurotoxic effects has been reported for a given pesticide is of course no guarantee that the pesticide is not neurotoxic, especially in regard to developmental exposure. A prudent evaluation of the evidence would therefore suggest that, if individual members of a chemical grouping of pesticides have been documented as neurotoxic, then all members of that group should be considered to be neurotoxic as well.

In addition to the problem of scarce – in many cases even non-existing – scientific evidence on developmental neurotoxicity of the pesticides in current use, some discrepancies exist between results of animal studies. An important factor in regard to apparent discrepancies is that the timing of exposure varies between studies. In some studies, animals are exposed prenatally, in other studies neonatally (during the first weeks of life), and in some studies both prenatally and neonatally. The timing of exposure may greatly influence the extent and type of neurotoxicity induced. Most animal studies have been performed in rodents, where brain development is mainly neonatal and spans the first three to four weeks of postnatal life 14195. Thus, although neurotoxic effects may be induced in rodents by only prenatal exposure, it is highly likely that these studies underestimate the neurotoxic effects, which may occur in response to prenatal exposure of humans, where the third trimester of pregnancy is a crucial period of brain development.

A further concern is that humans are very likely to be exposed to a number of pesticides and other neurotoxic compounds simultaneously. Because it is possible that some of these may have synergistic or additive effects, exposure to even very low doses during development may cause neurotoxic damage.

In addition to "direct" neurotoxicity, there is also evidence that several pesticides may indirectly cause neurotoxicity, e.g. by interference with thyroid function. Some 60% of all herbicides, in particular 2,4-D, acetochlor, aminotriazole, amitrole, bromoxynil, pendamethalin, and thioureas have been reported to interfere with thyroid function (reviewed in196). In addition, EBDC dithiocarbamates, organophosphates and synthetic pyrethroids are thought to interfere with thyroid function (reviewed in 197). A key concern with thyroid inhibitors is that impaired thyroid function may alter hormone-mediated events during development, leading to permanent alterations in brain morphology and function 127128. Other types of endocrine disruption can conceivably lead to neurobehavioural deficits, but this evidence has not been included here.

The current evidence can therefore be summed up as follows. A substantial proportion of pesticides in current use are known to be neurotoxic. However, neurotoxicity potentials of pesticides have not necessarily been examined, as legally mandated tests do not require specific assessment of neurotoxic potentials, apart from tests for peripheral neurotoxicity in hens required for OPs. A test battery for developmental neurotoxicity has only recently been completed by OECD, and very limited test data are available for pesticides. Because developmental neurotoxicity can occur at exposures much below those that cause toxicity to the adult brain, usage restrictions and legal limits for pesticide residues in food may not be sufficiently protective against developmental neurotoxicity. In addition, experimental, clinical and epidemiologic evidence supports the notion that neurotoxicity may be much more severe and possibly irreversible when the exposure occurs during early development.

Unless documentation exists for a particular pesticide to falsify this notion, all neurotoxic pesticides should be considered likely of inducing developmental neurotoxicity at low doses. The public health significance of this issue is illustrated by the epidemiologic observation of neurodevelopmental deficits at exposure levels that seem to be commonly occurring in the general population. Although the exact identity of the causative substances may be uncertain, pesticide contamination of foods is common in the EU, it often exceeds previously identified legal limits, and it involves substances that are known to be neurotoxic. Given the substantial impact of neurodevelopmental abnormalities in society and the likely impact of environmental aetiologies, prevention of pesticide exposure appears to be an obvious public health priority.