The magnitude of the problem

PCB levels in human populations

Twenty years after their production ban in most industrialised countries, PCBs are still present in the environment and in the food chain. Exposure in humans occurs mainly from this environmental contamination of food products. In addition, children are exposed to part of the body burden of their mothers during intrauterine development and through breast-feeding after birth. Several epidemiological studies in different parts of the world have tried to analyse possible adverse effects of this PCB exposure on neurodevelopment of children. Exposure levels in the collective studies ranged from very high and overtly toxic to low environmental levels. Reviews of these studies come to somewhat different conclusions. While Kimbrough et al. (2001) could see no conclusive evidence that environmental exposure to PCBs affects the neurobehavioral development of infants and children, Ribas-Fito and co-workers concluded that these studies suggest a subtle adverse effect of prenatal PCB exposure on child neurodevelopment, but do not allow the derivation of the degree of risk associated with current levels of exposure (Ribas-Fito et al. 2001). Schantz and coworkers concluded that there is growing evidence for PCB effects on neurodevelopment in humans, but also pointed out that many of the PCB congeners present in human tissues have never been studied in the laboratory, and their relative potency to produce nervous system effects is unknown (Schantz et al. 2003).

Concentrations measured in human populations differ depending on the age of the subjects, their nutrition, the origin of the samples and other factors. For detailed data of this aspect the reader is referred to the specific literature. In general, measurements of plasma and tissue levels in humans have indicated that exposure to these chemicals is on the decline since the production of PCBs has been significantly reduced. The three persistent PCBs measured in these studies, 2,2′,3′,4,4′,5-hexachlorobiphenyl (PCB 138), 2,2′,4,4′,5,5′-hexachloro-biphenyl (PCB 153) and 2,2′,3,4,4′,5,5′-heptachlorobiphenyl (PCB 180), are the most abundant in biological extracts. However, minor components may also be important because of their toxic potency. Mean concentrations of total PCBs in young humans (neonates to 25 years of age) are reported in the current literature to be in the range 0.56–1.2 ng/ml serum or approximately 150–200 ng/g lipid weight (Umweltbundesamt 1999, 2002; Becker et al. 2002; Lackmann 2002; Nawrot et al. 2002; Voorspoels et al. 2002).

Many experimental studies using commercial mixtures have been published to clarify the possible risk to humans who are mainly exposed via food. However, the exposure conditions for humans who are confronted with a mixture of persistent PCB congeners that have already been filtered through the ecosystem are very different from most experimental study situations in which animals are exposed to single congeners or commercially available mixtures of PCBs. Thus, data on single PCB congeners can considered to be relevant when the respective substance persists in the environment, whereas commercial mixtures (Aroclors, Clophens) are not truly representative of the PCB exposure encountered by humans today. Some authors, therefore, have used reconstituted PCB mixtures with compositions similar to the congener range found in human milk.

Animal studies have mainly been conducted in rodents (rats, mice). A few non-human primate collectives have also been studied, the results being published in several series of papers (Allen and Barsotti 1976; Arnold et al. 1993a, 1993b, 1997, 1999; Barsotti and Van Miller 1984; Levin et al. 1988; Mele et al. 1986; Mes et al. 1994; Rice 1999b; Rice and Hayward 1997; Schantz et al. 1989, 1991; Tryphonas et al. 1991). Although rhesus and cynomolgus monkeys can be considered good model species for humans, especially for the complex behavioural aspects, all monkey studies published so far have considerable drawbacks. For example, they used dosages that elicited obvious PCB toxicity (low pregnancy rate, infant death, skin lesions, hyperpigmentation) in either mother or offspring and thus are representative of highly exposed human populations only. In addition, the low numbers of offspring evaluated (usually below ten per group) and the use of feral monkeys with unknown age and reproductive history does not inspire great confidence in the results of these studies. We decided therefore not to include these studies. In this review article we briefly summarise the available experimental data on developmental toxicity of PCBs in rodents. Results with selected congeners are shown in Figs. 1, 2 and 3; for an overview see Fig. 4 and Table 2.

Fig. 1
figure 1

Three studies in two different rat strains were evaluated, two with prenatal and one with postnatal exposure of the dam. Reference numbers in square brackets, as well as strain, exposure period, daily dose and cumulative dose (in brackets) are listed on the left side of the figure. Brain PCB levels were measured only in the postnatal exposure study. Colored arrows indicate the tested parameters that were affected (m = male, f = female). The line ending in a bar (no arrow head) indicates that grip strength was unaffected by postnatal exposure through the milk. The most sensitive parameters in this set of studies (affected at the lowest dose tested) are marked with red dots

Full size image
Fig. 2
figure 2

Fourteen studies in three different strains of rats (SD = Sprague-Dawley, LE = Long-Evans, Wistar Rats) were evaluated, all of them with exposure through the mother animal. Reference numbers in square brackets, as well as strain, exposure period, daily dose and cumulative dose (in brackets) are listed on the left side of the figure. Colored arrows indicate the tested parameters that were affected (m = male, f = female). The most sensitive parameters in this set of studies (affected at the lowest dose tested) are marked with red dots. Exposure to 3 mg/kg/day during the fetal period reduced pup viability (dose field marked in red). All other dose regimens were compatible with postnatal survival of the offspring. Two studies provided PCB brain levels and one study found hydroxylated metabolites as the main PCB compounds in the fetuses. Growth reduction, delayed eye opening, alterations in anogenital distance in ales (decrease) and females (increase), increased sperm counts, decreased sexual receptivety in females and in an increase in brain dopamine/dopamine turnover were the most sensitive parameters in different studies

Full size image
Fig. 3
figure 3

Six studies in three different strains of mice (B6 = C57BL/6) were evaluated, two from them with direct postnatal exposure of the offspring. Reference numbers in square brackets, as well as strain, exposure period, daily dose and cumulative dose (in brackets) are listed on the left side of the figure. Colored arrows indicate the tested parameters that were affected (m = male, f = female). The most sensitive parameters in this set of studies (affected at the lowest dose tested) are marked with red dots. Dose levels of 8 mg/kg or higher were overtly toxic for the conceptus of the offspring (dose field marked in red) and therefore cannot be considered relevant for current human exposure. This induces the only study where behavioral parameters were tested after exposure in utero. No PCB tissue levels were available for any of the studies. NMRI mouse pups treated with a single low dose postnatally displayed a decrease of muscarinic receptor binding in the hippocampus in the preweaning period and an altered motor activity pattern when tested in the open field as adults. This was characterized by decreased initial exloratory activity and a lack of habituation over time. Publications considered are: [1] Agrawal et al. (1981); [2] d’Argy et al. (1987); [3] Marks et al. (1989); [4] Eriksson (1988); [5] Eriksson (1991); [6] Esser et al. (1994)

Full size image
Fig. 4
figure 4

Changes in behavioral parameters, brain biochemistry or electrophysiology have been observed for most of the PCB congeners and mixtures evaluated in this review. However, not all of the effects were detrimental. Spatial memory was found to be improved in rats that were exposed to the coplanar congeners 77 and 126 during devlopment. The learning impairment observed with other PCBs appears unrelated to the decrease in circulating thyroid hormone as it also appeared when plasma T4 was unaffected (PCB 28). Overall, the data available for individual compounds and mixtures are insufficient to derive patterns of action on the brain and the resulting behavioral changes. For references and more details see Table 2

Full size image

The amount of literature on PCB-induced developmental effects in rodents is enormous, and it did not seem possible to describe all papers in detail in this review without impairing the readability. Therefore, in addition to the printed comprehensive version of the review, data with other congeners and mixtures are summarised in a series of similarly designed figures, which are available as additional electronic material in the online version of this article at http://dx.doi.org/10.1007/s00204-003-0519-y; this material contains graphic overviews on different PCBs, which are linked with the relevant citations. All compounds and mixtures which are subject of the review and the supplementary electronic material are listed in Table 1.

Table 1 List of PCBs and PCB mixtures included in this review and in the supplementary electronic material, which is available online (http://dx.doi.org/10.1007/s00204-003-0519-y) and provides links to the detailed data of the individual compounds as well as the corresponding references
Full size table

Hormonal interactions

Although only relatively few of the 209 theoretically possible PCB congeners persist in the environment, they have been found to interact with a multitude of systems, probably mostly in the form of their hydroxylated metabolites. Of special interest are interactions of PCBs and/or their metabolites with hormonal systems, which themselves influence a variety of processes. PCBs may act through their structural similarities to thyroid hormones and other nuclear receptor ligands, through binding to the arylhydrocarbon receptor (AhR) by inducing AhR-dependent genes, and through the induction or inhibition of various metabolic enzymes. Moreover, it appears highly likely that not all interactions of PCBs with developmental processes have been identified yet. Since thyroid hormones, estrogens and androgens play an important role in the development of mammals, environmental PCB exposure during sensitive developmental periods may result in adverse effects on the growth and functional integrity of the organism, especially in the brain.

Thyroid hormone depletion by PCBs

One of the most consistent effects of PCB exposure in adult and neonatal rodents is a decrease in serum thyroxine (T4) concentration. One of the mechanisms discussed for this decrease is the ability of PCBs to induce hepatic microsomal enzymes UDP-glucuronosyltransferases (UGTs), which conjugate triiodothyronine (T3) and T4 before biliary excretion. In fact, an increased glucuronidation activity for T4 has been found in hepatic microsomes from pregnant rats treated with PCB 169 and PCB 77 on day 1 of pregnancy and days 2–18, respectively, as well as in their fetuses and offspring throughout the lactation period (Morse et al. 1993). Similar results were seen in pregnant and weanling rats exposed to Aroclor 1254 (A1254) from day 10–16 of pregnancy (Morse et al. 1996). However, circulating concentrations of thyroid stimulating hormone (TSH), which controls thyroid hormone synthesis, have not been found to be increased under conditions of PCB exposure and, consistent with this finding, a much smaller decrease than the one for T4 is seen for the active hormone T3, mostly with high doses that deplete T4 severely.

T3 plasma and brain concentrations are maintained in the normal range under PCB exposure by an upregulation in the activity of the type II thyroxine 5′-deiodinase, which produces T3 from T4 (Morse et al. 1993; Morse et al. 1996). Metabolic inactivation of T3 and its excretion have not been analysed in these studies. However, data from young adult male rats treated with A1254 in the diet (100 ppm) for 7 days show that although T4 and T3 serum concentrations were decreased by 67% and 21%, respectively, there was no increase in biliary excretion of T3 metabolites (Vansell and Klaassen 2002). Therefore, any depletion of T3 in the plasma appears to be mediated by an insufficient supply of T4 for conversion and/or by an increased tissue uptake, but not by an increased loss of T3 from the body.

Adult humans exposed to PCBs through their diet were found to have decreased plasma levels of thyroid hormones (Persky et al. 2001) but since the values were well within the normal range and well above the hypothyroidal range a toxicological relevance is questionable. No relationship between cord blood T4 and PCB levels could be found in a cohort of North Carolina neonates born between 1978 and 1982 (Longnecker et al. 2000).

There is mounting evidence that PCB effects are also mediated by PCB metabolites. Methylsulfonyl metabolites of tetra- and penta-chlorinated biphenyls, the major methylsulfonyl metabolites found in human tissue and milk, have been shown to decrease T4 and T3 serum levels in adult rats (Kato et al. 1999), and hydroxylated metabolites of PCBs with the OH-group in meta or para position are potent inhibitors of thyroid hormone sulfation by enzymes of the phenol sulfotransferase family in vitro (Schuur et al. 1998a, 1998b). Unfortunately it is difficult to compare relative potencies of PCBs and their metabolites in vitro because of the extensive adsorption of these substances to the plastic wells and glassware used in such experiments (Layton et al. 2002).

Interactions with thyroid hormone binding proteins and receptors

In the plasma of rodents, T4 is bound to transthyretin (prealbumin), which in addition to thyroxine transports the complex of retinol-binding protein and retinol. The more important T4 plasma transport protein for humans, thyroxine-binding globulin (TBG), is not present in rats and mice except during the period of postnatal thyroid development until 3 weeks after birth and in the hypothyroidal state (Savu et al. 1989; Vranckx et al. 1990). Although transthyretin is only a secondary carrier of T4 in human plasma, it is the primary T4 carrier in cerebrospinal fluid (Robbins 1996; White and Kelly 2001) and is implicated in the transfer of T4 to the fetal compartment (Meerts et al. 2002). PCB congeners and their hydroxylated metabolites with structural and conformational similarities to T4 bind to rodent and human transthyretin, many with affinity similar to or higher than the endogenous ligand. Hydroxylated metabolites bind strongly to human transthyretin and can displace T4 in vitro from this protein even where the parent congener is inactive in this respect. In contrast, TBG binds most PCBs very weakly or not at all, as does the thyroid hormone receptor β. The affinity of PCB metabolites for human TBG is at least two orders of magnitude lower than that of T4 (Cheek et al. 1999; Lans et al. 1994). It would appear therefore that plasma transport of T4 in humans is less sensitive to PCB exposure than that in rodents. However, the effects on T4 supply to the brain may be similar in humans and rodents due to the involvement of transthyretin in this process.

To complicate matters further, the planar PCBs 77, 126 and 169 (or their metabolites) and commercial mixtures that contain them (A1254) seem to possess not only antithyroid but also some thyreomimetic properties in developmental studies. Rat pups exposed to these substances or to A1254 open their eyes at an earlier time-point, an effect that is also elicited with an excess of l-thyroxine (Schantz et al. 1997a; Gray et al. 1999; Goldey et al. 1995a; Brosvic et al. 2002). Treatment of rat dams with A1254 during pregnancy and lactation at doses greater than 1 mg/kg per day increased RC3/neurogranin mRNA in their pups on postnatal day 15 (PND15) and restored myelin basic protein mRNA to the normal level. A lower dose of 1 mg/kg reduced myelin basic protein mRNA compared with that in controls (Zoeller et al. 2000). These findings show that A1254 contains antagonistic as well as agonistic components that can affect the transcription of these thyroid receptor-inducible neuronal protein genes. RC3/neurogranin is implicated in long-term potentiation and learning/memory processes and appears in rat brain around prenatal day 18 with a rapid increase during the first two postnatal weeks when thyroid hormone production is upregulated (Miyakawa et al. 2001). Another indication for a thyreomimetic function of PCBs could be the selective hearing loss that can be induced with A1254 and PCB 126 (Herr et al. 2001; Crofton and Rice 1999). While propylthiouracil-induced hypothyroidism increased auditory thresholds in developing rats over a range of frequencies (1–40 kHz), only low-frequency (1 kHz) hearing ability was affected in PCB-exposed animals, despite a similar reduction in T4 serum concentrations (Goldey et al. 1995a, 1995b). Histological examinations of a few PCB-exposed rats showed a mild-to-moderate loss of outer hair cells in the upper-middle and apical turns of the cochlea. However, such changes were only observed at doses that induced pronounced postnatal mortality (Crofton et al. 2000a).

Estrogenic and antiestrogenic activity of PCBs

PCBs can induce estrogenic as well as antiestrogenic effects, and evidence is accumulating that this is due partly to the generation of hydroxylated PCB metabolites (Layton et al. 2002). Estrogenic and antiestrogenic effects observed in cell cultures (Kramer et al. 1997) seem to be restricted mostly to readily metabolised substances, which are rapidly excreted in animals and humans and do not persist in the environment. Hydroxylated metabolites of PCBs bind to estrogen receptors ERα and ERβ with similar preference, but their binding affinities are at least 1000-fold lower than that of 17β-estradiol (E2), a measure that may be confounded by the preferential adsorption of the PCB metabolites to the test vials (Layton et al. 2002). Such metabolites have been shown to compete with E2 and to act as agonists or antagonists in various models. For example, 4-hydroxy-2′,4′,6′-trichlorobiphenyl and 4-hydroxy-2′,3′,4′,5′-tetrachlorobiphenyl stimulate transcriptional activity of both ERs but compared to E2 may induce a different preference of the receptor–ligand complex for specific estrogen response elements on the DNA and the preferential binding of other cofactors (Connor et al. 1997; Kuiper et al. 1998; Hall et al. 2002).

In addition to direct interactions with ERs that misdirect their translational or inhibitory actions, other influences on estrogenic mechanisms are possible: the local production of estrogen from testosterone is catalysed by the aromatase enzyme complex. This reaction is especially important for the development of sexual dimorphic brain regions, implicated in male and female specific behaviour. Both environmentally relevant PCBs and 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) have been shown to reduce aromatase activity in the brain of newborn male rats (Hany et al. 1999b; Ikeda et al. 2002). The mechanisms for this reduction remain to be clarified; however, for TCDD no direct effect on enzyme activity in vitro could be detected. On the other hand, hydroxylated PCBs inhibit members of the hydroxysteroid sulfotransferase family, as has been shown with human estrogen sulfotransferase, and thus may increase availability of endogenous estrogens (Kester et al. 2000).

Antiestrogenicity of PCBs may be mediated not only by an interaction with estrogen receptors but also by cross-talk of the arylhydrocarbon receptor (AhR) with ER pathways after activation by TCDD-like PCBs. Inhibitory dioxin-responsive elements have been found in promoter regions of some E2-responsive target genes (Safe et al. 1998) and can be expected to modulate ER-directed gene expression. In addition, data from rat hepatocyte cultures indicate that the AhR activates the transcription of several other nuclear transcription factors belonging to the hepatocyte nuclear factor and the CCAAT/enhancer-binding protein families, which then can activate or repress other genes not directly responsive to the AhR (Borlak et al. 2002).

Bonefeld-Jørgensen used several in vitro systems (MCF-7 cells, Chinese hamster ovary cells) and showed that three environmentally persistent congeners possess receptor-mediated antiestrogenic (PCB 138, PCB 153 and PCB 180) and antiandrogenic (PCB 138) effects (Bonefeld-Jørgensen et al. 2001).

Androgenic and antiandrogenic activity of PCBs

In mice and rats anogenital distance (AGD) is an indicator of prenatal androgenisation. Endogenous androgen production as well as the androgen concentrations provided by adjacent littermates in the uterus play a role in determining AGD. Higher androgen concentrations are associated with longer AGDs (Vandenbergh and Huggett 1995; Hernandez-Tristan et al. 1999). Prenatal differences in androgenisation also modulate social and sexual behaviour in these species and associations between AGD and behavioural parameters have been described (Drickamer et al. 1995; Drickamer 1996; Zehr et al. 2001).

PCBs have been shown to increase as well as decrease male AGD. Treatment of pregnant CD-1 mice with Aroclor 1016 at a dose of 0.1 mg/kg per day increased AGD and prostate size while decreasing epididymal weight in male offspring. This regimen also permanently increased the androgen receptor (AR) binding activity of the prostate. The findings on prostate growth and AR binding were reproduced in vitro (Gupta 2000). In another study, coplanar PCB congeners 77 and 126 given at single doses of 0.1 mg/kg and 0.01 mg/kg, respectively, to pregnant rats on day 15 of pregnancy reduced AGD in male pups and decreased circulating testosterone concentrations in adult male offspring (Faqi et al. 1998). Androgenic effects of PCB treatment on female offspring have been found after intraperitoneal exposure of the dams to PCB 77 or PCB 47 from day 7 to 18 of pregnancy. This treatment increased female AGD, but had no effects on male AGD. Dose levels that affected AGD in females also impaired female sexual receptivity in adult offspring (Wang et al. 2002).

Commercial mixtures of PCBs (Aroclor 1260, Aroclor 1242, Aroclor 1254, Aroclor 1248) as well as some individual congeners (PCB 31, PCB 42) have been shown to interfere with AR-mediated transcription in an in vitro reporter assay system. In addition to an antagonistic action on dihydrotestosterone-elicited transcription, Aroclor 1254 by itself was a weak AR agonist. Androgen ligand binding to the AR was reduced in the presence of PCBs 42, 128 and 138 (Portigal et al. 2002).

Effects on serum steroid concentrations

Concentrations of sexual hormones have been measured by several authors in rats after treatment with PCBs. A1254 or a reconstituted PCB mixture (4 mg/kg per day) decreased E2 and increased testosterone in exposed pregnant female Long-Evans rats. No effects on testosterone were seen in female offspring on PND 21 (testosterone levels were highly variable). Decreased E2 and testosterone in weanling female offspring of female Long-Evans rats exposed until term of pregnancy were found in another study. In adult male offspring both treatments decreased serum testosterone concentrations (Hany et al. 1999b; Kaya et al. 2002). Similar findings on serum testosterone were reported in adult male Wistar rats exposed in utero to PCB 77 or PCB 126 on a single day at the beginning of the fetal period; testosterone levels were highly variable (Faqi et al. 1998). A1254 had no effects on serum and testicular testosterone concentrations in F344 rats after oral treatment with doses up to 25 mg/kg per day from weaning for up to 15 weeks (Gray et al. 1993).

PCB 126 dose-dependently increased E2 in gonadotropin-primed immature female Sprague-Dawley rats while progesterone levels were decreased (Gao et al. 2000).

Vitamin A depletion by PCB exposure

An oral dose of 10 mg/kg per day of A1254 reduces hepatic vitamin A levels in mice and rats (Hallgren et al. 2001) and PCB 77 decreases rat serum levels (Brouwer et al. 1988). One of the reasons may be an increased catabolism due to the PCB-related induction of cytochrome P450 enzymes. In addition, in contrast to T4, which does not interfere with the binding of the retinol-binding protein–retinol complex to transthyretin, the plasma transport of retinol can be disturbed by PCB congeners/hydroxylated metabolites (Brouwer and Van der Berg 1986). In general, such a reduction appeared to be insufficient to induce malformations indicative of vitamin A deficiency. Except for morphological abnormalities of the inner ear in rats and probably in mice, gross morphological changes, mostly cleft palate, have been observed only in mice treated with the planar PCBs 77 and 169 (d’Argy et al. 1987; Marks et al. 1989; Zhao et al. 1997). However, in rat studies embryofetolethality may mask the occurrence of teratogenic effects at doses that induced malformations in mice, and malformed fetuses may have been eliminated by prenatal loss or by maternal cannibalism.

Behavioural effects of PCBs in animal studies

Most of the studies that have been conducted with PCB mixtures or specific congeners suffer from shortcomings. Very few of them comply with current regulatory requirements for reproductive and developmental hazard identification studies. Main problems for the reviewer are listed below.

–:

Group sizes employed in studies in which the mother animals were treated in order to expose the conceptus through the placenta or the newborn through the milk are mostly in the range of less than ten pregnant or lactating females. Such animal numbers are appropriate for a dose-range finding study but not for a main investigation on hitherto unknown outcomes.

–:

Selecting one pup per sex per litter for behavioural studies must be considered insufficient when exposure occurs through the mother animal. This study design assumes that behavioural effects (if present) are 100% penetrant so that each member of a litter is equally well suited to detect the effect. However, it is well known from morphological effect studies that malformations often occur in some of the offspring in a litter while other littermates appear normal. If functional deficiencies follow a similar distribution, effects are likely to be missed with only one litter representative for each sex.

–:

Dose–response or dose–effect relationships cannot be evaluated in many studies because they used only one or two PCB-treated groups.

–:

Plasma and tissue levels of PCBs or their metabolites are frequently not available. However, as the kinetic behaviour of the compounds differs considerably in rodents and man, blood and tissue concentrations must be considered for any risk assessment.

–:

Basic information that must be available for any rational evaluation of a developmental toxicity study, such as adverse effects on the mother, litter sizes, birth weights, postnatal weight gains, and pup mortality, is lacking in many publications on PCBs. Detailed criticism of such studies is included in the supplementary electronic material to this article. Thus, the functional effects in the offspring that are reported cannot be considered in relation to other toxicities.

–:

The studies in rats and mice do not provide individual data. Only in monkey studies are individual performances sometimes reported in the form of a graph. However, studies in non-human primates have even lower group sizes than rodent studies and, for this reason, were not considered in this review. This lack of individual data can make it difficult to evaluate the biological relevance of a finding, especially for behavioural parameters, where the use of mean values may hide the presence of a subgroup of affected animals within a treatment group.

Bearing these difficulties in mind, the reader is directed to a series of diagrams that summarise the effects of individual PCB congeners or commercial and environmentally relevant mixtures on functional development in rodents: Figs. 1, 2 and 3. The emphasis is on single congeners because commercial mixtures have little relevance for current environmental exposure scenarios. Unfortunately, for those persistent congeners (PCBs 138, 180), which together with PCB 153 are most frequently measured in humans, no studies of their effects in animals could be found. Studies have been conducted mainly with different strains of rats with various exposure periods ranging from before mating of the females to the adult age of the offspring. Studies in mice, with a few exceptions, have been performed by only one laboratory where 10-day-old pups were exposed directly by oral gavage (Eriksson 1988; Eriksson and Fredriksson 1996, 1998; Eriksson et al. 1991).

Behavioural aspects that have been studied after PCB exposure fall into four main categories: (1) learning and memory, (2) motor activity, (3) gender dimorphic behaviour, and (4) sexual behaviour. In addition, neurochemical, electrophysiological and thyroid hormone measurements have been performed for several of the compounds in order to elucidate possible mechanisms of behavioural impairment.

Learning and memory

PCBs appear to have detrimental as well as enhancing effects on learning performance. While PCBs 28, 118 and 153 slowed spatial alternation learning in female rats (Schantz et al. 1995; Figs. 1, 4; and Table 2), congeners 77, 95 and 126 had no effects on T-maze performance in either males or females. They increased acquisition of a working memory task in the radial-arm maze spatial learning test, and decreased the error rate in this task (Schantz et al. 1996, 1997b; Figs. 2, 4; and Table 2).

Table 2 Effects described with various PCBs and PCB mixtures. Abbreviations: LOAEL lowest observed adverse-effect level, NOAEL no observed adverse-effect level, pc postcoitum, pp postpartum (exposure of dams), pn postnatal (exposure of offspring), LTP long-term potentiation, T4 thyroxine, DA dopamine, DOPAC l-3,4- dihydroxyphenylacetic acid,LH luteinising hormone, FSH follicle-stimulating hormone, NMDAR NMDA receptor
Full size table

Adverse effects of PCBs on learning in animals and humans have been discussed in the light of the ability of these substances to decrease thyroid hormone plasma levels and the possible decrements in brain function that could arise from hypothyroidism. However, from the data obtained with different PCB congeners, it is unlikely that the learning problems observed in these rat studies are caused by a lack of thyroid hormones. The effects on T-maze performance were similar with PCBs 28, 118 and 153, but only congeners 118 and 153 also affected T4 plasma levels. Plasma concentrations of thyroxine were normal in animals exposed to PCB 28 (Ness et al. 1993).

Another possible mechanism by which learning could be impaired is interference of PCBs and their hydroxylated metabolites with the synthesis of nitric oxide (NO). Endothelial nitric oxide synthase (NOS) in the hippocampus has been implicated in long-term potentiation and learning (references cited in Sharma and Kodavanti 2002). Ortho-substituted PCBs and hydroxylated metabolites, also of non-ortho PCBs, inhibited especially cytosolic NOS (nNOS) in cerebellar, hippocampal and hypothalamic preparations in vitro. The inhibitory concentrations in the assay (10–50 μM) were similar to the concentrations observed in biological fluids and tissues from environmental exposure (Sharma and Kodavanti 2002).

Besides these mechanisms, other neurological systems can be affected by PCBs and may contribute to changes in learning and memory performance: e.g. N-methyl-d-aspartate (NMDA) receptors in the cortex (Altmann et al. 2001). Furthermore, indirect influences via effects on steroids are possible because it is well-known that most transmitter systems, such as serotonin, dopamine, acetylcholine and glutamate, can be influenced by sexual hormones (Auger 2001; Cyr et al. 2001; Haberny et al. 2002; Luconi et al. 2002; Nilsson and Gustafsson 2002; Partridge et al. 2002; Rissman et al. 2002; Rupprecht et al. 2001; Sanchez et al. 2002; Song et al. 2002).

In mice, no improved performance in learning tests was observed with any of the tested PCBs. On the contrary, learning seemed to be impaired. Whereas PCB 118 had no effects on Morris water maze learning in mice exposed on PND10, PCBs 28 and 126 reduced the escape latency after the submerged platform of the Morris water-maze had been moved into a new position, a finding which indicates that treated animals had memorised the original position less consistently than control mice. PCB 77 prolonged latency in an avoidance learning test but at a dose that was overtly toxic to the offspring (Eriksson and Fredriksson 1996, 1998; Tilson et al. 1979; Figs. 3, 4). Impaired performance of passive avoidance after exposure to PCB 77 has been described in rats in the absence of overt toxicity in dams or offspring (Weinand-Härer et al. 1997).

Motor activity

It has been shown that severe reduction of T4 and T3 plasma levels during perinatal and preweaning development by propylthiouracil leads to preweaning decreased or delayed motor activity and to persistent, postweaning hyperactivity in rats (Brosvic et al. 2002). However, in contrast to thyroid hormone levels, motor activity has not been examined as consistently in rat offspring exposed to PCBs (Fig. 4). Data obtained in offspring exposed to PCBs 126 or 153 are consistent with an effect on activity mediated by a decrease of T4 concentration during the first two postnatal weeks (Holene et al. 1998). Adult male offspring exposed to PCB 47 and PCB 77 showed hyperactivity in the open field (Figs. 1, 4). Treatment with PCB 95 during the late embryonal and early fetal period in contrast decreased locomotor activity in an open-field test conducted in adult offspring (Schantz et al. 1997b).

After treatment with PCB 77 from day 7 to 18 of pregnancy, rats showed increased latencies in the haloperidol-induced catalepsy test on postnatal day 100. Similar, but non-significant differences compared to controls, were observed after treatment with PCB 47 or a mixture of these congeners. The endpoint of this test was the time rats required to draw one or both paws from a vertical grid; in modified versions of the test (paws positioned on a bar or block), no differences were detected between PCB-treated rats and controls (Hany et al. 1999a). In another publication by this group, significant effects were reported in another strain of rats, although this time differences were not observed in the “vertical grid test” but in the “bar test” only (Weinand-Härer et al. 1997).

Activity testing in mice exposed on PND10 has produced similar results for four PCB congeners (PCBs 28, 52, 77 and 126). The mice were initially hypoactive compared with controls but displayed an impaired habituation response with time. Therefore, when later time periods of the test are considered, PCB-exposed animals were hyperactive since there was little or no decrease in their ambulation (Eriksson and Fredriksson 1996, 1998; Fig. 3, 4). Changes in activity patterns appeared consistently at the lowest effective doses. Possible mechanisms, however, have not been elucidated.

Gender dimorphic and sexual behaviour

During brain development sexual hormones induce gender-specific differentiation in different brain regions of males and females. Masculinisation or feminisation of the brain leads to different behavioural patterns in adult animals with respect to interactions with possible sexual partners or in behaviour that is unrelated to reproduction. Sweet preference is a sexual dimorphic behaviour that has been used to analyse sex-specific development in the brains of PCB-exposed rats. Females usually display a higher preference for sweetened drinking water than males. Female sweet preference was decreased by PCB 77 (masculinisation of response) whereas male sweet preference was unaffected (Amin et al. 2000). A commercial mixture (Aroclor 1254) was reported to increase female consumption of sweet solutions by approximately 30%, but this was not statistically significant (Hany et al. 1999b). An environmentally relevant mixture increased sweet preference in males at the highest dose tested (Hany et al. 1999b; Kaya et al. 2002; Figs. 2, 4).

Female mating behaviour (lordosis, a measure of receptivity) was found to be decreased with congeners 47 and 77. Both substances also increased anogenital distance in females, which indicates an increased exposure to androgens during the late fetal period (Wang et al. 2002; Figs. 2, 4).

Summary and recommendations

Behavioural effects of PCBs in children have been described in numerous studies. To support the biological plausibility of such effects it has often been argued that corresponding changes have been observed in various animal species; however, many of these animal experiments show methodological flaws. When behavioural parameters are compared with other endpoints for sensitivity across studies, it becomes obvious that only for 4 of the 13 compounds or mixtures with adequate data (PCB 77, 118, 126, and 153) are the behavioural tests at the most sensitive end of the testing battery. For the other cases, changes in enzyme activity, organ weights, postnatal growth or T4 plasma levels can be detected at dose levels below those at which behavioural changes become apparent. As a result of our extensive literature survey we came to the conclusion that, despite the enormous amount of literature, several aspects of PCB-induced developmental toxicity have not been adequately studied so far. For further research in this area the following recommendations may help to close these data gaps:

  1. 1.

    Of those environmentally persistent PCBs that are found preferentially in human tissues and have become standards for PCB exposure (PCBs 138, 153, 180) only PCB 153 has been tested for a number of different endpoints including behaviour. No comparable studies have been found for the other two compounds, which have only been tested as part of mixtures. In order to determine the contribution, if any, of these substances to the adverse effects of PCBs on brain function the individual congeners PCB 138 and PCB 180 should be tested.

  2. 2.

    With respect to environmental exposure, commercial mixtures are not representative. Therefore, it is important that environmentally relevant mixtures are studied to a greater extent. The available data on the reconstituted mixture that was evaluated in this review are incomplete because data on thyroid hormones, metabolic enzymes or behavioural parameters, such as locomotor activity, learning and memory or social/sexual behaviour, are lacking. Additional studies need to be done to characterise the spectrum of effects elicited by mixtures of persistent PCBs.

  3. 3.

    Studies in which exposure is through the dam, either prenatally or postnatally, need to be large enough (group sizes, number of dose groups, number of evaluated offspring per litter) to inspire confidence in the results. In addition to all the parameters normally recorded in regulatory reproductive toxicity studies (maternal body weight gain and clinical signs, implantation sites, live litter size, pup weights at birth and during development, pup mortality data and developmental landmarks), a widespread selection of additional endpoints covering behaviour (learning/memory, activity, sexual and gender-specific), endocrine functions, neurochemistry and measurements of internal exposure should be included. It should be underlined that a general risk exists in studies with small number of animals that the control group is not a valid and representative sample of non-exposed animals, even if the animals are randomly selected. This is especially true for behavioural testing when the variability within and between litters has not been ascertained beforehand.