A review of the evidence for endocrine disrupting effects of current-use chemicals on wildlife populations

Abstract

This review critically examines the data on claimed endocrine-mediated adverse effects of chemicals on wildlife populations. It focuses on the effects of current-use chemicals, and compares their apparent scale and severity with those of legacy chemicals which have been withdrawn from sale or use, although they may still be present in the environment. The review concludes that the effects on wildlife of many legacy chemicals with endocrine activity are generally greater than those caused by current-use chemicals, with the exception of ethinylestradiol and other estrogens found in sewage effluents, which are causing widespread effects on fish populations. It is considered that current chemical testing regimes and risk assessment procedures, at least those to which pesticides and biocides are subjected, are in part responsible for this improvement. This is noteworthy as most ecotoxicological testing for regulatory purposes is currently focused on characterizing apical adverse effect endpoints rather than identifying the mechanism(s) responsible for any observed effects. Furthermore, a suite of internationally standardized ecotoxicity tests sensitive for potential endocrine-mediated effects is now in place, or under development, which should ensure further characterization of substances with these properties so that they can be adequately regulated.

Keywords:

• Endocrine disruption
• legacy chemicals
• current-use chemicals
• wildlife
• populations

1. Introduction and methods

Two recent global reviews of endocrine disruption (ED) in wildlife and humans (Bergman et al. 2012; Kortenkamp et al. 2012) have comprehensively addressed the issue of whether ED is damaging individuals and populations. Here we focus only on the evidence for effects on wildlife. For the purposes of this review, we consider that ED is occurring when a chemical has interfered directly with the endocrine system, either by interaction with hormone receptors or by alteration of hormone synthesis or metabolism, and has thereby caused adverse effects which are observable in wildlife. This is in line with the widely accepted World Health Organisation definition (WHO/IPCS 2002), although it should be noted that the Endocrine Society considers interference with any aspect of hormone action to constitute ED (Zoeller et al. 2012). In broad terms, both Bergman et al. (2012) and Kortenkamp et al. (2012) followed the WHO definition.

Bergman et al. (2012) concluded inter alia that:

Wildlife populations have been affected by endocrine disruption, with negative effects on growth and reproduction. These effects are widespread and have been primarily due to POPs [Persistent Organic Pollutants]. Bans of these chemicals have reduced exposure and led to recovery of some populations.

Kortenkamp et al. (2012) also concluded that some wildlife populations had been harmed by ED, and they provide a list of so-called chemicals of concern, some of which are not POPs as listed by the Stockholm Convention, but which have been the subject of other regulatory action [e.g. tri-organotins; alkylphenols (APs)]. None of these substances were restricted specifically because they present an unacceptable risk from ED properties, as such, but simply because their fate and adverse effects (which may or may not have been caused by an endocrine mechanism) in the environment were deemed unacceptable. All can be described as legacy chemicals, which have been subject to regulatory action resulting in withdrawals from use or sale, but in some cases they are still widespread in the environment due to their persistence (e.g. Rasmussen et al. 2015), and may cause harm in some sensitive species.

There is no doubt that some legacy substances have caused (and are still causing) adverse endocrine-related effects (i.e. ED) in wildlife populations. The reader can find further information in various publications in addition to Kortenkamp et al. (2012) and Bergman et al. (2012), as shown in Table 1. Yet, the most important question to be answered is if, and to what extent, current-use (non-legacy) chemicals are causing ED in the environment.

Table 1. Summary information on impacts in wildlife populations caused by legacy substances with endocrine disrupting properties.

Legacy substance or process	Summary of adverse endocrine-mediated effects in wildlife	References
Tri-organotins	The effects of tri-organotins on mollusks are probably one of the best documented examples of wildlife damage caused by legacy chemicals with ED properties, although the precise mode(s) of action is still not entirely clear. In essence, tri-organotins [especially tributyltin (TBT) derived from antifouling paints, but also triphenyltin (TPT) to a lesser extent], caused masculinization of aquatic mollusks, which led to widespread population declines and local extinctions. The secondary effects also damaged whole invertebrate communities and impacted other phyla such as crustaceans and ascideans. Recovery in many mollusk communities began soon after TBT was globally banned by the International Maritime Organisation from use in antifouling paints on shipping in 2008, but some effects are still detectable near marinas and harbors due to the remaining residues in sediments.	Bryan and Gibbs (1991) Matthiessen and Gibbs (1998) Alzieu (2000) Oehlmann et al. (2007) De Mora (2009) Matthiessen (2013)
Polychlorinated biphenyls (PCBs), polychlorinated dibenzo dioxins (PCDDs), polychlorinated dibenzo furans (PCDFs)	PCBs (the production of which was banned globally under the Stockholm Convention in 2001) are probably now a significant contributor to ED in wildlife. PCDDs and PCDFs are also listed under the Stockholm Convention for reduction of unintentional releases. However, PCBs are still entering the environment through leaching from old equipment such as electrical transformers. Some members of this family of very persistent and bioaccumulative substances, as well as the similar PCDDs and PCDFs, have been implicated in reproductive failures in top predators ranging from mustelids, cetaceans and pinnipeds to reptiles and birds. Significant residues and effects are present in Arctic wildlife resulting from long-distance transport of PCBs and other organochlorines. Some congeners are able to produce estrogen-like effects, while others interfere with the immune and thyroid systems.	De Swart et al. (1996) Murk et al. (1996) Grasman et al. (1998) Robertson and Hansen (2001) Fossi and Marsili (2003) Basu et al. (2007) Sonne (2010) Jepson and Law (2016) Jepson et al. (2016)
Polybrominated diphenyl ethers (PBDEs)	These persistent and bioaccumulative compounds are similar to PCBs and have been widely used as flame retardants, although several congeners have been listed for elimination under the Stockholm Convention since 2009. They are widely present in the fatty tissues of a range of wildlife, from fish to mammals, although there is no direct field evidence for adverse apical effects. However, they have been associated with thyroid hormone imbalances in wild seals and eagles, and laboratory experiments have shown that environmentally-relevant PBDE concentrations can cause thyroid disruption leading to a variety of adverse effects such as (inter alia) delayed metamorphosis in amphibians, reduced eggshell thickness and abnormal courtship behavior in birds, and damaged neurological functioning in mustelids.	Hall et al. (2003) Birnbaum and Staskal (2004) Yogui and Sericano (2009) Cesh et al. (2010)
Perfluorinated compounds (PFCs)	PFCs have a variety of uses, but examples of particular environmental concern are perfluorooctanesulfonic acid (PFOS) and perfluorooctanoic acid (PFOA) which are used to make textiles water-resistant and as components of fire-fighting foams. Production and use of PFOS have been restricted under the Stockholm Convention, with some exemptions. PFCs are widely found in vertebrate top predators, sometimes at concentrations thought to be biologically active against the thyroid system on the basis of laboratory experiments, although there is little evidence for their adverse effects in the field.	Giesy and Kannan (2002) Houde et al. (2006) Peden-Adams et al. (2009) Stock et al. (2010) Lindstrom et al. (2011)
DDT/DDE and other organochlorine pesticides	The widespread use of DDT for insect control in the 1950s and 1960s caused eggshell thinning in birds, particularly some raptors such as sparrowhawks (Accipiter nisus), white-tailed eagles (Haliaetus albicilla) and ospreys (Pandion haliaetus). This led to frequent reproductive failure when the birds tried to incubate their eggs. It was later discovered that the active moiety was a metabolite of DDT, p,p’-DDE. The precise mechanism of action is almost certainly the ability of p,p’-DDE to inhibit synthesis of a prostaglandin* hormone in the eggshell gland, which subsequently interferes with normal calcium metabolism. Since DDT was withdrawn from use in most developed countries in the 1970s, raptor populations have largely returned to their pre-DDT levels, although part of this recovery is possibly attributable to additional factors such as improved bird conservation measures. p,p’-DDE has also been implicated as one cause of sexual abnormalities (male to female sex reversal) in aquatic reptiles such as alligators (Alligator mississippiensis). DDT-use is still permitted under the Stockholm Convention for disease vector (e.g. mosquito) control in the tropics, and is continuing to cause impacts on certain wildlife around the globe. Many other organochlorine pesticides (e.g. dieldrin; mirex; chlordane) are able to interact in a damaging fashion with the endocrine systems of wildlife, generally causing various types of feminization, and have themselves been banned under the Stockholm Convention.	Ratcliffe (1967, 1970) Lundholm (1997) Grasman et al. (1998) Dawson (2000) Guillette et al. (2000) Henny et al. (2010)
Alkylphenols and alkylphenol ethoxylates	Alkylphenol ethoxylates (APEs) were widely used in certain types of detergent and other products, but have now largely been withdrawn from use or voluntarily phased out in regions such as the European Union and the USA. Both APEs and their degradation products (e.g. nonylphenol and octylphenol) are weakly estrogenic, and can induce a variety of sexual abnormalities in aquatic vertebrates such as fish. They used to form a significant part of the estrogenic load in sewage and industrial effluents, and as such once contributed to the widely observed estrogenic effects on fish populations, which included the occurrence of widespread intersex in several species. Estrogenic abnormalities caused by alkyphenols have also been observed in other vertebrate populations, e.g. birds.	Sheahan et al. (2002) Dods et al. (2005) Kovarova et al. (2013)

* Unlike systemic hormones, prostaglandins act locally in tissues, but they are nevertheless part of an organism’s chemical signaling system. As such, disruption of prostaglandin signaling is considered to fall within the definition of ED.

Substances with ED properties which have not already been restricted are now subject to specific regulatory action in several jurisdictions such as the European Union and the United States (EU 2009, 2012; USEPA 2015). New testing methods specifically developed to identify and evaluate potential endocrine disrupting substances (EDSs) are available from the Organisation for Economic Cooperation and Development (OECD 2016), Japan and US-EPA. Few current-use chemicals have yet been systematically subjected to the whole testing suite (e.g. under the US-EPA’s Endocrine Disruptor Screening Program – EDSP), as this is an animal-intensive, long and costly endeavor (ca. 1 million US dollars per substance). Nevertheless, studies looking at endocrine activity or endocrine-mediated effects are routinely performed in mammals and conducted, when triggered, in ecotoxicological test species (e.g. using the Fish Short Term Reproduction Assay – OECD TG 229 – and the Amphibian Metamorphosis Assay – OECD TG 231 for screening or in more extensive higher tier tests). However, many current-use or non-legacy chemicals are subject to comprehensive environmental hazard and risk evaluations (e.g. registered biocides and pesticides), including partial and sometimes full lifecycle tests, which probably detect the adverse apical effects of many ED modes of action without necessarily identifying the chemical as endocrine-active (e.g. Weltje et al. 2013). In the interests of improving our general understanding of the level of protection afforded, it is worth investigating how ED effects can be attributed to current-use chemicals, and the extent to which the problem may be restricted to legacy chemicals.

This paper therefore aims to review the reports cited in Kortenkamp et al. (2012) and Bergman et al. (2012) in which population-level effects in wildlife under field conditions have been attributed to current-use chemicals, with only a brief survey of the effects of legacy chemicals. However, some additional studies, mainly those published since 2012, will also be discussed where informative. This review, while not claiming to be exhaustive, covers many animal groups, from invertebrates (crustaceans and mollusks), through lower vertebrates (fish, amphibians, and reptiles), to higher vertebrates (birds and mammals). Types of ED discussed include sexual disruption, stress-response perturbation, and thyroid system disruption. The papers referenced by Kortenkamp et al. (2012) and Bergman et al. (2012) had already been screened for scientific quality and reliability, but additional publications here have been evaluated informally using the principles of Klimisch et al. (1997). Any known quality shortcomings in papers referenced in this review are discussed in the text.

As with almost all epidemiological and field evidence, certainty in establishing the causes of observed effects is hard to obtain. The causality issue is generally approached iteratively by testing hypotheses in laboratory and field experiments and then returning to the field to confirm whether controlled experimental findings agree with the field response. Of course, this is sometimes difficult or impossible if large or rare species are the focus of attention. Experimental confirmation is one of the criteria proposed by Hill (1965) for assessing the strength of association between a disease and a putative cause. Others include the scale of the effect, its consistency or reproducibility, its specificity to the putative cause, the temporal sequence observed (i.e. did exposure occur prior to the effect), the presence or absence of a gradient of effect (i.e. a dose-response), the biological plausibility of the effect, the effect’s agreement with current knowledge, and finally, whether removal of the exposure leads to recovery.

Hill’s criteria are a valuable method for establishing causality; however, rarely can all be addressed in a given case, and as such they should be treated flexibly considering the question in hand. However, they allow an assessment of the strength of an association between possible cause and effect. They have therefore been adopted with minor changes in wording for use in the current review (see Section 3). For further discussion on the strengths and weaknesses of the Hill criteria for ED assessment see Bergman et al. (2015).

The purpose of this review is to gain a broad picture of the severity of endocrine-related effects of current-use chemicals in wild populations in comparison with those caused by legacy chemicals, and to assess whether stricter regulatory testing regimes based on new EDS-sensitive assays are likely to detect potential environmental problems which would otherwise be missed.

It should be noted that one area of “current-use” that will not be dealt with in this review concerns the continuing effects of naturally-occurring phyto-estrogens (e.g. β-sitosterol), phyto-androgens (e.g. stigmastanol), and other natural materials in pulpmill effluents. These have a range of adverse effects on downstream fish populations, and the reader is referred to various reviews (Servos et al. 1996; Pait and Nelson 2002; Jobling and Tyler 2003; Van den Heuvel et al. 2010) for more information.

2. Evaluation of reports of ED which may be linked to currently used substances

Research on ED in wildlife has been repeatedly reviewed over the past 20 years (e.g. Ankley and Giesy 1998; Tyler et al. 1998; Van Der Kraak 1998; deFur et al. 1999; IPCS 2002; Jobling and Tyler 2003; Matthiessen 2003; Zala and Penn, 2004; Mills and Chichester, 2005; Porte et al. 2006; Bergman et al. 2012; Kortenkamp et al. 2012). This paper does not attempt to be another exhaustive review. Its purpose is to evaluate field studies which have identified apparent ED (or at least endocrine activity which might be causing adverse effects) resulting from current-use chemicals, to consider the strength of the evidence, and to assess the likely population impacts of the effects in comparison with those caused by legacy EDS. Some laboratory experiments which help to interpret the field data are also referred to.

2.1 Impacts on invertebrates

A major difficulty with the study of ED in invertebrates is that their hormone systems are poorly understood by comparison with the vertebrates (DeFur et al. 1999). Much more research is therefore needed on how invertebrate hormones operate, and in the meantime our knowledge of ED in this large and disparate group will remain limited. However, there is no doubt that at least some invertebrates are subject to ED, as the cases of growth regulators in insects and of tri-organotins in mollusks make clear (see Table 1). Nevertheless, even in the latter well-studied case, there remains a lack of clarity about the precise mode of endocrine action (Oehlmann et al. 2007).

An exception to this general concern can be found in the insects whose endocrine systems have been closely studied in order to assist with culturing beneficial insects (e.g. silkworms and honeybees) and for the development of insecticides. In particular, the functions of the insect hormones (hydroxy-)ecdysone and juvenile hormone are well-understood, and various insecticides known as insect growth regulators (IGRs) have been specifically developed to disrupt these systems (Weltje 2013). This is not the place to discuss the effects of these insecticides on target arthropod pests, but it would be relevant to summarize their unintended effects on non-target insects and other arthropods such as crustaceans which share similar hormone systems.

However, there are almost no reports of such effects in the field, although whether this is absence of evidence or evidence of absence is unclear. There is no doubt, however, that several IGRs are able to interfere not only with insect development, but also that of crustaceans (McKenney 2005). In one case (Walker et al. 2005), it was suspected that the juvenile hormone antagonist methoprene, which had been used for mosquito control, might have been responsible for a large decrease in lobster (Homarus americanus) catches in Western Long Island Sound (WLIS) in the USA. Laboratory experiments established that it is highly toxic to stage II lobster larvae, reducing survival significantly over 72 h at a nominal concentration of 1 µg/L. However, Walker et al. (2005) presented no data on methoprene concentrations in WLIS, and the mosquito control campaign in the area had used malathion, resmethrin, and sumethrin simultaneously with methoprene, making attribution of causes very difficult. Biggers and Laufer (2004) found other substances with vertebrate endocrine activity in sediments and lobsters of WLIS, and speculated that the effects seen in lobsters may have been caused by the combined effects of several contaminants, but it is not possible to draw conclusions about whether the decrease in lobster catches was a result of ED, any other type of chemical exposure, or non-chemical factors.

Other mosquito control campaigns have used methoprene sprayed over wetlands such as mangrove swamps and salt marshes, and some studies have been conducted of ecological side-effects (e.g. Lawler et al. 1999; Russell et al. 2009). However, at the rates of methoprene application used for mosquito control (∼10–20 g/ha), the side-effects reported for crustaceans and insects have been minor or non-existent, although the available data are sparse.

The main body of information about probable ED in invertebrates in the field, with the exception of that related to organotins, concerns crustaceans and mollusks exposed to estrogenic substances including ethinylestradiol (EE2), mainly via sewage treatment works (STW) discharges. These studies were conducted on the assumption that invertebrate hormone systems might have some similarities with those of vertebrates (Jobling et al. 2004), although in fact they are quite different. It should be borne in mind when studying this work that although mollusks seem to be responsive to vertebrate sex steroids, they do not appear to contain the vertebrate estrogen receptor (Scott 2012, 2013), so the mode of action of estrogens in these taxa remains speculative.

Moore and Stevenson (1991) described the presence of intersexuality in benthic harpacticoid copepod crustaceans (Paramphiascella hyperborea) found in the Firth of Forth, Scotland. Twenty-eight out of 30 individuals (93%) collected near a STW discharge were found to be intersex on the basis of morphological indicators, but the condition was found to be very rare in other samples from the Forth and elsewhere. Unfortunately, no firm conclusions can be made about the possible causation of this effect as measurements were not made of vertebrate estrogenic activity at the study site, and no attempt has been made to replicate the effect under laboratory conditions.

A similarly inconclusive report (Sangalang and Jones 1997) described the presence of intersexuality in lobsters (H. americanus) caught in the coastal waters of Nova Scotia, Canada. Intersexuality was manifested as the presence of oocytes in testes. Although some of the affected animals were caught near sewage discharges, no measurements were made of vertebrate estrogenic activity, and the authors of the study were unable to conclude whether the effects were a natural background phenomenon or related to estrogen exposure. Once again, no attempt was made to investigate these effects in the laboratory.

A more conclusive study was made by Chesman and Langston (2006) and Langston et al. (2007) of the bivalve Scrobicularia plana collected from the Avon estuary in southwest England. This gonochoristic species was collected from the estuary at monthly intervals for 17 months, and histological examinations were made of the gonads. Between July and August, the proportion of males fell to 28–35%, a statistically significant alteration of the sex ratio. A mean of 21% of the individuals were intersex (males with ovotestis), while samples from six other estuaries in southwest England showed no effect on sex ratio. The authors (Langston et al. 2007) also reported results from 10 other estuaries in the area, in which 17 out of 23 populations displayed intersex, with up to 60% of males being affected. It was speculated that the effects in the Avon estuary might have been caused by an estrogenic sewage discharge, but the data cannot be directly used to support this conclusion as no measurements of estrogen contamination were made. However, Langston et al. (2007) also conducted experiments in which they exposed maturing S. plana for 1 month to sediments spiked with mixtures of vertebrate estrogens and their mimics (nominal concentrations of 100 µg/kg wet wt. estradiol [E2] and EE2, plus 1000 µg/kg wet wt. octylphenol (OP) and nonylphenol – NP), after which the treated animals were transplanted to the Avon estuary for a further 4 months. There was no confirmation of these exposure concentrations, and the laboratory exposures were well above those found under most environmental conditions. However, this dosing regime led to a statistically significant mean of 44% of the males showing signs of intersex compared with 0–7.7% of controls, and the treatment also caused small (14–27%) but statistically significant increases in mean oocyte diameter. It therefore appears possible that the longer exposures experienced by wild S. plana may have been responsible for the levels of intersex observed in these populations.

The subject of biomarkers of estrogen exposure in mollusks has been reviewed by Porte et al. (2006), and this review will only describe the more significant field studies. Mollusk data of a more convincing nature than those discussed above have been obtained from bivalve populations in Canada. Wild freshwater mussels (Elliptio complanata) (10 per site) were collected by Gagné et al. (2011) from upstream and downstream of two treated sewage effluent discharges in the St Laurence River. At both effluent discharge sites, the proportion of females increased significantly from 30% upstream to 80% downstream, and male vitellogenin (VTG)-like protein (measured as alkali-labile phosphate (ALP), which is not entirely specific for VTG) was elevated downstream by between 50 and 66%. Elevation of VTG in males is a well-known marker of estrogen exposure in fish and amphibians (see below), and it has been presumed that the same applies to mollusks, although this has been challenged (Scott 2012, 2013). VTG-like protein (ALP) in females was also apparently increased at the downstream site, but this was not statistically significant. Female gonads downstream were in early vitellogenesis, but egg production was not underway and downstream gonadosomatic indices (GSI) were low. Experimental work (Gagné et al. 2001) has been able to reproduce some of these effects in the laboratory, and an attempt was also made to study caged E. complanata at the field sites where effects had been noted (Bouchard et al. 2009). Unfortunately, this latter study was flawed due to very high (60%) mortality in the downstream cages, and ALP was induced at only one of the two sites. Gagné et al. (2001) also conducted a study with E. complanata held downstream from a single STW discharge and showed elevations of ALP of >50% compared with the upstream reference, but mortality rates in the caged mussels were not reported. They also showed that E. complanata has specific estrogen binding sites on cytosolic proteins, which may be the receptor mediating the increase in ALP.

Sounder experimental work was conducted by the same research group (Quinn et al. 2004) using freshwater zebra mussels (Dreissena polymorpha) held in tertiary treated sewage effluent for 112 days during gametogenesis and all remained in good health. ALP increased in both sexes by up to 100% in males and over 300% in females. The volume of testicular interstitial tissue also increased, although no adverse apical effects were reported. Analysis of the effluent showed that E2, EE2, and bisphenol A (BPA) were all present, although a definitive quantification was not conducted.

A field study has also been conducted of possible estrogenic effects in European marine bivalves (Ortiz-Zarragoitia and Cajaraville 2010). Samples of 15–20 mussels (Mytilus galloprovincialis) were taken every 2 months over 15 months from two Spanish estuaries, the Oka and the Abra. Both have a history of urban and industrial pollution, but the Oka was also affected by polycyclic aromatic hydrocarbons (PAH) from the Prestige oil tanker spill in the year prior to sampling. There was no true reference site although the Abra estuary has been remediated to some extent. Up to 75% oocyte atresia was observed in the Oka and up to 25% in the Abra, although it should be noted that atresia can be a natural occurrence at certain stages of the life cycle, and it may not solely be the result of ED. ALP in females only varied as expected with the reproductive cycle, and there were no differences in male ALP between the two sites, except for one date when ALP in Oka males was nearly double that in the Abra males. Given that ALP was only elevated in males on one occasion in the more contaminated estuary, and that estrogen concentrations were not reported, these data do not provide clear evidence of ED. The severe oocyte atresia in the Oka may have been related to ED, but seems more likely to have been due to PAHs interacting with the aryl hydrocarbon (AH) receptor rather than the result of an estrogenic effect.

In summary, the field evidence for the impact of insect hormone regulators on crustaceans is essentially non-existent, and the same applies to reliable studies on the possible effects of vertebrate estrogens on that group. However, there is some evidence that bivalve mollusks have shown various forms of feminization and consequent presumed reproductive impairment as a result of exposure to presumably estrogenic sewage effluents or other sources of estrogens. The causal evidence for this is rather weak, and mechanisms of action have not been clarified, but laboratory experiments (in addition to those of Langston et al. (2007) which used very high exposure concentrations) in which bivalves have been exposed to vertebrate estrogens provide limited support for the hypothesis. For example, Nice et al. (2003) ran an experiment in which Pacific oyster (Crassostrea gigas) larvae (7–8 d post-fertilization) were exposed for 48 h to NP concentrations (nominal: ranging from 1 to 100 µg/L) and then were reared in clean water for 10 months, during which reproduction and development of the next generation took place. Nice et al. (2003) showed that up to 30% of the surviving adults were functional hermaphrodites, while no hermaphrodites appeared in the controls. These changes in sex ratio resulted in reduced gamete viability, which led to poor embryonic and larval development in the next generation. However, the reliability of this study is questionable given that measured concentrations of NP (from <1 to 2–9 µg/L) were much lower than the nominal values and sample sizes appeared small.

Despite these observations, no studies have been made of damage to invertebrate populations exposed to EDSs other than organotins, so it is unknown whether the biochemical, physiological, and histopathological changes reported above have led to population-level impacts in practice. This situation may change when the endocrinology of invertebrates becomes more thoroughly understood, and it should not be assumed that this large and important group of organisms is unaffected by currently-used substances.

2.2 Feminisation of fish, amphibians, and reptiles related to sewage discharges

Feminisation in fish and other aquatic vertebrates has been intensively studied since it was first discovered in the 1980 s that sewage discharges were able to induce VTG in caged male rainbow trout (Oncorhynchus mykiss) (Purdom et al. 1994). Most STW effluents contain a complex mixture of natural and synthetic estrogens, some of which (e.g. E2) are steroidal and some non-steroidal (e.g. BPA) (Rutishauser et al. 2004). Most that have been identified do not fall under the category of current-use chemicals, being either natural substances like E2 or partially phased out and/or very weakly potent substances such as NP and BPA. An exception to this is EE2, which is still a major component of oral contraceptives and appears in most treated STW effluents at low ng/L concentrations. On the basis of in vitro assays, Rutishauser et al. (2004) report EE2 to have a potency relative to E2 of only 1.19, but the situation is different in vivo. Thorpe et al. (2003), for example, showed that EE2 is 11–27 times more potent than E2 at inducing VTG in juvenile rainbow trout (O. mykiss), while E2 is in turn 2.3–3.2 times more potent than estrone (E1). Furthermore, EE2 is considerably more resistant to degradation than most other estrogenic steroids. This means that for many treated STW effluents, EE2 is likely to be the major contributor to overall estrogenic activity by comparison with the other steroids. Non-steroidal substances such as NP are likely to contribute only a very small fraction of overall activity, again because their potency at the estrogen receptor is extremely low (by at least a factor of 10⁴) in comparison with E2 and EE2 (Jobling and Sumpter, 1993). For the purpose of this review, it is therefore considered appropriate to examine reports of estrogenic effects in wild fish and amphibians living near sewage discharges, and to ascribe a significant proportion of these effects to EE2. The opposite effect, i.e. masculinization of female fish, has been reviewed by Matthiessen and Weltje (2015) but this phenomenon appears to be uncommon and of lower importance compared to feminization.

The major body of work on estrogenic effects in fish emanates from the research teams at Brunel and Exeter Universities, United Kingdom, who have thoroughly studied the relationship of estrogenic effects in roach (Rutilus rutilus) and exposure to treated sewage. Jobling et al. (1998) took roach from eight UK rivers and showed that in five, the proportion of intersex males was between 40 and 100% compared with <20% in reference site fish. Intersex was defined as the presence of ovotestis or feminized gonoducts. Furthermore, fish from all but one site had more marked intersex severity than reference fish. Mean male plasma VTG was elevated by factors of up to 100 in intersex fish, and by up to >10 in histologically normal males. Even more severe feminization was observed in two other rivers, the Aire and Nene (Jobling et al. 2002a). In subsequent work (Jobling et al. 2002b), it was shown that only 17.4% of moderately intersex fish and 33.3% of severely intersex male fish were able to produce milt, compared to 97.6% in intersex-free reference male fish. If milt was produced by intersex males, its volume was reduced by 50% in comparison with histologically normal males, and sperm motility and fertilization success were also reduced by 50%. Although this was a significant degradation of reproductive ability, it should be noted that most roach populations do not appear to be under threat (Freyhof and Kottelat 2008; Johnson and Chen 2017).

However, although estrogen concentrations were not simultaneously measured at the time of sampling, Jobling et al. (2006) showed that both severity and incidence of roach intersex at 45 sites were positively correlated with the predicted estrogenic risk modeled on the basis of expected estrogen inputs from upstream STW discharges. Furthermore, a range of experiments with roach and other species exposed to STW effluent dilutions have been able to replicate many of the effects reported from the field (Rodgers-Gray et al. 2001; Bjorkblom et al. 2009; Lange et al. 2011). In addition, Harris et al. (2011) showed in an elegant experiment with roach taken from estrogen-contaminated locations that reproductive performance (fry production) in semi-natural breeding groups was negatively correlated with individual intersex severity, and reproductive performance was reduced by up to 76% in the most severely intersex individuals. It has also been convincingly demonstrated (Geraudie et al. 2010) in 474 roach sampled over 18 months from a site where there was no detectable estrogenic or mutagenic activity that no intersex fish were present, and mean male plasma VTG was only 24 ng/ml (rising to 120 ng/ml during the spawning period). The sex ratio did not deviate from 1:1. This clearly shows that the natural background rate of estrogenic abnormalities in roach can be very low, although this may not be the case in all fish species (Bahamonde et al. 2013).

Finally, although the severity of intersex in roach appears to increase with age, it should be pointed out that mildly feminized roach exposed to estrogens for less than a full life cycle are probably not reproductively compromised. This was demonstrated by Hamilton et al. (2015) who showed experimentally that the male offspring of STW effluent-exposed females had only weakly feminized testes after exposure to 100% STW effluent for up to 3 years and 9 months, and they were able to reproduce normally. Furthermore, there was no evidence that the exposure history of the females had any influence on the reproductive performance of their male offspring. It is also worth noting that exposure to estrogens can even enhance reproductive success in fish by increasing fecundity, although it is not known if this can also lead to increased population size (Parrott et al. 2017).

Field studies with other fish species caught near STW discharges around the world have produced similar results to those obtained with roach (e.g. Aravindakshan et al. 2004; Kavanagh et al. 2004; Kirby et al. 2004; Game et al. 2006; Leusch et al. 2006), although effects may be hard to detect in some places (e.g. Pottinger et al. 2011). Furthermore, species vary considerably in their sensitivity to estrogens (e.g. Caldwell et al. 2012). In several such studies, estrogen concentrations in the sampled environment were measured, thus making the link between estrogen exposure and reproductive effects stronger (Vethaak et al. 2005; Vajda et al. 2008). For example, Vethaak et al. (2005) reported median estrogen concentrations in waters receiving Dutch STW effluent in 1999 of 0.4 ng EE2/L, 1.0 E2/L, 1.0 ng E1/L, 45 ng BPA/L, 300 ng OP/L and 990 ng NP/L, and Vajda et al. (2008) reported that mean E2-equivalents for a receiving water in the USA in 2003 and 2005 were 3.4–11 ng/L. At 35 sites on five UK rivers, Johnson and Chen (2017) predicted similar mean E2-equivalents (0.6–3.2 ng/L). Taken overall, these combined concentrations of estrogens are largely able to explain reported incidence of feminization in fish. Finally, a meta-analysis of VTG induction field data from 13 fish species from around the world (Desforges et al. 2010) showed that human population size (a surrogate for estrogen and other anthropogenic contamination) upstream of the sampled fish populations explained 28% of the variation in male VTG.

There have been many other reports of putative estrogenic effects in wild fish (e.g. Hashimoto et al. 2000; Nagler et al. 2001; De Metrio et al. 2003; Scott et al. 2006, 2007; Körner et al. 2007), but these cannot all be directly linked to sewage discharges and EE2, due to often remote sampling locations. Evidence suggests that some of these effects may be related to the bioaccumulation of persistent estrogenic materials (such as some organochlorines), and as these have generally been phased out of use (even though they still exist in the environment), they fall outside the scope of this review.

Despite this large body of work with fish, much of it considered robust and reliable, the question still remains about whether the undoubted estrogenic effects on reproductive variables that have been observed are a threat to fish populations (Johnson and Chen 2017). An approach to this question was made by Kidd et al. (2007) and Palace et al. (2002, 2009) who dosed an experimental lake in Canada (Lake 260) with EE2 three times weekly for 3 years during the ice-free season, and followed the fathead minnow (Pimephales promelas) population for 7 years. Annual mean measured levels of EE2 ranged between 4.8 and 6.1 ng/L, with a maximum weekly mean of 8.1 ng/L (Park and Kidd, 2005), but it should be pointed out that these concentrations are approximately an order of magnitude higher than those generally found in rivers downstream of treated sewage discharges. Two comparable lakes were used as EE2-free reference sites. VTG levels in whole-body homogenates of male fatheads rose from ∼0.5 µg/g wet wt. to 2000–12,000 µg/g after 7 weeks. In Year 1 of the treatment, male testes displayed delayed spermatogenesis, widespread fibrosis and malformed gonoducts. In Year 2, male GSI was only 0.40 compared with 1.39 and 2.27 in the reference lakes. By Year 3 of the treatment, four out of nine males were shown to have ovotestis (the small sample size was due to a population crash). By Year 4 (i.e. 1 year after treatment had ended) all reproduction had ceased and catch per unit effort had dropped from between 5 and 100 before treatment to 0.7, and the population was almost extinct.

Interestingly, the populations of three other fish species in the lake experiment with EE2 did not collapse, though two declined, and there were indirect secondary consequences for some invertebrate populations in the ecosystem (Kidd et al. 2014). The authors speculate that the differences in fish species response could be due to inherent differences in sensitivity, differences in exposure due to habitat requirements, differing longevity, and differing stages of development during the treatment. Furthermore, the relationship between sex ratio and reproductive success (the so-called mating function) varies widely between fish species, and this also has an impact on the population response to endocrine disruptors (White et al. 2017). It is apparent from these results that VTG induction does not automatically indicate that the population will collapse, but its predictive value for population-level effects is enhanced if combined with measures of abnormal gonadal histopathology.

Apart from the experiment in Lake 260, there have not been any other unequivocal demonstrations of fish population declines or collapses that can be firmly attributed to EE2 or to estrogens generally. This was pointed out over 10 years ago by Mills and Chichester (2005) and little has changed. This is at least partly due to the difficulty of measuring fish reproduction and population variables under field conditions and making firm associations with causative factors. The lack of evidence for population declines is probably also partly attributable to the fact that some fish populations which are targeted by anglers (e.g. roach in the UK) are re-stocked regularly. Despite their expense and complexity, there is clearly a need for further large-scale experiments like that of Kidd et al. (2007), and it would also be helpful if improved population models for important fish species could be employed to allow more confident extrapolation from laboratory data. However, there is little doubt that the estrogens in treated sewage discharges are able to cause adverse effects in some downstream fish populations with potential implications for their stability, and that EE2, due to its high relative potency compared with E2, plays a major role in this.

There is no evidence for estrogenic effects in wild amphibians related to sewage discharges, although it is known from laboratory experiments that in males VTG can be induced by estrogen exposure and they can experience a range of effects on reproductive potential. In one of these (Sowers et al. 2009), northern leopard frogs (Rana pipiens) were exposed from egg to metamorphosis (∼2 months) to 0, 10, 50, and 100% treated STW effluent. The total mean measured estrogenic activity of the effluent was 1.7 ng E2 equiv/L, with a mean EE2 concentration of 0.21 ng/L. Relative potencies for this calculation were obtained from in vitro data by Rutishauser et al. (2004). 50 and 100% effluent caused a 4 and 7 d delay in metamorphosis, respectively, as well as producing 37 and 64% ovotestis, respectively. There were no effects on sex ratio or female gonadal development, but nonspecific effects on thyroid histology. It is likely that a 7 d delay in metamorphosis, which may or may not have been due to ED, could have significant consequences for survival and reproductive success in adult amphibians (e.g. Semlitsch et al. 1988).

The only field-relevant data for effects in amphibians caused by estrogen exposure derives from the EE2 experiment in Lake 260, Canada (Park and Kidd 2005). As well as showing that caged amphibian larvae were adversely affected by the EE2 concentrations in the lake (reduced hatching success in green frogs Rana clamitans between 4.8 and 6.1 ng/L), they found that larvae of wild mink frog (Rana septentrionalis) sampled in 2001, 2002, and 2003, possessed intersex gonads with an incidence of 2.4, 0 and 28.6%, respectively. No intersex gonads were observed either before EE2 dosing started, or in the reference lakes, and there were no effects on sex ratio in Lake 260. It is not known to what extent intersex testes in mink frogs would be damaging for reproductive success, but by analogy with fish (e.g. Harris et al. 2011) one might expect severe intersex to have such effects.

As with amphibians, there are almost no field studies of reptiles exposed to estrogen-containing sewage effluent. Tada et al. (2007) studied VTG induction in 320 male freshwater turtles (Chinemys reevesii) caught from four pond sites, three of which were contaminated with sewage-derived estrogens (0.52–1.7 ng/L E2 equivalents, determined by a yeast estrogen screen). Only five turtles showed elevated serum VTG levels (1.1–5.9 µg/ml) compared with the other males from the contaminated sites (0.1–0.74 µg/ml), and overall there were no significant differences between the contaminated and reference ponds. It appears from other studies (e.g. Irwin et al. 2001) that aquatic reptiles may be insensitive to estrogen concentrations in the surrounding water.

2.3 Feminisation of amphibians related to intensive agriculture

It has been suggested that amphibian populations in the USA have become feminized in areas of intensive agriculture (Hayes et al. 2002). More specifically, it has been hypothesized that exposure to the triazine herbicide atrazine, which is very widely used for control of weeds in intensively grown maize and soybean crops in certain countries, may be causing elevations in aromatase enzyme activity, which could lead to inappropriate increases in E2 titers in male amphibians and consequent feminization (Hayes et al. 2002).

Hayes et al. (2003) conducted both laboratory experiments and field studies on this issue. In the experiments, newly-hatched leopard frogs (R. pipiens) which developed from eggs collected in the field (Sensiba Marsh, Wisconsin) were treated with atrazine in a static-renewal design at nominal concentrations of 0, 0.1, and 25 µg/L. Actual concentrations were reported to have been measured but no data were presented. All males were sexually differentiated at metamorphosis, but 36 and 12% at 0.1 and 25 µg/L, respectively, were claimed to be showing testicular dysgenesis syndrome (underdeveloped testes, poorly structured testicular lobules, and low numbers of germ cells). However, the data underpinning this claim are not presented, and no statistics are presented. An additional claim, that 29% of the frogs at 0.1 µg/L and 8% at 25 µg/L showed “varying degrees of sex reversal”, i.e. oocytes in the testis, is also not backed up with statistics. The data did not conform to a monotonic concentration-response, although that does not necessarily mean that they are erroneous.

In the fieldwork by Hayes et al. (2003), 800 R. pipiens were sampled from each of eight sites across the central USA in 2001, the aim being to take newly metamorphosed individuals. Sex and gonadal histopathology were recorded from a subset of individuals. No atrazine was detected at a reference site, but all other sites contained atrazine, at 0.2–6.7 µg/L. As with the laboratory experiment, no statistics were presented, but it was claimed that no testicular disorders were present at the reference site, while ovotestis at incidences of 10–90% occurred at the other sites. One site also showed gonadal dysgenesis (poorly developed testicular lobules), but no abnormalities were seen in females. There was no apparent relationship with atrazine as the highest incidence of testicular abnormalities occurred at 0.2 µg/L. Analyses for other pesticides (simazine, hexazinone, diuron, and norflurazon and several others) were negative in all but one case (0.39 µg/L metolachlor).

Hayes et al. (2003) suggested that atrazine is likely having a significant impact on amphibian populations, but their data as reported are unable to support this conclusion. Nevertheless, their work triggered a large number of studies aimed at investigating the matter further.

Attempts to replicate the experimental results of Hayes et al. (2003) have not been successful. For example, Jooste et al. (2005) conducted an outdoor microcosm experiment in South Africa in which 4 d post-hatch clawed frogs (Xenopus leavis) were exposed to atrazine in 1100 L microcosms for up to 10 months. Nominal atrazine concentrations (three replicates/concentration) were 0, 1.0, 10, and 25 µg/L, and weighted mean measured concentrations were 0, 1.4, 12.1, and 30.8 µg/L. Test solutions were replaced after 80 d. At metamorphosis (larval stage 66), mean incidence of ovotestis (i.e. oocytes present within the testicular tissue) was 57, 57, 59 and 39%, respectively, and mean numbers of oocytes per male were 9.5, 9.8, 8.5, and 11.1, respectively. Mean numbers of eggs in the resulting juvenile males at 10 months were low (≤2 per individual). There were no statistical differences between treatments and control and no relationship between atrazine concentrations and ovotestis.

Du Preez et al. (2008) ran a well-conducted experiment in which Xenopus laevis were raised in mesocosms from 96 h post-fertilization in measured concentrations of atrazine (0, 1.1, 10.4, and 24.8 µg/L, static-renewal) and then allowed to breed while continuing to be exposed. Treated male frogs were mated with both treated and untreated females, and a range of reproductive endpoints were measured in both F1 and F2 groups. In summary, there were no treatment-related effects, either on clutch size of the F1 adults, or on hatching success, time to metamorphosis or sex ratio of the F2 offspring. There were low background incidences of segmented and single testes, and ovotestis [termed testicular ovarian follicles (TOF) 5–15% in F2 frogs], but these occurred randomly and were not treatment-related.

In addition, two thorough, blinded experiments with X. laevis larvae (Kloas et al. 2009) exposed to atrazine concentrations up to 100 µg/L from 8 days post-fertilization (dpf) to 83 dpf or metamorphosis failed to show any effects on sex ratio or incidence of intersex. Thus, the experimental work by Jooste et al. (2005), Du Preez et al. (2008), and Kloas et al. (2009) has consistently failed to find reproductive effects from atrazine in clawed frogs.

In other reliable studies, a variety of experiments with the same species as that used by Hayes et al. (2003), R. pipiens, have also failed to find significant effects of environmentally-relevant atrazine concentrations on a range of variables including hatchability, survival, time to metamorphosis, metamorphosis success, growth, sex ratio, GSI or gonadal abnormalities such as ovotestes (e.g. Allran and Karasov, 2001; Orton et al. 2006; Langlois et al. 2010; Knight et al. 2013).

However, considerable attention has been focused on amphibian populations living in areas of intensive agriculture. Reeder et al. (1998) sampled 341 cricket frogs (Acris crepitans) from up to 8 pond sites in Illinois, USA, in 1993–95, of which 2.6% overall showed ovotestis. At some sites, the incidence of ovotestis was up to 25%. A long list of contaminants in water was analyzed for, but only atrazine, metolachlor, cyanazine and chlorpyrifos were detected. Concentrations of atrazine ranged between 1 and 70 µg/L. There was no significant relationship between presence of ovotestis at a site and atrazine (p = 0.07). However, there was a significant relationship between sex ratio alteration (excess females) and polychlorinated biphenyls/polychlorinated dibenzofurans (PCBs/PCDFs), so it should be considered that legacy substances such as some organochlorines may be an underlying cause of these reproductive anomalies in amphibians.

Hecker et al. (2004) sampled adult X. laevis from eight sites in South Africa, some in maize-growing regions exposed to atrazine and others in non-maize-growing regions, with between 10 and 20 males and females taken per site. Water samples were analyzed for simazine, atrazine and its major metabolites, and terbutylazine. The frogs were analyzed for plasma testosterone (T) and E2, gonadal aromatase activity, and GSI. Atrazine concentrations ranged from <0.1 to 4.1 µg/L, and combined atrazine metabolites were ≤2 µg/L. There were no correlations between gonadal aromatase activity or GSI and concentrations of any measured agrochemicals, but there were negative correlations between atrazine and/or its metabolites and plasma T in females and males. There was also a negative correlation between atrazine and a metabolite and E2 in females. These effects on steroid hormones were of questionable biological significance (1.7–3.5-fold reductions in these hormones). Overall, atrazine or co-applied pesticides did not appear to be damaging sex steroid homeostasis, although this possibility could not entirely be ruled out.

Murphy et al. (2006a) sampled green frogs (R. clamitans) and other frog species for 2 years from three areas of intensive maize growing and three nonagricultural areas in Michigan. Atrazine levels in water were generally low (≤2 µg/L), although at one site the maximum measured concentration was 250 µg/L. Biological variables measured included GSI, plasma T, E2, 11-ketotestosterone (11-KT), and gonadal aromatase activity. Atrazine concentrations were not correlated with any biological variable. Plasma hormone levels varied between areas, e.g. the E2/T ratio was elevated in adult males and females in agricultural areas in 2002, but not in 2003, while in juvenile males, E2/T was elevated in 2003. There were insufficient male aromatase data for statistical analysis, but female aromatase was elevated at agricultural sites in 2002 and not in 2003. In a second paper from the same study, Murphy et al. (2006b) reported the incidence of hermaphrodites (i.e. intersex individuals) in three species, R. pipiens, R. clamitans and R. catesbeiana. There was a low incidence of testicular oocytes at both agricultural and nonagricultural sites, with greatest incidence in juvenile R. pipiens. There were no consistent differences between agricultural and nonagricultural areas. However, in one year (2003) the incidence of ovotestis in juvenile R. pipiens was higher at agricultural than nonagricultural sites. Ovotestis overall was not correlated with mean atrazine levels, but ovotestis in juvenile R. pipiens was correlated with maximum atrazine levels in 2003 only.

Smith et al. (2005) sampled 207 adult clawed frogs (X. laevis) in autumn 2002 from maize-growing areas (MGA) and non-MGA (NMGA) in South Africa, the sampling sites being those described by Hecker et al. (2004). A male secondary sexual characteristic (laryngeal mass), and testicular histology were measured. There were low incidences of ovotestis in MGA (2%) and NMGA (3%), and no differences in laryngeal mass or testicular cell volumes (especially spermatocytes and spermatozoa) between areas. There were no correlations with the presence of atrazine.

Northern leopard frogs (R. pipiens) and green frogs (R. clamitans) were sampled by McDaniel et al. (2008) every autumn for 3 years from up to 33 sites in intensive agriculture areas (maize and soy), two sites in agricultural reference areas, and four nonagricultural reference sites in SW Ontario, Canada. A large suite of pesticides was analyzed for in the water (including atrazine), and frogs were analyzed for plasma VTG, T, 11-KT, and E2; gross morphology, and male gonad histopathology. Ovotestis (termed testicular ovarian follicles or TOF) was significantly more prevalent in intensive agriculture areas (42%) compared with reference sites (7%), but VTG was only detected in one male from an intensive area, and no other significant differences were observed. The effect on ovotestis did not correlate with atrazine alone, but did correlate with a mixture of pesticides and nutrients (including atrazine), and the numbers of pesticides per site also correlated with effects. Median atrazine levels were 0.068–0.78 µg/L in intensive agricultural areas, 0.045–0.39 µg/L in reference agricultural areas, and LOD-0.090 µg/L in nonagricultural reference areas. The authors concluded that the data provide only limited evidence for estrogenic activity, and testicular oocytes were generally present in individuals from all areas at low levels.

Other studies have not explicitly looked for the effects of atrazine, but have simply reported levels of abnormalities in amphibians from agriculturally-intensive and non-intensive areas without analyzing for pesticide residues. For example, Mosconi et al. (2005) found relatively minor biomarker effects in R. esculenta from an intensive farming area in Italy by comparison with a relatively pristine area. McCoy et al. (2008) found somewhat greater effects in Florida toads. 20 or more cane toads (Bufo marinus, now known as Rhinella marina) were collected over 2 years from each of five areas with agricultural activity ranging from 0 to 97% of the area within a 5.6 km² zone around the sampling point. The number of male gonadal abnormalities increased from 2 to 5 along the agricultural gradient, and the frequency of intersex gonads increased from 0 to 40%. Male T titers (but not E2) decreased, and secondary sexual characters were either feminized (skin mottling score up from 4.5 in nonagricultural site males to eight in intersexes) or demasculinised (forearm width reduced from 1.5 to 1.3 cm in intersexes; number of nuptial pads reduced from 2.5 to 1.5 in intersexes). Overall, males from agricultural areas had hormone levels and secondary sexual characters that were intermediate between intersex toads and toads from nonagricultural areas.

Reeder et al. (2005) tackled the issue from a new direction by measuring the presence of intersex in archived frog samples. Archived gonads of 814 cricket frogs (A. crepitans) sampled in Illinois between 1852 and 1996 were examined for ovotestis. This showed the highest incidence of intersex (11.1%) occurred in 1946–59, declining to 2.7% in 1980–96, a level similar to 1852–1929 (1.2%) before the widespread use of synthetic chemicals. The paper does not include data on levels of contaminants in the frogs, but the authors suggest their results may be explained by the high levels of dichlorodiphenyltrichloroethane (DDT) and PCB use in the 1940 s and 50 s. They also point out that intersex levels were high before atrazine was introduced on maize in Illinois (1959), and that intersex had declined to near background levels in 1980–96, even though atrazine was being very heavily used by then in Illinois. It was also noted that intersex was most common in urban and industrial areas, a fact which also suggests that atrazine (or other agricultural chemicals) was not the main causative factor.

Knutson et al. (2004) took a strictly ecological approach and addressed the issue of whether intensive farming operations are affecting amphibian populations. The study was conducted in 2000/2001 in 40 farm ponds in Minnesota, comparing the impact of land use in their immediate vicinity (intensive maize/soybean growing, livestock rearing, or low intensity agriculture) on their amphibian populations. Total species richness over a season and reproductive success (numbers of eggs and larvae), were measured over the 2 years. Up to 10 species were recorded, including R. pipiens, which had been claimed by Hayes et al. (2003) to be sensitive to the effects of atrazine. Knutson et al. (2004) showed that species richness was statistically indistinguishable in ponds from intensive agriculture areas compared with reference areas, while livestock rearing caused some reduction in reproductive success, probably due to disturbance and excessive nutrient inputs (although it seems possible that natural and synthetic steroids may also have played a part). Unfortunately, measurements made of pesticide contamination in the ponds were not published, although levels of atrazine were reported to be low in a sub-set of ponds (<0.1 –0.5 µg/L). Furthermore, the presence of ovotestis and other abnormalities were not recorded. However, it is evident that if sub-lethal biological effects in pond amphibians were being caused by the chemicals used in intensive agriculture, they were not translated through to population or community damage.

Finally, Pickford et al. (2015) studied nine breeding sites of common toads Bufo bufo in England and Wales. Using passive samplers and in vitro assays, they detected weak androgenic and estrogenic activity in water at some locations, but toad hatching rates and low levels of intersex were not correlated either with local agricultural operations or with levels of endocrine activity. There was, however, a negative correlation (r² = 0.45) between the proportion of male toads present after extended exposure and the amount of estrogenic activity as measured by the Yeast Estrogen Screen. It was concluded that substances other than plant protection products were responsible for this effect, although operational difficulties limited any further conclusions.

An assessment of all the available field evidence does not reveal a clear and consistent picture of endocrine activity in amphibians which can be linked to atrazine or intensive maize-growing alone. Furthermore, there are significant concerns about the repeatability of the early work reported by Hayes et al. (2003) – see Van Der Kraak et al. (2014). Some studies (e.g. Knutson et al. 2004; Reeder et al. 2005; Smith et al. 2005; Murphy et al. 2006a, 2006b) have failed to find abnormalities which can be consistently linked to agricultural intensification, while others have reported generally low levels of sexual abnormalities (e.g. Pickford et al. 2015). In some places (McDaniel et al. 2008) there is a correlation between effects, such as incidence of ovotestis, and a complex of different pesticides rather than any single one. However, severity of ovotestis is generally low (one or two oocytes per testis) even though incidence can exceed 40% (McDaniel et al. 2008). Some studies (e.g. McCoy et al. 2008) also report elevated ovotestis incidence in intensively farmed areas without providing evidence for a link with atrazine. In others (Hecker et al. 2004), some correlations of endocrine activity with the presence of atrazine have been found, but they generally involve relatively minor changes in hormonal variables such as plasma T and E2. Correlations do not, of course, prove causality, and the evidence of laboratory experiments with amphibians exposed to atrazine suggests that concentrations of this herbicide found in natural waters pose a minimal risk to amphibians (Van Der Kraak et al. 2014).

Even though there is not a strong link between atrazine concentrations and amphibian sexual abnormalities observed in the field, the effects that have been observed in some places may be partly due to the presence of other substances such as organochlorines, which could be causing mixture effects in certain locations (Reeder et al. 1998), or to other factors related to modern agriculture, which have not been implicated to date or quantified. However, the weight of evidence suggests that even where such effects are occurring in the vicinity of intensive maize-growing, they tend to be of relatively low severity, such as small changes in hormone titers or mild ovotestis. Crucially, there are indications from several species living in farm ponds that these effects are not being translated into damage at the population level (Knutson et al. 2004), and good evidence from one species (A. crepitans; Reeder et al. 2005) that the incidence of ovotestis peaked long before the era of agricultural intensification and use of atrazine.

In a recent weight-of-evidence review by Van Der Kraak et al. (2014) it has been suggested that there are many more likely causes than atrazine-related ED for the global decline in amphibian populations, including the fungal disease chytridiomycosis, habitat loss, and climate change. They also draw attention to the fact that regional withdrawals of atrazine use (e.g. in 2004 in the European Union) have not led to recoveries of amphibian populations.

2.4 Impaired stress response in fish, amphibians, birds and mammals

2.4.1 Fish and amphibians

The response to stress in vertebrates is hormonally mediated by the hypothalamus-pituitary-inter-renal/adrenal (HPI/A) axis. In brief, stressors such as physical threats trigger synthesis of corticotropin-releasing hormone by the hypothalamus, which in turn stimulates release of adrenocorticotropic hormone (ACTH) by the pituitary. The ACTH then causes the inter-renal tissue in the head kidney of fish to produce the active hormone cortisol (corresponding to the closely-related corticosterone in higher vertebrates, which is produced by the adrenal gland). Cortisol/corticosterone together with catecholamines such as adrenalin helps the animal cope with stress by modifying its physiology. Cortisol prepares fish to adapt to the stressor by instigating a suite of changes including stimulation of heart rate, glucose mobilization, fat metabolism and osmoregulation. For a general discussion of these processes, see Pottinger (2003). Altered cortisol levels can cause immunosuppression and may have additional effects on growth and reproduction (Hontela 1998).

Research conducted in Canada in the 1990 s by the research group of Hontela showed that this cortisol response could be damaged in fish and amphibians by exposure to pollutants, and the early work has been reviewed (Hontela 1998). A wider review of pollution damage to the hormonal stress response in vertebrates as a whole was later published by Pottinger (2003). Both reviews made it clear that although interference with the hormonal stress response is likely to have implications for fitness, the precise functional significance of alterations in cortisol titers remained to be uncovered. To date this situation has not materially changed.

Hontela (1998) found that chronic exposure in the field to mixtures of pollutants or single substances can damage the ability of fish and amphibians to make the normal cortisol response to stress. At least some of these effects appear to be a specific impact on the hormone system caused by relatively low contaminant concentrations interfering directly with the production of cortisol rather than a secondary effect caused by a systemic toxic response, and as such they could be regarded as a form of ED (provided an adverse effect is caused). Fish (yellow perch Perca flavescens; Northern pike Esox lucius) and amphibians (Necturus maculosus) caught in areas polluted by mixtures of heavy metals (Cd, Hg, Zn), PCBs and PAHs, or by heavy metals or pulpmill effluent alone, produced 20–50% less plasma cortisol than reference organisms when subject to brief confinement stress. It was also found in ACTH challenge tests that the ability of the inter-renal tissue of the head kidney to produce cortisol had been reduced in field-sampled N. maculosus by 30–70%. The consequences of this effect were not explored but, while not adverse in themselves, it is expected that the ability of fish and amphibians to make metabolic, vascular and immune compensations for stress would be impaired, and that there might therefore be secondary consequences for growth and reproductive success.

Subsequent work has thrown more light on the widespread nature of this response to pollutants in lower vertebrates, and found it occurring at many different types of polluted site. Similar observations to those observed in yellow perch and northern pike from Canada (i.e. a diminished ability to produce cortisol after transient stress) have been made in marbled sole (Pleuronectes yokohamae) from polluted areas of Tokyo Bay (Kakuta 2002), and in caged rainbow trout (O. mykiss) exposed in the field in Canada to treated STW effluent for 14 d (Ings et al. 2011). Further work with yellow perch (Laflamme et al. 2000) also showed that fish from Canadian metal-contaminated lakes (up to: 16 µg/L Zn; 0.6 µg/L Cu; 0.18 µg/L Cd) exhibited a depressed cortisol response. Several of these studies also conducted ACTH challenge experiments with fish and confirmed that the cortisol-producing inter-renal tissues had been damaged (Laflamme et al. 2000; Kakuta 2002).

However, some studies have failed to find these effects in fish from other polluted areas, such as rainbow trout (O. mykiss) and brook trout (S. fontinalis) from selenium-contaminated streams in Canada (Miller et al. 2009a), and white suckers (Catastomus commersoni) from Canadian agricultural drains contaminated with selenium (0.4–26.7 µg/L) and ten pesticides (<0.005–7.3 µg/L) (Miller et al. 2009b).

Other studies have only looked at the basal cortisol level in unstressed organisms. For example, Hopkins et al. (1997) found that basal cortisol levels in male toads Bufo terrestris were elevated by factors of 5–9 in areas contaminated with coal ash waste by comparison with reference sites. A similar increase in basal cortisol titers occurred when uncontaminated toads were transplanted for up to 12 weeks to the ash-contaminated sites. Again, while changes in basal cortisol levels are not adverse in themselves, it is to be expected that large changes would have implications at the apical level.

More recent work on the damaged stress response in fish has been conducted in the UK by Pottinger et al. They have studied three-spined stickleback (Gasterosteus aculeatus) populations living in English rivers receiving treated sewage effluent. This work shows that perturbations of the cortisol response to brief confinement stress are related to proximity to sewage discharges (Pottinger et al. 2013, 2016). Furthermore, long-term water quality data suggest that sewage is not the only source of this perturbation. Variation in 14 water quality determinands explains 30–60% of the variation in hormonal stress reactivity, irrespective of whether the river is sewage polluted or not (Pottinger and Matthiessen 2016b). One of the chemical variables that might be responsible in part for the effects is nitrate, although this remains to be proven, and it appears likely that mixtures of several contaminants may contribute. Experiments in which female fish from areas where the stress response was damaged were held in clean water for 5 months and then stress-tested showed that the damage is unchanged over this period, but it is currently unknown if this is due to genetic or other causes (Pottinger and Matthiessen 2016a).

It should be noted that the mechanism of damage to the stress response system in fish is probably more complicated than simple impairment of the ability of the inter-renal tissues to produce cortisol when stimulated by ACTH. For example, there is some evidence that the pituitary corticotrope cells which produce ACTH itself are atrophied in fish from polluted environments in Canada (Hontela et al. 1992). As pointed out by Pottinger (2003), this could be due to direct toxic effects of pollutants on the pituitary, or to prolonged negative feedback suppression by corticosteroids, or to some other mechanism. It is, therefore, not necessarily the result of a direct interaction with the endocrine system.

2.4.2 Birds

The situation in higher vertebrates such as birds and mammals has some similarities with that in fish, although fewer data are available. For example, Wada et al. (2009) sampled tree swallow nestlings (Tachycineta bicolor) from mercury (Hg) contaminated and reference rivers in Virginia, USA, where relative mean levels of Hg contamination in nestling blood were 354 µg/L compared with 17 µg/L at reference sites. The plasma corticosterone response to confinement stress varied with nestling age. The strongest relationship with Hg was observed in late-stage nestlings (13–17 d) where the corticosterone baseline in contaminated areas was elevated by 103% (from ∼2.5 to 5 ng/ml plasma), while the stress-induced level was depressed by 27% (25 ng/ml at reference site, 18 ng/ml at Hg site) in comparison with reference nestlings. However, in younger nestlings, the corticosterone response in the Hg area appeared to increase by comparison with reference sites, suggesting that the mercury had to bioaccumulate to a certain level before effects occurred. It was not made clear if the reduced corticosterone response to stress was statistically significant.

Mayne et al. (2004) studied nestling tree swallows (Tachycineta bicolor) and eastern bluebirds (Sialia sialis) in 2000/2001 from pesticide-treated and reference orchards in Ontario, Canada. Nests were sprayed with up to 7 individual pesticide products and 5 pesticide mixtures containing a total of 19 active substances, but it was also noted that eggs of swallows and bluebirds from treated orchards contained p,p′-DDE mean levels of 1.14 mg/kg wet wt. and 25.0 mg/kg, respectively, compared to 0.22 and 1.35 mg/kg at the reference sites. In tree swallows, there were no effects on basal corticosterone or on stress-induced levels. However, ACTH challenge produced a statistically significantly higher level of corticosterone response in potentially exposed swallows (mean 82.6 ng corticosterone/ml blood compared with 67.2 ng/ml). There were no correlations of these effects with pesticides. In contrast, potentially exposed bluebird chicks were significantly less responsive to ACTH challenge than reference chicks (mean 45.1 ng/ml blood compared with 58.2 ng/ml), a possible sign of abnormality in the adrenal tissue, and this was negatively associated with p,p′-DDE in eggs. Other studies of adrenal abnormalities in birds have produced less clear-cut results (e.g. Baos et al. 2006).

2.4.3 Mammals

The work of Mayne et al. (2004) on tree swallows and eastern bluebirds is noteworthy, because they showed that residues of a legacy substance (p,p′-DDE), and not a suite of currently used pesticides, were linked to the effects on ACTH responsiveness. Work on adrenal effects in mammals also seems to link endocrine changes with legacy organochlorines. Detailed research, summarized by Bergman (1999) and first described by Bergman and Olsson (1985), described postmortem adrenal hyperplasia in 3 species of seal, mainly 159 gray seals (Halichoerus grypus), found dead around the Baltic Sea between 1977 and 1996. Prevalence of hyperplasia varied from 0 to 100%, and was most common and severe in older animals (>15 years). It was suggested that this condition was related to organochlorine contamination, although residue data were not given in this paper. However, later work strongly suggested that the condition was associated with high body burdens of PCBs and DDT residues, and Bergman et al. (2012) state that the pathological signs are suggestive of Cushing’s Disease (hyper-secretion of corticosterone), although no measurements of corticosterone have been made in wild seals. This pathology has not been observed in seals outside the Baltic, but has been seen in beluga whales (Lair et al. 1997).

It should be noted that studies of adrenal (and other) pathology and its possible links to contaminants in marine mammals should be treated with a degree of caution, because it is impossible to be sure about cause and effect in the stranded animals which form almost the entire sampled population. For example, Kuiken et al. (1993) studied postmortem adrenal pathology and organochlorine contaminant levels in 28 stranded harbor porpoise (Phocoena phocoena) found on the shores of Britain in 1990–1991. Adrenal hyperplasia was present, but it was not associated with elevated levels of organochlorines (HCH, DDE, DDT, dieldrin, PCBs) in blubber. The authors suggested that the hyperplasia probably resulted from the chronic stressors which had led to death rather than from the organochlorines in their tissues. Nevertheless, given the similarity of the cortisol/corticosterone stress response system in all vertebrates, it is not unreasonable to conclude that the effects seen in some fish, amphibians and birds are also potentially occurring in mammals.

In summary, there is a weight of evidence from many sites that the ability of fish, birds, and probably amphibians and mammals to make a normal hormonal response to stress, can be damaged by exposure to a wide range of contaminants acting at low concentrations, some of which are current-use materials (such as heavy metals, and mixtures present in sewage and pulpmill effluent) and some legacy pollutants (such as various organochlorines), although the causal and mechanistic evidence implicating specific substances is rather weak. Furthermore, damage to the cortisol response in individual fish appears to be permanent. The major question which remains to be answered, however, is whether any of these effects are actually causing adverse apical damage in wildlife populations. At present, there is no evidence that this is occurring, so the effects may not constitute ED as such, although it seems reasonable to speculate that the fitness to survive and reproduce of affected individuals may be compromised as a direct result of chronic modulation of their stress hormone systems. It should also be noted in passing that the effects described above are entirely distinct from the suppressive effect which acute stress-induced elevated levels of cortisol can have on circulating steroid hormone titers (e.g. Pankhurst and Van Der Kraak 2000).

2.5 Thyroid disruption in fish, amphibians and birds

The thyroid hormone system in vertebrates is mediated through the hypothalamo-pituitary-thyroid (HPT) axis (Blanton and Specker 2007; Zoeller et al. 2007). Following environmental stimuli acting on the hypothalamus, the pituitary synthesizes thyroid-stimulating hormone (TSH) which in turn triggers the synthesis of thyroxine (T4) and some triiodothyronine (T3) by the thyroid gland. However, the majority of T3 (the active hormone) is synthesized from circulating T4 in peripheral target tissues via the action of 5′-iodothyronine deiodinase. The complexity of this system potentially allows EDSs to impact at many points, including on hormone synthesis, transport, and peripheral activation (Crofton 2008). In fish and amphibians, thyroid hormone is essential for the control of early development and for metamorphosis. In addition, it has a major influence on growth and reproduction. With relatively minor differences, thyroid hormone plays similar roles in the higher vertebrates.

Most of the contaminants which have been implicated in damage to the thyroid system (PCBs; polychlorinated dibenzo dioxins – PCDDs; DDT and other chlorinated pesticides; polybrominated diphenyl ethers – PBDEs) are legacy chemicals which fall outside the scope of this review (see Boas et al. 2006). For example, detailed early work by Leatherland et al. (reviewed by Leatherland 1998) showed that polychlorinated hydrocarbons including various PCB congeners were mainly responsible for thyroid perturbations in fish and predatory birds from the Great Lakes. The link between PCBs and thyroid effects in Great Lakes herring gulls (Larus argentatus) is particularly clear (McNabb and Fox 2003). A more recent example (out of many) involved a study of fish (shiner surfperch Cymatogaster aggregata and Pacific staghorn sculpin Leptocottus armatus) in San Francisco Bay, USA, which showed reductions in plasma T4 and perturbations in T3/T4 ratios that were strongly correlated with PCB body burdens (Brar et al. 2010).

However, some thyroid-active substances are still in use (e.g. perchlorate; some heavy metals such as mercury etc.) and will be considered further. The clearest evidence is that related to perchlorate, an oxidizing agent mainly derived from ammonium perchlorate used in solid rocket fuels. Perchlorate is known to inhibit iodide uptake by the thyroid gland, thus interfering with normal synthesis of T4, which in turn abrogates negative feedback on TSH, thereby leading inter alia to thyroid follicular hyperplasia and hypertrophy. Mean perchlorate concentrations in contaminated streams in Texas studied by Theodorakis et al. (2006) ranged from 1.45 to 23.09 µg/L (max. 150 µg/L), with nothing detected in the reference rivers. They sampled fish, the central stoneroller (Campostoma anomalum) and amphibians, the cricket frog (A. crepitans), in three perchlorate-contaminated streams, and two reference streams in 2001–2003. Fish from the contaminated sites had increased thyroid follicular cell hyperplasia (in up to 20% of follicles), follicle epithelial hypertrophy, and depletion of T4-rich colloid in the follicles (in up to 12%), and these variables changed seasonally. In contrast, no colloid depletion or hyperplasia were evident in the frogs, but some hypertrophy was present at the most contaminated sites, and hypertrophy was correlated with mean perchlorate in water across all sites. This is strong evidence for inhibition of thyroid hormone synthesis, and is supported by many laboratory experiments with fish and amphibians exposed to perchlorate (e.g. Goleman et al. 2002; Bradford et al. 2005).

Perchlorate has caused similar effects in other amphibian populations (Carr et al. 2003). Bullfrog tadpoles (R. catesbeiana) collected from a perchlorate-contaminated site showed a five-fold lower hindlimb/snout-vent length ratio than reference tadpoles. The thyroid gland volume was 2.5-fold smaller in the contaminated larvae, probably because the reference animals had developed more rapidly to metamorphic climax. Western chorus frog tadpoles (Pseudacris triseriata) living in an ephemeral pond contaminated with perchlorate showed gross abnormalities of the thyroid gland including colloid depletion and follicular cell hypertrophy. Overall, these frogs showed delayed development of thyroid-hormone sensitive structures. It is to be expected that severe developmental delays could lead to adverse effects at the population level, but this is difficult to measure and has not been demonstrated to date.

Other contaminants which have been associated with thyroid modulation in the field include mercury, other heavy metals, and PAHs derived from partial combustion of oil products. Zhou et al. (2000) studied a fish, the mummichog (Fundulus heteroclitus), in Piles Creek, New Jersey, which was polluted with mercury and petroleum hydrocarbons. Mean mercury levels in sediment were 11.2 mg/kg (max. 200 mg/kg), although Cu, Zn and Cd were also elevated (625, 628, and 5.8 mg/kg sediment, respectively) (Khan and Weis 1993; cited in Smith and Weis 1997). Piles Creek fish had enlarged thyroid follicles (×6.7) in comparison with a reference site, follicular hyperplasia, and a 60% elevation of plasma T4, but no differences in T3. These effects could be simulated by exposing fish to Piles Creek sediments in the laboratory for 1 month. It seems likely that mercury was mainly responsible for the thyroid effects, because Smith and Weis (1997) showed that the fish had reduced growth and longevity compared with reference sites, prey capture was slower, and predator avoidance poorer. These changes are symptomatic of the developmental and neurological damage which mercury can cause, partly via its effects on the thyroid.

However, other work suggests that some PAHs are also thyroid-active and may have contributed to the effects in Piles Creek. Gentes et al. (2007) sampled nestling tree swallows (Tachycineta bicolor) from nest boxes in three areas of wetlands contaminated with tar sands mine tailings in Alberta, Canada. Mean contaminant levels were 140–207 ng total PAH/g sediment, 1010–2273 ng alkylated PAH/g sediment, and 10.3–68 mg/L naphthenic acids in water, compared with 81.5 ng/g, 175.9 ng/g and 0.3 mg/L, respectively, at the reference site. Plasma T3 was slightly elevated in swallows at the study sites (reference: 1.60 ng/ml; study: 1.37–2.57 ng/ml). Plasma T4 was not significantly affected, but T4 was elevated in thyroid gland by ∼100%. The results suggest that T4 synthesis by the thyroid was increased (the opposite effect to that caused by perchlorate), as was deiodination of T4 to T3 in peripheral tissues. The authors indicate that these changes could have negative effects on metabolism, behavior, feather development and molting, but it is not clear whether apical changes such as these would actually have occurred. Although PAH exposure could have been responsible for the effects, it was also suggested that environmental factors such as food availability may have been a factor.

The involvement of PAHs in thyroid modulation in wildlife was given support by Hersikorn and Smits (2011). They raised wood frog (Lithobates sylvaticus) tadpoles in enclosures on reclaimed oil sands wetlands of different degrees of maturity, and compared them with controls raised on reference sites. Metamorphosis was delayed (24%) or halted in tadpoles raised on freshly produced tailings, while development proceeded normally on old tailings and reference sites. The T3/T4 ratio was lowest in tadpoles raised on fresh tailings (max. reduction of 45%). No measurements were made of PAH concentrations, but it is reasonable to suppose, based on the evidence of Gentes et al. (2007), that they were elevated.

Heavy metals other than mercury have also been associated with thyroid effects. Kulczykowska et al. (2007) studied 32 white stork nestlings (Ciconia ciconia) near a copper smelter in Poland and compared them with 48 sampled near a town and in a reference area. Near the smelter, serum T4 was significantly reduced, by 58%, while serum melatonin (a free radical scavenger) was increased by 150%. Levels of Zn, Mg and Cd in nestling blood near the smelter were 16.60, 2933, and 5.06 mg/L, respectively, compared with 9.38, 1469, and 2.57 mg/L in blood of nestlings near the town and reference area. It is not clear which of these metals, if any, was responsible for the thyroid effects, but there was obviously a relationship with the presence of the smelter itself.

Finally, Sowers et al. (2009) and Mosconi et al. (2005) have demonstrated thyroid abnormalities in amphibians raised in diluted sewage effluent, or sampled from an intensive agricultural area, respectively. Sowers et al. (2009) showed that 50 and 100% STW effluent caused a 4–7 d delay in metamorphosis of northern leopard frog larvae (R. pipiens). There were nonspecific effects on thyroid histology – no hyperplasia or hypertrophy was observed, but small differences in follicular cell shape were recorded. This possibly suggests that the effect was not endocrine-mediated, but represented some nonspecific systemic toxicity. The frogs (R. esculenta) from an area of mainly cereal production in Italy (Mosconi et al. 2005) exhibited T3 and T4 titers which were increased by a factor of two in summer by comparison with a reference area. However, no data on thyroid histopathology or contaminant levels were presented, so it is impossible to conclude whether these frogs were experiencing endocrine influences as the fluctuations in hormone titers were probably within the natural range.

Considering the thyroid dataset as a whole, there is no doubt that some wildlife populations are experiencing chemical-related perturbations of the thyroid system, although in some cases the effects are probably nonspecific systemic toxicity. In other cases, however, it is likely that substances, such as perchlorate, mercury, and PAHs are causing more or less direct modulation of the thyroid cascade in a variety of vertebrates from fish to birds. Unfortunately, only in one case (fish exposed to Hg and PAHs: Smith and Weis 1997; Zhou et al. 2000) do we have evidence that this modulation has been causing adverse apical effects at the population level, but it seems likely that such effects, due to a variety of substances, are more widespread. For example, biologically significant delays in amphibian metamorphosis and interference with normal neurological development in a range of species will have adverse consequences for individuals and probably populations. More research is required to measure the extent of such effects in wildlife populations experiencing thyroid modulation.

3. Discussion and conclusions

Based on the Hill criteria (Hill 1965), Table 2 summarizes the strength of evidence for ED resulting from current-use chemicals, while Table 3 does the same for legacy chemicals. Effects (and their putative causes) have only been included if they have been demonstrated in the laboratory for related taxa. The strength of evidence (in terms of the amount of data considered reliable) has been graded 2 for strong, 1 for weak, and 0 for unknown. We acknowledge that this grading is somewhat subjective, but nevertheless consider it to be a helpful way of summarizing the data. Particularly relevant factors include whether there is a well-established endocrine mechanism, whether population damage has occurred, and whether there has been any experimental confirmation. Other important criteria include whether or not the putative endocrine effects are reproducible (i.e. consistency of effect), and whether the effects are clearly linked to the putative endocrine cause (i.e. specificity of association). The remaining criteria are also useful but often cannot be addressed due to lack of data.

Table 2. Strength of evidence for endocrine effects of current-use chemicals in the field, using criteria modified after Hill (1965). Strength of evidence: 2 = strong/large; 1 = weak; 0 = unknown.

	Evidence for:
	An endocrine mechanism	Population damage	Experimental confirmation	Consistency of effects	Specificity of association	A temporal sequence of cause/effect	Gradients of effect	Wide geographic extent	Decreased effect after reduced exposure	Total
1. Invertebrates – possible estrogenic effects of EE2 and other estrogens in sewage effluent	0	0	1	1	1	0	1	1	0	5
2. Fish (amphibians/reptiles) – estrogenic effects of EE2 and other estrogens in sewage	2	1	2	2	2	0	2	2	1	14
3. Amphibians – possible estrogenic effects of intensive agriculture, including atrazine	0	0	1	1	1	1*	1	1	0	6
4. Vertebrates – impaired endocrine stress response to multiple contaminants	1	0	1	2	1(multi-factorial causes)	0	1	1	0	7
5. Vertebrates – thyroid effects linked to perchlorate and other contaminants	2(perchlorate)1(other)	1	2(perchlorate)1(other)	1	1(multi-factorial)	0	1	1	0	97

* Some evidence for stronger effects before the use of atrazine in intensive agriculture.

Table 3. Strength of evidence for endocrine effects of legacy chemicals in the field, using criteria modified after Hill (1965). Strength of evidence: 2 = strong/large; 1 = weak; 0 = unknown.

	Evidence for:
	An endocrine mechanism	Population damage	Experimental confirmation	Consistency of effects	Specificity of association	A temporal sequence of cause/effect	Gradients of effect	Wide geographic extent	Decreased effect after reduced exposure	Total
Tri-organotins and mollusks	1	2	2	2	2	2	2	2	2	17
Polychlorinated biphenyls (PCBs), polychlorinated dibenzo dioxins (PCDDs), polychlorinated dibenzo furans (PCDFs)and vertebrates	1 (multi-factorial)	2	1	1	2 (multi-factorial)	1	2	2	1	13
Polybrominated diphenyl ethers (PBDEs) and vertebrates	1	0	1	2	2 (multi-factorial)	0	0	2	0	8
Perfluorinated compounds (PFCs) and vertebrates	1	0	1	1	1	0	0	2	0	6
DDT/p,p’-DDE and other organochlorine pesticides and vertebrates	1	2	2	2	2	2	2	2	2	17
Alkylphenols/alkylphenol ethoxylates and vertebrates	2	1	2	2	2	0	2	2	1	14

Taking Table 2 on current-use chemicals first, it is apparent that the case for EE2 in sewage effluents, in combination with natural estrogens, causing adverse effects in wild fish (and to a much lesser extent, amphibians, and reptiles) is reasonably strong. There is even evidence of damage to the breeding capability of some affected fish populations, although convincing evidence for actual population declines or extinctions only exists for the large Canadian lake experiment with EE2 (Kidd et al. 2007) in which the exposure concentrations were relatively high compared with most locations downstream of sewage effluent discharges. Other missing evidence includes the lack of data from before the era when estrogens were discharged in sewage (i.e. a temporal sequence is lacking), and the availability of only weak evidence for recovery when estrogenic discharges are remediated. Such remediation has hardly begun, mainly due to the high cost.

The other cases summarized in Table 2 for field effects of current-use chemicals are weak by comparison with estrogens and fish. Perhaps the least convincing is the idea that intensive agriculture, particularly that employing the herbicide atrazine, has caused feminization in wild amphibians. Probably the most compelling evidence against this hypothesis is that archived amphibian testes show prevalence of intersex which peaked long before atrazine was used, and declined during the period when atrazine-use was high at the end of the twentieth century to levels seen in the nineteenth century. Also, populations have not recovered in areas where atrazine was withdrawn from wide-scale use, and none of the other causality criteria appear to support the hypothesis.

The case of thyroid effects in vertebrates is interesting. There is some evidence that interference with thyroid function in fish can be caused by exposure to mercury and possibly also PAHs, and that this can lead to a series of apical impacts on development and behavior that are probably damaging at the population level. Furthermore, the thyroid effects of perchlorate on wild fish and amphibians are clear, even though there is no evidence that populations have been affected, and no examples of recovery have been observed. Overall, however, the strength of association between various current-use chemicals and thyroid effects in wild populations is at present rather weak, and more data are required.

Another weak case, perhaps again due to lack of data, concerns so-called estrogenic and other effects in mollusks and crustaceans. There is no doubt that exposure to estrogenic effluents causes elevations of ALP in some mollusks, but the mode of action is unknown (it may not even be ED), and there is no evidence for resulting adverse apical effects at the population level. There is weak experimental evidence that estrogen exposure in bivalves can cause reproductive damage, and only limited field evidence of this. Surprisingly, there is also no hard evidence for the adverse effects of ecdysone- and juvenile-hormone active insecticides on non-target insects and crustaceans, although this is not necessarily evidence of absence. More research in this field would be desirable.

Finally, there is a presently weak case for damage to the hormonal stress response in vertebrates caused by a range of unrelated contaminants, some of which are current-use chemicals. There is little experimental evidence, rather sparse information on modes of action, and no data about possible adverse apical effects on populations. However, this area is under-researched in comparison with the huge body of work on sex steroids and thyroid hormones, and it seems probable that damage to the stress response may be widespread in the vertebrates. The effects that have been observed can occur at low concentrations and some, at least, appear to result specifically from interference with the HPI/A axis, so they seem unlikely to be the result of systemic toxic action. It remains to be seen, though, whether the fitness of affected individuals to survive and reproduce is being impacted.

In only one of these cases, that of fish and amphibians exposed to EE2 in sewage, would it be true to state that the effects are widespread – indeed, they appear to be global wherever dilution of sewage is limited, although the precise threshold for effects is unclear. The remaining examples of effects due to current-use chemicals seem to be locally distributed, although more research may substantially widen any cause for concern. It should also be pointed out that establishing the existence of population declines and linking these to a likely cause is extremely difficult, so in many cases the best that can be expected is to identify effects on fitness, which are likely to result in population damage.

Turning to Table 3, which summarizes the strength of evidence for endocrine effects in the field caused by legacy substances, it is clear that this group of chemicals has a greater body of evidence supporting population effects than the current-use group of chemicals. The two strongest cases are those of tri-organotins in mollusks and organochlorine insecticides in predatory birds, both of which undoubtedly caused effects on a wide scale, which resulted in population and community crashes. The case for the PCBs and related chlorinated hydrocarbons having caused, and still causing, adverse population-level effects in many vertebrate species is also strong, only being limited by the logistical and ethical objections to the conduct of experiments in top predators, and by the fact that continuing exposure in some cases has probably prevented substantial population recoveries.

The evidence in the case of PBDEs and PFCs is somewhat less robust, but widespread exposure to, and bioaccumulation of, these chemicals combined with their toxicological properties suggests that they are contributing to the population-level effects on top-predators, which have been fairly well established for the PCBs. There is also little specific field evidence on the APs, but although it is reasonable to suppose that they contributed to a limited extent at one time to the estrogenic effects of many STW and other discharges, current aquatic concentrations of APs (at least in the USA) now appear to be below harmful levels (Coady et al. 2010).

On an optimistic note, it is considered that the current regulation of chemicals in general, and specific programs aimed at screening for and characterizing ED properties in particular, may be leading to further environmental improvement. Many of the legacy chemicals have POP-like properties, which is not the case for current use or new chemicals. Since POPs are being phased out globally, chemicals possessing POP characteristics are typically screened out during product development. Further, the general increase in environmental toxicity testing requirements for chemicals is likely to detect more sensitive apical adverse effects that would not have been captured previously. For instance, in Europe, the guidance document for aquatic ecotoxicology of pesticides went from 62 pages (European Commission 2002) to 267 pages (EFSA Panel on Plant Protection Products and their Residues 2013) in about a decade and further guidance documents are in development. The results of the additional testing and consequent risk assessment procedures are lowering acceptable exposures, and are tending to result in safe uses that protect the environment from both non-endocrine and endocrine-mediated adverse effects. In addition, the development of new ecotoxicological screens and tests with diagnostic sensitivity to EDSs has enhanced the regulatory tool box. Some of the OECD’s internationally-standardized in vivo toxicity screens and tests with fish, amphibians, and mammals, as well as in vitro assays, now have specific diagnostic ability to detect substances that can perturb the (anti)estrogenic, (anti)androgenic, steroidogenic and thyroid systems (OECD 2016). Additional diagnostic assays for use with other taxa, including invertebrates, are under development. As screening programs (e.g. US-EPA’s EDSP) and other legislation begin to implement these, our ability to identify and regulate substances with ED properties will be further improved. However, it is too early to state that availability of the new EAS-sensitive diagnostic assays will actually lead to improvements in environmental protection.

In conclusion, with the exception of the estrogenic effects on fish of EE2 and related estrogenic substances in treated sewage, it appears that legacy chemicals have caused, and in some cases are still causing, much more severe and widespread damage to many wildlife species than current-use chemicals. This conclusion must, of course, be accompanied by a significant caveat concerning the need for continued monitoring and research on these issues. Furthermore, the advent of improved regulatory testing does not imply that releases of new chemicals with side-effects including endocrine activity are necessarily a thing of the past, although they will likely become less common as they will be picked up earlier (i.e. in the substance discovery phase).

Declaration of interests

The authors’ employment affiliations are as shown on the cover page. The lead author is an independent consultant in ecotoxicology, while the other authors are employed as ecotoxicologists in the crop protection industry. The preparation of the paper was the professional work product of the authors and the synthesis of information, conclusions drawn and recommendations are the exclusive positions of the authors and not necessarily those of their employers. None of the authors have appeared in any legal or administrative proceedings related to the contents of the paper. CropLife International supported and encouraged this project by funding the time of the lead author, but the other authors received no funding from CropLife. An earlier version of the paper was circulated to CropLife, the European Crop Protection Association, and the authors’ employers, but no significant comments were received. CropLife International (https://croplife.org/) is a group of companies and industry associations involved in developing and marketing crop protection products. It champions the role of agricultural innovations in crop protection and plant biotechnology to support and advance sustainable agriculture.

Acknowledgements

The authors gratefully acknowledge the useful comments of three reviewers selected by the editor and whose identity was not known to the authors. The authors also thank two additional reviewers: Dr Mike Roberts (independent consultant, formerly UK Department for Environment Food & Rural Affairs) and Professor Glen Van Der Kraak (University of Guelph, Canada) whose suggestions helped improve the paper.