Environ. Sci. Technol. 2006, 40, 5478-5489
Modeling Effects of Mixtures of Endocrine Disrupting Chemicals at the River Catchment Scale JOHN P. SUMPTER* Institute for the Environment, Brunel University, Uxbridge, Middlesex UB8 3PH, U.K. ANDREW C. JOHNSON AND RICHARD J. WILLIAMS Centre for Ecology and Hydrology, Wallingford, Oxfordshire OX10 8BB, U.K. ANDREAS KORTENKAMP AND MARTIN SCHOLZE Centre for Toxicology, The School of Pharmacy, University of London, 29/39 Brunswick Square, London WC1N 1AX, U.K.
For endocrine disrupting chemicals in the environment, concerns arise primarily from the effects that may be induced in wildlife. A well studied example is estrogenic chemicals in the aquatic environment and their effects on fish. Directly measuring effects, in fieldwork studies, is an expensive and time-consuming approach that is fraught with many difficulties, ranging from study design right through to data analysis and interpretation. An alternative approach would be to predict the scale of effect(s) using suitable modeling techniques. We have attempted to do this using estrogenic chemicals as an example. We chose this group of aquatic pollutants because of the current considerable interest in them and the wealth of biological data available on them. Using the established GREAT-ER hydrological model, we have first predicted the concentrations and then the estrogenic effects on fish, of estrone, estradiol, ethinyl estradiol, and nonylphenol individually throughout an entire river catchment. We then show that knowledge of the biological responses of fish to mixtures of these chemicals can be used to predict the effect of environmentally realistic mixtures of them. To determine the degree of risk posed by this group of chemicals, it was necessary to take into account mixture effects: assessment on a chemical by chemical basis led to underestimations of the risk. Finally, we show that the approach can be used to predict how the risk will be affected by changes in the concentration of one chemical in the mixture. Although we have used only one endpoint (vitellogenin induction as an estrogenic response) and one group of similarly acting chemicals, we suggest that this general approach could prove extremely useful to regulatory authorities and other parties charged with protecting aquatic wildlife from adverse effects caused by chemicals in their environment.
Introduction Endocrine disruption has become a major research topic in the last 10 years (1). Within this broad topic, one area that * Corresponding author e-mail:
[email protected]. 5478
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
has received a lot of attention is the presence of estrogenic chemicals in the aquatic environment and their effects on aquatic organisms, in particular fish (reviewed in ref 2). There is very good evidence to support the contention that vitellogenin induction in wild fish living in rivers receiving effluent from sewage treatment works (STWs) is due to the presence of estrogenic chemicals in the effluents (3-5). In contrast, although another indicator of endocrine disruption, intersexuality, also occurs primarily in fish living downstream of STWs (4-6), it is less clear whether it is also a consequence of the presence of estrogenic chemicals in the water. This is because much less is known about the mechanisms underlying intersexuality than is known about vitellogenesis; the latter is strictly under the control of estrogens. Even less is known about the cause of the ‘feminization’ of the reproductive duct that can be observed in fish exposed to STW effluent (7, 8), though as with intersexuality, exposure to estrogenic chemicals could play a contributory, or even major, role. What is very clear, however, is that endocrine disruption in fish is occurring throughout the world (reviewed in ref 2) and that regulatory authorities would like to reduce or eliminate this problem (9, 10). Reasonable progress has been made toward identifying and quantifying the main estrogenic chemicals present in STW effluents (and hence, presumably, rivers). Desbrow et al. (11) made a significant advance when they identified the natural estrogens estrone (E1) and estradiol (E2), and the synthetic steroid estrogen ethinyl estradiol (EE2), as the major contributors to the estrogenic activity of some STW effluents. Many other studies (e.g. refs 12-15) have subsequently supported these conclusions. However, there are some locations, often apparently associated with particular industries such as the textile industry, where alkylphenolic chemicals (such as nonylphenol and its ethoxylates) appear to contribute significantly to the total estrogenicity of a river or effluent (e.g. refs 16-19). In summary, based on current knowledge, it seems very likely that steroid estrogens and/or alkylphenolic chemicals are the cause of the endocrine disruption reported in wild fish in most situations. There is currently no strong evidence for the widespread involvement of other estrogenic chemicals. Based on the above information, many researchers have used a variety of analytical approaches to directly measure the concentrations of steroid estrogens and alkylphenolic chemicals (especially NP) in STW effluents and receiving waters. A consistent pattern has emerged from these studies. Steroid estrogens are present at low concentrations (nanograms per liter), with the concentrations of E1 and E2 higher than those of EE2 (e.g. refs 11 and 20-22). In contrast, NP and its ethoxylates are present at higher concentrations, ranging from the upper nanogram per liter to tens and even hundreds of micrograms per liter (e.g. refs 12, 16, 19, 23). These concentrations are within or near to the ranges shown to cause biological effects on fish (24-27). Measuring actual concentrations of chemicals in the aquatic environment can be a time-consuming and expensive operation. This is particularly true if the chemical(s) of interest is present at a low concentration and/or present in a highly complex matrix, such as an effluent (28, 29). An alternative approach is to model, rather than measure, the concentrations of endocrine disrupting chemicals in the aquatic environment. Models capable of predicting the environmental concentrations of pollutants have been developed and been applied to predict influent, effluent, and river concentrations of steroid estrogens (30, 31). Given that these models can incorporate parameters such as flow rates and 10.1021/es052554d CCC: $33.50
2006 American Chemical Society Published on Web 08/03/2006
degradation rates, they may provide better average estimates of concentrations than can be obtained from studies involving actual measurements and hence possibly provide better predictions of long-term (whole lifetime) exposure of aquatic organisms. Although modeling (predicting) concentrations of pollutants in the aquatic environment is now possible, it is biological effects, not chemical concentrations, that are of most concern. For example, the main driver for endocrine disruption in the aquatic environment is the presence of elevated vitellogenin levels and intersexuality in fish, not the mere presence of estrogenic chemicals in rivers. Hence, modeling (predicted) effects, rather than chemical concentrations, is what is ideally required. Further, because many effects are likely to be a consequence of exposure to mixtures of chemicals, rather than to an individual chemical, predictive modeling of effects should try to incorporate this factor. There is now quite strong evidence that estrogenic chemicals act additively when inducing vitellogenin synthesis in fish (26, 27, 32) and that the effects of a mixture of estrogenic chemicals can be accurately predicted using well-established mathematical models (27, 33, 34). As long as the dose-response to each individual chemical in the mixture is established, then the response to any mixture containing those chemicals can be accurately predicted. The concept of modeling the effects of mixtures of pollutants on fish is not new. As early as 1968, Brown (35) took the then novel approach of calculating the acute toxicity (mortality) of “mixtures of poisons” to fish. Solbe´ and colleagues (36, 37) then applied this toxic units concept to the ‘real world’ and developed cumulative probability graphs of toxic units throughout both a small stream (36) and an entire river catchment (37). A similar, albeit more advanced, approach was taken by Cook et al. (38), who used an additive toxicity model (based on toxicity equivalence concentrations) for aryl hydrocarbon (Ah) receptor agonists such as dioxins to assess temporal changes in the status of the Lake Trout population in one of the Great Lakes of North America. More recently still, Jobling and colleagues have used modeling approaches to predict the incidence and severity of intersexuality in fish in U.K. rivers (2) and demonstrate that predicted concentrations of steroid estrogens correlate with the degree of sexual disruption in wild fish populations (39). We have built on these earlier attempts to explain changes in biological effects over time and/or space by combining hydrological, toxicological, physiological, and mathematical expertise to produce predictive effect maps for estrogenic chemicals and their mixtures that could be used in many ways to address risk assessment and risk management issues.
Materials and Methods Modeling Concentrations of Estrogenic Chemicals. The Rivers Aire and Calder flow eastwards, cutting across the Carboniferous rocks of the Pennines and the Triassic Sandstones of the Vale of York, North West England, before joining the River Ouse and discharging into the Humber Estuary, and the North Sea (40). The catchment has a population of approximately 1.9 million within an area of 1932 km2 and a population density of 983 inhab/km2 and receives about 80% of Yorkshire’s industrial effluents (41). The annual mean flow from the downstream limit of the catchment at Beal is 35.8 m3/s, thus for each individual in the catchment their 24 h excretion is diluted by 1628 L. We used the GREAT-ER model (Geo-referenced Regional Exposure Assessment Tool for European Rivers) to predict concentrations of the target chemicals (E1, E2, EE2, and NP). This model calculates the distribution of predicted environmental concentrations (PECs) of chemicals in surface waters, for both individual river reaches as well as entire catchments. Details of the model can be found in refs 42 and
43. Basically, the model requires estimates of the discharge concentrations of the chemicals of interest (in this case the main estrogenic chemicals) from STWs, which are then combined with flow data to give a distribution of river concentrations. Degradation of the compounds within the river is allowed and follows first-order decay kinetics. The values of the in-river half-lives ascribed to E1, E2, EE2, and NP were 5 days, 3 days, 17 days, and 8.5 days, respectively. The outputs from the STWs were based on measured data for NP (see below for further details) and on human excretion and STW removal efficiencies for E1, E2 and EE2 (see ref 31, for full details). The GREAT-ER 1.0 model has been validated for both boron and a surfactant, LAS (linear alkylbenzene sulfonate), in six study areas. The results of this validation exercise demonstrated that GREAT-ER can provide very accurate simulations of concentrations of these chemicals in a river system, provided reliable data sets and accurate hydrological and chemical fate models are used (43). However, to date its use to accurately predict concentrations of estrogenic chemicals has not been validated. Much of the development of GREAT-ER was based on data from the River Aire-Calder catchment in Yorkshire, U.K. and hence we chose to base our studies on this catchment. The output from GREAT-ER is in the form of color-coded river maps, with each color indicating a different concentration range which are predefined by the user. Usually these color-coded maps are used to identify locations (“hot spots”) where predicted environmental concentration (PEC) values are of concern (adverse biological effects might be expected). In our case we have taken this further, first to produce complete ‘effect maps’ for each chemical of interest and then to combine these to produce a “total risk” map which combines the effect of each individual chemical into a single effect of the mixture (see below for details). The effect maps are, of course, theoretical and are based on the responses of naı¨ve fish in laboratory studies. Like all models, GREAT-ER has its limitations. Probably the most serious is that the quality of the input data determines the quality of the output (the maps of predicted concentrations and effects). We used predicted concentrations of E1, E2, and EE2 in STW effluents to predict the concentration of these three chemicals throughout the catchment. In the case of NP, we used actual measured values for the River Aire-Calder catchment. In April-May, 2003, the U.K. Environment Agency conducted a sampling campaign for 25 STWs across 5 regions (covering much of England). This comprised two grab samples taken on separate days, with representative flow measurements also being taken at the same time. These samples were analyzed (under contract) for NP. There were a large number or detections above the 1 µg/L detection limit. NP concentrations were highest in the Yorkshire region (where the River Aire-Calder catchment is), the mean effluent NP concentration being 4.5 µg/L. This latter finding was perhaps not surprising, given that this catchment has historically had the highest NP concentrations found in England (17, 23). For effluents that were not monitored in the 2003 exercise, we used an NP concentration that was the mean of all measured values from the catchment. However, concentrations of alkylphenolic chemicals in this catchment are decreasing (44), as the use of these surfactants is restricted, and hence lower concentrations than those measured (in 2003) might be expected soon, if not now. To address this likely, and probably more representative, scenario, we also used a second data set of NP concentrations in effluents. This was based on that collected as part of the European-wide COMPREHEND project; full details are available in ref 45. A mean NP concentration across six European countries of 0.46 µg/L, with a SD of 0.39, was found. This concentration was applied to all STW effluents in the catchment, to provide a ‘low NP’ scenario. VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5479
TABLE 1. Division of the Dose-Response Curves into Classesa class
vitellogenin induction (percent of maximum)
descriptor
1 2 3 4 5
0-one-tenth of EC 10 one-tenth of EC10-10 10-50 50-95 95-100
no effect low effect medium effect high effect severe effect
a The dose-response curve for each chemical was divided into five sections (or classes), ranging from no effect to severe effect. The same class boundaries were used for all individual chemicals and mixtures.
FIGURE 1. Dose-response curves for the induction of vitellogenin in the male fathead minnow caused by exposure to either ethinyl estradiol (EE2), estradiol (E2), estrone (E1), or p-tert-nonylphenol (NP). Although the data are plotted using nominal concentrations of chemicals, actual (measured) concentrations were very similar. Actual plasma vitellogenin concentrations were normalized using a log10 transformation. The results are presented as the percentage of the maximum induction. Data were taken from refs 27 and 46. Modeling Effects of Individual Chemicals. We chose vitellogenin induction as the effect to model because there is a wealth of information available of vitellogenin synthesis in fish in response to waterborne exposure to the main estrogenic chemicals of concern, and because the model of concentration addition has been validated only for the endpoint of vitellogenin induction. Ideally, full doseresponse curves needed to be available for each individual chemical and for a mixture of all the chemicals. Further, these data all needed to come from a single species of fish, preferably an indigenous one present in the catchment modeled. Finally, all dose-response curves had to have been obtained from experiments in which known (measured) concentrations of the chemical(s) of interest were present in the water (because the major route of exposure to these chemical is via the water). Fortunately, a single data set (27) fulfilled most of these conditions and hence was used in our study. It was complemented by the data on estrone (a chemical not tested by ref 27) taken from ref 46. All these data were obtained from well-controlled laboratory experiments which used the fathead minnow, a North American species. Unfortunately, data of comparable quality for a European species of fish are not available. To construct predicted effect maps for each individual chemical, we used log 10-transformed vitellogenin levels which were normalized to give a scale from 0 to 100% (see ref 27 for further details). This effect scale was then subdivided into five classes, with the following boundaries: no effect, below 10% (low effect), 10-50% (medium effect), 50-95% (high effect), 95-100% (severe effect). Concentrationresponse relationships for each chemical (Figure 1), together with information on their predicted concentrations in reaches of the catchment, were utilized to derive the corresponding effects classes (Table 1). However, given the lack of generally accepted criteria for the distinction of adverse effects, the class distinction between “no effect” and “low effect” is problematic: the estimation of low effect concentrations is associated with a great uncertainty, which is influenced not least by the biological variation underlying vitellogenin induction in fish. Given this variation, it is very hard to distinguish any effect below 10% from zero-effect, i.e., 5480
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
concentrations of chemicals resulting in effects below 10% are often on the borderline of statistical significance. As a pragmatic approach, we used therefore 1/10 of the concentration producing a 10% effect (EC10) as borderline between “no effect” and “low effect”. Modeling Effects of Mixtures. To be able to predict (model) the effect of a mixture of chemicals, it is necessary to know how they interact. For example, do they act synergistically, additively, or antagonistically? It is feasible that different chemicals interact in different ways (e.g. X and Y act additively, whereas X and Z act synergistically), and also possible that the type of interaction depends on the effect studied. However, in the case of vitellogenin induction in fish in response to exposure to estrogenic chemicals, it has recently been shown conclusively that chemicals (both steroids and xenoestrogens) act dose additively and that the response can be very accurately predicted by the model of concentration addition (26, 27). This information made it a relatively straightforward task to predict the degree of vitellogenin induction in response to any mixture of the estrogenic chemicals of interest. The following mathematical approach was used. A combined effect according to the concept of concentration addition is defined for a multicomponent mixture of n substances by n
ci
∑ ECx ) 1 i)1
(1)
i
(47), with ci being the concentrations of the individual substances present in a mixture with a total effect of x %, ECxi are the equivalent effect concentrations of the single substances, i.e., those concentrations that alone would cause the same quantitative effect x as the mixture. However, if we only want to know whether the expected mixture effect exceeds or falls below a fixed effect level x, then eq 1 can be used in order to quantify the factor k by which the given mixture concentration exceeds (or falls below) the mixture concentration ECxmix: n
ci
∑ ECx ) k i)1
(2)
i
If k ) 1, the mixture is expected to produce exactly the mixture effect x, if k > 1, the expected mixture effect is higher than x (correctly, the mixture exceeds the ECxmix by a factor of k), and if k < 1 the expected mixture effect is smaller than x. Thus, for the defined effect boundaries for vitellogenin induction it is only necessary to know the EC10, EC50, and EC95 values for each compound in order to classify the expected mixture effects. The evaluation rules for the five classes are the following: k1 e 0.1 (“no effect”), 0.1 < k1 e 1 (“low effect”), k2 e 1 < k1 (“medium effect”),
FIGURE 2. Predicted concentrations of estrone (E1, in pg/L) throughout the River Aire-Calder catchment. Only E1 concentrations in the two major rivers have been predicted, not concentrations in the smaller tributaries (which are shown in blue). The locations of all major sewage treatment works are indicated by the b symbol. k3 e 1 < k2 (“high effect”), k3 > 1 (“severe effect”)
with k1 ) n
ci
∑ EC10 , i)1
i
n
k2 )
ci
∑ EC50 , i)1
i
n
and k3 )
ci
∑ EC95 i)1
i
Results Vitellogenin Synthesis by Estrogenic Chemicals. All four chemicals of interest (E1, E2, EE2, and NP) induced vitellogenin synthesis in a dose-dependent manner, although their potencies varied by about 10 000-fold (Figure 1). EE2 was 25-fold more potent in vivo than E2 and induced vitellogenin synthesis at concentrations in the high picogram per liter range: 50% induction occurred at a concentration below 1 ng/L. E2 induced vitellogenin synthesis in the ng/L range, with 50% induction occurring at 20 ng/L. E1 was around 2.5-fold less potent than E2 and hence also effective at inducing vitellogenin synthesis in the ng/L range. NP was the least potent of the estrogenic chemicals; µg/L concentrations are required to induce vitellogenin synthesis, with around 6 µg/L needed to induce 50% of the maximum response. Predicted Concentrations of Estrogenic Chemicals. Concentrations of all four test chemicals throughout the River Aire-Calder catchment were predicted using GREAT-ER. We chose to map the 90%ile concentrations (concentrations exceeded only 10% of the time), which would be representative of typical summer river flows. Figure 2 shows a colorcoded map of the predicted concentrations of E1. These were highly variable and ranged from very low (less than 2 ng/L) in the upper reaches of the rivers to much higher (over 20 ng/L) in some of the lower reaches, including reaches of some tributaries that received effluent from STWs. However, although it was generally the case that predicted E1 concentrations increased the further downstream one went, there were short reaches of river in the upper reaches of the catchment where high E1 concentrations were predicted. Similar maps predicting concentrations of E2, EE2, and NP were produced (results not shown).
Predicted Effect of Each Individual Chemical. Based on the predicted concentrations of each chemical and actual vitellogenin induction in response to measured concentrations of each chemical (shown in Figure 1), it was possible to predict the degree of vitellogenin induction due to each individual chemical through the entire catchment. E1 on its own was predicted to have no effect in most of the upper reaches of both the River Aire and the River Calder, although even in these reaches a low effect was predicted in short reaches of both rivers (Figure 3). However, in the majority of both rivers, E1 was predicted to induce vitellogenin synthesis, with the degree of induction ranging from low to high. A relatively similar picture was predicted for E2 (Figure 4), although the overall effect of E2 was predicted to be less than that due to E1; no part of the catchment was predicted to show a high effect due to E2 alone (unlike the situation for E1). In marked contrast to the two natural steroids (E1 and E2), the synthetic steroid EE2 on its own was predicted to cause some effect over almost the entire catchment (Figure 5). Although the predicted concentrations of EE2 were lower than those of E1 and E2 (data not shown), this was more than compensated for by the much higher biological potency of EE2 (see Figure 1). Over approximately half the length of both the rivers Aire and Calder, predicted concentrations of EE2 were high enough to cause either medium (primarily) or high (occasionally) effects. However, nowhere in the catchment were predicted concentrations high enough to cause a severe effect (Figure 5). The situation with NP using the higher input scenario was unique in that over most of the catchment a low effect was predicted, but in a few, relatively short, reaches, medium, high, and even (in two reaches) severe effects were predicted (Figure 6). This was the only chemical predicted to cause severe effects on its own. Predicted Effects of Mixtures of Chemicals. The predicted effect of a mixture of the three steroids (E1, E2, and EE2) is shown in Figure 7. This shows that with the exception of a short stretch of the upper River Aire, the total steroid mixture was predicted to cause vitellogenin induction throughout VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5481
FIGURE 3. Predicted effect of oestrone (E1) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentrations of E1. See the legend to Figure 2 for further details.
FIGURE 4. Predicted effect of estradiol (E2) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentration of E2. See the legend to Figure 2 for further details. the catchment. The degree of effect increases from the upper reaches to the lower reaches. A medium effect was predicted over about half the entire catchment, with high effects in some short reaches. However, the total steroid mixture was not predicted to cause severe effects anywhere in the catchment. The ‘total risk’ was predicted when the contribution due to NP was added to that of the steroid mixture; this is shown in Figure 8. As expected, given the fact that NP alone caused effects through much of the catchment, combining NP with 5482
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
the three steroids produced an effect map suggesting a more pronounced effect than was produced by the steroids alone. Now along about three-quarters of the two rivers the predicted effect was medium or greater (Table 2). Quite significant reaches of both rivers making up the catchment were predicted to have a high effect, and in two short reaches a severe effect was predicted. The Effect of Lower NP Concentrations. As mentioned in the Materials and Methods, the recently measured NP concentrations we used in our modeling are higher than
FIGURE 5. Predicted effect of ethinyl estradiol (EE2) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentrations of EE2. See the legend to Figure 2 for further details.
FIGURE 6. Predicted effect of nonylphenol (NP) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on recently measured concentrations of NP. See the legend to Figure 2 for further details. might be expected and higher than in more typical rivers in England and elsewhere. Further, recent severe restrictions on the use of nonylphenol ethoxylates are likely to lead to continuing falls in the concentrations of NP and its ethoxylates and carboxylates in rivers throughout Europe. To assess the consequence of this decrease in NP concentrations, we predicted the effect of a “low NP” scenario (Figure 9). This
substantially reduced the predicted effect due to NP on its own throughout the catchment (compare Figures 6 and 9). At ‘low’ NP concentrations, no reaches of river were predicted to have medium or higher effects. A total risk map was then produced using the effect of ‘low NP’ combined with the effects due to the three steroid hormones (Figure 10). This shows that in most of the VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5483
FIGURE 7. Predicted effect of the mixture of steroids (E1, E2, and EE2) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentrations of the steroids. See the legend to Figure 2 for further details.
FIGURE 8. Predicted effect of the mixture of all four estrogenic chemicals (E1, E2, EE2, and NP) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentrations of the steroids but recently measured concentrations of NP. See the legend to Figure 2 for further details. catchment the overall effect was predicted to be either low or medium, although in three short reaches a high effect was predicted. A comparison of Figures 8 and 10 demonstrates the consequence of a lower NP concentration throughout the catchment. The effect in some reaches in the middle of both rivers was predicted to fall from medium to low, a shorter length of river was predicted to produce a high effect, and 5484
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
no longer were any reaches predicted to produce a severe effect. Predicted Mixture Effects. The results predict mixture effects, which occur when the effect of a mixture is greater than that of the effect due to any of the individual chemicals. Table 2 shows what proportions of the entire catchment are in what effect categories (low, medium, high, etc.) for both
TABLE 2. Lengths (m) and Proportions (%) of the Entire Catchment in the Various Effect Bandsa no effect
E1 E2 EE2 NP (high) total steroids total mixture (high NP) NP (low) total mixtures (low NP) a
low effect
medium effect
high effect
severe effect
m
%
m
%
m
%
m
%
m
%
49469 146733 7466 10656 7466 7466 179810 7466
22 65 3 5 3 3 80 3
166482 76370 108799 185806 83271 48953 45120 81009
74 34 48 83 37 22 20 36
7152 1826 101753 19488 125214 140043 0 127476
3 1 45 9 56 62 0 57
1826 0 6912 2067 8979 21555 0 8979
1 0 3 1 4 10 0 4
0 0 0 6912 0 6912 0 0
0 0 0 3 0 3 0 0
Results for both individual chemicals and various mixtures are shown.
FIGURE 9. Predicted effect of ‘low’ concentrations of nonylphenol (NP) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The concentrations of NP used were one-tenth of the measured concentrations, bringing them more into line with average NP concentrations in effluents in Western Europe. See the legend to Figure 2 for further details. individual chemicals and various mixtures of them. For example, most individual chemicals (E1, E2, NP-high) are predicted to cause either no, or a low, effect in the various majority (nearly 90%, or more) of the catchment. Only EE2 is predicted to cause significant effects over appreciable lengths of the catchment, but even with this chemical (the most active by far), less than half the catchment is predicted to show an effect greater than ‘low’, with only 3% of the catchment showing a high effect due to this one chemical. However, when the effect of a mixture of these chemicals (E1, E2, EE2, NP-high) is considered, a medium or greater effect is predicted to occur in 75% of the entire catchment. Ten percent of the catchment now shows a high effect, whereas no individual chemical was predicted to cause a high effect in more than 3% of the catchment. Put another way, a further 14 643 m of river is now in the high effect category, due to the combined effects of the individual chemicals.
Discussion It is important to keep constantly in mind that we have taken a theoretical approach only to developing effect maps. We have not attempted to validate this approach by collecting chemical and/or biological data and then comparing these to our predictions. However, we believe that our maps are likely to provide good estimates of both the chemical and biological parameters we have modeled. This supposition is based on the facts that the predicted concentrations of steroid estrogens agree very well with measured concentrations both within the U.K. (48) and elsewhere (e.g. refs 49-51) and that these in turn have been shown to be positively correlated with endocrine disruption in wild fish (39). Given the current difficulties associated with accurately measuring the very low (yet biologically active) EE2 concentrations in the aquatic environment, we consider it unlikely that an extensive analytical program, to measure E1, E2, EE2, and NP concentrations throughout the catchment, would provide a VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5485
FIGURE 10. Predicted effect of the mixture of all four estrogenic chemicals (E1, E2, EE2, and ‘low’ NP) on vitellogenin induction in fish throughout the River Aire-Calder catchment. The map is based on predicted concentrations of the steroids and ‘low’ concentrations of NP. See the legend to Figure 2 for further details. robust test of our predictions. Nor do we consider that testing the biological predictions of our model, by sampling fish throughout the catchment, would be justified on ethical grounds (it would involve killing a large number of wild animals). Our objective was not to produce a cast iron model of estrogenicity in a river catchment, nor was it to validate models, but rather to demonstrate how combining existing knowledge can provide thought-provoking results of wide interest and applicability. The results established that not only was it possible to model the effect of a chemical (in contrast to concentrations) but that it was also possible to model the effect of a mixture of similarly acting chemicals. Thus we have established the principle of combing an accepted mathematical model of mixtures toxicity with a hydrological model of an entire catchment as a first attempt at predicting the effect of mixtures of the main estrogenic chemicals known to be present in rivers. Doing so was made possible by (1) the fact that a well-established hydrologically based model (GREATER) for an entire river catchment existed (42), (2) that it had already been established that such a model could be used to accurately predict concentrations of some chemicals (although not yet estrogenic ones), if measured values were not available (43), (3) that biological dose-response relationships were available for each individual chemical (27, 46), and (4) that it had been shown that these chemicals interact in a dose additive manner, which can be modeled by concentration addition, when inducing vitellogenin synthesis (26, 27). Although these conditions are rarely met currently, they are likely to be met more often in the future. Hence, our approach could, at least in theory, be applied to any chemical, or mixtures of chemicals, of interest, as long as the dose-response relationship(s) is known. We predict that all four chemicals could contribute to the effect of the mixture, albeit to different degrees. It appears that E2 plays the least significant role; environmental concentrations are not high enough to induce pronounced 5486
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
vitellogenin synthesis, due in part to the ease with which this chemical is degraded (52, 53). Of the three steroids, EE2 plays the most significant role, a consequence of its extremely high potency (26) and relative resistance to degradation (53). NP can contribute very significantly to the overall estrogenic activity if it is present at “high” concentrations, though even these are nowhere near as high as the concentrations of NP and NP ethoxylate present in this river catchment in the past (23). Very probably NP and some of its metabolites were the main reason for the extremely high estrogenic activity of at least some reaches of this river up the 1990s (3, 17, 44). If the concentrations of the alkylphenolic chemicals continue to fall, as a consequence of reduced use of the chemicals, then our results suggest that they will play a progressively decreasing role in the overall oestrogenic activity of the rivers and that the steroids will dominate. According to this analysis, their removal would not eliminate endocrine disruption in fish from U.K. rivers. Fish could be at high, or even severe, risk (of inappropriate vitellogenin synthesis) in some reaches of river. In fact, over half the catchment presents a medium or higher risk. This conclusion is supported by the results of earlier fieldwork (fish sampling) from the River Aire, which has shown that the wild fish living in this river can have very high blood vitellogenin levels (4, 54). It is important to realize that without considering mixture effects, it would be possible to underestimate the degree of risk posed by the chemicals. Our modeling studies have identified reaches of river where each individual chemical poses little or no risk, but where their combined (mixture) effect is greater, and the effect becomes medium, or even high. Our predictions at the whole catchment level are in agreement with the results of laboratory in vitro (34, 55) and in vivo (27) experiments, which have shown that the effects of individual estrogenic chemicals are dose additive in mixtures; that is, the effect of the mixture is greater than the effect caused by any of the individual chemicals. In fact, “something from nothing” is possible,
where no individual chemical is present at a high enough concentration to cause a measurable effect, but a mixture of them does (55, 56). There is no reason such a situation could not occur in the ‘real world’, rather than in a laboratory. There are reasons to think that our modeling studies might actually have underestimated the degree of effect. One reason for this is because the GREAT-ER model does not have the ability to account for the possibility that one estrogenic chemical might degrade into a second (different) estrogenic chemical, rather than completely lose its biological activity as it degrades. It is well documented that in the aquatic environment E2 rapidly degrades to E1 (53), which is somewhat more persistent. Hence, whereas GREAT-ER treats the rapid degradation of E2 as loss of estrogenic activity, in fact it is degrading into another estrogenic chemical that is not that much less potent. Future refinements of the GREATER model to enable it to cope with such eventualities would be beneficial. A related, though more complex, situation is occurring with the alkylphenolic chemicals. NP is a degradation product of NP polyethoxylates, which can degrade by shortening of the ethoxylate chain, prior to loss of all ethoxylate groups and the formation of NP (57, 58). The short chain ethoxylates (e.g. NPIEO, NP2EO) are also estrogenic, though not as potent as NP (25). These short chain ethoxylates are typically present in the aquatic environment at appreciably higher concentrations than NP (59). Hence, there is no doubt that we have not included all the estrogenic chemicals known to be present in the aquatic environment in our modeling studies. Further, these NP ethoxylates will degrade to NP, hence adding to the amount of NP in the water and increasing the estrogenic activity (because NP is more estrogenic than its ethoxylates). The second reason is that there are likely to be other estrogenic chemicals present in the aquatic environment that we have not considered but which add to the total estrogenicity. These chemicals will include nonylphenol ethoxylates and carboxylates (25, 59), octylphenol and its ethoxylates (23, 25), various other steroid estrogens (of both human and animal origin), and possibly many xenoestrogens (e.g. bisphenol A, various pesticides, and various parabens). Although most evidence to date points to steroid estrogens and nonylphenolic chemicals contributing the majority of the estrogenic activity in STW effluents (e.g. ref 11), nevertheless there will probably be specific situations/locations where one or more of these other estrogenic chemicals does contribute significantly. For example, a toxicity identification and evaluation approach was used recently to show that a phytoestrogen, genistein, contributed most of the estrogenic activity to one particular effluent in Japan (60). To do our modeling, we used biological data from a single species of fish (the fathead minnow). We do not know how readily our predictions can be extrapolated to other species. However, it seems likely that different species of fish are relatively similar in their sensitivities to the estrogenic chemicals we chose; for example the fathead minnow (26, 61), the species we based our modeling on, the zebrafish (62), the Medaka (63), and rainbow trout (64) all seem equally sensitive to EE2. Hence, we believe that use of data from one species (which may sometimes be all that are available) are probably adequate to produce predictions that apply to other species. In contrast, different effects (e.g. intersexuality) may show different sensitivities compared to vitellogenin synthesis, and thus that would need to be known before extrapolating across different endpoints that are caused by the same chemical(s). Further, provided sufficient data of high quality are available, there is no reason effects of chemicals on other groups of aquatic organisms (e.g. invertebrates) could not be modeled. Having established the principle that effects maps (for both single chemicals and mixtures) can be produced, how
can such tools be used? In terms of catchment management, they could be used to target STWs, so that those STWs predicted to have the most severe effects on aquatic organisms were improved first. This would produce the most rapid improvement in the quality of the aquatic environment. For scientists and regulators, they aid greatly in ranking the chemicals responsible for a particular effect; in this case, EE2 and NP appear to contribute approximately equally to vitellogenin induction in wild fish in this catchment, whereas E1 and E2 play minor roles. Given that it would not be easy to reduce the concentration of EE2 in effluent, without major improvements in the STWs, it would seem sensible to try and reduce the concentration of NP. In fact, the recently introduced severe restrictions on the use of NP and its ethoxylates in the European Union have done just that. Our comparison of the effects of ‘low’ and ‘high’ NP scenarios illustrates the effectiveness of this strategy. Further decreases in the concentrations of NP and its ethoxylates in STW effluents are likely, as use of these chemicals continues to fall, which we predict will lead to lower levels of vitellogenin induction (and possibly other effects) in wild fish. As far as biological uses are concerned, maps such as those we have produced could be used for targeted fieldwork, which in turn should provide a better picture of the general situation in a catchment, yet with less sampling of wild animals. Another factor whose influence could readily be investigated is seasonality and its consequences on biological effects. Obviously flow-rate varies greatly, depending on weather conditions. High flows are likely to be associated with lower concentrations of estrogenic chemicals (due to their increased dilution) and therefore should lead to reduced effects. In contrast, low flows in summer will lead to higher concentrations of these estrogenic chemicals and hence potentially more pronounced effects. As better, and/or site specific, data on parameters such as degradation rates become available, these can easily be incorporated into the model, to improve it. It also ought to be possible to model other effects of interest, as long as the appropriate data to do so were available. Even changes at the highest ecological levels, such as species composition within a catchment, could, in theory, be predicted if these were known to be dependent on one or more environmental parameters (e.g. oxygen level, nutrient level, water temperature, etc.) for which there was sufficient information for meaningful modeling.
Acknowledgments We thank NERC, through its CEH Science Budget, and the European Commission, through ‘ACE’ (contract EVK1-200100091), for supporting this work.
Literature Cited (1) Matthiessen, P. Historical perspective on endocrine disruption in wildlife. Pure Appl. Chem. 2003, 75, 2197-2206. (2) Jobling, S.; Tyler, C. R. Endocrine disruption in wild freshwater fish. Pure Appl. Chem. 2003, 75, 2219-2234. (3) Harries, J. E.; Sheahan, D. A.; Jobling, S.; Matthiessen, P.; Neall, P.; Sumpter, J. P.; Tylor, T.; Zaman, N. Estrogenic activity in five United Kingdom rivers detected by measurement of vitellogenesis in caged male trout. Environ. Toxicol. Chem. 1997, 16, 534-542. (4) Jobling, S.; Nolan, M.; Tyler, C. R.; Brighty, G.; Sumpter, J. P. Widespread sexual disruption in wild fish. Environ. Sci. Technol. 1998, 32, 2498-2506. (5) Kavanagh, R. J.; Balch, G. C.; Kiparissis, Y.; Niimi, A. J.; Sherry, J.; Tinson, C.; Metcalfe, C. D. Endocrine disruption and altered gonadal development in white perch (Morone americana) from the lower Great Lakes region. Environ. Health Perspect. 2004, 112, 898-902. (6) Aravindakshan, J.; Paquet, V.; Gregory, M.; Dufresne, J.; Fourrier, M.; Marcogliese, D. J.; Cyr, D. G. Consequences of xenoestrogen exposure on male reproductive function in Spottail Shiners (Notropic hudsonius). Toxicol. Sci. 2004, 78, 156-165. VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5487
(7) Rodgers-Gray, T. P.; Jobling, S.; Kelly, C.; Morris, S.; Brighty, G.; Waldock, M. J.; Sumpter, J. P.; Tyler, C. R. Exposure of juvenile roach (Rutilis rutilis) to treated sewage effluent induces dosedependent and persistent disruption in gonadal duct development. Environ. Sci. Technol. 2001, 35, 462-470. (8) Beresford, N.; Jobling, S.; Williams, R.; Sumpter, J. P. Endocrine disruption in juvenile roach from English rivers: a preliminary study. J. Fish Biol. 2004, 64, 1-7. (9) Environment Agency. Endocrine-disrupting substances in the environment: the Environment Agency’s strategy. The Environment Agency: Bristol, U.K., 2000. (10) Gies, A. Government view of endocrine disruption in wildlife. Pure Appl. Chem. 2003, 75, 2563-2574. (11) Desbrow, C.; Routledge, E. J.; Brighty, G. C.; Sumpter, J. P. Waldock, M. Identification of estrogenic chemicals in STW effluent. I. Chemical fractionation and in vitro biological screening. Environ. Sci. Technol. 1998, 32, 1549-1558. (12) Spengler, P.; Ko¨rner, W.; Metzger, J. W.; Substances with estrogenic activity in effluents of sewage treatment plants in south western Germany. I. Chemical analysis. Environ. Toxicol. Chem. 2001, 20, 2133-2141. (13) Ko¨rner, W.; Spengler, P.; Bolz, U.; Schuller, W.; Hanf, V.; Metzger, J. W. Substances with estrogenic activity in effluents of sewage treatment plants in south western Germany. 2. Biological Analysis. Environ. Toxicol. Chem. 2001, 20, 2142-2151. (14) Furuichi, T.; Kannan, K.; Giesy, J. P.; Masunaga, S. Contribution of known endocrine disrupting substances to the estrogenic activity in Tama River water samples from Japan using instrumental analysis and in vitro reporter gene assay. Water Res. 2004, 38, 4491-4501. (15) Nakada, N.; Nyunoya, H.; Nakamura, M.; Hara, A.; Iguchi, T.; Takada, H. Identification of estrogenic compounds in wastewater effluent. Environ. Toxicol. Chem. 2004, 23, 2807-2815. (16) Sole´, M.; Lo´pez de Alda, M. J.; Castillo, M.; Porte, C.; LadegaardPedersen, K.; Barcelo, D. Estrogenicity determination in sewage treatment plants and surface waters from the Catalonian area (E. Spain). Environ. Sci. Technol. 2000, 34, 5076-5083. (17) Sheahan, D. A.; Brighty, G. C.; Daniel, M.; Kirby, S.; Hurst, M. R.; Kennedy, J.; Morris, S.; Routledge, E. J.; Sumpter, J. P.; Waldock, M. J. Estrogenic activity measured in a sewage treatment works treating industrial inputs containing high concentrations of alkylphenolic compounds - a case study. Environ. Toxicol. Chem. 2002, 21, 507-514. (18) Ce´spedes, R.; Petrovic, M.; Raldu ´a, D.; Saura, U.; Pina, B.; Lacorte, S.; Viana, P.; Barcelo´, D. Integrated procedure for determination of endocrine-disrupting activity in surface waters and sediments by use of the biological technique recombinant yeast assay and chemical analysis by LC-ESI-MS. Anal. Bianal. Chem. 2004, 378, 697-708. (19) Quiros, L.; Ce´spedes, R.; Lacorte, S.; Viana, P.; Raldu ´a, D.; Barcelo, D.; Pina, B. Detection and evaluation of endocrine-disruption acitivty in water sample from Portuguese rivers. Environ. Toxicol. Chem. 2005, 24, 389-395. (20) Baronti, C.;, Curini, R.; D’Ascenzo, G.; Di Corcia, A.; Gentili, A.; Samperi, R. Monitoring natural and synthetic estrogens at activated sludge sewage treatment plants and in a receiving river water. Environ. Sci. Technol. 2000, 34, 5059-5066. (21) Kuch, H. M.; Ballschmiter, K. Determination of endocrinedisrupting phenolic compounds and estrogens in surface and drinking water by HRGC-(NCI)-MS in the picogram per litre range. Environ. Sci. Technol. 2001, 35, 3201-3206. (22) Servos, M. R.; Bennie, D. T.; Burnison, B. K.; Jurkovic, A.; McInnis, R.; Neheli, T.; Schnell, A.; Seto, P.; Smyth, S. A.; Ternes, T. A. Distribution of estrogens, 17β-estradiol and estrone in Canadian municipal wastewater treatment plants. Sci. Total Environ. 2005, 336, 155-170. (23) Blackburn, M. A.; Waldock, M. J. Concentrations of alkylphenols in rivers and estuaries in England and Wales. Water Res. 1995, 29, 1623-1629. (24) Purdom, C. E.; Hardiman, P. A.; Bye, V. J.; Eno, N. C.; Tyler, C. R.; Sumpter, J. P. Estrogenic effects of effluents from sewage treatment works. Chem. Ecol. 1994, 8, 275-285. (25) Jobling, S.; Sheahan, D.; Osborne, J. A.; Matthiessen, P.; Sumpter, J. P. Inhibition of testicular growth in rainbow trout (Oncorhynchus mykiss) exposed to estrogenic alkylphenolic chemicals. Environ. Toxicol. Chem. 1996, 15, 194-202. (26) Thorpe, K.; Cummings, R. I.; Hutchinson, T. H.; Scholze, M.; Brighty, G.; Sumpter, J. P.; Tyler, C. R. Relative potencies and combination effects of steroidal estrogens in fish. Environ. Sci. Technol. 2003, 37, 1142-1149. (27) Brian, J.; Harris, C. A.; Scholze, M.; Backhaus, T.; Booy, P.; Lamoree, M.; Pojana, G.; Jonkers, J.; Runnalls, T.; Bonfa`, A.; 5488
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 17, 2006
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35) (36)
(37)
(38)
(39)
(40) (41)
(42)
(43)
(44)
(45)
Marcomini, A.; Sumpter, J. P. Accurate prediction of the response of freshwater fish to a mixture of estrogenic chemicals. Environ. Health Perspect. 2005, 113, 721-728. Gomes, R. L.; Avcioglu, E.; Scrimshaw, M. D.; Lester, J. N. Steroid estrogen determination in sediment and sewage sludge: a critique of sample preparation and chromatographic/mass spectrometry considerations, incorporating a case study in method development. Trends Anal. Chem. 2004, 23, 737-744. Rodriguez-Mozaz, S.; Lopez de Alda, M. J.; Barcelo, D. Picogram per litre level determination of estrogens in natural waters and waterworks by a fully automated on-line solid-phase extractionliquid chromatography-electrospray tandem mass spectrometry method. Anal. Chem. 2004, 76, 6998-7006. Williams, R. J.; Ju ¨ rgens, M. D.; Johnson, A. C. Initial predictions of the concentrations and distribution of 17β-oestradiol, oestrone and ethinyl oestradiol in 3 English rivers. Water Res. 1999, 33, 1163-1671. Johnson, A. C.; Williams, R. J. A model to estimate influent and effluent concentrations of estradiol, estrone, and ethinylestradiol at sewage treatment works. Environ. Sci. Technol. 2004, 38, 3649-3658. Thorpe, K.; Hutchinson, T. H.; Hetheridge, M. J.; Scholze, M.; Sumpter, J. P.; Tyler, C. R. Assessing the biological potency of binary mixtures of environmental estrogens using vitellogenin induction in juvenile rainbow trout (Oncorhychus mykiss). Environ. Sci. Technol. 2001, 35, 2476-2481. Bo¨deker, W.; Altenburger, R.; Grimme, L. H. Synopsis of concepts and models for the quantitative analysis of combination effects: from biometrics to ecotoxicology. ACES 1992, 4, 4553. Payne, J. P.; Rajapakse, N.; Wilkins, M.; Kortenkamp, A. Prediction and assessment of the effects of mixtures of four xenoestrogens. Environ. Health Perspect. 2000, 108, 983-987. Brown, V. M. The calculation of the acute toxicity of mixtures of poisons to rainbow trout. Water Res. 1968, 2, 723-733. Solbe´, J. F. de L. G. The relation between water quality and the status of the fish population in Willow Brook. Water Treat. Exam. 1973, 22, 41-61. Alabaster, J. S.; Garland, J. H. N.; Hart, I. C.; Solbe´, J. F. de L. G. An approach to the problem of pollution and fisheries. Symp. Zool. Soc. London 1972, 29, 87-114. Cook, P. M.; Robbins, J. A.; Endicott, D. D.; Lodge, K. B.; Guiney, P. D.; Walker, M. K.; Zable, E. W.; Peterson, R. E. Effects of aryl hydrocarbon receptor-mediated early life stage toxicity on Lake Trout populations in Lake Ontario during the 20th Century. Environ. Sci. Technol. 2003, 37, 3864-3877. Jobling, S.; Williams, R.; Johnson, A.; Taylor, A.; Gross-Sorokin, M.; Nolan, M.; Tyler, C. R. van Aerle, R.; Santos, E.; Brighty, G. Predicted exposures to steroid estrogens in U.K. rivers correlate with widespread sexual disruption in wild fish populations. Environ. Health Perspect. 2006, 114, Suppl. 1, 32-39. Jarvie, H. P.; Neal, C.; Robson, A. J. The Geography of the Humber catchment. Sci. Total Environ. 1997, 194/195, 87-99. National River Authority. River Aire Catchment Management Plan; National Rivers Authority: Northumbria Yorkshire Region, Leeds, U.K., 1993. Feijtel, T. C. J.; Boeije, G. M.; Matthies, M.; Young, A.; Morris, G.; Gandolfi, C.; Hansen, B.; Fox, K.; Holt, M.; Koch, V.; Schroder, F. R.; Cassini, G.; Schowanek, D.; Rosenblom, J.; Niessen, H. Development of a Geography-referenced Regional Exposure Assessment Tool for European Rivers - GREAT-ER. Contribution of GREAT-ER # 1. Chemosphere 1997, 34, 2351-2374. Schowanck, D.; Fox, K.; Holt, M..; Schroeder, F. R.; Koch, V.; Cassani, G.; Matthies, M.; Boeije, G.; Vanrolleghem, P.; Young, A.; Morris, G.; Gandolfi, C.; Feijtel, T. C. J. GREAT-ER: a new tool for management and risk assessment of chemicals in river basins. Contribution to GREAT-ER # 10. Water Sci. Technol. 2001, 43, 179-185. Sheahan, DA.; Brighty, G. C.; Daniel, M.; Jobling, S.; Harries, J. E.; Hurst, M. R.; Kennedy, J.; Kirby, S. J.; Harris, S.; Routledge, E. J.; Sumpter, J. P.; Waldock, M. J. Reduction in the estrogenic activity of a treated sewage effluent discharge to an English river as a result of a decrease in the concentration of industrially derived surfactants. Environ. Toxicol. Chem. 2002, 21, 515519. Johnson, A. C.; Aerni, H.-R.; Gerritsen, A.; Gibert, M.; Giger, W.; Hylland, K.; Ju ¨ rgens, M.; Nakari, T.; Pickering, A. D.; Suter, M. J.-F.; Svenson, A.; Wettstein, F. E. Comparing steroid estrogen and nonylphenol content across a range of European sewage plants with different treatment and management practices. Water Res. 2005, 39, 47-58.
(46) Panter, G. H.; Thompson, R. S.; Sumpter, J. P. Adverse reproductive effects in male fathead minnows (Pimephales promelas) exposed to environmentally relevant concentrations of the natural oestrogens oestradiol and oestrone. Aquat. Toxicol. 1998, 42, 243-253. (47) Berentaum, M. C. The expected effect of a combination of agents: the general solution. J. Theor. Biol. 1985, 114, 413-431. (48) Williams, R. J.; Johnson, A. C.; Smith, J. J. L.; Kanda, R. Steroid estrogen profiles along river stretches arising from sewage treatment works discharges. Environ. Sci. Technol. 2003, 37, 1744-1750. (49) Huang, C.-H.; Sedlak, D. L. Analysis of estrogenic hormones in municipal wastewater effluent and surface water using enzymelinked immunosorbent assay and gas chromatography/tandem mass spectrometry. Environ. Toxicol. Chem. 2001, 20, 133139. (50) Baronti, C.; Curini, R.; D’Ascenzo, G.; Di Corcia, A.; Gentili, A.; Samperi, R. Monitoring natural and synthetic estrogens at activated sludge sewage treatment plants and in a receiving river water. Environ. Sci. Technol. 2000, 34, 5059-5066. (51) Ko¨rner, W.; Bolz, U.; Triebskorn, R.; Schwaiger, J.; Negele, R.D.; Marx, A.; Hagenmaier, H. Steroid analysis and xenosteroid potentials in two small streams in southwest Germany. J. Aquat. Ecosyst. Stress Recovery 2001, 8, 215-229. (52) Ternes, T. A.; Kreckel, P.; Mueller, J. Behaviour and occurrence of estrogens in municipal sewage treatment plants - II. Aerobic batch experiments with activated sludge. Sci. Total Environ. 1999, 225, 91-99. (53) Ju ¨ rgens, M. D.; Holthaus, K. I. E.; Johnson, A. C.; Smith, J. J. L.; Hetheridge, M.; Williams, R. J. The potential for estradiol and ethinylestradiol degradation in English rivers. Environ. Toxicol. Chem. 2002, 21, 480-488. (54) Van Aerle, R.; Nolan, M.; Jobling, S.; Christiansen, L. B.; Sumpter, J. P.; Tyler, C. R. Sexual disruption in a second species of wild cyprinid fish (the gudgeon, Gobio gobio) in UK freshwaters. Environ. Toxicol. Chem. 2001, 20, 2841.2847. (55) Silva, E.; Rajapakse, N.; Kortenkamp, A. Something from “nothing” - eight weak estrogenic chemicals combined at concentrations below NOECs produce significant mixture effects. Environ. Sci. Technol. 2002, 36, 1751-1756. (56) Tinwell, H.; Ashby, J. Sensitivity of the rat uterotrophic assay to mixtures of estrogens. Environ. Health Perspect. 2004, 112, 575582.
(57) Ahel, M.; Giger, W.; Koch, M. Behaviour of alkylphenol polyethoxylate surfactants in the aquatic environment - II. Occurrence and transformation in rivers. Water Res. 1994, 28, 1143-1152. (58) Dussault, E. B.; Sherry, J. P.; Lee, H.-B.; Burnison, B. K.; Bennie, D. T.; Servos, M. R. In vivo estrogenicity of nonylphenol and its ethoxylates in the Canadian environment. Hum. Ecol. Risk Assess. 2005, 11, 353-364. (59) Bennie, D. T. Review of the environmental occurrence of alkylphenols and alkylphenol ethoxylates. Water Qual. Res. J. Can. 1999, 34, 79-122. (60) Kawanishi, M.; Takamura-Enya, T.; Ermawati, R.; Shimohara, C.; Sakamoto, M.; Matsukawa, K.; Matsuda, T.; Murahashi, T.; Matsui, S.; Wakabayashi, K.; Watanabe, T.; Tashiro, Y.; Yagi, T. Detection of genistein as an estrogenic contaminant of river water in Osaka. Environ. Sci. Technol. 2004, 38, 6424-6429. (61) La¨nge, R.; Hutchinson, T. H.; Croudace, C. P.; Panter, G. H.; Siegmund, F.; Schweinfurth, H.; Hampe, P.; Sumpter, J. P. Effects of the synthetic oestrogen 17R-ethinylestradiol over the lifecycle of the fathead minnow (Pimephales promelas). Environ. Toxicol. Chem. 2001, 20, 1216-1227. (62) Rose, J.; Holbech, H.; Lindholst, C.; Nørum, U.; Povlsen, A.; Korsgaard, B.; Bjerregaard, P. Vitellogenum induction by 17βestradiol and 17R-ethinylestradiol in male zebrafish (Danio rerio). Comp. Biochem. Physiol., Part C: Pharmacol., Toxicol. Endocrinol. 2002, 131, 531-539. (63) Metcalfe, C. D.; Metcalfe, T. L.; Kiparissis, Y.; Koenig, B. G.; Khan, C.; Hughes, R. J.; Croley, T. R.; March, R. E.; Potter, T. Estrogenic potency of chemicals detected in sewage treatment plant effluents as determined by in vivo assays with Japanese medaka (Oryzias latipes). Environ. Toxicol. Chem. 2001, 20, 297308. (64) Thomas-Jones, E.; Thorpe, K.; Harrison, N.; Thomas, G.; Morris, C.; Hutchinson, T.; Woodhead, S.; Tyler, C. R. Dynamics of estrogen biomarker responses in rainbow trout exposed to 17βestradiol and 17R-ethinylestradiol. Environ. Toxicol. Chem. 2003, 22, 3001-3008.
Received for review December 21, 2005. Revised manuscript received June 5, 2006. Accepted June 8, 2006. ES052554D
VOL. 40, NO. 17, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5489
