Case study one: Binary mixture
Two freshwater arthropods, Hyalella azteca and Chironomus dilutus (previously named C. tentans), were exposed to mixtures of fluoranthene and pentachlorobenzene in aqueous exposures without changing or measuring the water concentrations (Schuler et al. 2009). The 2 chemicals have similar log octanol–water partition coefficients (KOW, 5.18–5.20) but different toxicokinetics, because fluoranthene is biotransformed whereas there is no measurable biotransformation of pentachlorobenzene (Schuler et al. 2004). The endpoint selected was mortality at 4 and 10 d based on body residue as the dose metric. Residues in the 2 invertebrates were similar at 4 and 10 d for both substances, suggesting that the concentrations in organisms were near steady state. The use of body residue as the dose metric overcomes the difficulty of interpreting the toxicokinetics but does require that biotransformation be explicitly addressed because metabolites can contribute to toxicity. Thus, organic extractable metabolites were included in the calculation of TU and the molar sum of body residues for toxic response. The chemical characteristics of the metabolites were not established. The assumption that the authors used was that organic extractable metabolites would contribute directly to the mortality endpoint at 4 and 10 d.
The toxicity of pentachlorobenzene and fluoranthene were also determined under the same conditions so that the toxic potential of each compound would be available for assessing mixture response. The lethal residue for 50% mortality (LR50) was 1.73 and 1.51 µmol/g as the molar sum of the components in the binary mixture in H. azteca at 4 and 10 d respectively. These values are similar to the expected concentration required to produce mortality by nonpolar narcosis. For C. dilutus, the molar concentrations for the LR50 for the binary mixture were somewhat lower at 0.93 and 0.75 µmol/g, suggesting that the toxicity may not be due strictly to nonpolar narcosis. The lower body residue was driven by the sensitivity of C. dilutus to fluoranthene likely due to the presence of a nonnarcotic or polar narcotic metabolite (Schuler et al. 2006). When the individual potencies were addressed using a TU model, the LR50 values for both organisms were near 1 and showed additivity in TU contributions to the observed mortality. However, for C. dilutus, the molar contribution of fluoranthene and its metabolites to the toxicity was lower than that of the pentachlorobenzene for the same TU. Thus, in this study, there was a clear dose metric that produced monotonic dose-response relationships.
In the case of H. azteca, either molar additivity or TU would have been an acceptable model for examining causality so long as the contributions of organic extractable metabolites were included. Both of these approaches demonstrated additivity. For C. dilutus, the TU model addressed differences in toxic potentials of the 2 compounds, again incorporating the importance of organic extractable metabolites. For C. dilutus, the mode of toxicity was not as clear, but the contribution of the 2 compounds to the toxicity was clear. Thus, in both cases causality was relatively clear except that the chemical identity of the toxic metabolites was not identified and their potencies were not determined.
Measuring the toxic thresholds of the mixture's components was critical to interpreting the mixture interactions. It was particularly critical to understand the complication of biotransformation and account for the toxicity of metabolites to interpret the toxicity of these compounds. A negative feature was the failure to identify the specific toxic metabolite and/or metabolites. However, a TU model including total concentrations of extractable metabolites demonstrated additive interactions of the 2 compounds. The additivity of the 2 compounds was well demonstrated, and their contribution to the toxicity well understood, except that the identity of the fluoranthene metabolites was missing. Because the toxicity was based on body residues for the dose metric, translating the results to environmental concentrations for regulation of water concentrations would require using bioconcentration factors (Schuler et al. 2004, 2006). However, direct measures of environmental residues would be interpretable.
Case study two: Moderately complex mixture
The toxicity to marine copepods of 10 polycyclic aromatic hydrocarbons (PAH), characteristic of those in the heavy fuel oil released by the 2002 Prestige oil spill was determined for narcosis and mortality endpoints by Barata et al. (2005). The authors specifically assumed that PAH were responsible for most of the toxicity of spilled oil (Neff et al. 1976). Exposures to aqueous concentrations were for a period of 48 h to single PAH or equitoxic combinations of PAH mixtures. Chemical analyses of exposure media were done at the beginning and end of the exposure and were restricted to 1 and 3 exposure levels for single-substance and mixture-toxicity assays, respectively. The measured aqueous concentrations declined over time, particularly for the more volatile PAH. The concentrations used for assessment assumed that PAH concentrations that were not measured declined in the same manner. Thus, the concentration for assessment used the average value between the beginning and end of the exposure as the exposure concentration with a similar extrapolation for the nominal concentrations assuming the same relative decline for each compound. The authors also compared the toxicity of each compound to a nonpolar narcosis model fit to the KOW. A TU approach was used to evaluate the combined toxicity of the mixture components for compounds that produced measurable lethal concentrations for 50% mortality (LC50) values within the solubility limit. Dimethylphenanthrene was not included in the mixture study because it did not produce an LC50 or effective toxic concentration for 50% response (EC50) within its solubility limit.
No tissue residue measurements were made, resulting in the assumption that the compounds were taken up in proportion to their aqueous concentration. Measuring tissue residues would have been difficult for such small organisms, but it would have enabled a more direct dose-response determination and would have allowed inclusion of the contribution of more hydrophobic contaminants to the mixture study. By using the aqueous concentration, contributing biotransformation products were implicitly included but not specifically determined; thus, understanding causality was incomplete. A similar approach for evaluating PAH has suggested that metabolites are part of the contribution to the mortality endpoint (DiToro et al. 2000).
Although the authors focused on those mixture components that were toxic below the water solubility limit in individual tests, they did examine the ability of those compounds to fit a quantitative structure activity (QSAR) model for nonpolar narcosis. With the exception of methylphenanthrene, all of the individual compounds fit the QSAR model for nonpolar narcosis, supporting narcosis as the mode of action for most of the components in the mixture. Furthermore, the components of the mixture, including the methylphenanthrene, showed additivity to produce dose-response curves that had 50% response at total TU values that were not different from 1.0. The approach used here does not allow examination of the potential toxicity contribution of less soluble PAH to mixture toxicity. High molecular weight PAH undoubtedly contribute to the toxicity of spilled oil (Neff 2002; Hodson et al. 2007). The failure to explicitly examine body residues or the potential impact of biotransformation products may have contributed to the limited understanding of the mode of action for methylphenanthrene because marine copepods apparently are able to metabolize and excrete PAH rapidly (Sole and Livingston 2005; Berrojalbiz et al. 2009). Because the exposures were short and the mean exposure concentration was used as the dose metric, extrapolation to environmental conditions is possible.
Comparing the experimental results to known modes of action through the use of a QSAR model and the demonstration of additivity was an excellent approach for establishing the contribution of the individual components to the toxicity and to establishing causality for most of the components in the mixture. A negative feature was the failure to measure body residues of PAH and biotransformation products. Measures of body residues might have led to better understanding of the mode of toxic action for the methylphenanthrene and allowed the evaluation of contributions from compounds that do not exhibit acute toxicity within their solubility limit but may contribute to the same mode of toxic action. The investigation was also limited because the authors assumed that PAH were the causative agents for toxicity of spilled oil. However, this does not account for all components of oil and perhaps misses important contributions from other compounds. Although the compounds did fit an expected QSAR model except for methylphenanthrene, there is a need to examine a broader range of compounds from oil that may be contributing either as additive to the narcosis mechanism or producing toxicity by separate mechanisms.
Case study three: Environmental complex mixture
Mixture components contribute to toxicity in a variety of ways and with different interactions. When the number of compounds contributing to environmental mixtures is large and there are multiple mechanisms and/or modes of toxic action, the difficulty in establishing the causative agents or even groups of compounds for each specific endpoint is challenging.
To advance an approach for establishing the contribution of different organic chemicals in a contaminated sediment to mixture toxicity, Sundberg et al. (2005, 2006) used a bio-effect directed fractionation approach to measure the contribution of several organic chemicals to the toxicities of toluene extracts of sediments collected from a reference and contaminated site. The toluene extracts were cleaned up and fractionated into 3 primary fractions that contained aliphatic and monoaromatic hydrocarbons (MAC), diaromatic hydrocarbons (DAC) mostly polychlorinated biphenyls (PCBs), and polycyclic aromatic compounds (PAC = PAH). The PAC fraction was further subfractionated into 10 fractions that contained different mixtures of PAH. The concentrations of PCBs and PAHs in the total extract and all fractions and the amount dosed to the eggs were measured. The toxicity of each fraction was evaluated by injection into rainbow trout (Oncorhynchus mykiss) eggs. Endpoints included egg and larval mortality, teratogenicity, aryl hydrocarbon receptor (AhR) mediated toxicity, hepatic ethoxyresorufin O-deethylase (EROD) activity, and larval malformations. Injection into eggs provided a clearly defined single dose but bypassed potential bioavailability limitations and absorption kinetics expected in environmental exposures. Dose-response relationships were used as part of the evaluation of differences between fractions and to compare the contributions of the fractions relative to total extract response.
The toxicity of the reference site extract and the carrier control were not different at the low exposure concentration and were similar to exposure to a reference compound (benzo[a]pyrene) dosed at 2900 µg/kg egg. The total extract of the reference site sediment containing 210 µg total PAH/kg egg and the contaminated site sediment containing 840 µg PAH/kg egg showed deviation from control, with the reference site only significant for edema, and larval mortality, whereas the contaminated site extract produced high egg and larval mortality and significant responses for all teratogenic endpoints.
When the contaminated site sediment extract was fractionated, the MAC subfraction was not mutagenic and did not induce EROD. The sum of EROD induction demonstrated by the individual DAC and PAH subfractions was greater than that of the total extract and gave dose-response relationships that were parallel, indicating nonadditive effects when eggs were exposed to the whole mixture. All 10 PAH subfractions induced significant EROD activity and 2 subfractions were mutagenic. The greatest induction was produced by the subfraction that contained 5- and 6-ring PAH and the fractions containing 3- and 4-ring and 4- and 5-ring PAH were mutagenic. The DAC fraction induced a higher EROD activity than a similar dose of the PAH fraction but was less mutagenic and teratogenic than the PAH fraction. Thus, there was not a clear relationship between EROD induction and mutagenicity or teratogenicity in these fish larvae (Sundberg et al. 2005).
Because most larval malformations occurred with the PAC subfractions, the eggs were exposed to a synthetic mixture of 17 PAH, and responses were compared to those elicited by exposure to the sediment PAH fractions (Sundberg et al. 2006). The synthetic PAH mixture, composed primarily of high molecular weight nonalkylated PAH, could not explain all the toxic effects and could explain only 4% of the EROD induction demonstrated by the sediment PAC fraction. Thus, not all PAH were acting with equal potency to produce the observed toxicity or there were other unmeasured chemicals (possibly alkyl-PAH and PAH degradation products) in the sediment organic extracts that contributed to toxic effects. Specific PAH contributing to the observed toxicity were not identified, although inferences can be made based on the contributions of different fractions or the synthetic PAH mixture to observed toxic effects.
The use of bio-effect directed fractionation helped to identify important groups of toxic compounds in complex environmental mixtures but was limited, in this case, in that only subfractions of the complex mixture of compounds producing the results could be isolated, not the specific causative compounds. The approach demonstrated the importance of compound interactions, e.g., the lower EROD induction for the whole extract compared to the fractions. However, the findings were limited because the potential limitations of bioavailability of sediment-associated contaminants and the corresponding toxicokinetics of uptake were bypassed. Although the injections provided a clearly defined dose, the exposures could not be directly related to field exposures but might be surrogates for tissue–residue concentrations. Finally, the work demonstrated that all PAH are not equally toxic with respect to different measurement endpoints; thus, more effort is required to identify causative compounds among PAH for specific endpoints.
Case study four: Simulated environmental complex mixture
Heintz et al. (1999) carried out laboratory experiments in 1993 using cylindrical vertical columns containing gravel to which different amounts of weathered Alaska North Slope crude oil had been applied to simulate the exposure of pink salmon (Oncorhynchus gorbuscha) eggs in a tidal environment to hydrocarbons leaching from weathered crude oil in intertidal sediments. Alternating freshwater and seawater, to mimic tidal mixing of fresh and salt water in the salmon stream, were pumped upward through the columns to produce oil in water fractions that served as exposure media for the fertilized eggs.
Two types of weathered oil were used. The first was an oil that had been artificially weathered by heating at 70° C overnight (AWO) The second was a very weathered oil (VWO) on gravel reused (without separation of the VWO from the gravel) from a similar 1992 study (Marty et al. 1997) after the oil on gravel had been water washed in columns for 9 months and then stored outside for 3 summer months.
For the AWO treatments, 7 AWO-dose levels on gravel were used, ranging from 0 to 2450 mg/kg dry wt total extractable organic matter (oil). For the VWO, there was a single treatment using the further degraded oil on gravel from 1992. The initial 1992 oil concentration of this VWO on gravel had dropped from 4510 mg/kg dry wt in 1992 to 2860 mg/kg dry wt in 1993 (Heintz et al. 1995) and after approximately 1 y of natural biodegradation and oil loss, had a very different hydrocarbon composition from the AWO. In addition, the different dose levels of AWO gravels had different initial relative PAH compositions (EVOSTC 2009).
Eggs were placed directly in the gravel for the VWO treatment and all AWO treatments. In addition, there were 3 indirect exposures to AWO at doses of 0, 74, and 717 mg/kg oil on gravel, in which eggs were place on a perforated aluminum plate suspended over the gravel.
Measurements of PAH exposure (as PAH concentration in exposure water and egg–alevin tissues) and egg–alevin responses to the different exposures were made at different times after initiation of the approximately 200-day exposures. Response measurements included egg mortality at eyeing, larva mortality at emergence, and selected sublethal effects in emergent larvae (alevins) (Heintz et al. 1995, 1999). Exposure measurements included the concentration of total extractable organic matter (parts per million, mg/kg) in the gravel at the start of the experiment and the concentrations of alkanes, total polycyclic aromatic hydrocarbon (TPAH), and individual PAH concentrations in water, tissue, and oiled gravel samples taken at time intervals during each experiment. These data are available in the public Exxon Valdez Oil Spill Trustee Council Hydrocarbon Database (EVOSTC 2009). There were no measurements of microbial degradation products, biotransformation products produced by eggs–alevins, or other organic chemicals in column effluents or fish tissues.
Total PAH concentration in the column effluent decreased rapidly during exposure, and tissue PAH concentrations also decreased from an (unknown) initial peak value. Loss of the more volatile compounds was greater at lower treatment concentrations (EVOSTC 2009). Thus, the exposures were time variable in terms of both the concentration and composition of PAH in the effluent water such that a true dilution series of the exposure mixture did not occur.
The time-variable nature of each exposure was equivalent to a single addition exposure with a rapid decline in exposure concentration and changes in composition over time in each treatment, with mixtures of different initial compositions and concentrations over all treatments. Evaluation of the response should have been based on a dose metric that specifically addressed the contribution of the individual ingredients of the mixture with a TU model and/or models for each endpoint. Such TU model and/or models could have been based on 1) the concentrations of each chemical in the mixture in the effluent water using the initial concentration, 2) an integrated external dose approach using peak concentration of mixture components in embryo tissues (the body residue approach), or 3) some measure integrating these 2 exposure metrics. Development of a TU model would have required estimation or measurement of threshold concentrations for the various components of the mixture.
The exposure metric for estimating the dose-response was the loading of total extractable oil on the gravel in the columns, as shown in Figure 2 of Heintz et al. (1999), which is clearly not appropriate as the loading does not reflect the differences in the resulting PAH concentration and composition in exposure water or tissues among the treatments. More important, this dose metric does not reflect differences in the apparent bioavailability of different hydrocarbons in the AWO and VWO treatments (Heintz et al. 1999), even though tissue PAH concentrations were measured. The authors then based conclusions on the initial total PAH concentration in the water as the surrogate exposure dose even though such measures were not used to establish a dose-response (Heintz et al. 1999).
Figure 2. Dose-response curves for Heintz et al. (1999) day 36 embryo mortality data with the dose given as (A) initial water alkyl-phenanthrenes concentration (EVOSTC 2009), and (B) day 36 egg alkyl-phenanthrenes concentration (EVOSTC 2009). Using tissue or water TPAH as the dose metric yields similar graphs. Because the toxicant concentration (PAH or alkyl-phenanthrenes) for the VWO experiment is in the same range as those showing no toxic effect, the toxicity observed in the VWO experiment was not PAH-related.
Download figure to PowerPoint
Tissue residue data were used subsequently to evaluate the dose-response (Barron et al. 2004). However, the time point selected for evaluation of the Heintz et al. (1999) body residue data was the PAH concentration in the eggs after 36 d exposure. This concentration probably was below the peak concentration, which likely occurred before 36 d, thus likely underestimating exposure to the un-metabolized PAH (Mathew et al. 2008) and did not incorporate contributions from biotransformation products.
Heintz et al. (1999) reported that an initial 18 µg/L aqueous TPAH concentration was toxic to embryos in the AWO treatment and a 1 µg/L aqueous TPAH was toxic to embryos in the VWO treatment, where toxicity was expressed as mortality and growth reduction. They concluded that toxicity of the AWO and VWO was likely caused by the higher molecular weight PAH (alkyl-phenanthrenes and chrysenes) because the relative concentration (fraction of total PAH) of these higher molecular weight PAH was higher in the VWO than the AWO water. Actually, measured concentrations of all PAH, including alkyl-phenanthrenes, were much lower in the 1 µg/L aqueous dose of VWO than in the 18 µg/L toxic dose of AWO (EVOSTC 2009). As a result, the subsequent evaluation of tissue data did not yield a monotonic dose-response relationship when the VWO treatment was included (Barron et al. 2004). An attempt to replicate the results of the VWO exposure, using gravel columns containing naturally weathered oil collected from the shore of Prince William Sound after the Exxon Valdez spill, demonstrated that the concentrations of aqueous TPAH in the effluent from the naturally weathered oil at 8.27 µg/L did not exhibit mortality or elevated blue sac disease (Brannon et al. 2006). The egg and larvae mortality reported by Heintz et al. (1999) in the VWO treatment occurred at tissue PAH concentrations (EVOSTC 2009) below those showing no mortality for AWO treatments (Figure 2), indicating a lack of a dose-response relationship for the VWO treatment. Alkyl-phenanthrene concentrations, which are approximately proportional to TPAH concentrations, are plotted as the dose in Figure 2 because Heintz et al. (1999) assumed, but did not prove, that they represent the toxic components of the VWO.
Causality of the observed toxicity was not established. The authors seem to have assumed that dissolved PAH were the only toxicants of concern and that all target PAH were acting with equal potency for all endpoints investigated, based on use of TPAH in their conclusions. The authors did acknowledge the potential for different modes of toxicity by different PAH in the mixture but failed to follow-up with experiments or models to support this possibility, including demonstration of dose-response. They also suggested that metabolites and biodegradation products could be contributing to toxicity but did not measure or estimate metabolite concentrations in the embryos or even all organic components in the column effluents, including other hydrocarbons, phenols, and microbial biodegradation products.
If the apparent dose-response for the AWO is evaluated independent of the VWO, there is a suggested monotonic dose-response relationship when using either total PAH or total alkyl-phenathrenes (see Figure 2 for the alkyl-phenanthrene curve). However, the observation of an apparent dose-response for the AWO exposures does not prove causality. In the work of Barron et al. (2004), where selected models were examined to attempt to establish causality, it is clear that there were not high enough concentrations in the tissue to cause mortality by narcosis for total PAH. The best model was an alkyl-phenanthrene model based on retene as the denominator for creating the TU. This model may have overestimated the TU of some of the alkyl-phenanthrenes because retene is the most toxic of this class, whereas other alkyl-phenanthrenes are less toxic based on their EC50 values (Turcotte et al. 2011). Thus, if the species specific toxicities of each of the alkyl-phenanthrenes were used, the number of TU may have been lower, which would have altered the assessment of the contribution of these potential causative agents. However, these models lay a foundation for establishing that alkyl-phenanthrenes were substantial contributors to the toxicity for the AWO studies. This conclusion would need to be supported with confirmative toxicity tests containing only alkyl-phenanthrenes. Furthermore, the use of models that explain most of the toxicity with only a portion of the PAH signature, alkyl-phenanthrenes, demonstrate that conclusions cannot be made on the basis of TPAH in either water or tissues, because the composition of the PAH mixtures vary substantially and TPAH would not necessarily retain the same proportion to the alkyl-phenanthrene fraction for all potential environmental mixtures. Finally, the results of the modeling would still be deficient for the VWO exposure, which seems to be an outlier for all the models and suggests that other compounds or stressors than just PAH or alkyl-phenthrenes in the gravel column effluents were participating in the VWO toxicity.
It is particularly difficult to meet the 2 fundamental criteria for mixture studies with complex mixtures and exposures. Failure to fully characterize the mixture composition and concentration over time including biotransformation products in tissues as well as exposure water and to establish an appropriate dose metric limits the interpretation of the experimental results. Finding that some of the treatments do not fall on a dose-response curve when applying the appropriate dose metric, e.g., aqueous or tissue concentrations, prevents interpreting the results from that outlying treatment relative to the dose metrics used. This emphasizes the need to consider all stressors that may contribute to the observed responses. In addition, the failure to establish dose-response for all treatments based on a consistent and meaningful surrogate and to establish causality, limits the conclusions that can be drawn and makes extrapolation of the data to environmental conditions impossible. Furthermore, correlation creating a dose-response relationship only provides inference of toxicity and does not confirm causality; thus, it is critical to perform definitive tests to determine actual causality rather than making assumptions about which compounds are contributing to specific toxic responses.