In the experiments described by Carls et al. (1997, 1999), Pacific herring (Clupea pallasi) eggs were exposed to seawater effluents from artificially weathered crude oil coated onto gravel and placed in columns through which seawater was pumped. The exposures comprised 2 experiments: a 16-d exposure of herring embryos to effluents from the freshly prepared oiled-gravel columns, called “less weathered oil” (LWO), followed by a 13-d period when water flow through the columns was turned off, and then a 16-d exposure of a different set of herring embryos to effluents from the same oiled-gravel columns, called “more weathered oil” (MWO), was initiated. There were 4 treatments in each experiment: High, Mid (middle), Low, and Trace oil loading to the gravel along with a control.
The toxicity of the effluents was evaluated based on embryo and larval mortality and 10 sublethal toxicity parameters (Carls et al. 1999). Total polycyclic aromatic hydrocarbons (PAH) concentrations in water, measured at the beginning of each experiment, served as the selected dose metric. Tissue PAH residues were measured in embryos at different times during exposure, but not for all treatments or at the same frequency in all exposures. Water PAH concentrations were measured at 6 to 8 time intervals for different treatments of the LWO experiment and at 5 time intervals for the MWO experiment treatments, except controls. Tissue PAH concentrations also were measured at 3 time intervals during the exposure period for only the LWO-High and LWO-Control treatments and at 3 to 5 time intervals during the exposure period for all MWO treatments. Tissue PAH concentrations were also measured at 4 time intervals postexposure for the LWO-High and 3 time intervals postexposure for the LWO-control. No postexposure data were collected for the MWO experiment. PAH concentrations were not measured in herring embryos from the LWO-Mid, Low, and Trace treatments (EVOSTC 2009).
As demonstrated in the theoretical discussion, the initial total PAH (TPAH) concentration in water provides a conservative description of exposure; however, peak tissue dose will vary in time and magnitude depending on the initial water concentration, rate of loss of different target PAH from the water, and the toxicokinetics of each PAH in the organism. In the case of TPAH, as shown below, exposure dynamics also depend on the composition of the mixture. Thus, selecting initial water concentrations of TPAH as a dose metric may not be an accurate or appropriate approach for assessing dose (Landrum et al. 2012) and would not be appropriate for comparisons among experiments or spill sites unless the loss rate and mixture compositions in exposure water were identical or at least very similar. The extent of conservatism will depend on the actual kinetics of the decline in water concentrations, the composition of the mixture, and the toxicokinetics of the different mixture components in the organism.
We investigated the Carls et al. (1997, 1999) experiments based on toxicokinetics and simulations that included treatments where no tissue concentration measurements were taken to better interpret the exposure conditions represented by the treatments. Three analyses were performed for unsubstituted PAH, alkylated PAH, and TPAH. In the first analysis, because of variability in the wet to dry weight data (EVOSTC 2009), toxicokinetics were explored to assess the significance of this variability to further analyses. The data were fit to a first order toxicokinetic model (Eqn. 1) by numerical integration using Scientist (Micromath, St Louis, MO). The model included changes in water concentration with cessation of exposure after 16 d. This first order equation assumes that the elimination rate constant includes both passive loss from the organism and loss due to biotransformation.
In the second analysis, simulations of bioaccumulation were carried out based on toxicokinetics determined by Mathew et al. (2008), using data from Carls et al. (1999) and using the best representation of the change in water concentration from the data (EVOSTC 2009). The toxicokinetic parameters determined by Mathew et al. (2008), where the loss rate constant due to passive elimination based on partitioning and the biotransformation rate constant, were combined to provide the total loss rate constant. Our simulation results were compared to measured tissue concentrations from the same data source.
In the third analysis, because the work of Mathew et al. (2008) did not include TPAH as part of the analysis, the LWO-High treatment data (n = 10), for which both water and tissue concentrations were measured, were fit to a first order toxicokinetic model (Eqn. 1). Kinetic constants were used to simulate the other treatment concentrations for TPAH bioaccumulation.
The modeling analyses involved TPAH, naphthalene, C1(mono-alkyl)-naphthalenes, phenanthrene, C1-phenanthrenes, dibenzothiophene, C1-dibenzothiophenes, chrysene, and C1-chrysenes. These PAH covered the full range hydrophobicity modeled by Mathew et al. (2008).
PAH concentrations in exposure water
Concentrations of individual and TPAH in exposure water showed different trends over time, ranging from little or no change during the 16-d exposure period to double exponential declines. Most of the data sets followed a single first order decay model (Table 1). However, for naphthalene and C1-naphthalenes, none of the models was adequate for describing trends in water concentration over time. For these 2 compounds, water concentration trends typically showed declines that were best described by a first order decay model for 2 individual time periods, each of which was used to model contaminant uptake by the herring eggs.
Table 1. Models for the change in water concentration (ng/mL) over the 16-d exposure duration for the 6 treatments examined for toxicokinetics from the data of Carls et al. (1999) as found in EVOSTC (2009) a
|Naphthalene||0–4 d; 10.7e−0.67t||1.37e−0.641t||0.211e−0.32t||0.028e−0.88t + 0.015e−0.11t||0.0051 ± 0.0008||0.0098e−0.045t||0.00968e−0.046t|
| ||4–16; d 0.47e−0.15t|| || || || || || |
|C1-naphthalenes||0–2 d; 40.2e − 0.76t||10.0e−0.29t||0.50e−0.20t||0.56e−0.43t||0.0067e−0.031t||0.012e−0.043t||0.014e−10.06t|
| ||2–16 d; 8.2 ± 2.0|| || || || || || |
|Phenanthrene||1.38e−0.033t||1.14–0.72t||0.17e−1.57t + 0.14e−0.21t||0.18e−0.30t||0.0035 ± 0.00023||0.0086e−0.16t||0.0036e−0.055t|
|C1-phenanthrenes||1.04 ± 0.14||1.08–0.055t||0.74e−0.22t||0.40e−0.20t||0.016e−0.029t||0.0084e−0.25t||0.0025–0.00019t|
|Dibenzothiophene||1.09e−0.03t||0.86–0.054t||0.11e−1.82t + 0.089e−0.23t||0.22e−0.20t||0.0011 ± 0.00042||0.00097e−0.047t||0.0011e−0.059t|
|C1-dibenzothiophenes||5.02 ± 0.074||0.536–0.0254t||0.33e−0.22t||0.12e−0.14t||0.011e−0.03t||0.0054e−0.15t||0.0024e−0.11t|
|Chrysene||0.010 ± 0.003||0.0.011 ± 0.0041||0.0096e−0.036t||0.015 ± 0.0026||0.0093 ± 0.0011||0.0064e−0.017t||0.0031e−0.065t|
|C1-chrysenes||0.010 ± 0.006||0.0073 ± 0.0020||0.0069e−0.031t||0.013 ± 0.0040||0.0058 ± 0.0010||0.0073e−0.028t||0.0040e−0.023t|
|TPAH||51.3e−1.14t + 34.7e−0.02t||33.7e−0.15t||8.20e−0.027t||7.53e−0.17t||0.74e−0.021t||0.40e−0.12t||0.11e−0.057t|
Trends in declining water TPAH concentrations, used by Carls et al. (1999) as the representative dose metric, were dominated by the rates of decline for the lower molecular weight compounds, particularly for the LWO-High treatment. Concentrations of higher molecular weight compounds, such as chrysene, remained nearly constant during the tests. The traces of 4- through 6-ring PAH in the exposure water are likely associated almost exclusively with micro-oil droplets (Faksness et al. 2004; Redman et al. 2012) that have a PAH composition similar to that in the oiled-gravel columns. These PAH are depleted from the oiled gravel columns slowly, because of their very low aqueous solubilities and high log Kowvalues (Schleup et al. 2002), explaining their low, but relatively stable concentrations in the exposure water.
Wet versus dry weight for modeling
Because there was substantial variability in the embryo tissue dry/wet weight ratio in the Carls et al. (1999) herring embryo study (EVOSTC 2009), the appropriate condition for modeling the toxicokinetics was in question. The LWO-High 4-d tissue sample had the lowest dry/wet weight ratio (0.06 = 94% water); the ratio for embryos from all other exposure concentrations ranged from 0.12 to 0.19, suggesting that the relationship was otherwise not extremely variable. The average dry/wet weight ratio for embryos from all exposure concentrations and sampling times, except control, was 0.146 ± 0.027 (mean ± SD, n = 25).
The comparison between modeling the kinetics using wet or dry weight was based on the LWO-High data set, which had the most data as well as the highest variation in dry/wet weight ratio. TPAH was a useful data set to model because it contained the most detailed data for the series of treatments and provided the kinetics for later use for simulations.
The model (Eqn. 1) for the toxicokinetics of TPAH on a dry weight basis yielded a coefficient of determination (r2) of 0.95 with a ku of 616.5 ± 49.4 mL g−1dry weight d−1 and a ke of 0.20 ± 0.01 d−1 (Figure 3A). On a wet weight basis, the r2 was 0.98 with a ku of 85.4 ± 13.1 mg g−1wet weight d−1 and a ke of 0.18 ± 0.03 d−1 (Figure 3B). The ke values yielded similar elimination half-lives of 3.4 to 3.8 d for the 2 determinations. The 2 uptake constants only differed by a dry/wet weight ratio of 0.139, which fit within the range of most of the data, and which was not statistically different from the average across all treatments. Based on the similarity of the data fits, toxicokinetics were evaluated on a wet weight basis as per Mathew et al. (2008).
Figure 3. Model of the toxicokinetics of accumulation of TPAH for the LWO-High treatment on a dry weight (A) and wet weight basis (B) using data taken from EVOSTC (2009).
Download figure to PowerPoint
Evaluating toxicokinetics for consistency across treatments
If all treatments had adequate tissue data available for comparison, the toxicokinetics for each treatment could have been determined and a comparison of the exposure conditions across treatments could have been easily assessed. Unfortunately, the treatments LWO-Mid, -Low, and -Trace did not have any measured tissue data, and there were few and inconsistent data points for the MWO treatments (EVOSTC 2009).
However, assuming that the toxicokinetics did not vary across treatments, it was reasonable to apply the toxicokinetic parameters from Mathew et al. (2008) to each treatment to represent the toxicokinetics across all treatments (J McGrath, HydroQual, Mahwah, NJ, USA, personal communication). The kinetic constants from Mathew et al. (2008) for specific compounds were fit using available water data (Table 1). Naphthalene, C1-naphthalenes, phenanthrene, C1-phenanthrenes, dibenzothiophene, C1-dibenzothiophenes, chrysene, and C1-chrysenes were examined. Of these, naphthalene, C1-naphthalenes, and C1-chrysenes could not be fully examined because problems were identified with the toxicokinetics constants provided by Mathew et al. (2008). Specifically, problems were found with the elimination rate constant for naphthalene and C1-naphthalenes and both the uptake and elimination constants for C1-chrysenes. In the case of the naphthalenes, the reported elimination constants were substantially larger than those found by fitting the LWO-High naphthalene and C1-naphthalene data independently; this likely is the result of the finding in Mathew et al. (2008) that the rate constant determined for metabolism was negative, indicating that their model was incorrect for these 2 compounds.
To correct for this issue, the elimination constants for naphthalene and C1-naphthalene were determined by fitting the LWO-High data independently to provide the best estimate for their elimination rate constants. Fitting these data independently led to estimates for the uptake constants that were essentially the same as the Mathew et al. (2008) published values (1.03 ± 0.05 L g−1 lipid d−1 for naphthalene and 10.0 ± 1.8 L g−1 lipid d−1 for C1-naphthalenes). Statistical significance could not be determined because Mathew et al. (2008) did not publish the estimates for the error terms for the rate constants. However, because the values were similar, the uptake constant published by Mathew et al. (2008) was retained for modeling and the estimate of the elimination from the independent model was substituted (Table 2).
Table 2. Toxicokinetic parameters used for modeling tissue uptakea
|Compound||ku (L g−1 lipid d−1)||ke (d−1)b|
|Naphthalene||0.99||0.158 ± 0.008c|
|C1-naphthalenes||7.39||0.252 ± 0.046c|
|C1-chrysenes||0.518 ± 0.074c||0.133 ± 0.046c|
|TPAH||1.15 ± 0.178c||0.18 ± 0.029c|
For the C1-chrysenes, both the uptake and elimination rates of Mathew et al. (2008) were too large compared to an independent fit of the C1-chrysene data for LWO-High. Their published values resulted in model simulations of the kinetics that grossly overestimated the measured values. This discrepancy may have resulted from the approach used to model the water concentration, which varied from an initial estimated value of 3.3 ng L−1 to values as high as 20.76 ng L−1. As discussed above, most of the C1-chrysenes (and other high log Kow PAH) were likely associated with microdroplets in the exposure water, with very little in the more bioavailable dissolved form, perhaps explaining the unrealistic uptake rates. We elected to estimate the water concentration as a constant using the average concentration, which may have been different from the approach that Mathew et al. (2008) selected to model their water concentrations. To allow for model comparison to measured data, we employed both the uptake and elimination rate constants (Table 2) determined by fitting the uptake to the first order model from the LWO-High C1-chrysenes data (EVOSTC 2009).
Total elimination rate used for the modeling was the combination of the loss from passive elimination rate (kr) and biotransformation rate (km) (Mathew et al. 2008). Consistency across treatments was evaluated using the model data for phenanthrene (Figure 4), C1-phenanthrenes (Figure 5), dibenzothiophene (Supplemental Figure 1), and C1-dibenzothiophenes (Supplemental Figure 2). These data represent both a parent and an alkylated PAH. It was important to examine the alkylated PAH as well as the unsubstituted PAH because the alkylated PAH may be more potent for some of the endpoints (Rhodes et al. 2005; Turcotte et al. 2011).
In general, modeled data agree with the empirical data. The relative percentage differences (RPDs) between the modeled and measured data were 46.9 ± 43.7% for phenanthrene and 50.4 ± 40.6% for C1-phenanthrenes. The RPDs for dibenzothiophene and C1-dibenzothiophenes were 38.6 ± 24.4% and 33.9 ± 30.8%, respectively. These comparisons indicate that the modeled data and empirical data were comparable within about a factor of 2, indicating that the model was robust, and that estimates of the toxicokinetic parameters did not vary across the treatments for individual compounds. This finding supports the hypothesis that a simulation of the toxicokinetics, particularly where there were no measured tissue values, could estimate absorbed doses for similar exposure conditions, allowing tissue residue evaluation across all treatments. The toxicokinetics for naphthalene, C1-naphthalenes, chrysene, and C1-chrysenes (Supplemental Figures 3–6) did not fit the model as well, because of the complications with the toxicokinetics parameters discussed above.
The toxicokinetics determined from the LWO-High data set for TPAH were used for modeling across all treatments. The comparison of the TPAH across treatments was generally not as good as for the individual PAH discussed above and tended to show greater deviation at the lower concentrations for the MWO treatments (Supplemental Figure 7). The fit of the TPAH to obtain toxicokinetic parameters for modeling was dominated by naphthalene and C1-naphthalenes for LWO-High, possibly explaining a general overprediction of accumulation for the MWO treatments where these compounds represented a lower proportion of the water concentration.
Analytical data variability was assessed relative to the modeling results. There were few replicate measurements and only 1 triplicate tissue PAH concentration measurement (EVOSTC 2009) in the Carls et al. (1999) study. The triplicate tissue residue measurement on day 15 of the LWO-High treatment allowed for calculation of percentage standard deviation (PSD), which ranged from 5.8% for naphthalene to 30.9% for chrysene with an average of 16.9 ± 7.8% for the PAH selected for evaluation, including TPAH. The RPD determined for duplicate tissue measurements ranged from 0.6% for chrysene to 39.1% for naphthalene with an average of 10.5 ± 11.8%. Variability increased as the TPAH concentration declined. In the MWO-High treatment the RPD ranged from 5.6% for chrysene to 113.1% for naphthalene with an average across the substances examined of 51 ± 42%. There were triplicate samples of exposure water collected on day 0 of the LWO-Mid treatment that yielded a range of PSD from 11.5% for C1-naphthalene to 70.2% for chrysene with an average of 30.5 ± 21.9% for TPAH and different individual PAH. The RPD ranged from 0.14 to 15.8% for all the PAH and conditions where there were duplicate samples taken. When considering the variation in both water and tissue PAH concentrations, a factor of 2 between the modeled and measured values was considered reasonable. Thus, the simulations are expected to reasonably represent the exposure of the herring eggs.
Comparison of exposure using simulated toxicokinetics
If the data had been more abundant both temporally and across treatments, direct comparisons could have been made. However, given that toxicokinetics for individual compounds are likely constant across the treatments, based on the above evaluation, and that the simulations reasonably reflected the measured data, expected exposures across all treatments were compared based on the modeling, focusing on phenanthrene, C1-phenanthrenes, dibenzothiophene, and C1-dibenzothiophenes. The High, Mid, and Low concentrations for these PAH were examined for both the LWO and MWO experiments as these were the conditions under which Carls et al. (1997, 1999) observed a wide range of toxicological responses.
Based on the exposure of the herring eggs across treatments to phenanthrene (Figure 6), it is clear that, whereas the toxicokinetic parameters were constant, the exposure conditions led to different predicted temporal patterns and concentrations of phenanthrene in tissues. Specifically, the predicted peak tissue concentration shifted from day 9 for LWO-High (Figure 6A) to approximately day 2 and 3 for LWO-Low (Figure 6A) and MWO-High (Figure 6B), respectively. The LWO-Low and MWO-High treatments had similar exposure patterns although tissue phenanthrene appears to have been eliminated faster in the LWO-Low treatment, reflecting a more rapid decline in the water concentrations. Thus, for phenanthrene exposure to be evaluated at the peak exposure concentration, day-2 data would have been needed for some of the exposures; however, Carls et al. (1997, 1999) took their first tissue sample on day 4 of all treatments except MWO-Mid and, thus, were unable to measure the peak tissue concentration. Tissue samples were collected on days 1, 2, 4, 8, and 16 of the MWO-Mid treatment; tissue phenanthrene concentrations were similar on days 1, 2, and 4 (7.9, 8.0, and 9.9 ng/g dry wt, respectively) and were only slightly higher than in day 2 MWO-Control embryos (7.6 ng/g). Thus, tissue phenanthrene concentration reached a peak between day 2 and 4 in the MWO-Mid treatment. The low frequency of tissue sampling also precluded measurement of the peak tissue concentration of phenanthrene for the LWO-High treatment. This example illustrates the need for frequent sampling when comparing across treatments with different source-loss rates, particularly so that peak concentrations and trends are adequately measured. The fact that tissue concentrations peaked at different time intervals suggests that integrated exposures, even with similar peak concentrations, can occur at different developmental points in early life stages of aquatic animals, such as the different embryonic and larval stages of herring.
Depending on the endpoint and its timing relative to the peak exposure time, magnitude, and mechanism of action, exposures that peak at different times, even if the peak is of the same magnitude, may lead to different responses because of differences in time for repair processes to take place. When both peak tissue concentration and timing differ, particularly for different ingredients in a complex mixture, such as weathered petroleum, determination of the causes of different lethal and sublethal endpoints will be difficult if not impossible. However, estimating exposure based on peak-tissue concentration or the maximum concentration attained within the exposure time selected for the response is expected to lead to a reasonably conservative estimate of the maximum expected response. This expectation assumes that the tissue residue is proportional to the concentration at the site of toxic action and that the substance that is measured is the toxic agent or proportional to it. Even if a specific metabolite of the parent substance is the toxic agent, it should be proportional to the accumulated dose assuming that biotransformation rate is first order within the range of exposure conditions.
C1-phenanthrenes, dibenzothiophene, and C1-dibenzothiophenes (Supplemental Figures 8–10) show a similar variation in the timing of the peak concentration, depending on the source-loss kinetics. The modeled earliest peak tissue concentration for these PAH occurs at day 4, but some of the peak concentrations occur much later and, for C1-phenanthrene LWO-High, the modeled data reflect the measured relatively constant-exposure concentration that varied between 0.7 and 1.1 µg L−1 during the 16-day exposure. Measured tissue residues of C1-phenanthrene declined by approximately 38% between day 4 and day 15 in the LWO-High exposure, possibly reflecting induction of CYP1A enzymes that occurs on approximately day 7 postfertilization in Pacific herring embryos (Incardona et al. 2009). These comparisons illustrate the fact that concentrations of different substances peak in the organism at different times after the beginning of the exposure, dependent on both the source kinetics and the toxicokinetic parameters for the particular substance. This fact complicates interpretation of exposure conditions that lead to a toxic response. Even if time-variable response data were available to provide an estimate of toxic equivalents, the fact that timing of uptake varies along with concentration creates a difficult, if not impossible, condition for interpreting mixture interactions for the observed responses. Clearly, data collected from testing, such as conducted by Carls et al. (1997, 1999), are specific to those test conditions and generally cannot and should not be applied to other test conditions or field data without specific consideration of the exposure dynamics.
Complexity increases when considering TPAH rather than individual PAH in the complex mixture, because the proportion and concentration of the PAH in the mixture contributing to the observed response vary along with the variation in peak concentration timing and intensity (Supplemental Figure 11). What appears as a set of time-variable kinetics is actually the summation of the kinetics of the individual substances that make up the particular mixture and the individual decline in source concentrations for each substance at each treatment level.
In each treatment, there is clear variation in source concentration (Table 1). Thus, TPAH do not represent any specific mixture at any point in time in the exposure or among treatments. Even assuming that response to a specific TPAH mixture were known, the variation in the composition and timing of the peak tissue concentrations of the individual components among the treatments would lead to an uneven evaluation of exposures if one selected TPAH as the dose metric based on tissue concentration. This would be the case even accounting for the differences in timing and concentration of the peak value. Modeling of TPAH detailed above largely overestimated exposure (Supplemental Figure 7) because of both differences in the contributions of the various substances in the mixture to the TPAH profile and the use of estimated values from the LWO-High treatment. Clearly, TPAH should not be used as a dose metric in either water or tissue. This complicates the interpretation of the response data and indicates the need for investigation of specific causative agents and the temporal thresholds for such responses. Combining the impact of multiple compounds in mixtures through approaches such as a toxic unit analysis would help focus on both individual compound potency and would assist in sorting out compounds acting by different mechanisms of action.