Survival time models quantitatively can predict lethal effects
of pulsed and different duration exposures to wateraccommodated fraction PAH from spilt oil
A Final Report Submitted to
The Coastal Response Research Center
Submitted by
Dr. Michael C. Newman
Dr. Michael A. Unger
Department of Environmental and Aquatic Animal Health
Virginia Institute of Marine Science
College of William and Mary
P.O. Box 1346
Gloucester Point, VA 23062
Project Period: February 1, 2005 to January 31, 2007
Submitted: August 10, 2007
This project was funded by a grant from NOAA/UNH Coastal Response Research Center.
NOAA Grant Number(s): V710640. Project Number: 05-956
Abstract
The CRRC 2004 RFP identified the need to quantitatively predict injury from realistic conditions
including pulsed, short term and long term exposures. The research described herein addresses
this 2004 priority area.
Toxicity data derived from conventional concentration-effect test designs predict effect at a
single exposure time. A few test durations might be used in some tests to coarsely predict how
mortality changes with exposure time. Even in tests employing several exposure times, the
selected concentration treatments generate optimal data for only one of the exposure durations.
Predictions of effect at different durations are unavoidably gross because mortality information
was not collected optimally during the entire exposure. Also, the conventional concentrationeffect approach does not quantify mortality that potentially could occur after exposure stops.
Such post-exposure mortality can be quite high.
These shortcomings can be avoided by noting mortality in test treatments through time including
post-exposure mortality, and applying survival time modeling to the resulting data. Survival time
models allow inclusion of covariates such as exposure concentration, resulting in models that
include both exposure concentration and duration. We parameterized survival time models with
data generated for six representative polycyclic aromatic hydrocarbons (PAH) associated with
the water-accommodated fraction of oil. Survival models incorporating exposure concentration
and duration were produced for the grass shrimp, Palaemonetes pugio, a common test species
and an ecologically important one in salt marshes and other coastal environments. The absence
of post-exposure mortality allowed use of models based only on mortality during the actual
exposure. The results were then used to demonstrate a QSAR approach to predicting survival
time model parameters for untested PAH.
Keywords: PAH, survival, intermittent exposure, grass shrimp
2
Acknowledgments
Funding was provided by the NOAA/UNH Coastal Response Research Center (V710640). Help
with animal maintenance and tests provided by John Carriger and Kyle Tom is much
appreciated.
3
Table of Contents
1.0 Introduction …………………………………………………………………………………..6
2.0 Objectives …………………………………………………………………………………….7
3.0 Methods ………………………………………………………………………………………8
4.0 Results ………………………………………………………………………………………..9
5.0 Discussion and Importance to Oil Spill Response/Restoration ……………………………..14
6.0 Technology Transfer ………………………………………………………………………...10
7.0 Achievement and Dissemination ..……………………………………………………….…15
8.0 References ...…………………………………………………………………………………16
Appendices ………………………………………………………………………………………17
Appendix A. ……………………………………………………………………………..18
Appendix B ……………………………………………………………………………...44
4
List of Figures and Tables
FIGURES
Figure 1. Cumulative mortality for amphipods (Hyalella azteca) exposed for 48 hours to four
copper concentrations (Zhao and Newman 2004). Substantial post-exposure mortality occurred,
reflecting damage incurred during exposure. Conventional toxicity test designs and associated
metrics do not quantify post-exposure mortality, leading to the possibility of a downward
inaccuracy in predicted mortality for field exposures.
Figure 2. Exposure tanks used in all shrimp survival time experiments.
Figure 3. Survival curves (bars indicate the estimated standard errors) from the phenanthrene
tests. With the termination of exposure (60 hr), mortality quickly dropped to a minimum level in
all concentration treatments. Greenwood’s formula was used to produce the standard errors
shown as vertical bars for each exposure time.
Figure 4. An illustration of the ability of survival time models to predict the mortality with
different combinations of exposure time and concentration. In this figure, contours for
percentage of exposed shrimp expected to die are shown. The 48 hr LC50 estimate is also shown
for comparison. Less than 5% of the mortality during the entire test was post-exposure mortality.
Figure 5. QSAR models for 48 hr LC50 (top panel) and parameter estimates (b1 and b2) for the
survival time models (middle and bottom panel).
Figure 6. Concentration-effect (log normal) model slopes for the six compounds estimated with
the Microtox7 toxicity assay. Brackets indicate 95% confidence intervals.
TABLES
Table 1. Concentration-effect models (48 hr exposure duration)(95% CI = 95% confidence
interval, SE = standard error)
Table 2. Parameter estimates for survival models. (95% CI = 95% confidence interval, SE =
standard error)
5
1.0 Introduction
The CRRC 2004 RFP identified the need to quantitatively predict injury from realistic exposures
including pulsed, short term and long term exposures. The research described here addresses this
priority area, illustrating an approach to combining exposure duration and concentration into
lethality predictions, and generating information applicable nationwide to oil spills. The general
approach might also be relevant to predicting injury when confounding factors change through
time, e.g., changes in PAH-induced mortality due to diurnal changes in UV light intensity or
seasonal temperature changes.
Most toxicity data are derived from conventional concentration-effect test designs that produce
an effect metric, such as the 96 hr LC50, at one set exposure time. Occasionally, a few test
durations might be used to coarsely predict how mortality changes with exposure time. Gross
prediction is inevitable if mortality information was not collected throughout the exposure. As a
further complication in the conventional concentration-effect approach, mortality potentially
occurring after exposure stops is not considered in predictions. In the few studies that quantified
such post-exposure (latent) mortality, it was found to vary widely and could be quite high.
Mosquitofish exposed to a concentration of sodium chloride that killed 15% of exposed fish by
144 hr, experienced an additional 44% mortality in the days following the pulsed exposure
(Newman and McCloskey 1996, 2002). Zhao and Newman (2004) measured high (15 to 35%;
copper sulfate; e.g., Figure 1) to minimal (2 to 5%; sodium pentachorophenol) latent mortality of
amphipods. The magnitude of latent mortality appeared to depend on the mode of action,
consequent residual damage, and recuperative capacity of the exposed individual.
The conventional approach is limited relative to predicting all mortality from pulsed exposures
differing in duration from conventional tests. These shortcomings are weighty impediments to
accurately predicting effects from spilt oil exposures that vary in both duration and concentration
through time, and that require prediction of all mortality resulting from exposure.
0.9
Exposure
Post Exposure
0.6 mg/L
0.8
0.7
Proportion Dead
0.6
0.4 mg/L
0.5
0.3 mg/L
0.4
0.3
0.2 mg/L
0.2
0.1
Control
0
0
20
40
60
80
100
120
Time (hours)
Figure 1. Cumulative mortality for amphipods (Hyalella azteca) exposed for 48 hours to four copper concentrations
(Zhao and Newman 2004). Substantial post-exposure mortality occurred, reflecting damage incurred during
exposure. Conventional toxicity test designs and associated metrics do not quantify post-exposure mortality, leading
to the possibility of a downward inaccuracy in predicted mortality for field exposures.
6
2.0 Objectives
The shortcomings just described can be avoided by noting mortality in test treatments through
time including any potential post-exposure mortality, and applying survival time models to the
resulting data. We applied survival time methods to data generated for representative PAH1
associated with weathered oil. Survival models capable of incorporating exposure concentration
and duration, and post-exposure mortality were produced using widely accepted methods from
health science, social sciences, engineering, and ecology (Allison 1995, Cox and Oakes 1984,
Kalbfleisch and Prentice 1980, Muenchow 1986, Miller 1981, Nelson 1972). Although
underutilized in ecotoxicology, the PI (Newman and Dixon 1996, Newman and McCloskey
1996, 2002, Zhao and Newman 2004) and others (Crane et al. 2002, Lee et al., 2002) applied this
approach successfully to other ecotoxicity test data and risk assessment information. We also
began addressing issues germane to predicting joint effects of the PAH in the WAF mixture.
As described in the references cited above, survival time models can predict mortality during and
after exposures of different durations and concentrations. Models were developed for the grass
shrimp, Palaemonetes pugio, exposed to key PAH associated with weathered oil. By conducting
toxicity tests with these PAH, we were later able to incorporate compound qualities (e.g., log
Kow) into predictive models. The selected PAH were 1-ethylnaphthalene, 2,6dimethylnaphthalene, dibenzothiophene, fluorene, naphthalene, and phenanthrene. The grass
shrimp was chosen because it is a common test species, and is an ecologically important one in
salt marshes and other coastal environments. Because of the areas it inhabits, it would also be
subject to periodic exposures of a pulsed nature, e.g., compounds released from deposited
weathered oil during tidal cycles, or varying in exposure concentration and duration due to spill
characteristics or dispersant application.
Specific objectives were the following:
1. Produce predictive survival time models for grass shrimp exposed to a pulse of six
representative PAH in weathered oil WAF. Models included predictions of pulse and
post-pulse mortality under realistic ranges of exposure duration and concentration. This
objective was met by generating models for each compound. The ease of model generation
was greatly improved by the observation of minimal post-exposure mortality for shrimp
exposed to these narcotics.
2. Produce survival time models that incorporate molecular qualities (e.g., log Kow) of
the six PAH. Such a model will allow a level of prediction by interpolation to other
untested PAH in spilt oil. Several chemical properties allowed generation of QSAR
models for conventional LC50 and survival time model parameters. In this report, we use
the conventional lipophilicity metric, log Kow, to illustrate the QSAR models.
3. Produce predictive survival time models for grass shrimp exposed to pulses of a PAH
mixture. Models include predictions of pulse and any potential post-pulse mortality at a
range of realistic exposure concentrations and durations. Several experimental designs
were assessed and a final design employing a mixture of two substituted PAH
1
The term, PAH, is used for clarity throughout this report; however, one compound (dibenzothiophene) was a
heterocyclic aromatic hydrocarbon.
7
(ethylnaphthalene and dimethylnaphthalene) and the heterocyclic aromatic hydrocarbon,
dibenzothiophene, was executed. The difficulties of simultaneously maintaining the dosing
concentrations for these three compounds resulted in concentrations producing too low
mortality in the single compound treatments so the test of combining effects could not be
done with the planned conceptual approach.
3.0 Methods
Because specific methodological details are provided in the attached manuscripts (Appendices A
and B), only general descriptions will be provided here. Grass shrimp were exposed (Figure 2,
26 shrimp/tank) to a series of concentrations of each of six PAH. The range of concentrations
was determined during pilot studies with the intent of encompassing concentrations above the
threshold of apparent lethal effect and below the PAH solubility in saltwater. Shrimp were kept
in separate glass tubes with free exchange of exposure tank solutions. Triplicate tanks were
randomly placed on two adjacent water tables for each of five (including controls) exposure
concentration treatments. Tanks were not aerated to reduce PAH volatilization. Exposure
solutions were renewed every 12 hours.
Figure 2. Exposure tanks used in all shrimp survival time experiments.
At death or the end of the experiment, shrimp were removed from tanks, weighed, and frozen.
Body concentrations of the exposure PAH were determined in subsets of shrimp with the intent
of estimating body burdens upon death, and fortuitously, estimating elimination rate constants
during the post-exposure period.
Exposure data for each PAH were fit to a log logistic survival model,
8
ln Time − to − Death = b1 + b2 (ln Concentration) + b3 (W )
(1)
where b1 = the model intercept parameter estimate, b2 = model parameter estimate for the effect
of PAH concentration, and b3 = a scaling parameter estimate. The W is associated with a specific
proportion dying. In this case of a log logistic model, W is the log odds, i.e., ln(P/(1-P)) where P
is the proportion of exposed individuals dying. As an example, W for predicting median time-todeath (P=0.5) produces a W = ln(0.5/(1-0.5) = 0. The units of concentration are ug/L.
Under the intense and relatively brief exposure conditions, the mode of action for all six
compounds was judged to be general narcosis. Consequently, one could begin by assuming that
models for joint similar action could be applied. Such models assume a common slope for all
similar acting compounds so that the proportion of exposed individuals dying during exposure to
a specified mixture of these compounds can be estimated with the separate probit models for the
individual toxicants:
Pr obit ( PA ) = Intercept A + Slope(log Concentration A )
(2)
Pr obit ( PB ) = Intercept B + Slope(log Concentration B )
(3)
For the conventional concentration-effect design, the log of relative potency can be calculated.
Log ρ B =
( Intercept B − Intercept A
Slope
(4)
The combined effect of two similar acting toxicants (PA+B) would then be predicted as the
following,
PA+ B = Intercept A + Slope(log Concentration A + ρ B (log Concentration B ))
(5)
The assumption can be assessed by projecting this reasoning to survival time models that similar
acting PAH should have similar b2 values. Also the survival time models for each PAH in a
mixture could be used together to predict the total proportion dying at any time.
4.0 Results
The survival time tests for the six PAH produced 48 hr LC50 estimates in addition to
parameterized survival time models (Tables 1 and 2). The conventional concentration-effect
model (with natural mortality) used to estimate 48 h LC50 values was the following:
PDead = PB + (1 − PB )Φ (a1 + a 2 Log 10 Concentration)
(6)
where PDead = predicted proportion dying by 48 h, PB = estimated proportion dead at 48 h due to
background mortality, a1 and a2 = estimated model intercept and slope parameters, and M() = the
normal cumulative distribution function.
9
Table 1. Concentration-Effect Models (48 hr exposure duration). (95% CI = 95% confidence
interval, SE = standard error)
Compound
Naphthalene
Fluorene
Dibenzothiophene
Ethylnaphthalene
Dimethylnaphthalene
Phenanthrene
PB (SE)
0.01
(0.01)
0.03
(0.02)
0.06
(0.02)
0.06
(0.05)
0.06
(0.03)
0.02
(0.01)
a1 (SE)
-60.3
(8.2)
-33.3
(3.6)
-35.4
(5.3)
-31.4
(10.7)
-20.4
(3.6)
-14.5
(2.4)
95% CI
-76.5 - -44.2
-40.3 – -26.2
-45.8 – -25.0
-52.4 – -10.4
-27.5 – -13.8
-19.2 – -9.8
a2 (SE)
18.2
(2.5)
11.9
(1.3)
14.8
(2.2)
12.7
(4.2)
7.6
(1.3)
5.7
(1.0)
95% CI
13.3 – 23.1
9.4 – 14.4
10.6 – 19.1
4.4 – 21.0
4.9 – 10.2
3.8 – 7.5
The 48 hr LC50 estimates and associated 95% confidence intervals are provided in the left side
of Table 2. These 48 H LC50 estimates were consistent with the literature as described below in
discussions of QSAR models. The background mortality in the exposure experiments averaged
4%: roughly 1 of 25 shrimp that died in an exposure tank died due to nontoxicant-related reasons
during the 48 to 60 hour experiments.
Table 2. Parameter estimates for survival time models (95% CI = 95% confidence interval, SE =
standard error)
Compound
LC50 95% b1 (SE)
CI
(µg/L)
Naphthalene
2111 205752.2
2161
(2.1)
Fluorene
616
59322.1
637
(1.6)
Dibenzothiophene
242
22817.4
253
(0.6)
Ethylnaphthalene
295
16215.3
331
(1.0)
Dimethylnaphthalene
500
46314.9
535
(1.2)
Phenanthrene
360
33313.0
402
(1.1)
95%
CI
48.056.3
18.925.2
16.118.7
13.417.3
12.517.3
10.715.2
b2
(SE)
-6.3
(0.3)
-2.8
(0.2)
-2.4
(0.1)
-2.0
(0.2)
-1.8
(0.2)
-1.5
(0.2)
95%
CI
-6.8- 5.8
-3.3- 2.3
-2.6- 2.2
-2.4- 1.7
-2.2- 1.4
-1.9- 1.2
b3 (SE)
0.64
(0.03)
0.41
(0.02)
0.31
(0.02)
0.26
(0.01)
0.34
(0.03)
0.38
(0.03)
95%
CI
0.600.70
0.370.46
0.280.35
0.230.29
0.300.40
0.330.45
Initial survival modeling was designed to either model survival during exposure, or survival
during and after (48 h post-exposure) exposure if warranted. Because of the narcotic mechanism
for action at these PAH concentrations and the test durations, latent effects were minimal. This
permitted the most parsimonious modeling to occur: latent mortality was minimal for the
10
modeled toxicant exposure scenarios and only mortality during exposure required estimation.
Figure 3 demonstrates this using the phenanthrene survival data.
Proportion dead
Post-exposure
Exposure
0.9
153 µg/l
0.8
0.7
224 µg/l
0.6
0.5
0.4
302 µg/l
389 µg/l
0.3
0.2
0.1
0
0
20
40
60
80
100
Hours
Figure 3. Survival curves (bars indicate the estimated standard errors) from the phenanthrene
tests. With the termination of exposure (60 hr), mortality dropped quickly to a minimum level in
all concentration treatments. Greenwood’s formula was used to produce the standard errors
shown as vertical bars for each exposure time.
The models defined in Table 1 predict the amount of mortality expected with a specified pairing
of exposure concentration and duration as illustrated for phenanthrene in Figure 4. The first
objective of the project was completed successfully with the generation of these models.
Time to death (hours)
60
90%
50
40
70%
30
50%
20
30%
10%
10
5%
TTD PHEN = e12.9659 e- 1.5409 ln C e0.3828W
0
300
350
400
450
500
550
600
Water concentration (µg/L)
Figure 4. An illustration of the ability of survival time models to predict the mortality with
different combinations of exposure time and concentration. In this figure, contours for
percentage of exposed shrimp expected to die are shown. The 48 hr LC50 estimate is also shown
for comparison. Less than 5% of the mortality during the entire test was post-exposure mortality.
11
The second goal of assessing whether QSAR models could be generated for these data was also
successfully achieved (see Figure 5).
100,000
LC 50 (ug/L)
10,000
1,000
Unger et al, in prep
100
y = 4E+06e-2.1064x
R2 = 0.9379
Unger et al, 2007
10
Tatem, 1975
1
2
2.5
3
3.5
4
4.5
5
3.5
4
4.5
5
4
4.5
5
Survival Model Estimate b1
60
50
40
y = -33.211x + 162.63
30
Unger et al, in prep
20
Unger et al, 2007
10
0
Survival Model Estimate b2
2
2.5
3
7
6
5
y = -3.9292x + 19.381
4
3
Unger et al, in prep
2
Unger et al, 2007
1
0
2
2.5
3
3.5
Log KOW
Figure 5. QSAR models for estimates of the 48 hr LC50 (top panel) and survival time model
parameter estimates (b1 and b2)(middle and bottom panel).
The LC50 data generated from this study were consistent with those of Tatem (1975) (Figure 5,
top panel). Toxicity increased for these oil-related PAH with increasing lipohilicity (as
quantified with the log Kow). Similarly, the decrease in survival model parameter estimates (b1
and b2) with increasing lipophilicity indicates that time-to-death decreases with increasing
12
lipophilicity. The ability to accurately predict survival time model parameters based on data
from these six compounds could be greatly enhanced by adding estimates for more key PAH
(3.5<log Kow<4) associated with oil spills. Regardless, these QSAR models provide proof of
concept: the second goal of this project was met.
0
Est i mat ed Sl ope
0. 5
1
1. 5
2
The third goal to predict combined effects of a subset of these compounds in mixture was not
met yet. In our opinion, the difficulty with maintaining the correct combination of three toxicant
concentrations in this particular system made quantification of such mixture effects problematic
by the intended, similar mode-of-action computations (Equations 2 and 3). Further, simple
models based on the assumption of similar action (see equations above) may not be sufficient for
this task. As discussed above, the approach to mixtures for similarly acting toxicants assumes
identical slopes for each toxicant concentration-effect model. But, as evident in Table 1, the
slopes (a2) vary approximately 3 fold (5.7 to 18.2) for the six compounds. Together, the range of
a2 estimates and overlap of confidence intervals make the assumption of identical slopes difficult
to evaluate at this point. For this reason, we used the simple bacterial bioluminescence assay
(Microtox7) to generate more precise concentration-effect model slopes for these six compounds
(Figure 6).
3. 37
4. 18
4. 31
4. 4
Log Kow
4. 49
4. 57
Figure 6. Concentration-effect (log normal) model slopes for the six compounds estimated with
the Microtox7 toxicity assay. Brackets indicate 95% confidence intervals.
Although narrower than those from the shrimp tests, the slopes from the bacterial
bioluminescence assay had sufficiently wide 95% confidence intervals to make any definitive
inferences difficult about slope equivalency. However, as hoped, the range of slopes (0.8 for
naphthalene to 1.14 for ethylnaphthalene) is only approximately 1.4 fold for the Microtox7
assay. The similarity of mode-of-action at the relevant concentrations and the bacterial assay
results suggest that the slopes might be assumed to be equivalent during a first order prediction
of mixture effects. However, the intended mixture models were inadequate for fitting the data
from the shrimp single toxicant and also mixture studies. Further, recent studies have proposed
conflicting mixture models for PAH aquatic toxicity effects, including joint independent action
(Olmstead and LeBlanc 2005) and simple concentration additivity (Barata et al. 2005). It is the
13
intent of the authors to continue exploring alternative models to better analyze these data, e.g., a
survival model using all data and log Kow values for the six compounds to predict survival during
exposure to mixtures. Another option is to combine the relative potency approach embedded in
Equations 4 and 5 for the concentration-effect approach into survival time models. Relative
potencies would be included and the resulting predictions compared to those generated during
the three-compound mixture experiment. A final option is to explore the models described by
Olmstead and LeBlanc (2005) and Barata et al. (2005).
5.0 Discussion and Importance to Oil Spill Response/Restoration
The survival time modeling approach was successful in that models were produced that can be
applied immediately to predict lethal consequences of PAH exposures differing in concentration
and duration. Such a predictive capacity extends well beyond that afforded by the conventional
LC50 approach. It is also needed to make predictions about realistic exposures that will vary in
concentration and duration.
Clearly, post-exposure mortality can be ignored during general predictions. This statement is
based on the minimal post-exposure mortality seen in the six compound toxicity exposure tests.
This greatly simplifies modeling and consequent predicting of lethal consequences. The minimal
post-exposure mortality experienced by shrimp exposed to these general narcotics is similar to
that reported for mosquitofish exposed to sodium pentachlorophenol, a toxicant with a reversible
mode of action (uncoupling of oxidative phosphorylation). The minimal post-exposure contrasts
with that associated with sodium chloride exposed mosquitofish (Newman and McCloskey 2000)
or the copper-exposed amphipods shown in Figure 1 that experienced extensive damage during
exposure. The extensive damage required significant time and energetic resources to repair. Like
sodium pentachlorophenol, the six compounds used in these tests also were eliminated very
quickly when shrimp were moved to clean water (see Appendix A for elimination rate
constants). Likely, this also contributed to the rapid recovery and minimal post-exposure
mortality. In the absence of direct evidence for each relevant PAH, it is a reasonable assumption
at this point that minimal post-exposure mortality will occur for untested PAH components under
exposure scenarios in which narcosis is the dominant mode of lethal action.
Relative to a sequence of pulsed exposures, one additional factor emerges as relevant, the
individual tolerance or individual effective dose theory (Zhao and Newman 2007). Briefly, this
theory holds that an individual has a unique or characteristic dose (or concentration) below
which it will survive but at or above which it will die: the distribution of effective
doses/concentrations are assumed to be log normally distributed in populations of exposed
individuals. This theory is pervasive in the ecotoxicology literature but it remains poorly tested.
The alternative, stochasticity theory (Newman and McCloskey 2000) assumes that all individuals
are equally sensitive and which die during an exposure is a matter of chance. The stochastic
process can be modeled with a skewed (e.g., log normal) distribution. With repeated exposures
as might happen with a tide-dominated movement of PAH into and out of a marsh, which theory
is correct becomes important to predictions of population consequences (Zhao and Newman
2007). Based on the research of Zhao and Newman (2007), and Newman and McCloskey (2000)
the conservative assumption of the stochastic theory seems most advisable at this time. The
14
methods of Zhao and Newman (2007) can also be applied to assessing whether shrimp sensitivity
to the lethal effects of PAH exposure changes significantly with repeated exposures.
6.0 Technology Transfer
The results of this study were presented annually at NOAA CRRC meetings and, upon request,
at NOAA-related workshops. For example, three presentations were made in 2006 at NOAA’s
request. One (August 15-16) was given at the CRRC PAH Toxicity Summit in Seattle, WA.
Another was given in September at the CRRC Coastal Modeling for Decision Makers meeting in
Durham, NH. A final presentation was made at another Durham meeting, CRRC Submerged Oil
Workshop (December). These opportunities allowed for researchers and oil spill responders to
interact and discuss the utility and applications of the CRRC research projects. Much discussion
at the Toxicity and Modeling workshops was centered on time dependence as a factor
influencing PAH toxicity and how dispersants and physical factors may affect the exposure
duration and concentration of toxicants in the water-soluble fraction from oil. The time to death
approach of evaluating PAH toxicity was of great interest to managers as a mechanism to better
understand and predict population mortality under changing conditions to help facilitate oil spill
response. Correctly so, these models were also seen as a viable approach to couple chemical fate
and toxicity models, a long-term goal of NOAA oil spill responders.
7.0 Achievement and Dissemination
Beyond the contributions to NOAA sponsored workshops, the results of this study have been
disseminated in a variety of formats including national and international meetings, and peerreviewed publication. Presentation at national and international (The Hague, The Netherlands)
meetings provided wide exposure of the environmental science community to the techniques and
results. Written transfer of the study techniques and findings was achieved primarily in the peerreviewed literature. The first part of the study was published in 2007 (Unger et al. 2007) as cited
in the reference section of this report. The second part of the research project is appended to this
final report as an “in preparation” manuscript that will soon be submitted to the international
journal, Environmental Toxicology and Chemistry. The third part of the study involving mixture
effects requires further analysis and possibly more experimentation. Although it is difficult to
anticipate the outcome, the results will minimally be presented at Society of Environmental
Toxicology and Chemistry annual meetings in combination with the rest of the CRRC-funded
study.
15
References
Allison, P.D. 1995. Survival Analysis Using the SAS System. A Practical Guide. SAS Institute,
Cary, NC.
Barata, C., A. Calbet, E. Saiz, L. Ortiz, and J.M. Bayona. 2005. Predicting single and mixture
toxicity of petrogenic ploycyclic aromatic hydrocarbons to the copepod Oithona davisae.
Environ. Toxicol. Chem. 24: 2992-2999.
Cox, D.R. and D. Oakes. 1984. Analysis of Survival Data. Chapman & Hall, London.
Crane, M., M.C. Newman, P. Chapman and J. Fenlon. 2002. Risk Assessment with Time-to-Event
Models. CRC Press LLC, Boca Raton, FL.
Kalbfleisch, J.D. and R.L. Prentice. 1980. The Statistical Analysis of Failure Time Data. John
Wiley and Sons, New York.
Lee, J-H, P.F. Landrum and C-H. Koh. 2002. Prediction of time-dependent PAH toxicity in
Hyalella azteca using a damage assessment model. Environ. Sci. Technol. 36: 3131-3138.
Muenchow, G. 1986. Ecological use of failure time analysis. Ecology 67: 246-250.
Miller, Jr., R.G. 1981. Survival Analysis. John Wiley & Sons, New York.
Nelson, W. 1972. Theory and applications of hazard plotting for censored failure data.
Technometrics 14: 945-966.
Newman, M.C. and P.H. Dixon. 1996. Ecologically meaningful estimates of lethal effect on
individuals. In: Newman, M.C. and C.H. Jagoe (Eds.), Ecotoxicology: A Hierarchical
Treatment. CRC/Lewis Publishers, Inc., Boca Raton, FL.
Newman, M.C. and J.T. McCloskey. 1996. Time-to-event analysis of ecotoxicity data.
Ecotoxicology 5: 187-196.
Newman, M.C. and J.T. McCloskey. 2000. The individual tolerance concept is not the sole
explanation for the probit dose-effect model. Environ. Toxicol. Chem. 19:520-526.
Newman, M.C. and J.T. McCloskey. 2002. Applying time-to-event methods to assess pollutant
effects to populations. In: Crane, M., M.C. Newman, P. Chapman, and J. Fenlon (Eds.) Risk
Assessment with Time-to-event Models. CRC Press LLC
Olmstead, A.W. and G.A. LeBlance. 2005. Joint action of polycyclic aromatic hydrocarbons:
predictive modeling of sublethal toxicity. Aquat. Toxicol. 75: 253-262.
Tatem, HE. 1975. Ph.D. Thesis, Texas A&M University, College Station, TX : 133 pp.
Unger M.A., Newman M.C., Vadas G.G. 2007. Predicting survival of grass shrimp
(Palaemonetes pugio) during ethylnaphthalene, dimethylnaphthalene, and phenenathrene
exposures differing in concentration and duration. Environ. Toxicol. Chem. 26:528–534.
Zhao, Y. and M.C. Newman. 2004. Shortcomings of the laboratory derived LC50 for predicting
mortality in field populations: exposure duration and latent mortality. Environ. Toxicol.
Chem. 23: 2147-2153.
Zhao, Y. and M.C. Newman. 2006. Effects of exposure duration and recovery time during pulsed
exposures. Environ. Toxicol. Chem. 25:1298-1304.
Zhao, Y. and M.C. Newman. 2007. The theory underlying dose-response models influences
predictions for intermittent exposures. Environ. Toxicol. Chem. 26:543-547.
16
Appendices
Appendix A. Unger, M.A., M.C., Newman, and G.G. Vadas. In prep. Predicting survival of grass
shrimp (P. pugio) for naphthalene, fluorine and dibenzothiophene exposures differing in
concentration and duration. Environ. Toxicol. Chem.
Appendix B. Unger, M.A., M.C. Newman, and G.G. Vadas. 2007. Predicting survival of grass
shrimp (P. pugio) during ethylnaphthalene, dimethylnaphthalene, and phenanthrene exposures
differing in concentration and duration. Environ. Toxicol. Chem. 26: 528-534.
17
PAH TOXICITY TO GRASS SHRIMP
Michael A. Unger
Department of Environmental and Aquatic Animal Health
Virginia Institute of Marine Science
College of William and Mary
P.O. Box 1346, Gloucester Point, VA 23062-1346, USA
804-684-7187
804-684-7793 FAX
munger@vims.edu
Total words: 4,535
18
PREDICTING SURVIVAL OF GRASS SHRIMP (P. pugio) EXPOSED TO AROMATIC
COMPOUNDS DERIVED FROM OIL
Michael A. Unger*, Michael C. Newman, and George G. Vadas
*Virginia Institute of Marine Science, College of William and Mary, P.O. Box 1346, Gloucester
Point, VA 23062-1346, USA
19
* To whom correspondence may be addressed (munger@vims.edu)
This paper is contribution XXXX of the Virginia Institute of Marine Science, College of
William and Mary.
20
Abstract- The composition and persistence of dissolved polycyclic aromatic hydrocarbons
(PAH) released to the water column during oil spills are altered by weathering, tidal transport
and the addition of dispersants. Conventional toxicity effect metrics such as the LC50 are
inaccurate predictors of mortality from all toxicant exposure duration/concentration
combinations likely to occur during spills. In contrast, survival models can predict the
proportions of animals dying as a consequence of exposures differing in durations and
intensities. Extending previous work with ethylnaphthalene, dimethylnaphthalene and
phenanthrene, survival time models were developed that include exposure duration and
concentration to predict time-to-death for grass shrimp, Palaemonetes pugio, exposed to two
additional PAH (naphthalene, fluorene) and a heterocyclic aromatic hydrocarbon
(dibenzothiophene). Preliminary explorations of these models confirmed that QSAR models
were possible for predicting survival model parameters from compound characteristics.
Conventional 48h LC50 values were also calculated for the compounds and were combined with
published LC50 values to predict relative PAH toxicity to P. pugio based on octanol:water
partitioning.
Keywords- Survival analysis, Oil spill, Polycyclic aromatic hydrocarbons, Toxicity, Grass
shrimp
21
INTRODUCTION
Tides, currents, weathering, and dispersant application alter both the exposure
concentration and duration an organism is exposed to water-soluble compounds from an oil spill
[1-3]. Conventional toxicity tests use a concentration-effect design to produce a metric such as
an LC50 at a set exposure time, e.g., 96 h, that is not well suited to predict toxic effects for
varied exposure scenarios. Although underutilized by environmental toxicologists, survival
models can include both exposure duration and concentration. Survival models have been
applied to predict the effects of a variety of contaminants [4-8] and potentially provide managers
a tool with which to make more effective response and remediation decisions about the potential
impacts from spills. In addition to predicting mortality under varying exposure time and toxicant
concentrations, the models are easily expanded to include post-exposure (latent) mortality effects
or other cofactors, such as organism size or sex, temperature or salinity [9].
The main objectives of this study were to add three more parametrized models to the
three built earlier [8] and then explore the possibility of building QSAR models for survival
model parameters. QSAR interpolation would permit prediction for exposures to untested oil
spill-related aromatic compounds. Exposed shrimp were also analyzed post mortem for
contaminant body burdens to determine if accumulated contaminants were consistent with a
critical body burden at time of death.
Acute toxicity experiments were done with the grass shrimp, Palaemonetes pugio,
exposed to two PAH (naphthalene, fluorene) and the heterocyclic aromatic compound
dibenzothiophene. Combined with the previous models for dimethylnaphthalene,
ethylnaphthalene and phenanthrene [8], the resulting six models represent the range of aromatic
compounds thought to be important contributors to the toxic effects of oil spills [2]. Results from
22
these exposure experiments were modeled with survival time methods [10] to produce models
for each compound that predicted the proportion of individuals dying during, and potentially,
after an exposure of specified duration and concentration. Survival models for six compounds,
representative of those found in the water soluble fraction derived from oil, were used to
examine the relationship between the toxicant octanol:water partitioning behavior and its
toxicity.
MATERIALS AND METHODS
Grass shrimp collection and maintenance
Grass shrimp collected locally from the York River (Virginia, USA) were maintained in
the laboratory in filtered York River water (salinity 19-20 0/00) for at least 2 weeks prior to being
used in the exposures. Shrimp were fed Tetramin® Tropical flake food (Tetra Holding,
Blacksburg, VA) daily. Individual shrimp with no outward signs of damage or disease were
gently placed into glass tubes and the exposure aquaria one day before exposures began.
PAH survival analysis experiments
The survival analysis experiments were conducted in August (naphthalene 1 and
dibenzothiophene), September (fluorene), and October (naphthalene 2) of 2006. The first
naphthalene exposure resulted in more mortality than anticipated by 24 hours so an additional
exposure experiment was conducted in October at lower concentrations. After graphical
23
comparison for consistency, the data from both tests were combined to develop a single
naphthalene survival model and to estimate the 48 h LC50. For each experiment, three replicates
of four PAH concentrations and a control were prepared from saturated solutions generated with
filtered York River water by techniques described in detail previously [8]. Briefly, a 7.5 cm x
59.2 cm generator column was used to produce the large volume of saturated PAH solutions
required to avoid the use of solvent carriers for the experiments.
To minimize toxicant volatilization, exposure chambers were not aerated during the experiment
and solutions were renewed every 12 h. All experiments were conducted under constant light
from standard fluorescent light fixtures that were approximately 1.5 m above aquaria.
Experimental chambers were constructed from 25 cm x 50 cm x 58 cm glass aquaria with glass
lids and were randomly assigned to positions on a wet table. A water-tight glass partition was
installed down the center of each aquarium to create tandem, 30 L exposure chambers 25 cm x
25 cm x 58 cm tall. This design reduced the surface area to volume ratio to minimize volatile
losses of the PAH. One chamber of each aquarium was used to expose the test organisms and the
other was used to prepare the new test solution every 12 h. The tandem 30 L test chambers
permitted the filling of one side with new exposure water and transferring of the shrimp in racks
from old to renewed water with minimal disturbance to the test organisms.
Individual grass shrimp (26 shrimp/replicate) were placed in 2.8 cm x 10.8 cm (40 ml)
glass vials with an open ended screw cap fitted with a stainless steel mesh screen. Shrimp were
not sexed but berried females were excluded from the tests. The vials were suspended in
randomly assigned exposure chambers on aluminum racks to facilitate monitoring the test
organism’s condition and to allow easy transfer of the test organisms to newly prepared solutions
every 12 h. Shrimp were monitored for mortality every 4 h and scored as dead if unresponsive to
24
repeated prodding and no appendage movement was apparent. All dead shrimp were removed,
weighed, and frozen. Shrimp surviving the exposure period (naphthalene 1, 24 h; naphthalene 2,
36 h; fluorene 60 h; and dibenzothiophene, 48 h) were transferred to filtered and aerated York
River water. The water was renewed every 12 h and shrimp were routinely monitored for postexposure mortality for 48 h, or until no further mortality was apparent. All shrimp were weighed
and frozen at the completion of the test (48 h post-exposure).
Water chemistry
Old and fresh test solution water chemistries were measured every 12h and included
temperature, dissolved oxygen, salinity, and pH. A HydrolabTM Surveyor 4a (Hydrolab Corp.,
Austin, TX) was used for these measurements. Unfiltered water samples were also collected and
frozen for ammonia analyses (phenol method).
PAH Analysis of water and tissue samples
Exposure waters were analyzed for PAH concentrations using a HPLC with a
fluorescence detector by methods described in detail previously [8]. Briefly, calibration
standards were made in York River water from stock solutions of internal standard (1-methyl
naphthalene) and the PAH analyte of interest. Calibration standards were injected on the HPLC
and the method calibrated for the specific analyte over a range from its quantitation limit to near
its aqueous solubility prior to running samples. The HPLC calibration was verified again with
fresh standards once the analyses for a particular exposure were completed. Shrimp tissues were
25
analyzed by GC-MS using selective ion monitoring (GCMS-SIM). Individual shrimp were
weighed, rinsed with DI water and placed into a 50 mL Teflon centrifuge tube containing 2.0 mL
of concentrated HCL and 500 ng of deuterated PAH surrogate standards. The shrimp were
homogenized with a spatula and ultrasonicated for 10 m. The aqueous homogenate was
sequentially extracted with two aliquots of hexane (2.0 mL each), centrifuging between
extractions to separate the layers. The combined hexane extracts were reduced to 0.1 mL under
dry nitrogen and 0.6 ug of p-terphenyl internal standard was added before analysis on a Varian
CP-3800 Gas Chromatograph with a Saturn 2000 GC/MS/MS ion trap mass spectrometer
operated in electron ionization mode (EI). Analytes and ions monitored were: p-terphenyl I.S.
[152+230], naphthalene-d8 [108+135+136], acenaphthene-d10 [160-165], phenanthrene-d10
[187-189], naphthalene [128+127+102], fluorene [163-166] and dibenzothiophene [184+139].
Six point calibration curves were generated for each analyte and identifications were based on
retention time and matches to library spectra. Shrimp tissue concentrations were corrected for
surrogate recovery relative to the internal standard (p-terphenyl). Surrogate standards were
naphthalene (naphthalene-d8), fluorene (acenaphthene-d10), and dibenzothiophene
(phenanthrene-d10).
Calculating LC50 values
The measured PAH exposure concentrations, number of dead shrimp at 48 h and total
exposed shrimp were fitted by maximum likelihood estimation (MLE) to a log normal model
with the PROBIT procedure in the SAS software (SAS Corp., Cary, NC). Spontaneous mortality
was included in the log normal model because low levels of mortality (≤ 6%) occurred in the
26
control aquarium shrimp during exposures. The 48 h LC50 values and associated 95% fiducial
limits were estimated.
Survival models
Survival time was modeled as a function of PAH concentration using mean
concentrations in each exposure tank and the SAS LIFEREG procedure (SAS Corp., Cary, NC).
The general approach was that described in detail in previous publications, e.g., [8, 10]. Survival
times noted at 4 h intervals were used directly in the model instead of incorporating interval
censoring because the fineness of the sampling used in these experiments minimized any
inaccuracies arising from the discreteness of the sampling [10]. The log logistic model was
chosen to predict survival based on exposure concentration (see [8] for details of model
selection),
TTD = eb1 eb2 (ln Concentration ) eb3W
where TTD = the predicted time-to-death for a specified proportion of the exposed shrimp, b1 =
intercept (MLE estimate) , b2 = the coefficient for the influence of ln of PAH concentration on
time-to-death (MLE estimate), b3 = the scale parameter (MLE estimate), and W = the response
metameter for the model distribution associated with the proportion dying (P) of the exposed
shrimp for which prediction is being made. The W can be generated by special functions within
most statistical or spreadsheet software, or taken from tables such as Appendix Table 7 in [9].
By changing W, the various combinations of exposure concentration and duration can easily be
found that result in P of the exposed shrimp dying. However, prediction is only recommended
27
within the range of concentrations and durations used in the tests from which these data were
generated.
RESULTS
Water chemistry
Temperature, salinity, dissolved oxygen, pH and ammonia were monitored during the
three exposure experiments for both freshly prepared solutions and 12 h-exposed test solutions.
Measured parameters remained within a narrow range for all exposure experiments. During the
two naphthalene exposure experiments temperature in the various treatments averaged 19.6-21.8
°C, salinity 18.8-21.2 ppt, dissolved oxygen 6.8-7.2 mg/L, pH 7.76-7.95, and ammonia 0.14-0.17
mg/L. During the fluorene exposure experiment temperature averaged 19.4-20.6 °C, salinity
18.4-19.8 ppt, dissolved oxygen 6.7-8.7 mg/L, pH 7.8-8.9, and ammonia 0.14-0.90 mg/L. During
the dibenzothiophene exposure experiment, temperature averaged 20.7-21.2 °C, salinity 21.321.7 ppt, dissolved oxygen 6.6-7.1 mg/L, pH 7.6-7.7, and ammonia 0.14-0.17.
Toxicant concentrations
The measured toxicant concentrations are summarized in Table 1. Saturated solutions
prepared by the generator column technique had the following mean concentrations: naphthalene
20,100 ug/L, fluorene 884 ug/L and dibenzothiophene 504 ug/L. These values are approximately
50-60% of the published values for the same compounds in fresh water at 25 oC [11, 12]. The
28
decreased solubility at lower temperature (20 oC) and high salinity (20 ppt) is expected based on
results with similar PAH and conditions [8, 13]. The PAH concentrations varied because they
were measured in both the newly prepared and the 12 h old solutions. Although tanks were not
aerated in order to minimize surface exchange losses, there was still some decrease of the lightest
compound, naphthalene. Biodegradation during the 12 h period might have contributed because
the rate of loss increased as the experiment progressed. Control PAH concentrations were below
detection limits (1 ug/L) during all exposures.
Survival analysis
Shrimp rapidly revived when placed in clean water and there was very little apparent
latent mortality. Conventional LC50 values and 95% fiducial limits were calculated,
naphthalene 2111 µg/L (2057-2161), fluorene 616 µg/L (593-637) and dibenzothiophene 242
µg/L (228-253). Survival data were fitted to accelerated failure time models with the log logistic
distribution [8]. Contours were developed from the models to depict shrimp mortality for various
combinations of exposure times and concentrations (Figure 2A-2C). Compound-specific models
and conventional LC50 values with 95% fiducial limits are also included in Figure 2. Estimated
values for model parameters (b1,b2,b3) for each compound tested are presented in Table 2 along
with those from previous work with three other PAH [8].
Tissue concentrations
Toxicant concentrations (wet weight) were measured in select whole shrimp that died
during the exposure experiments. Body burden concentrations at time of death spanned a
29
relatively wide range during each experiment. Mean concentrations are presented in Table 3. The
large range in concentrations was, in part, likely due to increasing body burdens with increasing
dose. Tissue concentrations measured in individual shrimp from the 300 µg/L and 500 µg/L
dibenzothiophene treatments are presented in Figure 2 for comparison. There is a trend of
increasing body burdens with increasing time of exposure and with increasing exposure
concentration.
Naphthalene and fluorene were eliminated rapidly from the shrimp after the exposures
ended. The dibenzothiophene was two-fold slower than the PAH to decrease in concentration. In
our previous study [8], phenanthrene body burdens were measured in exposed shrimp that died
during the depuration phase and they showed an exponential decrease in concentration with time.
Based on this earlier work with similar compounds, we assumed a first order decrease and
calculated elimination rate constants for each compound from the slope of the log transformed
concentration data. The elimination rate constants are shown in Table 3 with PAH data from the
previous study [8] included for comparison. Corresponding half-lives for the six compounds
ranged from 3.7 -17 h.
DISCUSSION
Conventional LC50 values calculated for P. pugio were naphthalene 2111 µg/L (20572161), fluorene 616 µg/L (593-637) and dibenzothiophene 242 µg/L (228-253). Tatem [14]
reported a 48 h LC50 concentration of 2600 µg/L (2340-2890) for P. pugio exposed to
naphthalene that is in general agreement with our results for naphthalene (2111 µg/L). There was
a trend of increasing toxicity with increasing molecular weight or lipophility. While individual
30
PAH toxicity data for P. pugio are not numerous in the literature, enough data were available
from two additional studies [8, 15] to examine the relationship between octanol-water
partitioning (Kow) and LC50 for the grass shrimp (Figure 3a). Data include LC50 values for
diverse compounds ranging form the single ring aromatic compound, benzene, to the three ring
compound, phenanthrene, and includes alkylated PAH as well as the heterocyclic compound,
dibenzothiophene. The results indicate good relationship between log Kow and LC50 over a range
of log Kow that includes the most abundant aromatic components found in the water-soluble
fraction of oil. Similarly, we looked at the relationship between log Kow and the estimates of the
survival model parameters (b1,b2) (Figure Xb, Xc). While survival data is currently too limited to
develop a robust relationship, the available data suggests that a similar approach could be used to
develop models for untested PAH compounds. Due to the increased predictive capability of
survival models over standardized LC50 values, additional work is warranted to develop the
QSAR approach further.
The PAH concentrations measured in whole shrimp at time of death showed a great deal
of variability and appears dependent on exposure concentration and duration. This is consistent
with the results from our earlier study examining the acute toxicity of dimethylnaphthalene,
ethylnaphthalene and phenanthrene to grass shrimp [8]. The high toxicant concentrations and
short exposure durations used in these acute experiments likely contribute to the variability in
body burdens at time of death. Lower doses with long-term exposures result in steady state body
burdens and less variance in tissue concentrations at time of death. Landrum et al [16] have
shown that critical body residues of various PAH are similar on a molar basis (7.5 ± 2.6 µmol g1
) when amphipods (Diporia sp.) were exposed for 28 days. When the mean lethal body burden
concentrations for P. pugio are compared on a molar basis there is a range of critical body
31
residues from 0.41 -1.2 µmol g-1 (Figure 4). There is much better agreement for the four
unsubstituted compounds tested (1.0 ± 0.19 µmol g-1) and this may be the result of the higher
toxicity of akylated PAH resulting in lower body burdens at time of death
Unlike the single LC50 value, the survival time models developed here can predict what
proportion of an exposed population would be killed for different exposure scenarios generated
from oil spill field measurements or computer simulations. The toxicants tested were eliminated
rapidly from the organisms post-exposure and little post-exposure mortality was observed,
allowing simplified survival models that just included mortality occurring during the exposure
period. Additional work is needed to better understand how the sub-lethal narcotic effects of
PAH exposure can affect survival of prey organisms due to increased predation under natural
conditions. The acute toxicity (LC50) of PAH to P. pugio is proportional to the log Kow of the
compound and provides the ability to predict, by interpolation, the toxicity of untested
compounds which may be present in the water soluble fraction derived from oil. Preliminary data
suggests that survival models can also be constructed for untested compounds based on their
partitioning behavior.
ACKNOWLEDGMENT
We thank John Carriger, Kyle Tom, Chris Prosser, J. Greene, and E. Harvey, for laboratory
assistance. Funding for this research provided by the Coastal Response Research Center
(www.crrc.unh.edu).
32
REFERENCES
1. Wang Z, Fingas MV. 2003. Development of hydrocarbon fingerprinting and identification
techniques. Mar Pollut Bull 47:423-452.
2. Barron MG, Podrabsky T, Ogle S, Ricker RW. 1999. Are Aromatic hydrocarbons the
primary determinant of petroleum toxicity to aquatic organisms? Aquat Toxicol 46:253-268.
3. Neff JM, Stout SA, Gunster, DG. 2005. Ecological risk assessment of polycyclic aromatic
hydrocarbons in sediments: identifying sources and ecological hazard. Integrated
Environmental Assessment and Management1:22-33.
4. Newman, M.C. and P.H. Dixon. 1996. Ecologically meaningful estimates of lethal effect on
individuals. In: Newman, M.C. and C.H. Jagoe (Eds.), Ecotoxicology: A Hierarchical
Treatment. CRC/Lewis Publishers, Inc., Boca Raton, FL, pp. 225-253.
5. Lingtian Xie and Klerks PL. 2003. Responses to selection for cadmium resistance in the least
killifish, Heterandria Formosa. Environ Toxicol Chem 22: 313-320.
6. Sildanchandra W and Crane M 2000. Influence of sexual dimorphism in Chironomus riparius
MEIGEN on toxic effects of cadmium. Environ Toxicol Chem 19:2309–2313.
33
7. Widianarko B and Van Straalen N. 1996. Toxicokinetics-based survival analysis in bioassays
using nonpersistent chemicals. Environ Toxicol Chem 15:402–406.
8. Unger MA, Newman MC, Vadas GG. 2007. Predicting survival of grass shrimp
(Palaemonetes pugio) during ethylnaphthalene, dimethylnaphthalene, and phenenathrene
exposures differing in concentration and duration. Environ Toxicol Chem 26:528–534.
9. Newman MC. 1995. Quantitative Methods in Aquatic Ecotoxicology. CRC/Lewis
Publications, Boca Raton, FL, USA. 137-153.
10. Dixon, PM and MC Newman. 1991. Analyzing toxicity data using statistical models for
time-to-death: an introduction. In: Newman MC and McIntosh AW (Eds.). Metal
Ecotoxicology. Concepts & Applications. CRC/Lewis Publishers, Boca Raton, FL, pp. 207242.
11. Lide DR. 2005, CRC Handbook of Chemistry and Physics, Internet version.
http://www.hbcpnetbase.com, CRC Press, Boca Raton, FL, USA.
12. Pearlman, RS, Yalkowsky, SH, Banerjee S. 1984. Water solubilities of polynuclear aromatic
and heteroaromatic compounds. J. Phys. Chem. Ref. Data, Vol. 13, No. 2. 555-562.
34
13 . Gordon JE, Thorne RL. 1967. Salt effects on non-electrolyte activity coefficients in mixed
aqueous electrolyte solutions-II. Artificial and natural sea waters. Geochem Cosmochem Acta
31:2433-2443.
14. Tatem, HE, Anderson JW. 1973. The toxicity of four oils to Paleomonetes pugio (Holthuis)
in relation to uptake and retention of specific petroleum hydrocarbons. American Zoologist
13, 261 (abstracts, 1307-1308).
15. Tatem, HE. 1975. Ph.D. Thesis, Texas A&M University, College Station, TX : 133 pp.
16. Landrum, PF, Lotufo GR, Gossiaux DC, Gedeon ML, Lee JH. 2003. Bioaccumulation and
critical body residue of PAHs in the amphipod, Diporeia sp. : additional evidence to support
toxicity additivity for PAH mixtures. Chemosphere. 51:481-489.
35
Figure legends
Figure 1. Response surfaces predicting mortality levels for P. pugio exposed to two polycyclic
aromatic hydrocarbons (PAH), A) naphthalene, B)fluorene, and the heterocyclic C)
dibenzothiophene. The survival models from which predictions are made are also given in the
figure. Lines indicate different proportions dying predicted with the models for different
combinations of exposure concentration and duration. The 48-h median lethal concentrations
(LC50) and the 95% fiducial limits are shown for comparison. TTD= time to death,
NAP=naphthalene, FLUOR=fluorene, DBT=dibenzothiophene.
Figure 2. Individual whole body concentrations (µg/g wet weight) of dibenzothiophene measured
in P. pugio at time of death. Shrimp from two exposure concentrations (300 µg/L and 550 µg/L)
are shown for comparison.
Figure 3. Acute toxicity of PAH to P. pugio predicted from compound log Kow, A) LC50, B)
survival model parameter b1= the maximum likelihood estimation (MLE)-estimated intercept, C)
survival model parameter b2 = the estimated coefficient for the influence of ln of PAH
concentration on time-to-death.
Figure 4. Mean molar body burden concentrations measured in P. pugio at time of death.
Average unsubstituted compound concentrations (1.0 ± 0.19 µmol g-1) compared to akylsubstituted PAH. NAP=naphthalene, FLUOR=fluorene, DBT=dibenzothiophene,
PHEN=phenenthrene, DMN=dimethylnaphthalene, ENAP=ethylnaphthalene.
36
Table 1. Measured toxicant concentrations in exposure aquariaa
Toxicant concentration (µg/L)
Treatment
Replicate
Naphthalene 24h
Naphthalene 48h
Fluorene
Dibenzothiophene
Conc. 1A
2430 ± 260, n=7
1430 ± 650, n=10
520 ± 110, n=8
180 ± 20, n=8
Conc. 1B
2330 ± 250, n=7
1400 ± 710, n=10
520 ± 140, n=10
190 ± 20, n=8
Conc. 1C
2430 ± 250, n=7
1330 ± 680, n=10
520 ± 130, n=8
190 ± 20, n=8
Conc. 2A
2710 ± 210, n=7
1600 ± 670, n=10
590 ± 110, n=9
250 ± 30, n=8
Conc. 2B
2540 ± 290, n=5
1590 ± 740, n=10
570 ± 120, n=7
240 ± 40, n=8
Conc. 2C
2620 ± 210, n=4
1550 ± 730, n=10
620 ± 150, n=8
280 ± 40, n=8
Conc. 3A
3050 ± 220, n=5
1630 ± 720, n=10
720 ± 110, n=8
330 ± 60, n=8
Conc. 3B
3030 ± 130, n=6
1700 ± 650, n=10
730 ± 90, n=9
330 ± 40, n=8
Conc. 3C
2950 ± 180, n=5
1700 ± 710, n=10
690 ± 80, n=8
330 ± 50, n=8
Conc. 4A
3290 ± 280, n=3
1870 ± 590, n=10
780 ± 120, n=9
430 ± 70, n=8
Conc. 4B
3390 ± 270, n=3
1770 ± 810, n=10
830 ± 110, n=9
440 ± 100, n=8
Conc. 4C
3310 ± 240, n=3
1930 ± 520, n=10
810 ± 80, n=7
440 ± 70, n=8
a
values presented as the mean ± standard deviation, number of samples)
37
Table 2. Survival models to predict time to death for P. pugio
Survival Models
TTD=eb1eb2 (ln Conc)eb3W
Compound
b1 (SE)
b2(SE)
b3 (SE)
Naphthalene
52.2 (2.1) -6.3 (0.3) 0.64 (0.03)
Fluorene
22.1 (1.6) -2.8 (0.2) 0.41 (0.02)
Dibenzothiophene
17.4 (0.6) -2.4 (0.1) 0.31 (0.02)
a
Ethylnaphthalene
15.3 (1.0) -2.0 (0.2) 0.26 (0.01)
Dimethylnaphthalenea 14.9 (1.2) -1.8 (0.2) 0.34 (0.03)
Phenanthrenea
13.0 (1.1) -1.5 (0.2) 0.38 (0.03)
TTD = the predicted time-to-death for a specified proportion of the exposed shrimp, b1 = MLEestimated intercept, b2 = the estimated coefficient for the influence of ln of PAH concentration
on time-to-death, b3 = the estimated scale parameter, and W = the response metameter for the
model distribution associated with the proportion dying (P) of the exposed shrimp for which
prediction is being made. (SE) standard error for the calculated parameter
a
data from Unger et al [8]
38
Table 3. Body burdens of toxicants in P.pugio at time of death
Cb mean ± s.d. Kelim (SE)
T1/2
(µm/g)
(hours)
Naphthalene
0.82 ± 0.4,
0.19(0.012),
3.7
n=12
n=8
Fluorene
150 ± 50, n=19
0.91 ± 0.3,
0.090(0.005),
7.7
n=19
n=7
Dibenzothiophene
220 ± 150, n=15 1.2 ± 0.8,
0.041(0.021),
17
n=15
n=7
Ethylnaphthalenea
65 ± 50, n=12
0.41 ± 0.3,
0.093(0.0066),
7.8
n=12
n=6
Dimethylnaphthalenea 70 ± 30, n=11
0.45 ± 0.2,
0.13(0.017),
5.3
n=11
n=10
Phenanthrenea
210 ± 180, n=15 1.2 ± 1.0,
0.12(0.0087),
5.8
n=15
n=6
Cb measured body burden concentration on a wet weight basis ± standard deviation, number of
individuals analyzed
Kelim calculated elimination rate constant (standard error), number of individuals analyzed
T1/2 half-life for body burden concentration
a
Data from Unger et al [8]
Compound
Cb mean ± s.d.
(µg/g)
110 ± 50, n=12
39
60
A
50
95%
40
90%
30
20
70%
50%
10
30%
.52.2
TTD NAP = e
0
1800
- 6.31 ln C
e
2000
0.643W
e
10%
5%
2200
2400
2600
2800
3000
Time to death (hours)
60
B
50
70%
40
50%
30
30%
20
10%
10
5%
TTD FLUOR = e22.078 e- 2.816 ln C e0.4103W
0
500
550
600
650
700
750
800
60
C
50
90%
40
70%
30
50%
20
30%
10%
5%
10
17.399
TTD DBT = e
0
200
250
- 2.405 ln C
e
e
0.3138W
300
350
400
450
500
Concentration (µg/L)
40
Body Burden Concentration
(ug/g)
Dibenzothiophene
700
600
Exposure
Post-exposure
500
300 ug/L
500 ug/L
400
300
200
100
0
12
24
36
48
72
96
Time (H)
41
100,000
10,000
y = 4.4 x 106e-2.1x
R2 = 0.94
1,000
This study
Unger et al [8]
Tatem [15]
100
2
2.5
3
3.5
4
4.5
5
log Kow
60
B
Toxicity (b1)
50
40
y = -33.211x + 162.63
30
20
This study
10
Unger et al [8]
0
2
2.5
3
3.5
4
4.5
5
7
6
Toxicity (-b2)
LC 50 (ug/L)
A
C
5
y = -3.9292x + 19.381
4
3
This study
2
Unger et al [8]
1
0
2
2.5
3
3.5
4
4.5
5
Log KOW
42
Tissue Conc. (µm/g wet weight)
2.5
2
This Study
Unger et al [8]
1.5
1
PHEN
NAP
0.5
0
120
DBT
FLUOR
DMN
ENAP
130
140
150
160
Molecular Weight
170
180
190
43
Appendix B
44
45
46
47
48
49
50
51