Environ. Sci. Technol. 2003, 37, 2236-2241
Uptake and Depuration of 4-Nonylphenol by the Benthic Invertebrate Gammarus pulex: How Important Is Feeding Rate? MELANIE Y. GROSS-SOROKIN,† ERIC P. M. GRIST,* MICHAEL COOKE, AND MARK CRANE School of Biological Sciences, Royal Holloway, University of London, Egham, Surrey TW20 0EX, U.K.
The major exposure and uptake route for soluble toxins by aquatic organisms is generally considered to be through the water column. In the case of hydrophobic chemicals, exposure and uptake through diet often take on greater importance as the chemicals adsorb onto organic sediments and food. A chemical that has recently come under close scrutiny because of its toxicity and possible endocrine disrupting effects in aquatic life is 4-nonylphenol (NP). It has been detected in environmental water and sediment samples and is a persistent and hydrophobic (log Kow ) 4.48) contaminant in many aquatic systems. In this study, the relative importance of NP uptake through accumulation from diet and water was examined for the detritus-feeding freshwater shrimp Gammarus pulex. Using a bootstrap nonlinear regression technique, the level of toxin present in G. pulex at any time during or after initial exposure was estimated. Heterogeneity, together with assumptions on feeding rate, was shown to affect the determination of NP uptake substantially. Because of its lifestyle as a benthic organism, the main exposure route was at first assumed to be through sediments and food. However, the results suggest that major uptake may also occur through water. The statistical and modeling methodology may be applied to uptake and depuration assessments for any aquatic organisms exhibiting a variable feeding phase.
than the ingestion of food and/or sediments. However, as the hydrophobicity of a chemical increases, the contribution of a chemical from food and/or sediment sources is likely to become increasingly important (6). Assessments have largely proceeded on the further assumption that individuals feed at a constant rate throughout the course of the exposure duration (e.g., refs 7 and 8). Feeding rate is either taken from the published literature or represented by an average value calculated over all replicates. This approach incorporates no information on variability of feeding rate through behavioral changes or individual heterogeneity, even though such variability in feeding rate has been observed over the time course of the experiment (8-10). In this paper, we determine the relative importance of aqueous and dietary uptake of NP by the freshwater epibenthic amphipod Gammarus pulex. G. pulex is a stream detritivore that feeds primarily upon coarse particulate organic material such as allochthonous decaying leaves material (11). Benthic invertebrates are exposed to chemicals through direct contact with sediment and pore water, through direct uptake from the water column, or by ingestion of sediment and/or food particles. The relative importance of aqueous and dietary uptake of contaminants for benthic organisms will depend on the exposure concentration in the water column and other phases, such as food, pore water, and sediments encountered by the organism. It is well-known from comparisons of NP concentrations detected in surface waters ( ) 0). The solution to eq 4 when feeding rate is modeled by eq 5 describes simultaneous uptake from aqueous and dietary sources and is given by
y)
( )
( )
CRG CRH (e-k2t - e-Bt) + (1 - e-k2t) + B - k2 k2 Cwk1 (1 - e-k2t) (6) k2
( )
The first and second terms on the right-hand side of eq 6 give the total NP accumulated from food, and the third term is from water (as previously in eq 3). Depuration. Depuration was modeled by an exponential decay function:
y ) Ie-k2t
(7)
where I is the initial concentration of NP in G. pulex at the start (at 48 h exposure) of the depuration phase (µg g-1). Half-lives (t1/2) were calculated using k2(t1/2 ) ln 2/k2) (17). Constant Feeding Rate Analysis (G ) 0). A constant feeding rate must be represented by an average value VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2237
FIGURE 1. Depuration data after 48 h of NP uptake together with best-fit curves (after log-linear transformation). (0 - -) after aqueous exposure 1; ([ s) after dietary exposure 2. calculated as the food ingested per gram of body mass (wet confidence intervals (CI)] and 83.4 µg/g ( 8.7 CI, respectively, weight) per hour of exposure duration. The calculation is in exposure 1 and 2.9 µg/L ( 0.9 CI and 82.9 µg/g ( 10.5 CI, not often specified in the literature, rather an “average value” respectively, in exposure 2. An initial flux of NP from the is usually quoted without any supporting computation (9). leaves into the water occurred within the first 8 h of the In fact, there are several ways in which such an average can experiment. However, no significant differences were found be computed, frequently yielding different values. Two typical between replicates and time points for water concentrations methods are as follows: (exposure 1: χ2 3 ) 2.69, p ) 0.44; exposure 2: χ2 3 ) 2.07, Model a (Mean Feeding Rate). Calculate the mean food p ) 0.56) and food concentrations (exposure 1: χ2 3 ) 3.61, mass ingested per gram of body mass for each set of three p ) 0.31; exposure 2: χ2 3 ) 5.82, p ) 0.12) from the 8-h replicates (8, 12, 24, 48 h), divide by the exposure duration measurement onward. Since no significant differences in each case, and then take the mean of these four values. between concentrations in water and food could be detected Model b (Linear Regression to Feeding Data). Regression after this time, the concentrations can be assumed to have of a straight line to the feeding data. This is equivalent to remained constant in both the exposure experiments. specifying G ) 0 in eq 5 and obtaining H as the slope of the Parameters were determined by separate nonlinear best-fit line by linear regression. regressions performed sequentially using the Gauss-Newton Variable Feeding Rate Analysis (G > 0, B > 0). If the method (e.g., ref 20) in the following order. feeding rate is assumed to vary with time as described by eq Depuration. The depuration rate constant k2 was obtained 5 and observed in exposure 2, the feeding rate parameters by regression of eq 7 to depuration data (Figure 1). Using the G, B, and H must be determined. As with the constant feeding data set from exposure 1, this gave k2 ) 0.01 (r 2 ) 0.87) with rate analysis, there are different ways of summarizing and a corresponding half-life (t1/2) of 62.4 h. Regression results incorporating the feeding data into the feeding description. using data from exposure 2 gave a similar value of k2 ) 0.01 We investigated two variable feeding rate descriptions, both (r 2 ) 0.90) and a corresponding half-life (t1/2) of 60.3 h. In incorporating the exponentially decreasing feeding rate both cases a good fit of the depuration data to the log-linear function of eq 5: depuration curve was achieved, as can be seen clearly in Model c (Nonlinear Regression to Feeding Data). Nonlinear Figure 1 after log-linear transformation of the respective bestregression of the integral of eq 5 to the feeding data. fit curves. Significantly, this suggests a single-compartment Model d (Nonlinear Regression to Feeding Rate Data). model is a suitable representation for the uptake and Direct nonlinear regression of eq 5 to time-averaged feeding depuration kinetics of the system (17). rate data. Bioconcentration. The concentrations of NP measured Separate nonlinear regressions of eq 6 were then perin shrimps at each time point following exposure 1 are formed in each case. If feeding rate is constant (Models a and presented in Figure 2 (circles). An uptake rate constant k1 b), eq 6 reduces to the standard constant feeding rate model was determined by nonlinear regression of eq 3 to the described by Spacie and Hamelink (17). bioconcentration data obtained in exposure 1. Using the Statistical Analyses. Differences in NP concentration in depuration rate constant k2 of 0.01 derived above, this gave leaf disks and water samples between replicates and time a k1 value of 38.38 h-1. points were tested with a Kruskal-Wallis one-way ANOVA. Bioconcentration and Food Bioaccumulation. The conEach experimental protocol generated three statistically centrations of NP measured in shrimps at each time point independent estimates of the NP present in the sample following exposure 2 are presented in Figure 2 (crosses). With obtained at each of the recorded time points. Replicate curves the constants k1 and k2 both derived, the assimilation rate were obtained by regression to data triplicates randomized constant R was determined by regression of eq 6 to the food at each time point by the block bootstrap (19). By bootstrap bioaccumulation and bioconcentration data obtained in resampling the data, pointwise percentile confidence interexposure 2. vals were generated for all regression curves and parameters. Uptake and Depuration Kinetics. A graphical overview Results of the complete uptake and depuration kinetics of NP together with data from exposures 1 and 2 is shown plotted in Figure Water and Leaf NP Concentrations. Mean measured con2. The upper (food bioaccumulation and bioconcentration) centrations of NP in water and leaf were 3.0 µg/L ( 0.9 [95% 2238
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 10, 2003
FIGURE 2. Uptake and depuration kinetics based on data sets for exposures 1 (b) and 2 (+) with best-fit regression curve replicates obtained by regression to data triplicates randomized at each time point by the block bootstrap (the lower and upper curve batches, respectively). Thirty replicate curves are shown in each case for uptake (until 48 h) and depuration (after 48 h) phases. The central batch of curves obtained by subtracting the lower from the upper curve batches represents uptake and depuration from bioaccumulation only. Feeding rate is described by Model c (as shown by the best-fit curve of Figure 3 and incorporated into the bootstrap regression results shown in Figure 4B,D). and lower (bioconcentration) batches of curve replicates were derived from regressions performed to resamples generated by the block bootstrap. The central batch of curves, obtained by subtracting the lower from the upper replicates, therefore represent the associated food bioaccumulation uptake and depuration profile. This estimation approach is a modification of that used by Pastershank et al. (7) in which water uptake concentrations in caddis fly larvae were subtracted from food and water uptake concentrations at steady state. Separate modeling of the different exposure routes and subsequent subtraction of the curves was required in the current study because steady-state concentrations were not attained in the short uptake phase. A higher body burden of NP was achieved in the gammarids by dietary uptake than from aqueous uptake. Influence of Feeding Rate. The influence of variable feeding rate was explored by comparing results obtained with constant feeding rate Models a and b with those of the variable feeding rate Models c and d. Model a (mean feeding rate) gave the mean feeding rate as 0.0047 g g-1 h-1 and an NP assimilation rate of R ) 0.82. Model b (linear regression to feeding data) gave the feeding rate as 0.0035 g g-1 h-1 and an associated NP assimilation rate of R ) 1.12. With a variable feeding rate assumed, Model c (nonlinear regression to feeding data) yielded an NP assimilation rate of R ) 1.13, whereas Model d (nonlinear regression to feeding rate data) yielded an NP assimilation rate of R ) 0.42. The respective regression line (Model b) and regression curve (Model c) of best-fit to the feeding data as mass of leaf ingested per gram of shrimp are presented in Figure 3. All parameter values, together with 95% bootstrap percentile CI, that were obtained with each model feeding description are presented in Table 1. For Models b (linear regression to feeding data) and c (nonlinear regression to feeding data), the estimated assimilation rate R was greater than unity and, therefore, not possible. In direct contrast, Models a (mean feeding rate) and d (nonlinear regression to feeding rate data) yielded realistic assimilation rate estimates of 82% and 42%, respectively. Therefore, incorporation of different feeding descriptions and/or data summary calculations substantially
FIGURE 3. Feeding data (O) obtained during the bioaccumulation phase in the dietary exposures together with plots of feeding Models b (constant feeding rate) and c (variable feeding rate) determined by linear and nonlinear regression, respectively.
TABLE 1. Uptake and Depuration Parameters Together with 95% Bootstrap Percentile Confidence Limits Determined in the Parameter Evaluationa 95% bootstrap percentile confidence limits parameter
k1 k2 (r 2 ) 0.90)
value 38.38 0.01
Model a constant feeding rate f 0.0047 R (r 2 ) 0.71) 0.82
lower
upper
32.74 0.009
44.21 0.013
0.0046 0.72
0.0047 0.93
G B H R (r 2 ) 0.70)
Model b (r 2 ) 0.69) 0 0 0.0035 0.0031 1.12 0.95
0.0038 1.31
G B H R (r 2 ) 0.70)
Model c (r 2 ) 0.69) 0.08 0.00 2.50 2.498 0.0026 0.0017 1.13 0.98
0.014 2.501 0.0035 1.34
G B H R (r 2 ) 0.74)
Model d (r 2 ) 0.32) 0.06 -0.038 0.32 0.00 0.0034 -0.0345 0.42 -0.08
1.52 0.69 0.0036 1.69
a See Constant Feeding Rate and Variable Feeding Rate Analyses section. All confidence intervals were calculated from B ) 5000 bootstrap resamples. Parameter values are shown to two decimal places (or less where necessary to enable comparison with upper and lower confidence limits), r 2 is the coefficient of determination calculated for the corresponding best-fit regression curve.
affect estimation of food bioaccumulation. As demonstrated in Figure 4, choice of feeding description [panels A (linear) B (nonlinear)] similarly influenced estimation of the associated uptake from food bioaccumulation and bioconcentration (panels C and D, respectively). In particular, the confidence limits (dotted lines) for the best fit regression curve (bold lines) were affected by both the inherent variability in the data and the choice of model.
Discussion Assimilation rates for xenobiotics have been calculated by a variety of methods. The assimilation rate of 82% obtained through using a constant feeding rate (Model a) in this study VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2239
FIGURE 4. Effect of feeding description on estimation of uptake of NP from food and water. Panels A and B show the best-fit regression curve (bold line) for each feeding model, with associated best-fit bioconcentration and food bioaccumulation curve (bold line) shown in the respective panels C and D. Thirty replicate curves (dashed lines) together with pointwise 95% upper and lower bootstrap confidence limits (B ) 5000) (dotted lines) computed by the block bootstrap are superimposed onto each panel. falls close to values reported for mercury uptake by Penaeus duorarum and Callinectes sapidus (72 and 76%, respectively) (21), whereas the assimilation rate of 42% obtained through using an exponentially decreasing feeding rate (Model d) falls within the range reported for PCB and PAH uptake by fish (23-50% and 2-75%, respectively) (22, 23) and for PCB uptake by mysids (29-41%) (9). However, estimates for R obtained by assuming Models b and c exceeded 1, which is impossible, while Models a and d resulted in estimates with wide confidence intervals. An overestimate of R could result if the product of Cwk1 is underestimated (9). In this study, R was determined in eq 4 by the insertion of k1 determined from the water bioconcentration model (eq 2). This assumes that k1 is the same in feeding and nonfeeding gammarids. However, it is possible that feeding activity may accelerate the flow of contaminant past the animal. This might have the effect of elevating k1 for feeding gammarids relative to nonfeeding gammarids and would be plausible if contaminant uptake were proportional to increases in respiration rate due to feeding activity (9). Indeed, gammarids in the food uptake experiment were more active than gammarids in the water uptake experiment, and uptake clearance rates from water are partially related to respiration rates in another crustacean, Mysis relicta (24). Since a realistic value for R was obtained by assuming a nonconstant feeding rate model by regression to time2240
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 10, 2003
averaged data (Model d) but with wide confidence intervals, the problems observed with estimating R from feeding descriptions in this study could lie with the raw feeding data being too scattered and variable. High variability in G. pulex feeding rates on chestnut leaves have been observed previously (8, 10). This variability can be due to a variety of factors such as variation in the quality of the food provided despite standardized procedures (25), nutritional status of the organisms, disease or parasitism, and natural differences in the appetites of individuals (10). The latter may be influenced by the physiological condition of the organism as, for example, observed prior to and during moulting (26). The result of such variability in feeding rate is that the determination of R from these data is subject to a large amount of uncertainty. This is consistent with the general finding that R is very sensitive to assumptions and stochastic variations in the feeding rate, as demonstrated by the adoption of four different feeding descriptions and by resampling of the data. Many dietary uptake studies rely on the use of published average rather than directly measured feeding rates (e.g., refs 7 and 18), and when feeding rates are measured, they are often highly variable (8, 27). Changes in feeding rate may be due to satiation or to increased contaminant accumulation and resulting toxic effects in the organisms. These factors are influential because they affect estimation of accumulation of xenobiotics from food as well as assimilation rate.
Our analysis confirms that a higher body burden of NP is achieved by dietary uptake in G. pulex (Figure 2) and might lead to the conclusion that diet is indeed the main uptake route. However, this evaluation does not take into account the higher source concentration in the leaf disks. When evaluated on these terms, uptake from water is unexpectedly high and is consistent with the assumption that water uptake is often the predominant uptake route because of the larger volume of water transferred across the gills to satisfy an organism’s oxygen demand. In addition, comparison of depuration rates after exposure to dietary and aqueous sources is often performed to determine the predominant uptake route (17, 28). In the few studies on bioconcentration of NP that have estimated uptake kinetics, k1 was reported in the range of 0.96-1.88 h-1 (2, 3), which is much lower than the value of 38 h-1 estimated in this study. Rapid depuration half-lives of 9 h for killifish (29), 7.2 h for mussels (2) and 96 h for salmon have been reported (3), which are consistent with the depuration half life of 62 h found for G. pulex here. Greater bioaccumulation of hydrophobic organic chemicals has been shown experimentally to be related largely to slower elimination kinetics rather than to increases in uptake kinetics (30). In this study, depuration rates after aqueous and dietary exposure were rapid and did not differ significantly. This lends further support to the conclusion that the higher body burden achieved in gammarids after dietary exposure is more likely due to the higher NP concentration in leaf disks relative to the water concentration rather than enhanced retention and accumulation from dietary sources. Nevertheless, this analysis indicates that significant uptake from food could occur at high dietary NP levels. In the aquatic environment, exposure of benthic detritivores to NP may occur through both dietary and aqueous sources, and experimental evidence supports the assumption of additivity of food and water contaminant sources (17). In the case of continuous NP discharge into a river, resulting in constant concentrations in the water column, uptake from water might therefore be predominant. However, sediment concentrations would also increase, resulting in higher body burdens in benthic organisms. In the case of intermittent NP discharge, exposure through dietary sources would take on greater importance as NP is expected to partition out of the aqueous phase into sediment and detritus. In addition, desorption of NP from sediment and detritus into water, as observed in this study, may occur when concentrations of NP in the water column decrease.
Acknowledgments E.P.M.G. was funded by Grant GST/02/2062 from the Natural Environment Research Council.
Literature Cited (1) Ekelund, R.; Bergman, A.; Granmo, A.; Berggren, M. Environ. Pollut. 1990, 64, 107-120. (2) McLeese, D.; Sergeant, D.; Metcalfe, C.; Zitko, V.; Burridge, L. Bull. Environ. Contam. Toxicol. 1980, 24, 575-581. (3) McLeese, D. W.; Zitko, V.; Sergeant, D. B.; Burridge, L.; Metcalfe, C. D. Chemosphere 1981, 10, 723-730.
(4) Liber, K.; Gangl, J. A.; Corry, T. D.; Heinis, L. J.; Stay, F. S. Environ. Toxicol. Chem. 1999, 18, 394-400. (5) Ahel, M.; McEvoy, J.; Giger, W. Environ. Pollut. 1993, 79, 243248. (6) Opperhuizen, A. In Organic Micropollutants in the Environment. Proceedings of the Sixth European Symposium; Angeletti, G., Bjorseth, A., Eds.; Kluwer Academic Publishers: Amsterdam, 1991; pp 61-70. (7) Pastershank, G. M.; Muir, D. C. G.; Fairchild, W. L. Environ. Toxicol. Chem. 1999, 18, 2352-2360. (8) Xu, Q.; Pascoe, D. Arch. Environ. Contam. Toxicol. 1994, 26, 459-465. (9) Landrum, P. F.; Frez, W. A.; Milagros, S. S. Chemosphere 1992, 25, 397-415. (10) Taylor, E. J.; Jones, D. P. W.; Maund, S. J.; Pascoe, D. Chemosphere 1993, 26, 1375-1381. (11) Welton, J. S.; Ladle, M.; Bass, J. A. B.; John, I. R. Oikos 1983, 41, 133-138. (12) Bennie, D. T.; Sullivan, C. A.; Lee, H.-B.; Peart, T. E.; Maguire, R. J. Sci. Total Environ. 1997, 193, 263-275. (13) Ahel, M.; Giger, W.; Schaffner, C. Water Res. 1994, 28. (14) Naylor, C. G.; Mieure, J. P.; Adams, W. J.; Weeks, J. A.; Castaldi, F. J.; Ogie, L. D.; Romano, R. R. J. Am. Oil Chem. Soc. 1992, 69, 695-703. (15) Johnson, A. C.; White, C.; Besien, T. J.; Jurgens, M. D. Sci. Total Environ. 1998, 210/211, 271-282. (16) Hale, R. C.; Smith, C. L.; deFur, P. O.; Harvey, E.; Bush, E. O.; La Guardia, M. J.; Vadas, G. G. Environ. Toxicol. Chem. 2000, 19, 946-952. (17) Spacie, A.; Hamelink, J. L. In Fundamentals of Aquatic ToxicologysMethods and Applications; Rand, C. M., Petrocellie, S. R., Eds.; Hemisphere Publishing Corp.: Washington, DC, 1985; pp 495-525. (18) van Hattum, B.; de Voogt, P.; van den Bosch, L.; van Straalen, N. M.; Joosse, E. N. G. Environ. Pollut. 1989, 62, 129-151. (19) Efron, B.; Tibshirani, R. J. An Introduction to the Bootstrap; Chapman and Hall: New York, 1993. (20) Bates, W.; Watts, D. Nonlinear Regression Analysis and Its Applications; John Wiley & Sons: New York, 1988. (21) Evans, D. W.; Kathman, R. D.; Walker, W. W. Mar. Environ. Res. 2000, 49, 419-434. (22) Opperhuizen, A.; Schrap, S. M. Chemosphere 1988, 17, 253262. (23) Fisk, A. T.; Norstrom, R. J.; Cymbalisty, C. D.; Muir, D. C. G. Environ. Toxicol. Chem. 1998, 17, 951-961. (24) Landrum, P. F.; Frez, W. A.; Simmons, M. S. J. Great Lakes Res. 1992, 18, 331-339. (25) Naylor, C.; Maltby, L.; Calow, P. Hydrobiologia 1989, 188/189, 517-523. (26) Willoughby, L. G.; Earnshaw, R. Hydrobiologia 1982, 97, 105117. (27) Weston, D. P. Mar. Biol. 1990, 107, 159-169. (28) Warnau, M.; Fowler, S. W.; Teyssie, J.-L. Mar. Pollut. Bull. 1999, 39, 159-164. (29) Tsuda, T.; Takino, A.; Muraki, K.; Harada, H.; Kojima, M. Water Res. 2001, 35, 1786-1792. (30) Spacie, A.; McCarty, L. S.; Rand, G. M. In Fundamentals of Aquatic Toxicology; Rand, G. M., Ed.; Taylor and Francis: London, 1995; pp 493-522.
Received for review May 8, 2002. Revised manuscript received December 6, 2002. Accepted December 12, 2002. ES020092N
VOL. 37, NO. 10, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2241
