Subscriber access provided by READING UNIV
Article
Modeling the toxicokinetics of multiple metals in oyster Crassostrea hongkongensis in dynamic estuarine environment Qiao-Guo Tan, Weitao Zhou, and Wen-Xiong Wang Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.7b04906 • Publication Date (Web): 13 Dec 2017 Downloaded from http://pubs.acs.org on December 19, 2017
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 26
Environmental Science & Technology
1 2 3 4 5 6
Modeling the toxicokinetics of multiple metals in the oyster Crassostrea hongkongensis in a dynamic estuarine environment Qiao-Guo Tan1,2, Weitao Zhou1, Wen-Xiong Wang1,2,3*
7 8 9 10 11 12 13 14 15 16
1. Key Laboratory of the Coastal and Wetland Ecosystems of Ministry of Education, College of the Environment and Ecology, Xiamen University, Xiamen, Fujian 361102, China 2. Center for Marine Environmental Chemistry and Toxicology, Xiamen University, Xiamen, Fujian 361102, China 3. Division of Life Science, The Hong Kong University of Science and Technology, Clearwater Bay, Kowloon, Hong Kong
17 18
*Corresponding author:
[email protected] 1 ACS Paragon Plus Environment
Environmental Science & Technology
19 20
Page 2 of 26
TOC Art
21
2 ACS Paragon Plus Environment
Page 3 of 26
Environmental Science & Technology
22
ABSTRACT
23 24
Metal contamination is a major problem in many estuaries. Toxicokinetic models are useful
25
tools for predicting metal accumulation in estuarine organisms and managing the associated
26
ecological risks. However, obtaining toxicokinetic parameter values with sufficient predictive
27
power is challenging for dynamic estuarine waters. In this study, we determined the
28
toxicokinetics of multiple metals in the oyster Crassostrea hongkongensis in a dynamic
29
estuary polluted by metals using a 48-day transplant experiment. During the experiment,
30
metal concentrations in oysters, water and suspended particles were intensively monitored at
31
3-d interval. The toxicokinetic parameters were then estimated using the Markov chain
32
Monte Carlo (MCMC) method. The calibrated model was capable of successfully simulating
33
the time-course of metal bioaccumulation in oysters and was further validated by predicting
34
the bioaccumulation at another site in the estuary. Furthermore, the model was used to assess
35
the relative importance of different pathways in metal bioaccumulation. With the MCMC
36
method, distributions instead of single values were assigned to model parameters. This
37
method makes the model predictions probabilistic with clearly defined uncertainties, and are
38
thus particularly useful for the risk assessments of metals in aquatic systems.
3 ACS Paragon Plus Environment
Environmental Science & Technology
39
Page 4 of 26
INTRODUCTION
40
In aquatic environments, the relationship between metal bioaccumulation by aquatic
41
organisms and the environmental exposure (dissolved and particulate metals) is complex and
42
dependent on a myriad of biotic and abiotic factors.1 Quantifying the relationship is important
43
for interpreting biomonitoring data in relation to the levels of contamination.2-4
44
Hypothetically, when equilibrium partitioning of metals is attained between organisms and
45
environmental compartments (e.g., water, suspended particles, and sediments), a simple
46
bioconcentration factor can adequately measure the aforementioned relationship.5 However,
47
reaching such equilibrium requires long-term exposure under stable conditions, including
48
metal concentrations and physicochemical and biological factors affecting bioavailability.
49
Therefore, such equilibrium in real aquatic environments is rare, if not impossible. To better
50
understand bioaccumulation in variable or dynamic aquatic environments, such as riverine
51
and estuarine waters, kinetic measures are often required.5-7
52
A good toxicokinetic model should be sufficiently sophisticated for simulating the major
53
bioaccumulation processes, but simple enough for convenient application. The
54
one-compartment first-order toxicokinetic model meets both criteria. The model is simple
55
with only four to five parameters and yet considers processes including aqueous uptake,
56
dietary uptake, elimination, and growth dilution.5, 8, 9 This model is also known as biokinetic
57
or biodynamic model in the area of metal ecotoxicological studies; the parameter values are
58
usually derived in laboratory under stable conditions.8, 9 Such laboratory-derived parameters
59
have been tested in a number of previous studies for predicting metal bioaccumulation in the
60
field and have shown good predictive powers.8, 10-12 Nonetheless, less precise predictions are
61
also reported in some studies.11, 13
62
Applying parameters calibrated under constant laboratory conditions to dynamic
63
estuarine waters is more challenging. In case of unsatisfactory prediction, the model
64
parameters need to be updated with the field data in a scientifically sound way, otherwise the
65
newly obtained biomonitoring data would not contribute to the enhancement of the model's
66
predictive power and be missed. Moreover, although the toxicokinetic model is kinetic in
67
nature, previous studies seldom used the model in its kinetic form for predicting the
68
time-course metal bioaccumulation.8, 10, 11 Instead, the steady-state concentration in organisms
69
was calculated by assuming a constant exposure for predicting metal bioaccumulation in the
70
field. This practice does not fully explore the predictive power of the toxicokinetic model and
4 ACS Paragon Plus Environment
Page 5 of 26
Environmental Science & Technology
71
may not be appropriate in dynamic environments (e.g., estuaries) where steady-state or
72
constant exposure is not possible.
73
In this study, we transplanted the oyster (Crassostrea hongkongensis) to a region of
74
Jiulong River estuary significantly contaminated by multiple metals. Our objective was to
75
calibrate and validate the toxicokinetics of multiple metals simultaneously in a real estuarine
76
environment. The parameter values reported in the literature were used as prior information.
77
During the 48-d oyster transplant, the diffusive gradients in thin films (DGT) technique was
78
used for measuring labile metal concentrations in water. Metal concentrations in suspended
79
particles were also measured periodically to reflect the dietborne exposure. The
80
bioaccumulation of metals in oysters were subsequently simulated with the one-compartment
81
first-order toxicokinetic model using metal concentrations in water and suspended particles as
82
driving variables. The Markov chain Monte Carlo (MCMC) method was used for parameter
83
fittings to provide distributions of parameters and better uncertainty analysis.
84 85
MATERIAL AND METHODS
86
Study area. The two sites for oyster transplant were in the north arm of the Jiulong River
87
estuary (Fig. S1), which is close to Xiamen City in the southeast of China and has a
88
subtropical climate. Site 1 was directly affected by the effluent discharged from Huyu
89
floodgate; site 2 was approximately 2 km downstream (see Fig. S1 for the map). Upstream of
90
Huyu floodgate is a drainage canal, which receives both domestic and industrial wastewater
91
from Jiaomei Town. Levels of multiple metals are high in the canal. The wastewater was
92
discharged into Jiulong River estuary during low tides irregularly but on average every 2 to 3
93
days. The wastewater discharge substantially elevated the metal concentrations in the
94
estuarine water nearby. The average concentrations of dissolved Cr, Ni, Cu, Zn measured
95
recently by Wang and Wang during their two sampling events were around 10 µg L-1, 30 µg
96
L-1, 10 µg L-1, and 10 µg L-1, respectively,7 which were similar to those reported by Weng
97
and Wang earlier.14
98 99
Oyster transplant and sampling. The 48-d oyster transplant experiment was
100
conducted between November 16th, 2013 and January 3rd, 2014. The oysters were juveniles
101
(2–3 months) of C. hongkongensis collected from an oyster farm in Jiuzhen Harbour
102
(24°2′15″N, 117°42′44″E), where no obvious metal contamination was detected.14 At each
103
site, around 300 oysters were placed in a mesh bag which was then fastened to a bamboo pole. 5 ACS Paragon Plus Environment
Environmental Science & Technology
104
The bamboo pole was inserted firmly into the sediment to keep the oysters suspended under
105
water during the whole tidal cycles. Losses of oysters due to mortality or other causes were
106
not counted but were estimated to be minimal. Of the ~300 oysters transplanted, 170 were
107
sampled for metal analyses, 85 were sampled for determining metal subcellular distribution
108
(results not presented in this study), and the others were discarded.
109
Page 6 of 26
Samples of 10 oysters were collected at 3 d intervals from each site. Oyster shells were
110
washed with on-site water to remove sediments. The oysters were then placed in clean
111
ziplock bags and transported to the laboratory in an icebox immediately. In the laboratory, the
112
oysters were opened with a stainless steel oyster knife. The soft parts were thoroughly rinsed
113
with deionized water to remove any visible sediment particles, dried at 80 °C for 48 h, and
114
weighed. The oysters were not depurated, following the methods commonly used by
115
environmental authorities in biomonitoring programs, e.g., marine monitoring programs of
116
China and the mussel watch program of the United States.15, 16 The measured bioaccumulated
117
concentrations of some metals may thus include contributions from gut content due to the
118
higher concentrations of these metals in the particulate phases than in the oysters 17 , but this
119
was considered negligible by theoretical calculation 18. The oyster tissues were digested with
120
concentrated HNO3 in 15-mL polyethylene tubes. Two mL of 65% HNO3 was added to each
121
tube and tubes were left at room temperature for 8-12 h until most solids were dissolved. The
122
tubes were further heated at 80 °C for 8 h using a dry block heater. After heating, the liquids
123
became clear with some lipid precipitates in the bottom of the tubes. The standard reference
124
material SRM1566b (oyster tissue) was digested together with each batch of the samples.
125 126
Samples of water and suspended particles. Water samples were collected each time
127
with the oyster sampling. Three replicates of 500 mL water were taken using polyethylene
128
bottles by immersing them in the water at a depth of around 10 cm with nitrile-gloved hands.
129
The bottles were previously acid washed (soaked in 5% HNO3 for at least 24 h) and rinsed 6
130
times with deionized water, and were placed in double ziplock bags before use. The bottles
131
were rinsed with site water 3 times before collecting water samples.
132
In the laboratory, salinity of water samples was measured with a salinity refractometer
133
(Suwei LS10T); pH and temperature were measured with a benchtop pH meter (pH 510,
134
EUTECH). Each water sample was filtered through a pre-weighed 0.45-µm polycarbonate
135
filter for analyzing the particulate metals. The filters were dried at 80 °C for 48 h and
136
weighed. Each filter together with the particles were digested in a mixture of 4.5 mL of 65% 6 ACS Paragon Plus Environment
Page 7 of 26
Environmental Science & Technology
137
HNO3 and 1.5 mL of 35% HCl using a microwave digestor (Preekem Coolpex). The
138
temperature program was 3 min at 150 oC, 180 oC and 200 oC, respectively, with another 30
139
min at 220 oC. The digested samples were centrifuged at 3000 g for 5 min, and the
140
supernatants were decanted into new polyethylene tubes for metal concentration analysis.
141
Three blank filters and the standard reference material (coastal sediment, GBW07314) were
142
also digested with each batch of samples and served as procedure blanks and quality control
143
samples, respectively.
144 145
DGT measurement. The DGT technique was used for measuring labile metal
146
concentrations in water. The DGT probes were purchased from DGT Research Ltd
147
(Lancaster, UK). A DGT probe was deployed at each site every 3 d during the oyster
148
transplant experiment following the method described by Dunn et al. and was retrieved ~3 d
149
later.19 The exact time of DGT deployment and the average temperature during the period of
150
deployment were recorded. In the laboratory, the retrieved DGT probes were first cleaned
151
with deionized water and then dissembled. The chelex gel layer was placed in a 2-mL
152
centrifuge tube using acid-washed plastic forceps and was eluted with 1 mL of 1 mol L-1
153
HNO3 for 24 h. Metal concentrations in the eluent were determined after appropriate dilution.
154
Labile metal concentrations in water were calculated using the equation and diffusion
155
coefficients provided by the DGT manufacturer (see Supporting Information for more details).
156
Time-integrated average metal concentrations during each 3 d were expected to be provided
157
by this method.
158 159
Metal concentration analysis. Metal concentrations in samples of oyster, suspended
160
particles and DGT eluents were determined using inductively coupled plasma mass
161
spectrometry (ICP-MS, Agilent 7700x). The internal standards, including 45Sc, 72Ge, 115In,
162
and 209Bi, were used for correcting instrument drift and matrix effects. A quality control
163
sample was measured every 10 samples. For the standard reference material GBW07314, the
164
recovery of Co, Ni, Cu, Zn and Cd was in the range of 90% to 110%; whereas the average
165
recovery of Cr and Pb were 80% and 84%, respectively. The lower recoveries of the two
166
metals were probably due to the pseudo-total (instead of the total) digestion method that we
167
used. The recoveries of all metals from the standard reference material SRM1566b were
168
within the range of 90 to 108%. The reported metal concentrations were thus not corrected
169
for recoveries. 7 ACS Paragon Plus Environment
Environmental Science & Technology
Page 8 of 26
170 171
Toxicokinetic modeling. Both water and food are potential sources of metals for the
172
oysters. By assuming first-order kinetics for the uptake and elimination processes, the
173
time-course of metal concentration in oysters can be described by a one-compartment
174
toxicokinetic model:8, 9 ௗ౪ (௧)
175
ௗ௧
= ݇୳ ∙ )ݐ( ୵ܥ+ ݇ ∙ ܥ ( )ݐ− (݇ୣ + ݃) ∙ ܥ୧୬୲ ()ݐ
(1)
176
where Cint (t) is the metal concentration in oysters (µg g-1 dry weight); Cw (t) is the
177
concentration of dissolved metals (µg L-1), which was measured using the DGT technique in
178
this study; Cf (t) is the metal concentration in food particles ingested by oysters (µg g-1); ku is
179
the uptake rate constant of dissolved metals (L g-1 d-1); kf is the uptake rate constant of metal
180
from food (g g-1 d-1); ke is the elimination rate constant (d-1); g is the growth rate constant
181
(d-1); and t is the time of exposure (d). The parameter kf can be considered as the product of
182
ingestion rate (IR, g g-1 d-1) and assimilation efficiency (AE, dimensionless) described in
183
previous studies.8, 9 In the present study, it is difficult to estimate IR and AE separately; kf is
184
thus used for simplicity and practicality.
185 186
Parameter estimation and model prediction. The growth rate constant (g) was
187
estimated based on the increases of individual dry weight of oysters at each site (Fig. S2).
188
The values of g were subsequently kept fixed during the estimation of the other three
189
parameters, i.e., ku, kf, and ke, for each metal. The parameters were estimated using
190
bioaccumulation data collected from site 1, while data of site 2 were used to validate the
191
model.
192
The parameters were estimated following the procedures proposed by Ashauer et al..20
193
Firstly, least-square fittings with the Marquardt algorithm were used to find the parameter
194
values that minimize the sum of the squared residuals. The starting parameter values were set
195
based on the literature on metal toxicokinetics (or wherein called biokinetics) in oysters and
196
are listed in Table S1.21-24 Subsequently, MCMC fittings with the Metropolis-Hastings
197
algorithm were applied to the bioaccumulation data to generate parameter samples of ku, kf,
198
and ke. The best-fit values from the least-square fittings were used as priors for the MCMC
199
fittings. Uniform distributions, with the lower and higher bound at 0.1 fold and 10 fold of the
200
best-fit values, were used as the prior distributions (see Table S2). A parameter sample is a
201
large number of accepted parameter combinations, from which correlation among parameters
202
can also be inferred (see Fig. S7-S13 for examples).25 The parameter sample thus can be 8 ACS Paragon Plus Environment
Page 9 of 26
Environmental Science & Technology
203
considered as a joint posterior distribution of the model parameters. The parameter samples
204
were further used to generate predictions of metal bioaccumulation in the oysters and the 95%
205
confidence intervals. All the model fittings and predictions were conducted with the software
206
openmodel (version 2.4.2) developed by Neil Crout of the University of Nottingham.
207 208
RESULTS AND DISCUSSION
209
Physicochemical parameters. The salinity of water samples varied markedly from 3 to
210
15 and 5 to 22 at the site 1 and 2, respectively (Fig. S3). Higher salinity was observed at site
211
2, which was closer to the sea end. Similarly, pH values also varied substantially. Most of the
212
measured pH were relatively low and ranged from 6.5 to 7.5. The low and even acidic pH
213
were probably caused by the wastewater discharge from the Huyu floodgate. The salinity and
214
pH were measured in grab samples taken at different tidal phases (but usually during low
215
waters) on different days; the observed variations were thus a combined effect of spring-neap
216
and low-high tide cycles.
217
Tide is semi-diurnal in the Jiulong River estuary. Salinity and pH were usually higher at
218
high tides. For example, adjacent to the sites of this study, it was recorded that salinity
219
increased from ~5 at low tides to ~25 at high tides; in the meantime, pH increased from
220
7.0-7.4 to 7.6-7.8.7 The concentrations of dissolved organic carbon (DOC) also varied with
221
tide and increased from around 0.3 mM C at high tides to more than 2 mM C at low tides.7
222
Bioavailability of metals is determined by multiple variables, including salinity, pH, and
223
DOC; therefore, metal bioaccumulation rates in waters with dynamic water chemistry, such
224
as Jiulong River estuary, are expected to vary with time in a very complex way even if metal
225
concentrations are stable.
226 227
Metal concentrations in water and suspended particles. Metal concentrations in
228
water (measured by DGT probes) and suspended particles also varied considerably during the
229
48 d of oyster transplant (Fig. 1; Fig. 2). The patterns of variation were similar for the two
230
study sites, driven by both tidal mixing and wastewater drainage. For the particulate metals,
231
similar patterns were observed among different metals (i.e., Cr, Ni, Cu, and Zn), suggesting
232
the same source of contamination (Fig. 2; Fig. S5). The elevated levels of Cr, Ni, Cu, and Zn
233
in Jiulong River estuary and the concurrence of these metals were also previously reported.26
234
In contrast, the correlations among metals in water were relatively weak (Fig. S4).
9 ACS Paragon Plus Environment
Environmental Science & Technology
235
Page 10 of 26
Dissolved metal. The concentrations of dissolved metals measured by DGT probes are
236
summarized in Fig. 1. The DGT technique provides time-integrated concentrations of labile
237
metals during the consecutive deployment periods of around 3 d each. Metal concentrations
238
were generally higher at site 1, which was closer to the floodgate, than at site 2. Compared to
239
other clean or contaminated estuaries (Table S4), the study sites were subjected to significant
240
contamination of Ni (site 1: 10–61 µg L-1; site 2: 6–35 µg L-1), Cu (site 1: 3–53 µg L-1; site 2:
241
2–19 µg L-1) and Zn (site 1: 7–35 µg L-1; site 2: 4–34 µg L-1). The concentrations of Co, Cd,
242
and Pb were slightly elevated, but were still at relatively low levels (Fig. 1; Table S4). The
243
measured Cr concentrations were quite low (site 1: 0.54 ± 0.43 µg L-1; site 2: 0.30 ± 0.28 µg
244
L-1); however, it should not be interpreted as low Cr contamination in this area but rather due
245
to the DGT technique (see discussion below). Actually, noticeably high Cr concentrations (2–
246
18 µg L-1) were detected in the 0.2-µm filtered water samples by Wang and Wang near the
247
study sites during two independent sampling events.7
248
The DGT probes used in this study measured only the fraction of dissolved metal species
249
which are potentially more bioavailable. In theory, they are cationic species that can bind to
250
chelex resin, including free metal ions, simple inorganic complexes, and small organically
251
complexed species.19 Therefore, the DGT-reactive proportion is metal-specific and dependent
252
on the conditions of water. In a study of estuarine waters, the DGT-reactive metal
253
concentrations as a fraction of 0.45 µm-filterable concentrations was 21 ± 2% for Cu, 29 ± 11%
254
for Pb, 28 ± 5% for Zn, and 27 ± 12% for Ni.19 In another study of near-pristine coastal
255
waters, the ratio of DGT-reactive to 0.45 mm-filterable concentrations was 51% for Cu, 108%
256
for Cd, 104% for Co, and 122% for Pb.27 The Cr species measured by the DGT probes were
257
cationic species (e.g., Cr3+, Cr(OH)2+); whereas the dominant species of Cr in estuarine
258
waters was Cr(VI) anionic species. For example, in the Columbia River and estuary, CrO42-
259
was found to account for more than 90% of the dissolved Cr; Cr(III) species had strong
260
tendency of particle adsorption, which limited their availability as dissolved species.28 This
261
explained why in this study Cr concentrations measured in DGT probes were 1 to 2 orders of
262
magnitude lower than the concentrations of filterable Cr.
263
Particulate metal. The concentrations of metals in suspended particles are summarized
264
in Fig. 2. Metal concentrations were similar between the two study sites. The particles also
265
showed elevated concentrations of Cr (90–1870 µg g-1; average: 245 µg g-1), Ni (24–834 µg
266
g-1; average: 94 µg g-1), Cu (52–658 µg g-1; average: 132 µg g-1), and Zn (99–519 µg g-1,
267
average: 251 µg g-1) (Table S5), consistent with the measurements in waters. The 10 ACS Paragon Plus Environment
Page 11 of 26
Environmental Science & Technology
268
concentrations of Co and Pb were comparable or slightly higher than those observed in other
269
rivers and estuaries.14, 29, 30 Interestingly, Cd concentration in the Jiulong River estuary (~0.1
270
µg g-1) was much lower than those observed elsewhere (Table S5). It is similar to the
271
concentrations (0.04–0.24 µg g-1) found in intertidal sediments in areas adjacent to Jiulong
272
River estuary,26 both suggesting low natural background and anthropogenic contamination of
273
Cd in this area.
274
Jiulong River estuary is a shallow water estuary and usually turbid. Therefore, a large
275
proportion of the sampled suspended particles may be resuspended surface sediments, in
276
addition to biotic and organic detritus. Filter-feeding bivalves, including oysters, can select
277
particles with variable efficiency before ingestion, preferring particles of higher nutritional
278
value.31 Therefore, the sampled particles should not be considered equal in composition to
279
those ingested by the oysters. Nevertheless, metal bioavailability in sediments was found
280
comparable to that in phytoplankton cells. For example, the assimilation efficiencies of Cd
281
(30%–40%), Se (26%–35%), and Zn (34–44%) from sediment particles were only slightly
282
lower or similar to those observed in phytoplankton cells; sediment particles thus contributed
283
an important fraction to metal bioaccumulation in oysters.22
284
The samples of suspended particles were grab samples taken every 3 d. Although taking
285
samples at 3 d intervals during the 48-d experiment was labor intensive, it was still uncertain
286
whether the samples were representative enough. Considering the dynamic nature of
287
estuarine waters and the complex composition of food ingested by oysters, it is challenging to
288
find a time-efficient and cost-efficient method to measure concentrations of metals in the
289
"real" food of oysters. Taking grab samples like what we did in this study is a matter of
290
expediency.
291 292
Metal accumulation in oysters and toxicokinetics. The bioaccumulation of most of
293
the studied metals in the oysters was well described by a simple toxicokinetic model, using
294
metal concentrations in water and suspended particles as driving variables (Fig. 3). Large
295
inter-individual variations were observed; therefore, median concentrations were used in the
296
parameter fittings to reduce the possible bias caused by extreme values. The estimated
297
parameter values and confidence intervals are listed in Table 1. The information on parameter
298
samples, including the posterior distribution of parameters and the correlation among
299
parameters, are presented in Fig. S7 to S13. Model parameters were calibrated with the
300
bioaccumulation data from site 1, and were further validated with the results observed at site 11 ACS Paragon Plus Environment
Environmental Science & Technology
Page 12 of 26
301
2. The time-course of oyster metal concentrations at site 2 was well predicted by the
302
calibrated model (Fig. S6), with most of the predictions within a factor of 0.5 to 2 of observed
303
concentrations (Fig. 5).
304
Generally, metals with more obvious bioaccumulation were better simulated with the
305
toxicokinetic model. Increases in the concentrations of Cr, Cu, and Zn were observed in the
306
oysters. Median Cr, Cu and Zn concentrations increased from 7 µg g-1 to ~20 µg g-1, 1200 µg
307
g-1 to ~9000 µg g-1, and 3400 µg g-1 to ~5000 µg g-1, respectively (Fig. 3). The increases of
308
Co and Ni concentrations were less conspicuous; in contrast, slight decreases were observed
309
for Cd and Pb, of which the modeling were relatively poor (Fig. 3).
310
Toxicokinetics. The dissolved uptake rate constants (ku) were estimated with DGT-labile
311
metal concentrations in this study (Table 1). In principle, they should be higher than ku values
312
reported in the literature, where kus were estimated with total dissolved (i.e., filterable) metal
313
concentrations. For example, ku of Zn estimated in this study was 5.6 ± 2.2 L g-1 d-1. If we
314
assumed a percentage of 28% for DGT-labile Zn to dissolved Zn,19 the ku based on dissolved
315
Zn should be 1.6 ± 0.6 L g-1 d-1, which was more comparable to those previously measured
316
for the same species of oyster, i.e., 1.15 to 2.40 L g-1 d-1 measured by Pan and Wang,23 and
317
2.05 L g-1 d-1 by Ke and Wang.22 Therefore, to use the ku values estimated in this study, metal
318
concentrations in water should either be measured using DGT probes or judiciously
319
converted from the measurements of other methods.
320
The dietary assimilation rate constant (kf) was the product of assimilation efficiency (AE)
321
and the specific ingestion rate (IR), i.e., kf = AE × IR. The values of AE are metal-, organism-,
322
and food-specific, and can be anywhere between 0 and 1, but is often observed between 0.1
323
to 0.9.32 The IR for oysters was also highly variable, usually assumed between 0.02 g g-1 d-1
324
and 0.6 g g-1 d-1 based on the laboratory measurements.22, 23 The IR should be rather uncertain
325
especially in dynamic estuarine waters. Therefore, using a simple parameter kf to relate
326
dietary assimilation proportionately to the metal concentration in food particles is necessarily
327
a simplification. The estimated kf values were very different between metals (Table 1), and
328
can be as high as 0.65 ± 0.25 g g-1 d-1 for Cu, suggesting that IR should be no less than 0.65 g
329
g-1 d-1 on average, since 0 ≤ AE ≤1. Consequently, it can be inferred that AEs of Cr, Co Ni,
330
and Pb were extremely low, i.e., lower than 1%, which was possibly due to that these metals
331
were mainly refractory species contained in inorganic particles. The low assimilation
332
efficiency of Cr (0–5%) is already well-known.32
333
The elimination rate constant (ke) of metals in the oyster was in the order: Ni (0.32 ± 0.18 12 ACS Paragon Plus Environment
Page 13 of 26
Environmental Science & Technology
334
d-1) > Co (0.025 d-1), Zn (0.012 d-1) > Cd, Cr, Pb, Cu (0.0017–0.0074 d-1) (Table 1). The kes
335
of Ni, Zn, and Cd were previously determined for C. hongkongensis (see Table S3).21-23 The
336
kes in the literature overlap the relatively wide ke ranges estimated in this study. The kes of the
337
other metals were not measured in C. hongkongensis but in a number of other oysters, using
338
various methods, generating kes of variable reliability. Not surprisingly, the literature ke
339
values were scattered and also overlaps the values estimated in this study (Table S3).
340
For Ni, Co, and Zn, the kes are higher than or similar to the growth rate constant (g,
341
0.015–0.018 d-1) of oysters, indicating that the efflux processes played an important role in
342
counteracting metal uptake and controlling tissue concentration of these metals. In contrast,
343
the kes of Cd, Cr, Pb, and Cu were much lower than g, indicating that growth dilution was the
344
major mechanism in balancing metal bioaccumulation. These four metals were difficult to be
345
eliminated once incorporated into oyster tissues.
346
MCMC simulation. With MCMC fittings, model parameters were assigned probability
347
distributions instead of the deterministic best-fit values. A parameter sample is a large
348
number (~10000 in this study) of combinations of parameter values that can lead to
349
acceptable agreement between model predictions and observations. The parameter samples
350
can thus be considered as posterior distributions of the parameters and are plotted in Fig. S7
351
to S13. Posterior distributions are also summarized by statistics calculated from the parameter
352
samples, including average, standard deviation, and 95% confidence intervals defined by the
353
2.5th to 97.5th percentiles (Table 1).
354
Due to the continuously varying water conditions (e.g., salinity, pH, DOC, temperature)
355
in estuarine waters, the biological variability, and other sources of uncertainty, the model
356
parameters are not expected to be single specific values but rather value distributions. In
357
addition, the parameters may be correlated, either truly correlated or correlated due to the
358
model design. The correlation between parameters cause difficulties in parameter estimation
359
and uncertainty analysis. These are where the MCMC simulations are particularly useful.
360
MCMC fittings yield a joint probability distribution of the model parameters, which describes
361
the simultaneous variation of all parameters and intrinsically considers the correlation
362
between parameters.25, 33 The joint distributions were then used to predict the confidence
363
intervals (shown in Fig. 3 and Fig. S6) and the relative importance of different pathways in
364
metal bioaccumulation (see below).
365 366
Metal accumulation pathways. Using parameter samples derived from the MCMC
367
fittings, it is convenient to assess the relative importance (and the associated uncertainties) of 13 ACS Paragon Plus Environment
Environmental Science & Technology
Page 14 of 26
368
water and food as the source of metals for oysters (Fig. 5). The bioaccumulation of Cr, Co, Ni,
369
and Cd were clearly dominated by the dissolved phase; whereas Pb might be mainly
370
assimilated from food. For Cu and Zn, both sources were similarly important. The low
371
bioavailability of Cr(III) in food particles was well documented.32 In this study, it is possible
372
that the particulate Cr was mainly Cr(III) species contained in mineral particles and thus
373
contributed negligibly to the Cr bioaccumulation. The low contribution of water to the
374
bioaccumulation of Pb is also expected, because Pb concentrations were low in water and
375
high in particles due to the high particle reactivity of Pb.
376
The quantitative information on the pathways (i.e., water or food) of metal
377
bioaccumulation is crucial for understanding metal biogeochemistry, ecological risks,
378
biomonitoring data.34-37 In previous studies, one well accepted method for delineating the
379
accumulation pathway is modeling. The model parameters (i.e., AE, ku, and ke) were firstly
380
quantified using well-controlled laboratory experiments, and were then used in the following
381
calculation in a deterministic way. In this study, we used posterior distributions (from MCMC
382
fittings) instead of fixed values of the model parameter values and provide better uncertainty
383
analysis with more clearly defined confidence intervals (Fig. 5).
384 385
Model limitations and implications. The model parameters were calibrated by
386
exposing oysters to multiple metals in a real estuarine environment. Theoretically, the
387
parameter values obtained in this study are conditional on that very specific scenario of
388
physicochemical conditions, metal concentrations, particle loads, and so on. Although the
389
model parameters were validated using data collected from another site with good
390
performance (Fig. S6), the two sites were rather similar in metal exposure conditions, which
391
weakened the confidence in the predictive power of the calibrated parameters. Nevertheless,
392
the parameter values (with uncertainties) should be more realistic and robust than those
393
calibrated under stable laboratory conditions, since they had been averaged over a wide range
394
of exposure conditions in the real estuarine environment. In addition, the MCMC method was
395
intrinsically convenient to incorporate new data input. The parameters can be updated to
396
simulate the new scenario of concern, where the parameters values from this study would
397
serve as valuable prior information.
398
The suspended particles sampled in this study were a mixture of algae, organic detritus,
399
suspended sediment particles, among others, which were not distinguished during metal
400
analysis and modeling. The composition of food particles may have important effects on 14 ACS Paragon Plus Environment
Page 15 of 26
Environmental Science & Technology
401
metal bioavailability to oysters.38, 39 Using one single toxicokinetic parameter (kf) for
402
modeling metal accumulation from a spectrum of different particles can be considered as a
403
simplification of the complex actual environment and may limit the accuracy of model
404
predictions.
405
The model can predict the ecological risks associated with metal bioaccumulation (if the
406
bioaccumulation results in adverse effects) and also can be used to back-calculate for setting
407
water quality criteria. For example, the water quality standard (or guideline) of China and
408
Australia on Cd (0.5 and 5 µg L-1) were tested with the model developed in this study, and
409
were found to be not stringent enough for producing oysters safe for human consumption, i.e.,
410
with Cd concentration less than 2 µg g-1 fresh weight. In order to produce ‘safe oysters’, the
411
criterion value of Cd needs to be lowered to 0.11 µg L-1 (Fig. S14).
412
When interpreting biomonitoring data of dynamic aquatic environments, the
413
toxicokinetic model provides a back-calculation of metal exposure integrated or averaged
414
over a certain timespan before the sampling. One important implication of the toxicokinetic
415
modeling is that for a metal with a longer biological half-life (t1/2 = 0.693/ke), the
416
biomonitoring data may reflect its exposure over a longer past period. The average biological
417
half-life estimated for the oyster is in the following order: Ni (2.2 d), Co (28 d), Zn (58 d), Cd
418
(94 d), Cr (105 d), Pb (114 d), and Cu (408 d). The very short half-life of Ni indicates that Ni
419
concentrations measured in the oysters can only reflect Ni exposure during the past several
420
days; whereas for other metals, the timespan reflected can be months or even longer.
421
Being kinetic (instead of assuming steady state) is both the strength and weakness of the
422
model. The toxicokinetic model can capture the whole time-course of bioaccumulation rather
423
than simply a steady-state concentration regardless whether a steady-state can be reached. In
424
addition, the toxicokinetic model can deal with fluctuating exposure, which is what happens
425
in the real environment, instead of assuming an unrealistic stable exposure. However,
426
calibrating or making full use of the toxicokinetic model demands much more variable input
427
(e.g., metal concentrations in water and food) and thus requires more intensive sampling.
428
Although, with the development of automatic in situ or on-site sampling and measurement
429
techniques,40 collecting data for the toxicokinetic model may become much easier in the
430
future.
431
In summary, we demonstrated the feasibility of using a simple toxicokinetic model for
432
simulating the bioaccumulation of metals in a complex and dynamic environment. We also
433
introduced the methodology of determining model parameters in the real estuarine 15 ACS Paragon Plus Environment
Environmental Science & Technology
Page 16 of 26
434
environment. Using distributions instead of single values for model parameters make the
435
predictions of the model probabilistic and with clearly defined uncertainties, and thus
436
particularly useful for risk assessments. Although this study was conducted with one specific
437
oyster species in one specific estuary, the methods we developed can be useful for other
438
biological species and other water bodies.
439 440
ACKNOWLEDGEMENT
441
We dedicated this paper to the memory of Dr. Guo Feng, who passed away right at the
442
beginning of this project. This study was supported by the National Natural Science
443
Foundation of China (21477099, 21237004), the General Research Fund (16140616) from
444
the Hong Kong Research Grants Council, and the Natural Science Foundation of Fujian
445
Province (2014J01165).
446 447
Supporting Information. Settings of initial and prior parameter values (Table S1, S2);
448
metal toxicokinetics of oysters reported in the literature (Table S3); metal concentrations in
449
estuarine waters and suspended particles reported in the literature (Table S4, S5 ); map of
450
study sites (Fig. S1); oyster growth (Fig. S2) and salinity and pH data (Fig. S3); pairwise
451
correlation between metal concentrations in water (Fig. S4) and suspended particles (Fig. S5);
452
metal bioaccumulation at site 2 (Fig. S6); posterior distribution of parameters and correlation
453
among parameters (Fig. S7-S13); Cd bioaccumulation in oysters under hypothetical
454
exposures (Fig. 14); information on DGT measurements and model implementation in
455
openmodel.
16 ACS Paragon Plus Environment
Page 17 of 26
Environmental Science & Technology
456
REFERENCES
457 458 459
1. Langston, W. J.; Spencer, S. K. Biological factors involved in metal concentrations observed in aquatic organisms. In Metal Speciation and Bioavailability in Aquatic System; Tissier, A.; Turner, D. R., Eds. John Wiley & Sons: Chichester, 1995.
460 461 462
2. Farrington, J. W.; Goldberg, E. D.; Risebrough, R. W.; Martin, J. H.; Bowen, V. T. U.S. "Mussel Watch" 1976-1978: an overview of the trace-metal, DDE, PCB, hydrocarbon and artificial radionuclide data. Environ. Sci. Technol. 1983, 17 (8), 490-496.
463 464 465
3. Farrington, J. W.; Tripp, B. W.; Tanabe, S.; Subramanian, A.; Sericano, J. L.; Wade, T. L.; Knap, A. H. Edward D. Goldberg's proposal of “the Mussel Watch”: Reflections after 40 years. Mar. Pollut. Bull. 2016, 110 (1), 501-510.
466 467
4. Rainbow, P. S. Biomonitoring of heavy metal availability in the marine environment. Mar. Pollut. Bull. 1995, 31 (4), 183-192.
468 469 470
5. Landrum, P. F.; Lydy, M. J.; Lee, H. Toxicokinetics in aquatic systems: model comparisons and use in hazard assessment. Environ. Toxicol. Chem. 1992, 11 (12), 1709-1725.
471 472 473
6. Balls, P. W.; Laslett, R. E.; Price, N. B. Nutrient and trace metal distributions over a complete semi-diurnal tidal cycle in the forth estuary, Scotland. Neth. J. Sea. Res. 1994, 33 (1), 1-17.
474 475
7. Wang, W.; Wang, W.-X. Phase partitioning of trace metals in a contaminated estuary influenced by industrial effluent discharge. Environ. Pollut. 2016, 214, 35-44.
476 477
8. Luoma, S. N.; Rainbow, P. S. Why is metal bioaccumulation so variable? Biodynamics as a unifying concept. Environ. Sci. Technol. 2005, 39 (7), 1921-1931.
478 479
9. Wang, W.-X.; Fisher, N. S.; Luoma, S. N. Kinetic determinations of trace element bioaccumulation in the mussel Mytilus edulis. Mar. Ecol. Prog. Ser. 1996, 140 (1/3), 91-113.
480 481
10. Pan, K.; Wang, W.-X. Validation of biokinetic model of metals in the scallop Chlamys nobilis in complex field environments. Environ. Sci. Technol. 2008, 42 (16), 6285-6290.
482 483
11. Roditi, H. A.; Fisher, N. S.; Sañudo-Wilhelmy, S. A. Field testing a metal bioaccumulation model for zebra mussels. Environ. Sci. Technol. 2000, 34 (13), 2817-2825.
484 485 486
12. Lee, J.-H.; Richards, R. G.; Birch, G. F. A novel coupled biokinetic-equilibrium model to capture oyster metal bioaccumulation in a contaminated estuary (Sydney estuary, Australia). Environ. Modell. Softw. 2016, 82, 152-166.
487 488 489
13. Kraemer, L. D.; Campbell, P. G. C.; Hare, L. Modeling cadmium accumulation in indigenous yellow perch (Perca flavescens). Can. J. Fish. Aquat. Sci. 2008, 65 (8), 1623-1634. 17 ACS Paragon Plus Environment
Environmental Science & Technology
Page 18 of 26
490 491
14. Weng, N.; Wang, W.-X. Variations of trace metals in two estuarine environments with contrasting pollution histories. Sci. Total Environ. 2014, 485, 604-614.
492 493 494 495
15. Apeti, D. A.; Johnson, W. E.; Kimbrough, K. L.; Lauenstein, G. G. National Status and Trends Mussel Watch Proram: Sampling Methods 2012 Update; NOAA National Centers for Coastal Ocean Science, Center for Coastal Monitoring and Asssessment: Silver Spring, MD, 2012; p 39.
496 497 498 499
16. General Administration of Quality Supervision, Inspection and Quarantine of P. R. China; Standardization Administration of P. R. China, The Specification for Marine Monitoring—Part 6: Organism Analysis (GB 17378.6—2007). In China Standards Press: Beijing, 2007; p 74.
500 501 502
17. Robinson, W. E.; Ryan, D. K.; Wallace, G. T. Gut contents: A significant contaminant of Mytilus edulis whole body metal concentrations. Arch. Environ. Contam. Toxicol. 1993, 25 (4), 415-421.
503 504
18. Wang, W.-X.; Griscom, S.B.; Fisher, N.S. Bioavailability of Cr(III) and Cr(VI) to marine mussels from solute and particulate pathways. Environ. Sci. Technol. 1997, 31, 603-611.
505 506 507
19. Dunn, R. J. K.; Teasdale, P. R.; Warnken, J.; Schleich, R. R. Evaluation of the diffusive gradient in a thin film technique for monitoring trace metal concentrations in estuarine waters. Environ. Sci. Technol. 2003, 37 (12), 2794-2800.
508 509 510
20. Ashauer, R.; Hintermeister, A.; Caravatti, I.; Kretschmann, A.; Escher, B. I. Toxicokinetic and toxicodynamic modeling explains carry-over toxicity from exposure to diazinon by slow organism recovery. Environ. Sci. Technol. 2010, 44 (10), 3963-3971.
511 512 513
21. Yin, Q.-J. Calibration of metals against salinity and high turnover of nickel in marine bivalves. Master Thesis, The Hong Kong University of Science and Technology, Hong Kong, China, 2017.
514 515 516
22. Ke, C.; Wang, W.-X. Bioaccumulation of Cd, Se, and Zn in an estuarine oyster (Crassostrea rivularis) and a coastal oyster (Saccostrea glomerata). Aquat. Toxicol. 2001, 56 (1), 33-51.
517 518 519
23. Pan, K.; Wang, W.-X. Reconstructing the biokinetic processes of oysters to counteract the metal challenges: Physiological acclimation. Environ. Sci. Technol. 2012, 46 (19), 10765-10771.
520 521
24. Pan, K.; Wang, W.-X. Biodynamics to explain the difference of copper body concentrations in five marine bivalve species. Environ. Sci. Technol. 2009, 43 (6), 2137-2143.
522 523
25. Hack, C. E. Bayesian analysis of physiologically based toxicokinetic and toxicodynamic models. Toxicology 2006, 221 (2), 241-248.
524
26. Tan, Q.-G.; Ke, C.; Wang, W.-X. Rapid assessments of metal bioavailability in marine 18 ACS Paragon Plus Environment
Page 19 of 26
Environmental Science & Technology
525 526
sediments using coelomic fluid of sipunculan worms. Environ. Sci. Technol. 2013, 47 (13), 7499-7505.
527 528
27. Munksgaard, N. C.; Parry, D. L. Monitoring of labile metals in turbid coastal seawater using diffusive gradients in thin-films. J. Environ. Monit. 2003, 5 (1), 145-149.
529 530
28. Ernstberger, H.; Zhang, H.; Davison, W. Determination of chromium speciation in natural systems using DGT. Anal. Bioanal. Chem. 2002, 373 (8), 873-879.
531 532 533
29. Benoit, G.; Oktay-Marshall, S. D.; Cantu, A.; Hood, E. M.; Coleman, C. H.; Corapcioglu, M. O.; Santschi, P. H. Partitioning of Cu, Pb, Ag, Zn, Fe, Al, and Mn between filter-retained particles, colloids, and solution in six Texas estuaries. Mar. Chem. 1994, 45 (4), 307-336.
534 535 536 537
30. Elbaz-Poulichet, F.; Garnier, J.-M.; Guan, D. M.; Martin, J.-M.; Thomas, A. J. The conservative behaviour of trace metals (Cd, Cu, Ni and Pb) and As in the surface plume of stratified estuaries: example of the Rhône River (France). Estuar. Coast. Shelf Sci. 1996, 42 (3), 289-310.
538 539
31. Evan Ward, J.; Shumway, S. E. Separating the grain from the chaff: particle selection in suspension- and deposit-feeding bivalves. J. Exp. Mar.Biol. Ecol. 2004, 300 (1), 83-130.
540 541
32. Wang, W.-X.; Fisher, N. S. Assimilation efficiencies of chemical contaminants in aquatic invertebrates: A synthesis. Environ. Toxicol. Chem. 1999, 18 (9), 2034-2045.
542 543 544 545
33. Weijs, L.; Yang, R. S. H.; Das, K.; Covaci, A.; Blust, R. Application of Bayesian population physiologically based pharmacokinetic (PBPK) modeling and Markov chain Monte Carlo simulations to pesticide kinetics studies in protected marine mammals: DDT, DDE, and DDD in harbor porpoises. Environ. Sci. Technol. 2013, 47 (9), 4365-4374.
546 547 548
34. Croteau, M.-N.; Luoma, S. N. Delineating copper accumulation pathways for the freshwater bivalve Corbicula using stable copper isotopes. Environ. Toxicol. Chem. 2005, 24 (11), 2871-2878.
549 550 551
35. Hédouin, L.; Metian, M.; Teyssié, J.-L.; Fichez, R.; Warnau, M. Delineation of heavy metal contamination pathways (seawater, food and sediment) in tropical oysters from New Caledonia using radiotracer techniques. Mar. Pollut. Bull. 2010, 61 (7–12), 542-553.
552 553 554
36. Munger, C.; Hare, L. Relative importance of water and food as cadmium sources to an aquatic insect (Chaoborus punctipennis): Implications for predicting Cd bioaccumulation in nature. Environ. Sci. Technol. 1997, 31 (3), 891-895.
555 556
37. Wang, W.-X.; Fisher, N. S. Delineating metal accumulation pathways for marine invertebrates. Sci. Total Environ. 1999, 237, 459-472.
557 558 559
38. Lee, J. H.; Birch, G. F.; Simpson, S. L. Metal-contaminated resuspended sediment particles are a minor metal-uptake route for the Sydney rock oyster (Saccostrea glomerata) — A mesocosm study, Sydney Harbour estuary, Australia. Mar. Pollut. Bull. 2016, 104 (1), 19 ACS Paragon Plus Environment
Environmental Science & Technology
Page 20 of 26
560
190-197.
561 562 563 564
39. Lee, J. H.; Birch, G. F.; Cresswell, T.; Johansen, M. P.; Adams, M. S.; Simpson, S. L. Dietary ingestion of fine sediments and microalgae represent the dominant route of exposure and metal accumulation for Sydney rock oyster (Saccostrea glomerata): A biokinetic model for zinc. Aquat. Toxicol. 2015, 167 (Supplement C), 46-54.
565 566 567
40. Buffle, J.; Horvai, G., In Situ Monitoring of Aquatic Systems: Chemical Analysis and Speciation. Wiley: New York, 2000; Vol. 6.
20 ACS Paragon Plus Environment
Page 21 of 26
Environmental Science & Technology
568
Tables and Figures
569
Table 1 The estimated values of the toxicokinetic parameters (ku, kf, and ke). The values are
570
mean ± standard deviation; values in parenthesis are 95% confidence intervals. The statistics
571
summarize the parameter samples generated by the MCMC fittings of the site 1 data (see Fig.
572
S7–S13 for more information of the parameter samples).
573
Metal Cr
Co
Ni
Cu
Zn
Cd
Pb
ku (L g-1 d-1)
kf (g g-1 d-1)
ke (d-1)
1.3 ± 0.2
0.00035 ± 0.00021
0.0066 ± 0.0012
(0.9, 1.6)
(0.00004,0.00078)
(0.0044, 0.0087)
0.071 ± 0.035
0.0032 ± 0.0018
0.025 ± 0.017
(0.011, 0.137)
(0.0003, 0.0074)
(0.002, 0.071)
0.39 ± 0.23
0.0039 ± 0.0026
0.32 ± 0.18
(0.07, 0.89)
(0.0006, 0.0100)
(0.05, 0.72)
6.3 ± 1.9
0.65 ± 0.25
0.0017 ± 0.0014
(3.8, 10.5)
(0.06, 0.95)
(0.0001, 0.0044)
5.6 ± 2.2
0.39 ± 0.15
0.012 ± 0.004
(0.4, 9.2)
(0.15, 0.74)
(0.003, 0.020)
1.5 ± 0.6
0.16 ± 0.09
0.0074 ± 0.0036
(0.4, 2.6)
(0.02, 0.31)
(0.0009, 0.0136)
0.11 ± 0.07
0.0011 ± 0.0006
0.0061 ± 0.0020
(0.01, 0.25)
(0.0003, 0.0024)
(0.0033, 0.0099)
574
21 ACS Paragon Plus Environment
Environmental Science & Technology
Page 22 of 26
575
Fig. 1 Labile metal concentrations in water measured at ~3 d intervals using DGT probes at
576
the two study sites during the 48-d oyster transplantation. 2.0
Cr 1.5 1.0 Site 1 Site 2
-1
DGT-measured metal concentration in water (µg L )
0.5 0.0 1.5
40
Co
1.2
Zn 30
0.9
20
0.6 0.3
10
0.0 80
0 0.4
Ni
Cd
60
0.3
40
0.2
20
0.1
0
0.0
60
0.6
Cu
50
Pb 0.5
40
0.4
30
0.3
20
0.2
10
0.1
0
0.0 0
577
10
20
30
40
Days of exposure
50
0
10
20
30
40
50
Days of exposure
22 ACS Paragon Plus Environment
Page 23 of 26
Environmental Science & Technology
578
Fig. 2 Metal concentrations in suspended particles collected at the two study sites during the
579
48-d oyster transplantation. Values are mean ± standard deviation (n = 3). 104
Cr 103
-1
Metal concentration in suspended particles (µg g )
102 Site 1 Site 2
10
1
102
103
Co
Zn
101
102
100 104
101 101
Ni
Cd
3
10
100
102
10-1
101
10-2
103
102
Cu
Pb
102
101
101 0
580
10
20
30
40
Days of exposure
50
0
10
20
30
40
50
Days of exposure
23 ACS Paragon Plus Environment
Environmental Science & Technology
Page 24 of 26
581
Fig. 3 Metal bioaccumulation in soft tissues of oysters at site 1 during the 48-d
582
transplantation. The × symbols represent measured concentrations; the red curves and the
583
shaded areas are average concentrations and 95% confidence intervals predicted by the model
584
based on the parameter samples from the MCMC fittings. See Fig. S7–S13 for the parameter
585
samples. 60
Cr 45
Metal concentration in oyster tissue (µg g-1 dry weight)
30 15
measured Modeled
0 4
10000
Co
Zn
8000
3
6000 2 4000 1
2000
0
0
160
40
Ni
Cd
120
30
80
20
40
10
0
0
15000
20
Cu
12000
Pb 15
9000 10 6000 5
3000 0
0 0
586
10
20
30
40
Days of exposure
50
0
10
20
30
40
50
Days of exposure
24 ACS Paragon Plus Environment
Page 25 of 26
Environmental Science & Technology
Fig. 4 The comparison between observed and model predicted metal concentrations in
588
oysters at site 2. The observed values are the measured median concentrations of ~10
589
replicated oysters. The predicted values are the average concentrations predicted by the
590
model based on the parameter samples from the MCMC fittings of data from site 1. The solid
591
line is the curve y = x; the two dashed lines are the curves y = 2·x and y=x/2, respectively.
-1
Observed concentration (µg g )
587
592
104
Cu
Zn
103 102 101 100
Cr
Ni
Pb Cd Co
10-1 10-1 100
101
102
103
104
Predicted concentration (µg g-1)
593
25 ACS Paragon Plus Environment
Environmental Science & Technology
Page 26 of 26
594
Fig. 5 The comparison between water and food (suspended particles) as the source of metals
595
accumulated by the oysters (C. hongkongensis) at study site 1 during the 48-d transplant. The
596
curves and the shaded areas are average concentrations and 95% confidence intervals
597
predicted by the model based on the parameter samples from the MCMC fittings of data from
598
site 1.
599
26 ACS Paragon Plus Environment