Copper detection in the Asiatic clam Corbicula fluminea: optimum valve closure response

Copper detection in the Asiatic clam Corbicula fluminea: optimum valve closure response

Aquatic Toxicology 66 (2004) 333–343 Copper detection in the Asiatic clam Corbicula fluminea: optimum valve closure response Damien Tran a,b,∗ , Elod...

136KB Sizes 0 Downloads 26 Views

Aquatic Toxicology 66 (2004) 333–343

Copper detection in the Asiatic clam Corbicula fluminea: optimum valve closure response Damien Tran a,b,∗ , Elodie Fournier a,b , Gilles Durrieu b , Jean-Charles Massabuau b a

Laboratoire de Radioécologie Expérimentale, Institut de Radioprotection et de Sˆureté Nucléaire 13115 Saint Paul-Lez-Durance, France b UMR 5805, Laboratoire d’Ecophysiologie et Ecotoxicologie des Systèmes Aquatiques. Université Bordeaux 1 and CNRS, Place du Dr. Peyneau 33120 Arcachon, France Received 9 January 2003; received in revised form 3 July 2003; accepted 9 July 2003

Abstract When exposed to a contaminant, bivalves close their shell as a protective strategy. The aim of the present study was to estimate the maximum expected dissolved copper sensitivity in the freshwater bivalve Corbicula fluminea using a new approach to determine their potential and limit to detect contaminants. To take into account the rate of spontaneous closures, we integrated stress problems associated with fixation by a valve in usual valvometers and the spontaneous rhythm associated with nycthemeral activity, to optimize the response in conditions where the probability of spontaneous closing was lowest. Moreover, we used an original system with impedance valvometry, using lightweight impedance electrodes, to study free-ranging animals in low stress conditions combined with an analytical approach describing dose-response curves by logistic regression, with valve closure reaction as a function of response time and concentration of contaminant. In C. fluminea, we estimated that copper concentrations >4 ␮g/l (95% confidence interval (CI95% ), 2.3–8.8 ␮g/l) must be detected within 5 h after Cu addition. Lower values could not be distinguished from background noise. The threshold values were 2.5 times lower than the values reported in the literature. © 2003 Elsevier B.V. All rights reserved. Keywords: Biosensor; Valve closure; Copper; Corbicula fluminea

1. Introduction The present work is part of a more general study aimed at estimating, by means of a series of laboratory studies, technical improvements and computation of confidence intervals, the limits and potential of bivalves to detect dissolved metals in the aquatic environment. It was performed on the Asiatic clam ∗ Corresponding author. Tel.: +33-4-42-25-27-87; fax: +33-4-42-25-27-87. E-mail address: [email protected] (D. Tran).

Corbicula fluminea and we studied its ability to detect dissolved copper. Copper is a metal that is widespread in aquatic systems. It is both natural and anthropic in origin. In natural freshwater systems, mean values of copper concentrations are variable and range from 0.04 to 294 ␮g/l (WHO, 1998; Neal and Robson, 2000; Mansour and Sidky, 2002; An and Kampbell, 2003) with extreme values up to 20 mg/l (Goodyear and McNeill, 1999). A study in Canada showed in drinking water, a concentration of Cu from 20 to 750 ␮g/l (WHO, 1998). In the last decades, numerous studies have been carried out to produce biosensors that

0166-445X/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.aquatox.2004.01.006

334

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

are able to quickly detect aquatic contamination by dissolved pollutants such as organic compounds (polychlorinated biphenyls, polycyclic aromatic compounds) or trace metals (Cd, Cu,. . . ). Some of these different systems used based on bivalve behaviour (Kramer et al., 1989; Ham and Peterson, 1994; Sluyts et al., 1996; Borcherding, 1997; Floch, 1994; Curtis et al., 2000; Markich et al., 2000), the basic idea being to use the bivalves’ ability to close its shell as an alarm signal when exposed to a contaminant. With the aim to improve the potential of these systems and the reaction threshold of various bivalve species, we recently reported a set of new developments using the freshwater bivalve C. fluminea (Tran et al., 2003). Three original features were introduced. First, a system of impedance valvometry was developed using lightweight impedance electrodes (20 mg each); this allows the bivalve to move freely with virtually no experimental constraints and artefactual stress, which strongly modify valve activity. Second, variations in the probability of spontaneous closing over the nycthemeral rhythm were taken into account by working during the period when the probability of spontaneous closing events was lowest. Both points were taken as a prerequisite to enhance our ability to approach absolute animal sensitivity. Third, a new analytical process was applied to compute the relationship between copper detection limit and response time together with the corresponding confidence intervals. It is suggested that this plural approach significantly optimizes (i) testing of the potential and the limitations of using bivalves as a rapid and/or sensitive biosensor for different contaminants as well as improving, (ii) the understanding of the effect of ventilatory changes on metal bioaccumulation (Tran et al., 2001, 2002), and (iii) the biology of the species.

2. Materials and methods 2.1. General conditions Experiments were carried out from May to June 2001 at 15 ± 0.5 ◦ C and pH = 8.2 ± 0.1 measured by a pH-CO2 stat (Consort R301). The water was air-equilibrated by bubbling, with a natural photoperiod (natural daylight, filtered through a blind). The water used was groundwater. Its ionic composition (in

meq/l) was as follows: NH4 + : 0.001; Na+ : 1.350; K+ : 0.092; Mg2+ : 0.327; Ca2+ : 0.470; HCO3 − +CO3 2− : 1.910; Cl− : 1.030; NO2 − : below limits; NO3 − : 0.002; SO4 2− : 0.073. In these experiments, 60 C. fluminea weighing 0.47 ± 0.01 g FW (mean ± 1 standard error, FW corresponding to the fresh weight of the animal without the shell) were studied. There were five bivalves per tank (36 cm × 15.5 cm, glass tank covered with alimentary plastic PLASTILUZ, France), containing 3 l of water in a flow-through system; they were fed continuously with the algae Scenedesmus subspicatus using a Gilson pump (Gilson, Villiers-Le-Bel, France). The substrate at the bottom was sand of quartz (SILAQ, France), which is a neutral substrate for the dissolved contaminants. Air was bubbled through, to ensure a uniform water column and distribution of trophic additions. The amounts of food continuously provided were adjusted in the range of 4–6 × 105 algae/ml, the concentration of algae was measured by optical density at 750 nm with a spectrophotometer (UV-1601 Shidmazu, Columbia, MD, USA). Within this concentration range, C. fluminea ventilatory activity is independent of algae concentration in normoxic water (Tran et al., 2000). 2.2. Impedance valvometry design 2.2.1. Description of the electrodes A total of 15 Corbicula were studied simultaneously for each set of experiments. Each animal was equipped with two light electrodes made of platinum (Ø: 3 mm; weight: 20 mg), fixed on both sides of the valve of the bivalve with cyanoacrylate glue (Loctite 431). The platinum disc was welded to a covered multistrand copper wire (Ø: 0.98 mm; length: 60 cm). The platinum wire weld and the external face of the electrodes were coated with resin (Altufix P10). After the electrodes were implanted, the animals were kept for at least 7 days in dim light conditions in isolated tanks equipped with an environment where they could move freely, dig, and hide. The free ends of the electrodes were connected to the measuring equipment. The animals with electrodes had at least a week to recover in the experimental apparatus before the study began. During the experiment itself, the animals were isolated from external vibrations in the laboratory using an anti-vibrating bench.

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

2.2.2. Experimental set-up The electrodes were linked to a laboratoryconstructed relay box, connected in turn to a laboratory-constructed multiplexer, which switched the current simultaneously to the different pairs of electrodes, and then to a laboratory-constructed impedance meter. This equipment was controlled by computer, via a data acquisition card (LAB PC 1200, National Instrument) using LabView software (National Instrument). With this equipment, the bivalves’ valve activity was recorded, and the signal processed. 2.2.3. Measurements The measuring system was based on the application of Ohm’s law, U = R × I (with the slight difference that resistance R is an impedance in alternating current). The HF impedance meter, the measuring apparatus, applies an alternating potential difference at high frequency (40 kHz) and constant voltage (500 mV) between the two electrodes. This apparatus measures current I which varies according to the distance between the electrodes. 2.3. Threshold sensitivity to copper The experiment was carried out on 12 groups of five C. fluminea. A total of five copper doses were tested. The initial dissolved Cu concentrations, added in a single step, in water were: 20, 50, 100, 200, and 500 ␮g/l. During the contamination period, the decrease of Cu concentration after 5 h of experiment was about 40%. As no organic matter was present in the sediment, the pore water in the quartz sand was assumed to be at most at equilibrium with the water column. These concentrations were obtained from a stock solution of Cu at 1 g/l (CuCl2 , Merck, Darmstadt, Germany). There were three replicates for [Cu]w = 20 and 50 ␮g/l, and two replicates for [Cu]w = 100, 200, and 500 ␮g/l (five animals per replicate). The experimental procedure consisted of recording valve activity over a period of 29 h: this consisted of 24 h prior to contamination, to define the valve behavior during reference state for C. fluminea, then the experimental units containing the animals were contaminated at t0 (tank of 3 l of water mixed by bubbling with normoxic air) with a single nominal addition of contaminant, in order to test the animal’s response to an instantaneous stimulus. Note that such a fast chang-

335

ing metal concentration from 0 to x ␮g/l represents in itself an extreme stimulation for any given contamination level. The total of valve reactions was calculated for various exposure periods, called integration times, ranging from 5 to 300 min to differentiate between fast and slow responses. The copper was always added at 12 noon. It is during this time that the percentage of natural closing in C. fluminea is minimal (Tran et al., 2003). 2.3.1. Calculating the sensitivity threshold This calculation was based on the detection of closing reactions, or more rarely, on a series of brief opening reactions followed by continuous closing. In our experimental conditions (long term acclimation periods, use of antivibrating benches), the animal’s reference status remains systematically in a very close range and these typical actions were never observed without contaminants (Tran et al., 2003). For each replicate of five bivalves, the number of events was counted, taking t0 as the moment of contamination. The bivalves’ response to the contamination was determined at different integration (exposure) time periods. First, the bivalves’ time response T, to react to a nominal copper concentration was calculated in minutes. T varies from 20 to 90%. For example, T50% , corresponds to the time for 50% of bivalves to react to a single concentration. To calculate the sensitivity threshold to copper, dose response models were performed and EC20% to EC80% were calculated, which corresponds to the concentration of contaminant that has an effect on 20–80% of the animal during the selected integration time. 2.4. Statistics Linear regression model based on least squares estimators is often used for quantifying the effect of several variables on one dependent continuous variable. Under usual parametric assumptions (normality and homoscedasticity of the error term ε), linear least squares regression estimator is the most efficient in the sense of minimum variance of all unbiased estimators. Unfortunately, when the method of least squares is applied to a model with binary response, parametric assumptions are not checked, and consequently estimator properties failed. So, for situations where the

336

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

dependent variable is qualitative, we have to use other statistical models such as logistic regression model given by 

p logit(p) := log 1−p

 = β0 + β1 log(x) + ε

where p(x) and 1 − p(x) denote the conditional probabilities of the response variable (equal to 1, 0, respectively) given x, β0 , and β1 are the unknown regression parameters estimated by maximum likelihood method. Hosmer and Lemeshow goodness-of-fit test was performed to check the significance of logistic models (Hosmer and Lemeshow, 2000). We are now using this model to estimate both the concentration and the time effects (denoted by ECp and Tp , respectively) and their (1 − α)% confidence intervals (α ∈ (0; 1)) and p ∈ (0%; 100%). From asymptotic normality of maximum likelihood estimators (Gourieroux and Monfort, 1981), it can be shown that 

 − log it(p) logit(p) Var(β0 ) + Var(β1 ) log2 (x) + 2cov(β0 , β1 ) log(x)

is asymptotically a standard normal distribution. And so, for every p value fixed between 0 and 1, we deduced that a (1 − α)% confidence interval of ECp and Tp , respectively is obtained finding the x values for which the equation

3. Results 3.1. Velocity of valve closure response Fig. 1 shows the percentage of response, characterised either by valve closure or by a series of brief opening reactions, followed by continuous closing, in Corbicula exposed abruptly to [Cu]w = 20, 50, 100, 200 and 500 ␮g/l. The response was modeled by a significant logistic regression. It is clear that the response was faster at higher [Cu]w and as all animals are closed, there is no more variation. Under controlled conditions ([Cu]w = 0 ␮g/l), 5.2% of the studied bivalves spontaneously closed their valves within 120 min, which is similar to 3.3% previously reported by Tran et al. (2003). Note importantly that it illustrates the stability and reproducibility of the reference Corbicula behavior, independent of time, in our laboratory conditions. In contrast, when [Cu]w = 500 ␮g/l was applied, 100% of the bivalves closed their valves within 10 min. The full description of the relationship between valve responses and time, using a logistic regression model, enables us to calculate the time T required to obtain from 20% (called T20% ) to 80% (T80% ) of responding animals in the studied sample (Table 1). For each T value, the 95% confidence interval (CI95% ) was computed and for each Cu concentration, the estimators of logistic regression parameters are given with standard error (S.E.). In reference conditions, ([Cu]w = 0 ␮g/l, i.e. during the 24 h period before Cu addition) T50% was 906 min, which

     0 )/β1  Z1−α/2  log(x) − (logit(p) − β ≤ Pr   =1−α   β1  2 Var(β0 ) + Var(β1 ) log (x) + 2cov(β0 , β1 ) log(x)  is verified since (logit(p) − β0 )/β1 denotes ECp or Tp estimators according to the x values (x being either the concentration or the time, respectively). Z1−α/2 is the (1 − α/2)-quantile of a standard normal distribution. All the computations were carried out with the SPLUS software (Splus Version 5, Venables and Ripley, 1999). Note that logistic regression model is often used in analysing dose-response data (Kerr and Meador, 1996; Hugues et al., 2001; Dongbin et al., 2002).

means that 50% of the studied individuals exhibited a valve closing reaction for 906 min. At the maximum test concentration of 500 ␮g/l, T50% was 4.6 min. For an ecologically relevant dissolved copper, [Cu]w = 20 ␮g/l, T20% = 12.9 min (CI95% , 5.7–20.1 min), T50% = 35.7 min (CI95% , 23.5–53.7 min) and T80% = 98.5 min (CI95% , 66.1–175.6 min). Thus, these first set of values give a first insight into the minimum delays, and their variability, required for a population

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

337

Fig. 1. Dose response curves of the percentage of valve response as a function of time, described by a logistic regression model, at different Cu concentrations (0, 20, 50, 100, 200, and 500 ␮g/l). Each value is a mean (n = 10–15, which correspond to two or three replicates of five bivalves) with standard error.

of Corbicula at 15 ◦ C to detect a minute change of [Cu]w from 0 to 20–500 ␮g/l. We then switch to data analysis aimed at determining the lowest detectable [Cu]w and its dependence on exposure period or integration time, as evidently a fast response can only be observed with the highest concentrations. 3.1.1. Looking for minimal detectable [Cu]w = f(t) In follow-up analyses, the EC values (effective concentration values) were calculated. They are the [Cu]w required to induce valve closure reactions for various exposure periods. To calculate these thresholds, for each [Cu]w increment applied at t0 , the number of

valve responses observed for various times ranging from 5 to 300 min was determined. For each integration time, the percentage of valve responses was represented as a function of five [Cu]w values studied (20, 50, 100, 200, 500 ␮g/l with 0 as the reference period). Data results and logistic significant models are presented in Fig. 2, which shows first that the velocity at which the dose response reached 100% increased with [Cu]w . Second, it shows that sensitivity was strongly dependent on integration time. Table 2 presents some selected effect concentrations, in ␮g/l, from EC20% to EC80% with their confidence intervals for various integration times. These models are significant for all integration times studied (P < 0.05). It

338

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

Table 1 Response time for 20, 50 and 80% of total valve closure response (T20% , T50% and T80% in minutes) of Corbicula (n = 10–15 depending on the number of replicates, see Section 2) exposed to various Cu concentrations ranging from 20 to 200 ␮g/l [Cu]w (␮g/l)

β0

β1

ρβ0 ,β1

S.E. −4.879 1.009

0.271

50

−1.994

0.843

0.697

0.206

−2.451

1.880

1.593

0.706

−3.550 2.514

2.764 0.038

200

T50%

T80%

35.7

98.5

5.7–20.1

23.5–53.7

66.1–175.6

2.1

10.6

55.1

CI95%

20

100

T20%

1.365

−0.965 −0.946

12.9

0.2–5.1 −0.948 −0.966

3.8–18.3

32.1–153.3

1.8

3.7

7.7

0–3.9

0.3–6.5

3.1–15.2

2.2 0–4.1

3.6 0–5.8

6.0 1.0–14.4

T and its 95% confidence interval, were computed by a logistic regression model. Maximum likelihood estimators of β0 and β1 with standard error and correlation coefficients between the two parameters (ρβ0 ,β1 ) are presented.

is worthwhile noticing that for an integration time of 10 min, EC50% = 52.5 ␮g/l (CI95% , 29.6–85.0 ␮g/l) but only 4.2 ␮g/l (CI95% , 2.3–8.8 ␮g/l), if responses were integrated over a period of 300 min, which represents a mean 12.5-fold increase in sensitivity.

3.1.2. Global relationship between sensitivity threshold and integration time To get a better insight into the ability of Corbicula to detect a change in copper concentration, and the error associated with their use in valvometer devices, EC50% was next plotted as a function of time

Table 2 Effect concentrations EC20% , EC50% and EC80% (␮g/l) causing 20, 50, and 80% of total valve closure response of Corbicula Time (min)

β0

β1

ρβ0 ,β1

S.E. 10

15

30

60

120

300

EC20%

EC50%

EC80%

52.5

111.5

CI95%

−7.288

1.840

2.196

0.532

−8.168

2.201

2.625

0.669

−6.173

2.015

2.150

0.609

−3.157

1.322

0.732

0.241

−2.649

1.268

0.587

0.221

−1.815 0.437

1.271 0.243

−0.983

24.7 6.9–39.9

−0.985

21.8 5.9–34.2

−0.975

10.7 4.0–18.7

−0.812 −0.730 −0.510

3.8

29.6–85.0

71.4–224.2

40.9

76.8

22.3–63.3

51.2–194.3

21.4

42.6

8.9–34.2 10.9

26.5–95.7 31.1

1.4–7.4

5.5–21.2

16.3–78.6

2.7

8.1

24.1

1.1–5.1

4.2–16.0

12.6–64.3

1.4 0.7–0.25

4.2 2.3–8.8

12.4 6.3–41.2

Ten to 15 Corbicula were exposed to various Cu concentrations ranging from 20 to 500 ␮g/l (depending on the number of replicates, see Section 2). EC20–80% and their confidence intervals, were computed by logistic regression model. Maximum likelihood estimators of β0 and β1 with standard error and the correlation coefficient between the two parameters ρβ0 ,β1 are presented.

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

339

Fig. 2. Dose response curves of the percentage of valve response as a function of Cu concentration, described by a logistic regression model, at different integrated times (10, 15, 30, 60, 120, and 300 min). Each value is a mean (n = 10–15, which correspond to two or three replicates of five bivalves) with standard error.

(Fig. 3). This shows that the distribution of points is in close agreement with an exponential model EC50% = 3.76 exp(91.91/(23.99 + time)), which accounted for 99.6% of total variability. It is important to note that the associated error decreased with time. Note also that detection times shorter than 45 min were associated only with concentrations ≥15 ␮g/l. This illustrates how sensitivity to Cu is associated with time. 4. Discussion The aim of the present study was to optimize testing of the potential and limitations of C. fluminea for

detecting acute [Cu]w changes. One of the objectives is using them as a rapid and/or sensitive biosensor for Cu. Importantly, present results are associated with confidence intervals to characterize precision of effect concentration estimators. The minimal Cu concentration able to induce a valve reaction was characterized by either a prolonged closure or a series of rapid closings and openings. By comparison with previously published data, we report here, valve responses as a function of either [Cu]w or response time. Indeed, fast responses can only be elicited by relatively high metal concentrations, while lower concentrations require much longer exposure periods in order to be de-

340

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

Fig. 3. Relationship between the threshold sensitivity of Corbicula to Cu (EC50% ) and the response time. The response is described by a non-linear regression, y = a exp(b/(x+c)), where y is EC50% and x is the response time. Each value is a calculated value (see Fig. 2) with confidence interval.

tected. It is estimated that the lowest detectable copper concentration, expressed as EC50% (time required to induce a reaction in 50% of the studied animals) is 4.2 ␮g/l (CI95% , 2.3–8.8 ␮g/l, Table 2) if the exposure period is 300 min. As they stand, present data can be used to predict the optimal potential of the species as a biosensor aimed at detecting Cu in valvometer devices. Note finally that, our computation overestimated the actual values, as we did not take into account the effective Cu concentration but the nominal value. 4.1. Critical analysis of the method In the literature, the methods used to determine a threshold concentration for any contaminant able to induce a significant change in valve behaviour vary considerably (Sloff et al., 1983; Mouabad and Pihan, 1993; Sluyts et al., 1996; Curtis et al., 2000). The first difference is the valvometric technique as all of them are using the bivalve’s ability to close their shell when exposed to a contaminant. Several valvometers were developed, such as Mossel Monitor (Kramer et al., 1989), Dreissena Monitor (Borcherding, 1997) or the Ifremer–Micrel valvometer (Floch, 1994). In these ap-

paratus, the bivalve is fixed by one of its shell, and cannot move freely. We showed in C. fluminea that, this could induce profound modifications of the natural valve behavior (Tran et al., 2003). The present non-invasive valvometric technique was developed to avoid these experimental artefacts and reduce the probability of abnormal closings. It evidently increases the measurable sensitivity of the tested animals as it reduces the background noise. The second common difference is the parameter chosen to diagnose a significant change in valve behaviour. Some authors used the frequency of valve movement activity (Sluyts et al., 1996), while others used the change in valve opening. In this latter category, some measured a change expressed in percentage (Sluyts et al., 1996; Curtis et al., 2000), while others used the duration of valve opening (Markich et al., 2000). Yet another group takes into account, as we do, the first occurrence of abnormal primary reactions after contamination, which is obviously the key parameter to measure response delays. Beyond data analysis problems, it is suggested that the ecophysiological approach developed in this study (use of anti-vibrating tables to reduce the occurrence of false responses and of lightweight impedance electrodes; study performed during the period when the probability of spontaneous closing events was lowest) brought significant improvements to the method of testing the potential and drew attention to the intrinsic limitations of using bivalves as a rapid response and/or sensitive biosensor. 4.2. Comparison with previous data Since copper is a biologically essential metal, various studies on copper contamination had already been published for bivalves. There are, for example, data on bioaccumulation (Davenport, 1977) and descriptions of a decrease in ventilatory activity and oxygen consumption in the presence of Cu in the blue mussel Mytilus edulis (Davenport and Manley, 1978; Manley, 1983). It is only in the past 20 years that, the technology of biosensors that use bivalves in a valvometer has been developed to detect water contamination. Different works on marine species are available. Bouget and Mazurié (1997) determined a sensitivity threshold of 25 ␮g/l in M. edulis. For Crassostrea gigas, contaminated with an aqueous solution of CuSO4 and CuCl2 , these authors reported threshold values of

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

40 and 60 ␮g/l, respectively. Regarding freshwater bivalves, to our knowledge, no data are available for C. fluminea. Existing results refer mainly to Dreissena polymorpha. Kramer et al. (1989) and Sloff et al. (1983) have found a detection threshold 10 and 30 ␮g/l of CuSO4 , respectively. Sluyts et al. (1996) showed that at 20 and 40 ␮g/l of CuSO4 , D. polymorpha significantly changed its valve opening status and at 80 ␮g/l, the bivalve changed both its valve opening amplitude and valve frequency movement. Unfortunately, in none of these studies, there were precise indication about the exposure times required to obtain a response to copper. It is only in a recent work, performed on the blue mussel M. edulis, that an EC50% = 135 ␮g/l of CuCl2 was determined for an integration time of 48 h (Curtis et al., 2000). However, note that this EC50% was again not directly comparable to the present work, as it corresponded to a reduction of 50% in the valve aperture and not the appearance of a first abnormal valve activity. The optimal EC50% of [Cu]w = 4 ␮g/l (CI95% , 2.3–8.8 ␮g/l) we report for C. fluminea is at least 2.5 times higher than the best sensitivity found in the literature, (10 ␮g/l, obtained with CuSO4 , Kramer et al., 1989) and 30 times, if the comparison is restricted to the use of CuCl2 (Curtis et al., 2000). 4.3. Using the relationship between [Cu]w and response time in biocaptors Modeling of the dose response curve by logistic regression according to the integrated time of signal analysis and the reaction fraction, allows a large spectrum of potential, in terms of early detection in water pollution. At least two new possibilities open up, in terms of metal threshold and probability of false response. First, the ratio between the reliability of response and rapidity can be adjusted. The user can choose a response time according to an expected sensitivity threshold, knowing that minimum detectable [Cu]w increases with integration response time i.e. length of the exposure period. Second, the effect concentration, can apply from 20 to 99% of the total reaction in the studied test population. Obviously, EC99% , the concentration at which 99% of clam reactions is observed, is associated with a more precise detection of contamination but a longer exposure period. If, on the other hand, pollution detection must be quicker,

341

an EC value ranging from EC20% to EC50% can be taken. 4.4. Response mechanism Fig. 2 shows the velocity at which a sample population of C. fluminea reacted to copper according to the exposure time. This is a statistical approach, which gives some insights into the detection mechanism developed by the species and the nature of the chemosensory system involved. As this relationship is described by a single equation in Fig. 2, this suggests the existence of a single mechanism whose activity is dose dependent, at least during 1 h of exposure. This type of reaction is different from what was observed for cadmium detection with C. fluminea (Tran et al., 2003). Indeed, the Cd curve was characterized by a “breakdown point” at 50 ␮g/l and 65 min, which strongly suggested a different type of change of physiological response mechanism in cadmium. First, there was a rapid detection mechanism in response to concentration >50 ␮g/l. Next, a second mechanism was involved, with longer response times, in the detection of concentrations <50 ␮g/l. It is hypothesised that either there were two types of receptors (e.g. external and internal), or a single type functioning with two mechanisms at different thresholds. In all cases, for cadmium detection capacities by Corbicula, it was clear that the mechanism with the low detection threshold was not triggered rapidly. A rapid response, occurring within <1 h, can only be obtained if one is dealing with Cd concentrations >45 ␮g/l. For Cu, it is clear that from 1 to 5 h, the sensitivity threshold was largely time independent. Waiting for more than 1 h is thus of least interest.

5. Conclusion In this study, the application of the valvometer by impedance was extended from the detection of cadmium (Tran et al., 2003) to the detection of copper. An original protocol and technique for the analysis of the response of Corbicula to copper allowed us to determine an optimal sensitivity threshold which significantly improved the copper sensitivity threshold previously reported in the literature.

342

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343

The methodology developed here with this biosensor, which uses (i) free-ranging (i.e. unconstrained) bivalves, (ii) takes into account the valve’s natural closing/opening rhythm (equivalent to background noise, described in Tran et al., 2003) and (iii) incorporates a mathematical description, using an analysis of the “curve dose-response” by logistic regression, which integrates time and confidence intervals, is very sensitive to detecting acute water contaminations. The present experimental protocol could be applied to other bivalves, either freshwater or marine origin, such as mussels and/or oysters, which are used as sentinel organisms. As it stands, it presents new data about species reaction to Cu contamination, but it is also certainly helpful in determining whether valvometry could be efficiently useful in the field, and its theoretical limits.

Acknowledgements The authors thank P. CIRET who developed all electronic aspects and analytical tools used in the present work. All the experiments presented in this paper comply with the current laws of France, where they were performed.

References An, Y.-J., Kampbell, D.H., 2003. Total, dissolved, and bioavailable metals at lake Texoma marinas. Environ. Pollut. 122, 253–259. Borcherding, J., 1997. The “Dreissena-monitor”—improved evaluation of dynamic limits for the establishement of alarm-threshold during toxicity tests and for continuous water control. In: Hill, I.R. (Ed.), Freshwater Field Tests for Hazard Assessment of Chemicals. Lewis Publishers, Boca Raton pp. 477–484. Bouget, J.-F., Mazurié, J., 1997. Dispositif de surveillance biologique de la qualité d’eau d’un site conchylicole estuarien utilisant un biocapteur valvaire muni d’huˆıtres et de moules. Tech. Sci. Méthodes 11, 71–79. Curtis, T.M., Williamson, R., Depledge, M.H., 2000. Simultaneous, long-term monitoring of valve and cardiac activity in the blue mussel Mytilus edulis exposed to copper. Mar. Biol. 136, 837– 846. Davenport, J., 1977. A study of the effects of copper applied continuously and discontinuously to specimens of Mytilus edulis (L.) exposed to steady and fluctuating salinity levels. J. Mar. Biol. Assoc. UK 57, 63–74.

Davenport, J., Manley, A.R., 1978. The detection of heightened sea-water copper concentrations by the mussel Mytilus edulis. J. Mar. Biol. Assoc. UK 58, 843–850. Dongbin, W., Aiqian, Z., Zhongbo, W., Shuokui, H., Liansheng, W., 2002. A case of logistic QSAR modelling methods and robustness tests. Ecotoxicol. Environ. Saf. 52, 143–149. Floch, A., 1994. Un capteur biologique original enregistrant les mouvements valvaires des mollusques bivalves. PhD Thesis, Ecole Vétérinaire de Nantes. pp 165. Goodyear, K.L., McNeill, S., 1999. Bioaccumulation of heavy metals by aquatic macro-invertebrates of feeding guilds: a review. Sci. Total Environ. 229, 1–19. Gourieroux, C., Monfort, A., 1981. Asymptotic properties of the maximum likehood estimator in dichotomous logistic regression models. J. Econometrics 17, 83–97. Ham, K.D., Peterson, M.J., 1994. Effect of fluctuating low-level chlorine concentration on valve movement behaviour of the Asiatic clam (Corbicula fluminea). Environ. Toxicol. Chem. 13, 493–498. Hosmer, D.W., Lemeshow, S., 2000. Applied Logistic Regression, second ed. Wiley, New York. Hugues, M.R., Bailer, A.J., Denton, D.L., 2001. Toxicant-and response-specific comparisons of statistical methods for estimating effective concentrations. Environ. Contam. Toxicol. 20, 1374–1380. Kerr, D., Meador, J., 1996. Modeling dose response using generalized linear models. Environ. Toxicol. Chem. 15 (3), 395–401. Kramer, K.J.M., Jenner, H.A., De Zwart, D., 1989. The valve movement response of mussels: a tool in biological monitoring. Hydrobiologia 188–189, 433–443. Manley, A.R., 1983. The effects of copper on the behaviour, respiration, filtration and ventilation activity of Mytilus edulis. J. Mar. Biol. Assoc. U.K. 63, 205–222. Mansour, S.A., Sidky, M.M., 2002. Ecotoxicological studies. 3. Heavy metals contaminating water and fish from Fayoum Governorate (Egypt). Food Chem. 78, 15–22. Markich, S.J., Brown, P.L., Jeffree, R.A., Lim, R.P., 2000. Valve movement response of Velesunio angasi (Bivalvia: Hyriidae) to manganese and uranium: an exception to the free ion activity model. Aquat. Toxicol. 51, 155–175. Mouabad, A., Pihan, J.-C., 1993. Le test comportemental de Dreissena polymorpha: un outil biologique de prevision et d’évaluation de la toxicité en milieu d’eau douce. Hydroécol. Appl. 51, 97–109. Neal, C., Robson, A.J., 2000. A summary of river water quality data collected within the land–ocean interaction study: core data for eastern UK rivers draining to the north sea. Sci. Total Environ. 251/252, 585–665. Sloff, W., De Zwart, D., Marquenie, J.M., 1983. Detection limits of a biological monitoring system for chemical water pollution based on mussel activity. Bull. Environ. Contam. Toxicol. 30, 400–405. Sluyts, H., Van Hoof, F., Cornet, A., Paulussen, P., 1996. A dynamic new alarm system for use in biological early warning systems. Environ. Toxicol. Chem. 15 (8), 1317–1323.

D. Tran et al. / Aquatic Toxicology 66 (2004) 333–343 Tran, D., Boudou, A., Massabuau, J.-C., 2000. Mechanism of oxygen consumption maintenance under varying levels of oxygenation in the freshwater clam Corbicula fluminea. Can. J. Zoolog. 78, 2027–2036. Tran, D., Boudou, A., Massabuau, J.-C., 2001. How water oxygenation level influences cadmium accumulation pattern in the Asiatic clam Corbicula fluminea. Environ. Toxicol. Chem. 20, 2073–2080. Tran, D., Boudou, A., Massabuau, J.-C., 2002. Relatioship between feeding-induced ventilatory activity and bioaccumulation of

343

dissolved and algal-bound cadmium in the Asiatic clam Corbicula fluminea. Environ. Toxicol. Chem. 21, 327–333. Tran, D., Ciret, P., Ciutat, A., Durrieu, G., Massabuau, J.-C., 2003. Estimation of potential and limits of bivalve closure response to detect contaminants: application to cadmium. Environ. Toxicol. Chem. 22, 116–122. Venables, W.N., Ripley, B.D., 1999. Modern Applied Statistic with S-Plus, third ed. Springer-Verlag, New York. WHO, 1998. Environmental health criteria. Copper. WHO, Geneva.