Effect of declining toxicant concentrations on algal bioassay endpoints

Abstract

Sorption, degradation, volatilization, and uptake by test organisms cause concentrations of many toxicants to decline during toxicity testing. Despite the recognition of this occurring, nominal, measured initial or time‐averaged concentrations are commonly used for the calculation of inhibitory or effect concentrations from toxicity test data. Because a premise of constant exposure is assumed but not met in these calculations, the toxicity of the test water may be significantly underestimated. Laboratory experiments using a 72‐h algal growth inhibition bioassay and copper as a toxicant are used to demonstrate that calculated inhibitory concentrations will often be underestimated twofold if losses of copper occurring over the 72‐h test duration are not considered. A simple model is presented for toxicant concentration decline and for the relationship between algal growth rate and toxicant concentration. This model is used to demonstrate the magnitude that calculated inhibitory concentrations may be underestimated if concentration declines are not considered for a series of different concentration decline scenarios. For a toxicant whose concentration declines exponentially to less than 5% of its original value within 36 h of a 72‐h test, the inhibitory concentration is shown to be underestimated by a factor of 50.

INTRODUCTION

In conducting toxicity tests on waters, the assumption is that the toxicant concentration of the collected sample is representative of the toxicant concentration at the collection site. It is well recognized, however, that over the duration of most toxicity tests, toxicant concentrations decline or the speciation of toxicants changes from that in the field [1–6]. Sorption, degradation, volatilization, and uptake by test organisms are common factors that cause concentrations to decline during testing [3,5].

Although presilanized borosilicate glass flasks or plastic test containers and short test durations are commonly used for minimizing adsorptive losses and degradation, significant losses are still common during water testing [6]. In algal bioassays, the use of lower initial cell densities has been shown to considerably reduce copper binding to algal surfaces and exudates and consequently to reduce copper losses from solution [7,8]. Flow‐through systems and water renewal have been used in some bioassays to better maintain toxicant concentrations; however, these techniques are difficult for algal bioassays, as cells must be filtered/centrifuged before replacement of solutions, and this can lead to a poor growth rate in the controls [3,9]. Continuous algal culture systems have been developed (e.g., chemostats); however, these techniques are laborious, and replication is difficult for use in statistically robust toxicity testing [3].

Despite the depletion of toxicant concentrations during test periods, it is still common practice to utilize either nominal, measured initial, or time‐averaged concentrations for the calculation of effect or inhibitory concentrations (EC50, IC50). As the premise of constant exposure is assumed, but often not met, in these calculations, the toxicity of the test water may be underestimated. This underestimation will be greatest for nonpersistent chemicals such as polycyclic aromatic hydrocarbons, chlorocatechols, where losses from solution occur rapidly.

To mathematically account for the effect that declining or fluctuating toxicant concentrations have on IC derived from toxicity test data, numerous approaches have been proposed [10–13]. The use of parametric [14] and toxicokinetic [12] models have been the more recent approaches. Extensions to the DEBtox model, which does not require biological assumptions about toxicokinetics and toxic effects, have also been shown to be applicable for these calculations [13,15]. None of these approaches have been adopted for routine use; this may be related to the complexity of the mathematics underlying these models.

This study investigates the effect of declining copper concentrations on algal growth inhibition bioassays and the calculated IC. The results from a standard bioassay procedure are compared with those from a new procedure where the decreases in toxicant concentration during testing are minimized. Simple models were developed for toxicant concentration decline and for the relationship between algal growth rate and toxicant concentration. Using a model developed for the bioassay experiments with copper as the toxicant, a series of concentration decline scenarios are modeled to illustrate the underestimation of calculated IC that would occur if the decline is not accounted for in IC50, IC25, etc., calculations. Suggestions are made as to how, during routine testing, errors associated with the toxicity testing of waters that contain toxicants whose concentrations decline significantly during test periods can be minimized.

MATERIALS AND METHODS

General methods and reagents

All glass and plasticware was cleaned by soaking in 10% (v/v) HNO₃ (Trace Pur, Merck, Darmstadt, Germany) for >24 h, followed by at least five rinses with deionized water (Milli‐Q, Millipore®, Sydney, Australia). All laboratory ware used for dissolved metals sampling and analysis were cleaned in a Class‐100 laminar flow cabinet (metal‐free HWS, Clyde‐Apac, Melbourne, Australia). High‐purity (18 M·cm⁻¹) Milli‐Q water was used to prepare all solutions. Clean seawater was collected from Cronulla, New South Wales, Australia, filtered immediately through a 0.45 μm filter (Sartobran‐P capsule with 0.65 μm prefilter; Sartorius Australia, Victoria, Australia), and stored at 4°C. Dilute copper standard solutions were prepared from a 1,000 mg/L stock solution of CuSO₄.7H₂O. All chemicals were analytical reagent grade or equivalent analytical purity.

Measurements of pH (calibrated against National Institute of Standards and Technology certified buffers, Orion Pacific, Sydney, Australia) were done with a pH meter (pH 320, WTW, Weilheim, Germany) equipped with a combination pH (Sure‐flow 9165BN, Thermo Orion, Beverley, MA, USA) probe. Salinity and temperature measurements used a conductivity meter (LF 320, WTW) with a probe (TetraCon 325, WTW). Dissolved oxygen measurements were undertaken using a meter (OXI 196, WTW) with an oxygen electrode (EO96, WTW) calibrated according to manufacturer's instructions.

Water samples for dissolved copper analysis were filtered into acid‐washed polyethylene vials using acid‐washed 0.45 μm membrane filters (Minisart, Sartorius) and acidified with HNO₃to 0.5% (v/v). Dissolved copper concentrations were determined by inductively coupled plasma atomic emission spectrometry (ICP‐AES; Spectroflame EOP, Spectro Analytical Instruments, Kleve, Germany) and by anodic stripping voltametry (ASV; 646 Trace Analyser, Metrohm, Hersiau, Switzerland). The ASV analysis was conducted in acid‐washed Teflon voltametric cells with deposition for 180 s at −600 mV versus Ag/AgCl. The ICP‐AES and ASV techniques had detection limits of 3 μg/L and 0.5 μg/L, respectively. Samples with copper concentrations < 10 μg/L were measured by ASV, whereas other samples were measured by ICP‐AES. Both instruments were calibrated with matrix‐matched standards and blanks and spike‐recovery tests run daily to monitor quality control.

Algal bioassays

The algal bioassay measured the decrease in growth rate and cell yield of the marine unicellular alga Phaeodactylum tricornutum in a 72‐h exposure. Phaeodactylum tricornutum was chosen for the study because it has previously been shown to be sensitive to copper [16], easy to count, and does not clump or adsorb to the walls of the test containers. The bioassay protocol [17] was based on the Organisation for Economic Cooperation and Development Guideline 201 [18].

Phaeodactylum tricornutum was obtained from the CNR Istituto di Biofisica, Pisa, Italy, and cultured in half‐strength f medium [19]. Cultures were maintained on a 12:12‐h light: dark cycle (Philips TL 40 W fluorescence daylight, 72 μmol photons/m/s) at 21°C. Cell density measurements were made using a particle analyzer with a 70‐ μm aperture (Coulter Multisizer II, Beckman Coulter, Fullerton, CA, USA) with correction for the background particle count.

Test results were considered acceptable if the mean growth rate for the controls was in the range 1.5 ± 0.5 doublings per day. Growth rate (cell division rate) and cell yield (N, biomass) endpoints were calculated according to standard methods [17]. Inhibitory concentrations (e.g., 72‐h IC50 values and their 95% confidence limits) were calculated by linear interpolation using TOXCALC 5 [20]. Data were tested for normality and homogeneity of variance using the Shapiro Wilk's test. Dunnett's multiple comparison test was then used to determine which concentrations were significantly different to the controls. The IC values were tested for significant difference using the method described in Sprague and Fogels [21].

Standard toxicity test procedure

Exponentially growing cells of P. tricornutum from a 4‐ to 6‐d‐old stock culture were used in the bioassays. For each test water, 55 ml of filtered seawater was dispensed into 250 ml glass Erlenmeyer flasks that had been presilanized with Coatasil® (Asia Pacific Specialty Chemicals, Sydney, Australia; BDH) to reduce metal adsorption to the glass walls. To all flasks, including controls, 0.55 ml of 2.1 g/L NaNO₃ solution and 0.55 ml of 0.22 g/L KH₂PO₄ solution were added as nutrients. Copper from a dilute standard copper sulfate solution was added to flasks to give ionic copper concentrations ranging from 0 to 50 μg/L. Following mixing, subsamples were taken from each flask for dissolved copper measurements before addition of the algal cells. Each flask was inoculated with 2 to 4 × 10⁴ cells/ml of a prewashed P. tricornutum suspension and incubated at 21°C on a 12:12‐h light:dark cycle at 140 μmol photons®m⁻²s⁻¹. Three test flasks were prepared for each copper concentration. At the beginning and end of each experiment, the pH of the test solutions was measured in one of the three replicates. Sub‐samples were taken from each flask at 0, 24, 48, and 72 h for cell density and dissolved copper measurements.

Tank bioassay procedure

Tank bioassays were carried out at the same time as the standard bioassays, matching the algal inoculum and light/temperature conditions as closely as possible. Polycarbonate tanks (27 L, 39 × 29 × 24 cm³) with removable, nonairtight lids were used for bulk test solutions, and polystyrene vials (80 ml) were used as test containers. Each vial had a 40‐mm diameter hole cut in the lid with a 47‐mm diameter membrane filter (5 μm, Millipore) secured when the lid was fitted. For 16 h before test commencement, the membranes and test vials were conditioned by soaking in solutions identical to the test solutions. To all tanks, 23 L of clean seawater was added, together with 23 ml of 21 g/L NaNO₃ solution and 23 ml of 2.2 g/L KH₂PO₄ solution. Copper was added to the tanks 1 h before tests commenced using a 5 mg/L standard solution to give ionic copper concentrations ranging from 0 to 50 μg/L. Once mixed, 40 ml of the freshly prepared tank mixture was added to each of the test vials after discarding the conditioning solution, and subsamples were taken from each tank for dissolved copper measurements before addition of the algal cells. One tank containing 12 replicate vials (pseudo replicates) was prepared for each copper concentration. Each vial was inoculated with 2 to 4 × 10⁴ cells/ml of a prewashed P. tricornutum suspension and incubated in the tank at 21°C on a 12:12‐h light:dark cycle at 125 μmol photons®m⁻²s Three replicate vials were taken from each tank at 0, 24, 48, and 72 h for cell density, dissolved copper, and pH measurements. Volumes in the test containers, measured at the end of the bioassays, were within 1% of the initial volume; therefore no correction for volume was made to the cell counts.

Modeling bioassay response with declining toxicant concentrations

The aim of modeling the bioassay response to waters with declining toxicant concentrations was to better estimate the true IC of the toxicants (e.g., 72‐h IC50 values). Toxicant concentration decline may occur through a number of mechanisms during a test period. Often concentration decline profiles include an initial, apparently instantaneous, decline followed by a decline that resembles anything from linear to exponential behavior. The rate of decline may also depend on the initial toxicant concentration and the duration of the test. To describe these concentration decline scenarios, the following simple mathematical model was used:

where Ct is the toxicant concentration at time t, C₀ is the initial toxicant concentration (t = 0), and t_max is the test period (72 h). The decline function allowed individually or additively for C₀ to decline instantaneously by fraction D₁, where the amount of decline changes linearly (D_1a) or exponentially (D_1b) with the initial concentration; C₀ to decline linearly at rate D₂; C₀ to decline exponentially at rate D₃ with the decline rate changing with time; and fraction D₄ of C₀ to decline exponentially at rate D_4a.

Relationship between algal growth‐rate parameter and toxicant concentration

Most algal bioassays measure the decrease in growth rate (cell division rate, μ) or the final cell biomass (cell yield, N) after a 48‐, 72‐, or 96‐h exposure to test waters. In the absence of a toxicant, and under constant light, temperature, and nutrient conditions (assuming no lag phase), algal growth is exponential [22] and the growth rate (μ_t) is constant. A plot of log cell density (log N) versus time is linear with a slope equal to μ_t (Eqn. 2).

where N_t is the number of cells (cell density) at time t, N₀ is the number of cells at t₀, and μ_t is the growth rate at time t.

Algal growth rate, μ_t, decreases with increasing toxicant concentration and was modeled using a four‐parameter logistic model [23] according to

where μ_t is the growth rate at time t; C_t is the toxicant concentration at time t; and α, β, δ, and Φ are constants chosen to vary the effect of C_t on μ_t.

In the model, C_t, and subsequently μ_t, were calculated at hourly intervals over the 72‐h test period, and algal cell densities were calculated according to Equation 4.

where N_t is the number of cells at any given time t, N_t‐1 is the number of cells at time t −1 (t −1 < t), and μ_t is the growth rate at time t (varying due to changing concentration, C_t).

Modeling the effect of toxicant concentration on algal growth

For the algal bioassay experiments with copper as the toxicant, a suitable fit to the measured copper concentration decline (from the standard bioassays) was provided by a model comprising an initial concentration drop followed by an exponential concentration decline at a rate that decreased with time (Eqn. 1). Using this concentration decline model, a model for the relationship between the algal growth‐rate parameter, μ_t, and copper concentration was then developed using the measured cell densities (from the standard bioassays) and Equations 3 and 4. The fit between the model and the measured data for copper concentrations and cell densities was not optimized further than visual inspection of the fit between the measured data and the model output. For the purpose of this study, this optimization procedure was sufficient.

To estimate true IC, or those IC that would be measured if no decline in toxicant concentration occurred during the test period, the developed model was used; however, copper concentrations were not allowed to decline from their initial concentration during the test period. (C_t = C₀ from Eqn. 1, and the growth rate parameter, μ_t, was constant). The model cell yield data and initial copper concentrations were then used to calculate true IC (IC50 values) using ToxCalc.

To extend the model to calculate IC for other scenarios of declining toxicant concentrations, a series of toxicant concentration decline scenarios were created using Equation 1, using appropriate combinations of the decay functions involving constants D₁ to D₄. Cell yield data were then calculated using Equations 3 and 4 using the algal growth‐rate parameters optimized from the copper bioassay experiments and the concentrations calculated for the specific decline.

RESULTS AND DISCUSSION

Copper‐sensitive algal bioassays

To experimentally examine the effect that declining copper concentrations have on the IC predicted from the algal bioassays, results from a standard bioassay method were compared with results from a more complex tank bioassay method designed to minimize copper losses. In the tank bioassay, equilibrium between a large external volume of test water and the water in test containers (containing the algae) was maintained through dialysis. This process enabled resupply of copper lost through adsorption to the container and algal surfaces.

Although the tank bioassay procedure did not fully eliminate the decline in copper concentration during the tests, the decline was reduced significantly relative to the standard bioassay. In Figure 1, mean concentration declines are shown for the two procedures. For both procedures, there was a rapid decline in copper concentration immediately following addition of test solutions, but before addition of the algae, which was attributed to adsorption to container surfaces. This amounted to a 15% loss at the highest concentration (40 μg/ L) in the standard bioassay and a loss of 11% in the tank bioassay, despite the fact that the containers had been pre‐equilibrated. Following the algal inoculation, the copper decline was greater for the standard bioassay procedure compared with tank bioassays, where “resupply” was possible through the 0.45‐μm membrane (Fig. 1). At low copper concentrations, losses of copper in the standard and tank bioassays were not significantly different (p > 0.05), indicating that the decline was dependent on the initial copper concentration.

In Figure 2, mean growth rates (expressed as percentage of control) are shown for the two algal bioassay methods for both 48‐ and 72‐h exposures. Algal growth decreased by greater amounts in the tank bioassays than in the standard bioassays at all copper concentrations for the 72‐h bioassay, although the differences were more significant for the lower copper concentrations. This is consistent with the higher copper concentrations measured during the tank bioassays (less decline in copper concentration, Fig. 1). For the 24‐h bioassays, the differences between the bioassay techniques were less significant.

Test endpoints (IC50, IC25, IC15) and 95% confidence intervals calculated using mean growth rates (Table 1) and mean cell yields (Table 2) are shown for the two algal bioassay methods using 0‐ to 48‐h and 0‐ to 72‐h data. The copper concentrations used in these calculations were those measured initially, prior to the algal inoculation, but after the initial rapid adsorption losses. Due to the greater decline in copper concentrations, IC values calculated from the standard bioassay data were greater than those of the tank bioassays, regardless of the calculation procedure used (growth rate or cell yield endpoints). Decreases in the IC values were more pronounced after 72 h (p < 0.05) than after a 48‐h (p > 0.05) exposure to copper, where differences were not statistically significant. This is consistent with the greater drop in copper concentration after 72 h.

Modeling algal bioassay results for copper

The decreases in copper concentration measured during the standard bioassay experiments and the relationship between algal growth rates and copper (toxicant) concentrations were modeled by optimizing parameters in Equations 1 and 3, respectively.

A suitable fit to the observed copper concentration decline in the experimental data was provided by a model comprising an initial concentration drop (which increased with increasing initial copper concentration) followed by an exponential concentration decline (at a rate that decreased with time), C_t = C₀ ‐ C₀ 0.08exp(0.01C₀) to 0.5C₀[1 ‐ exp(‐t/72)] (Eqn. 1). The fit between the model copper concentration declines and the measured data is shown in Figure 1.

The optimized model for the effect of copper (toxicant) concentration on algal growth rate, μ_t, was μ_t = 0.06 −0.054/[1 + 3exp( −0.19C_t] (Eqn. 3). In Figure 2, the mean growth rate (expressed as percentage of control) for the standard and tank bioassays are shown for the 48‐ and 72‐h algal growth data. Lines are shown for the model developed from the 72‐h standard bioassay data, where concentration decline occurs in accord with the model (Fig. 1), and for a model where no concentration decline occurs. In Table 3, IC calculated using model growth rate and cell yield data are shown.

Growth rate and cell yield endpoints

IC calculated using cell yield data are lower than those calculated using growth rate data (Tables 1 and 2). This is a consequence of the calculation procedure, rather than an indication of a greater sensitivity (i.e., lower IC) for cell yield endpoints compared with growth rate endpoints [22]. Inhibitory concentration values calculated using growth rate data are also independent of the algal bioassay duration, provided toxicant concentrations remain constant (Table 3). However, if the concentration of the toxicant declines during the test, then the algal growth rate will increase over the test duration and calculated IC values will increase (Tables 1–3).

In Figure 3, the mean cell yield data (expressed as percentage of control) for the tank bioassay are shown for 48‐and 72‐h test durations. Figure 3, along with the data in Tables 2 and 3, illustrates that IC values calculated using cell yield data decrease as the test duration increases. Nyholm [22] concluded that, from a theoretical point of view, growth rate is a better response variable than cell yield. For algal bioassays, European protocols [18] specify growth rate, whereas U.S. Environmental Protection Agency protocols [24] specify cell yield as the preferred method for endpoint calculations. For the purpose of comparing IC values between toxicants, growth rate‐derived IC values will be more useful because they are not dependent on the calculation procedure (test duration). Another alternative for calculating IC values is to use the area under the growth curve (the integral of log‐cell yield) [25]. This approach may be particularly useful for situations where irregular algal growth occurs and for interpreting data where exposure concentrations occur as pulses.

Application of the model to other toxicant loss scenarios

Often the purpose of the algal (or any) bioassay is to estimate the necessary dilution of an effluent in a receiving system. Here there may well be a naturally declining concentration resulting from losses to particulates in the field sample as well as other effects due to dilution. However, if there is a constant supply of contaminant, such as via a continuous discharge, then it is necessary to know the equilibrium concentration to be able to determine the appropriate IC.

Although the losses with metals are slow, that of many organic contaminants is rapid. This applies particularly to many hydrophobic organics such as polycyclic aromatic hydrocarbons, chlorophenols, or pesticides [17]. The available adsorptive surfaces in a flask bioassay situation will be far higher than a field situation (especially if there is a low suspended sediment concentration), which will be further exacerbated if high algal cell densities are used. In this situation it is often impossible to calculate any meaningful growth rate endpoint. Here the application of modeling scenarios offers considerable assistance.

Because toxicant concentration decline may occur through a number of mechanisms and will be affected by the properties and concentration of the toxicant, a large number of decline profiles are possible. In Table 4, the effect of declining toxicant concentrations on calculated IC values is shown for a series of concentration decline scenarios. For these calculations, the model toxicant affects algal growth analogous to that caused by copper. Concentration decline profile 1 (no decline) gives the true IC values for the toxicant. Scenarios 2 through 9 show the incorrect (overestimated) ICs that would be calculated for the concentration decline profile (depicted) if the decline was not accounted for in the calculations, i.e., concentrations were assumed to remain constant. In these calculations, the effect of the concentration decline profile on 95% confidence intervals of the IC values (empirically calculated by ToxCalc) was estimated by using triplicate growth rate data with values 95, 100, and 105% of the model‐calculated value.

For all scenarios, the degree of overestimation of the true IC value was dependent on the bioassay duration and the shape of the concentration decline profile (rate of decline through time). Comparison of ICs calculated for linear decline (scenario 5) with those where decline is only slightly curved (scenario 6) highlights the importance of considering the shape of the concentration decline profile. For toxicant concentrations that decline exponentially (scenarios 6‐9), the degree to which ICs are overestimated increases rapidly as the concentration decline rate increases. The overestimation of ICs for toxicants whose concentration declines exponentially, for example, light‐sensitive organics such as chlorocatechols and PAHs, represents the worst case for errors in calculated ICs. For a toxicant whose concentration declines exponentially to less than 5% of its original value within 36 h of a 72‐h test, the IC is shown to be underestimated by a factor of 50. In such a case, estimating the average toxicant concentration will be difficult unless continuous monitoring of the concentration was made. An average based on an initial and final concentration (i.e., −50% of initial concentration) will still result in an underestimation of the IC by a factor of 25.

Implications for toxicity tests

In the case of copper, these results show that the standard laboratory bioassay is likely to significantly underestimate the toxicity of copper by as much as a factor of 2. This underestimation can be reduced if the test duration is reduced, but the reason that 72‐h tests are preferred to 48‐h tests is that the precision of the algal cell density measurements increases with increasing cell density. The nature of the variability is seen in Table 1.

The model presented here is based on the presumption that in the field situation the toxicant concentration is constant, whereas most often it will be quite variable. The natural variability may be related to tidal flushing, pulse discharges, or changes in environmental conditions such as rainfall or sunlight, which cause pH and speciation changes. However, using the constant concentration scenarios is the best way to compare the effects of toxicants that decay to different extents, and a similar modeling approach may also be useful to deconvolute extreme variants in concentration‐time relationships where laboratory bioassays are more problematic. Studies in this area are currently in progress.

Another poorly addressed problem for bioassays is compounds (e.g., chlorocatechols) that photodegrade with time, forming new toxicants of similar or greater toxicity than the precursor compound. These secondary compounds may then be further degraded or be lost from solution through adsorption or uptake. Further studies are necessary to address difficult toxicants of this type. The additivity of joint toxicants is also a subject requiring more research [26].

For simple organisms like algae, the present study ignores physiological aspects of toxicant uptake and mode of action, and effects are directly related to water concentrations of contaminants. For higher organisms (e.g., fish), however, this assumption may not be valid, and more complex approaches that consider toxicokinetics may be necessary to account for both uptake, elimination, and recovery (postexposure) to accurately predict effects [12,15,27,28]. Mayer et al. [7] have shown that dose‐response relationships developed using internal concentrations (dose) of polychlorinated biphenyls in the alga Selenastrum capricornutum resulted in estimates of IC values with tighter ranges than those estimated using water concentrations. This observation is because internal concentrations effectively integrate the water concentrations over time and account for changes in toxicant concentrations with time. Franklin et al. [8,29] also show that internal dose‐response curves, rather than water concentration response curves, also account for competitive effects of H⁺, Ca²⁺, and metals at the cell membrane. However, intracellular toxicant cell concentrations are generally more difficult to measure than water concentrations.

For algal bioassays, the simple models presented here can be used routinely for estimating the true IC of toxicants that decline in concentration during testing. The only prerequisite for this approach is that accurate measurements are made of toxicant concentrations during testing. The modeling of toxicant effects also provides a simple means for the calculation of worst‐case IC where fluctuations are known to occur.

Acknowledgements

We thank David Fox of CSIRO Land and Water for useful discussions on the modeling component of the study. We thank Monique Binet for help with bioassay experiments and Karl Bowles for assistance with ASV analyses.

REFERENCES