BBP was purchased from Sigma-Aldrich (>98%, St. Louis, MO, USA). Methanol and dimethylsulfoxide (DMSO) were also purchased from Sigma-Aldrich (>98%). PDMS tubing (outer diameter 16.5 mm, inner diameter 12 mm, density 1.18 g/mL) was purchased from a local provider (Dongbang Silicone, Kimpo, Korea). The PDMS tubing was cut into a size fit in a well of a 24-well plate. The height was 14.6 mm and the mass of each PDMS was measured 1.726 g (n=20, standard deviation [SD]=0.0084 g). They were cleaned with n-hexane and methanol for 1.5 hours each in a Soxhlet extractor. Cleaned PDMS tubings were stored in methanol until use.

In order to load PDMS with desired concentration of BBP, partition coefficient of BBP between PDMS and the loading solution (methanol: deionized water = 6:4) was determined. It was found out that incubating for 48 hours in a shaking incubator (150 rpm) was sufficient for attaining equilibrium between PDMS and the loading solution. The partition coefficient between PDMS and the loading solution (K_{PDMS/loading solution}) was determined by measuring the concentration in PDMS (C_PDMS,0, mol/L) and the loading solution (C_{loading solution}, mol/L) independently.

The loading solution equilibrated was directly subjected to high-performance liquid chromatography (HPLC) analysis and the concentration in PDMS was quantified after extracting with 30 mL methanol. The concentration of BBP loaded into PDMS was then calculated from the concentration of BBP in the loading solution. After the loading, PDMS tubings were dried at room temperature for 1.5 hours to remove residual methanol swallowed by PDMS and placed into a well of a 24 well-plate for the determination of desorption kinetic parameters or acute toxicity tests.

The change in the concentration of BBP in the assay medium (0.032 M KCl and 0.051 M NaCl) [10] was monitored to evaluate the performance of the passive dosing technique. Each well is filled with 1.5 mL of the assay medium and BBP was delivered from the PDMS tubing (Figure 1). Applying a two-compartment model, the change in BBP concentration in the assay medium (C_medium) and in PDMS (C_PDMS) can be expressed by:

where k_a is the absorption rate constant (L_medium L_PDMS h^-1), k_d is the desorption rate constant (h^-1), and V_PDMS and V_medium are the volumes (L) of PDMS and the medium, respectively.

The analytical solution for C_medium of the above differential equations is given by [7]:

where K_PDMS/medium is the partition coefficient of BBP between PDMS and the assay medium, C_PDMS,0 is the initial concentration of BBP in PDMS. Because K_PDMS/medium is sufficiently high for BBP, it is reasonable to assume that the loss of BBP from the PDMS dosing phase during the experiment via delivery to the assay medium, sorption to the wall of the plate, or volatilization is negligible. By measuring C_medium at 0.083, 0.167, 0.33, 0.75, 1, 1.5, 2, 4, 10, 24, and 40 hours after the exposure to the PDMS dosing phase, values of K_PDMS/medium and kd were obtained using a non-linear regression (Equation 4) using Graphpad Prism 3.0 (La Jolla, CA, USA).

Desorption constant (k_d) in equation 4 can be estimated using a film diffusion model as follows:

where D_medium and D_PDMS are the molecular diffusion coefficients of the solute in the medium and in the PDMS, δ_medium and δ_PDMS are the thicknesses of the diffusion film in the medium and in the PDMS tubing, and A is the interface area between two phases. The resistance term in the assay medium is known to be much higher than in the PDMS dosing phase for highly hydrophobic chemicals such as BBP [7,9] due to high permeability of PDMS to organic chemicals [11], it is reasonable to simplify equation 5 neglecting the effects of the mass transfer resistance in PDMS side.

Because the assay medium used in this study can be regarded as water neglecting the effects of ionic strength (I=0.083 M), the molecular diffusivity can be estimated from the molecular weight (MW) of BBP [12]:

Thus, the thickness of the mass transport boundary layer of the medium (δ_medium) can be estimated with the measured kd value.

Prior to the toxicity test, organisms were prepared. The wildtype C. elegans Bristol strain N2 was used. C. elegans (Caenorhabditis Genetics Center, USA) were maintained on nematode growth medium (NGM) plates seeded with Echerichia coli strain OP50, at 20℃, using the standard method [13].

Acute mortality tests of BBP using C. elegans were conducted both using passive dosing and a conventional spiking method. For passive dosing, PDMS tubings loaded with desired concentrations of BBP as described earlier were placed in individual wells of 24-well plate containing the assay medium. The medium was exposed to PDMS tubings for 2 hours in an incubator at 20℃ to reach the equilibrium between the PDMS tubing and the solution. Approximately ten organisms were then added to the medium and live organisms were counted after 24 hours. For a comparative evaluation, BBP was spiked into the medium as dissolved in DMSO to prepare nominal concentrations of 0.3, 1, 3, 10, 30, and 100 mg/L. The concentration of DMSO was less than 0.1% (v/v) in the final solution. All experiments were conducted with three replicate wells for each concentration.

After 24 hours acute mortality test, approximately 0.5 mL medium was carefully taken and transferred to a glass vial to chek the equilibrium concentration of BBP in the medium. The sampled medium was directly subjected to HPLC quantification.

The concentration of BBP was quantified using an HPLC system equipped with Waters 600E pump (Milford, MA, USA), an autosampler (Waters 717+), and a UV detector (Waters 996). Deionized water:methanol (2:8) was used as an eluent in an isocratic mode with the flow rate of 1 mL/min. BBP was separated on a Thermo C18 column (4.6 × 150 mm, 5 µm particle size; Thermo Scientific) and detected at 230 nm.

Figure 2 shows the relationship between the concentration in the PDMS tubing and the loading solution in three concentration. The partition coefficient between PDMS and the loading solution (K_{PDMS/loading-solution}) was determined 4.05±0.09 from the slope of the regression. This value was then used for the calculation of the initial concentration of BBP in PDMS after loading without measuring concentration of BBP in individual PDMS tubings.

Figure 3 shows the release of BBP from PDMS tubing to the assay medium with the solid line representing the model fit using equation 4. The desorption rate constant (±standard error [SE]) and K_PDMS/medium (±SE) were 1.53 (±0.18) × 10^-4/h and 9,073 (±369), respectively. Time to reach 95% steady state was estimated 2.2 hours. Because the test duration of the mortality test using C. elegans was 24 hours, it is likely that a constant exposure condition was maintained during the test.

The effective thickness of the diffusion boundary layer in the medium (δ_medium) was estimated as 0.4 mm using equation 6 with the experimentally obtained kd value. The thickness of the unstirred water layer under non-agitated systems with different geometry has been reported approximately between 0.15 and 2.0 mm [14,15]. For an agitated system, it might decrease to less than 1 µm [16]. The value obtained in this study is slightly lower but approximately fall in the range reported in the literature.

Mortality of C. elegans was not observed for all doses tested regardless of the administration methods of BBP to the medium (Table 1). Because the maximum dose used in this study was as high as the limit of the solubility of BBP, 2.7 mg/L [17], it is likely that acute effects on C. elegans are not expected due to the exposure to BBP under any conditions.

Because BBP is delivered from a PDMS tubing, the maximum concentration in the medium cannot exceed the solubility in the medium. Figure 4 describes the measured BBP concentration in the medium after the mortality tests versus the intended concentration by the passive dosing method and the spiking method. The calculated values in the medium using K_PDMS/medium agreed well with the measured free concentration when BBP was administered using the passive dosing method. However, the discrepancy between the nominal concentration and the measured free concentration dramatically increased as the nominal concentration exceeded the solubility of BBP in water.

Due to the limited volume of test organisms used in this study, it was not possible to extract BBP from the organism and quantify the body-residue concentration to make sure that the equilibrium is obtained. Instead, it was checked that the concentration of BBP in the medium measured after the exposure agreed with the predicted values from the partition coefficient.

As shown in Figure 2 and Figure 4, the passive dosing method developed in this study was suitable for providing well-defined exposure condition for BBP and superior to active spiking methods. Although the prediction of the aqueous exposure concentration requires the pre-determined K_PDMS/medium, the free concentration in the medium can be well maintened once the K_PDMS/medium is known.

Based on the maginitude of the unstirred water boundary layer, time required to reach steady-state using different high-throughput platforms could be estimated. Figure 5 shows model calculation of the desorption kinetics of a hypothetic hydrophobic chemical (MW = 312.4 g/mol, D_medium = 1.65 × 10^-6 m/h, K_PDMS/medium = 9073). The conventional test conditions were estimated as shown in Table 2. As shown in Figure 5, 95% steady-state was obtained within 6 hours in a 6-well plate in which the equilibration time is the longest under the assumptions listed in Table 2. This simulation indicates that the passive dosing technique used in this study is very promising for maintaining constant exposure for high-throughput toxicity testings. In addition, it should be also noted that a slight agitation makes the unstirred water layer much thinner than modeled in this study and the resulting equilibration time can be much shorter.

The presence of organisms in the medium may slow the equilibration time. However, the total mass of C. elegans in the medium is less than 10^-4 g estimated from the size of the organisms with the asumption of unit density. This is too small mass to affect the overall desorption kinetics considering K_PDMS/medium value of 9073. The lipid content of C. elegans was reported approximaely 20% dry weight [18]. Thus, it is still reasonable that the exchange of BBP between C. elegans and the medium is much faster than the release from PDMS. In addition, it is likely that the motion of C. elegans in the medium (i.e., bioturbation) decreases the thickness of the unstirred water layer although the effects could not be quantified. Considering all these aspects together, well-defined exposure condition is highly likely to be maintained in the passive dosing method used.

A recent toxicity study for BBP using invertebrates showed that LC50 values were 0.46 and 1.23 mg/L for Hyalella azteca and Lumbriculus variegatus in a 10-day water only exposure test, indicating that the toxicity of BBP could be observed below the water solubility [19]. Because BBP is liquid in an ambient temperature, the expected chemical activity in the body (or the body-residue concentration) is likely to exceed the level causing baseline toxicity [20,21]. Thus, the non-observed toxicity of BBP in C. elegans may be due to the metabolic transformation of BBP although it needs further research.

Some recent studies still reported the acute toxicity of BBP even at the nominal concentration greater than the solubility in the medium. For example, Planello et al. [22] reported LC50 of 26.6 mg/L in a survival test for lavae of Chironomus riparius. This value obviously exceeds the limit of free concentration in the aqueous phase as shown in this study. Because overwelming assumption in ecotoxicology is that only freely available fraction penetrates the cell membrane and thus results in any toxic responses, it is very difficult to interprete those results without knowing well-defined exposure conditions. Incremental change in response is usually observed with increasing dose of a chemical at the level greater than its solubility in the assay medium. Two possibilities are that (1) the increasing nominal concentration still steadily increased the internal concentration (as shown in Figure 4) and thus toxic effects became apparent with the increasing free concentration and that (2) complicated interactions with co-solvents administered with the test substances in test organisms may result in a synergystic toxicity. For the evaluation of those hypotheses, further investigation is needed.