Estimating the Bioaccumulation Potential of Hydrophobic Ultraviolet Stabilizers Using Experimental Partitioning Properties

Although hydrophobic ultraviolet (UV) stabilizers are an emerging environmental concern because of their widespread occurrence, persistence, and bioaccumulation potential, experimental values of their partitioning properties required for risk assessment are scarce. In this study, n-octanol-water partition (Kow) and lipid–water partition constants (Klipw), which are key parameters for environmental risk assessment, were experimentally determined for five selected hydrophobic UV stabilizers (UV326, UV327, UV328, UV329, and UV531) based on third-phase partitioning among polydimethylsiloxane (PDMS), water, and n-octanol/lipid. The partition constants between PDMS and water (KPDMSw), obtained using the dynamic permeation method were used to derive Kow and Klipw. The obtained log Kow and log Klipw values were in the ranges of 7.08–7.94 and 7.50–8.34, respectively, indicating that the UV stabilizers exhibited a high bioaccumulation potential in aquatic environments. The experimental Kow and Klipw values obtained in this study provide valuable information for the evaluation of the fate, distribution, bioavailability, and toxicity of the UV stabilizers in aquatic environments.


Partition Constants between PDMS and Water (KPDMSw)
KPDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (CacceptorPDMS) is given by Equation (1), where C0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.
Equation (2) can be obtained by rearranging Equation (1), and the mass transfer-rate constant (k) can be determined using a linear regression between ln 1 and t. Mass transfer resistance in aqueous solution is dominant for highly hydrophobic chemicals because PDMS is an amorphous polymeric material. Thus, KPDMSw (Lw LPDMS −1 ) can be derived using Equation (3), and the aqueous diffusion coefficient (Dw in m 2 s −1 ) can be estimated from the relationship with the molar mass of the five UV stabilizers using Equation (4) [24]. (3)

Partition Constants between PDMS and Water (KPDMSw)
KPDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (CacceptorPDMS) is given by Equation (1), where C0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.

Partition Constants between PDMS and Water (KPDMSw)
KPDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (CacceptorPDMS) is given by Equation (1), where C0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.

Partition Constants between PDMS and Water (KPDMSw)
KPDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (CacceptorPDMS) is given by Equation (1), where C0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.

Partition Constants between PDMS and Water (KPDMSw)
KPDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (CacceptorPDMS) is given by Equation (1), where C0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.
Equation (2) can be obtained by rearranging Equation (1), and the mass transfer-rate constant (k) can be determined using a linear regression between ln 1 and t. Mass transfer resistance in aqueous solution is dominant for highly hydrophobic chemicals because PDMS is an amorphous polymeric material. Thus, KPDMSw (Lw LPDMS −1 ) can be derived using Equation (3), and the aqueous diffusion coefficient (Dw in m 2 s −1 ) can be estimated from the relationship with the molar mass of the five UV stabilizers using Equation (4) [24]. (3)

Partition Constants between PDMS and Water (K PDMSw )
K PDMSw values were measured using the aqueous boundary layer (ABL) permeation method. The theory and experimental procedure for this method have been described in detail in existing literature [24,26]. Briefly, by neglecting the mass loss in the aqueous solution between the two PDMS disks because of the high holding capacity of PDMS for HOCs, the concentration in the acceptor disk (C acceptorPDMS ) is given by Equation (1), where C 0 (mM) is the initial concentration of chemicals in the donor PDMS disk and k (s −1 ) is the mass transfer-rate constant through the ABL.
Equation (2) can be obtained by rearranging Equation (1), and the mass transfer-rate constant (k) can be determined using a linear regression between ln 1 − 2C acceptorPDMS C 0 and t. Mass transfer resistance in aqueous solution is dominant for highly hydrophobic chemicals because PDMS is an amorphous polymeric material. Thus, K PDMSw (L w L PDMS −1 ) can be derived using Equation (3), and the aqueous diffusion coefficient (D w in m 2 s −1 ) can be estimated from the relationship with the molar mass of the five UV stabilizers using Equation (4) [24].
where δ w is the thickness of the aqueous boundary layer (m), A PDMS is the surface area of PDMS (m 2 ), V PDMS is the volume of PDMS (m 3 ), and MW is the molecular weight of the chemical (g mol −1 ). A donor PDMS disk (diameter = 6 mm, thickness = 1 mm) was loaded with each UV stabilizer using methanol as the solvent for 24 h (Figure 1a). Donor and acceptor PDMS disks were placed at opposite ends of a custom-made glass well filled with water. The aqueous solution between the two disks was agitated with a stainless steel disk (diameter = 5.08 mm, thickness = 0.635 mm) at 300 rpm using a VP710F tumble stirrer (V&P Scientific Inc., San Diego, CA, USA) ( Figure 1b). The thickness of the ABL was set to 12.5 µm because the experimental system used in this study was the same as that used in previous studies [24,26]. After a designated time, the disks were removed and rinsed with 1 mL methanol:water (6:4, v/v) and 1 mL water. The UV stabilizers in the disks were extracted using 1 mL acetonitrile (ACN) for 24 h. The mass transfer rate was calculated by measuring the changes in the concentrations of both donor and acceptor disks (Equation (2)). K PDMSw was derived using the mass transfer-rate equation (Equation (3)). All experiments were conducted at 25 • C. A schematic diagram for determining K PDMSw is shown in Figure 1a,b. where δw is the thickness of the aqueous boundary layer (m), APDMS is the surface area of PDMS (m 2 ), VPDMS is the volume of PDMS (m 3 ), and MW is the molecular weight of the chemical (g mol −1 ). A donor PDMS disk (diameter = 6 mm, thickness = 1 mm) was loaded with each UV stabilizer using methanol as the solvent for 24 h (Figure 1a). Donor and acceptor PDMS disks were placed at opposite ends of a custom-made glass well filled with water. The aqueous solution between the two disks was agitated with a stainless steel disk (diameter = 5.08 mm, thickness = 0.635 mm) at 300 rpm using a VP710F tumble stirrer (V&P Scientific Inc., San Diego, CA, USA) ( Figure 1b). The thickness of the ABL was set to 12.5 μm because the experimental system used in this study was the same as that used in previous studies [24,26]. After a designated time, the disks were removed and rinsed with 1 mL methanol:water (6:4, v/v) and 1 mL water. The UV stabilizers in the disks were extracted using 1 mL acetonitrile (ACN) for 24 h. The mass transfer rate was calculated by measuring the changes in the concentrations of both donor and acceptor disks (Equation (2)). KPDMSw was derived using the mass transfer-rate equation (Equation (3)). All experiments were conducted at 25 °C. A schematic diagram for determining KPDMSw is shown in Figure 1a,b.

n-Octanol-Water and Fish Oil-Water Partition Constants
The partition constants between n-octanol or fish oil and PDMS (Koctanol-PDMS or Kfish oil-PDMS) were obtained by measuring the equilibrium concentrations of n-octanol or fish oil and PDMS in a batch system. Then, the Koctanol-PDMS or Kfish oil-PDMS values were used to calculate Kow or Klipw, assuming that the activity coefficients in PDMS were the same in water and in n-octanol or fish oil. A PDMS disk (diameter = 10 mm, thickness = 1 mm) was loaded with each UV stabilizer, dissolved in a 500 or 1000 mg L 1 methanol, for 24 h (Figure 1a). The loaded PDMS disks were rinsed with 1 mL methanol:water (6:4, v/v) and 1 mL water and placed in vials containing 125, 166, and 250 μL of n-octanol or fish oil (Figure 1c). Each vial was agitated at 25 °C and 150 rpm in the dark in a shaking incubator. Preliminary experiments showed that 24 h and 48 h were sufficient to attain equilibrium

n-Octanol-Water and Fish Oil-Water Partition Constants
The partition constants between n-octanol or fish oil and PDMS (K octanol-PDMS or K fish oil-PDMS ) were obtained by measuring the equilibrium concentrations of n-octanol or fish oil and PDMS in a batch system. Then, the K octanol-PDMS or K fish oil-PDMS values were used to calculate K ow or K lipw , assuming that the activity coefficients in PDMS were the same in water and in n-octanol or fish oil. A PDMS disk (diameter = 10 mm, thickness = 1 mm) was loaded with each UV stabilizer, dissolved in a 500 or 1000 mg L −1 methanol, for 24 h (Figure 1a). The loaded PDMS disks were rinsed with 1 mL methanol:water (6:4, v/v) and 1 mL water and placed in vials containing 125, 166, and 250 µL of n-octanol or fish oil (Figure 1c). Each vial was agitated at 25 • C and 150 rpm in the dark in a shaking incubator. Preliminary experiments showed that 24 h and 48 h were sufficient to attain equilibrium in the octanol-PDMS and fish oil-PDMS systems, respectively. After equilibrium was established, the PDMS disk was collected from the vial, rinsed using 1 mL methanol:water (6:4, v/v) and 1 mL water, and extracted using 1 mL ACN. Fish oil and n-octanol were diluted in ACN. The values of K octanol-PDMS or K fish oil-PDMS were obtained using a linear regression between the concentration in n-octanol or fish oil and that in PDMS. Finally, the values of K ow and K lipw were calculated using the third-phase equilibrium method using Equations (5) and (6):

Instrumental Analyses
The concentrations of the five tested chemicals were quantified using a Waters AC-QUITY ultra performance liquid chromatograph (UPLC) with a photodiode array (PDA) detector. The absorbance was measured at 221 nm (UV326), 204 nm (UV327 and UV328), 218 nm (UV329), and 287 nm (UV531). The UV stabilizers were separated using a C18 column (2.1 mm × 50 mm, 1.7 µm, Waters) at 35 • C. The mobile phase comprised 95% acetonitrile and 5% water in isocratic mode with a flow rate of 0.2 mL min −1 . Figure 2 shows the mass transfer kinetics of the selected UV stabilizers from the donor PDMS to the acceptor PDMS for the determination of K PDMSw . The desorption rate constants (k) for the five test chemicals were obtained using Equation (2). Except for data for UV531 (R 2 = 0.81) and UV328 (R 2 = 0.76), the experimental data fit well (R 2 > 0.9). The K PDMSw values were then calculated using Equation (3), and log K PDMSw ranged from 6.14 to 7.22, and the uncertainties of the K PDMSw were calculated as the 95% upper and lower limits of the error propagation from the standard error of regression and the estimated uncertainties of the thickness of the ABL (Table 2). These values were used to estimate the K ow and K lipw values. in the octanol-PDMS and fish oil-PDMS systems, respectively. After equilibrium was established, the PDMS disk was collected from the vial, rinsed using 1 mL methanol:water (6:4, v/v) and 1 mL water, and extracted using 1 mL ACN. Fish oil and n-octanol were diluted in ACN. The values of Koctanol-PDMS or Kfish oil-PDMS were obtained using a linear regression between the concentration in n-octanol or fish oil and that in PDMS. Finally, the values of Kow and Klipw were calculated using the third-phase equilibrium method using Equations (5) and (6):

Instrumental Analyses
The concentrations of the five tested chemicals were quantified using a Waters AC-QUITY ultra performance liquid chromatograph (UPLC) with a photodiode array (PDA) detector. The absorbance was measured at 221 nm (UV326), 204 nm (UV327 and UV328), 218 nm (UV329), and 287 nm (UV531). The UV stabilizers were separated using a C18 column (2.1 mm x 50 mm, 1.7 μm, Waters) at 35 C. The mobile phase comprised 95% acetonitrile and 5% water in isocratic mode with a flow rate of 0.2 mL min 1 . Figure 2 shows the mass transfer kinetics of the selected UV stabilizers from the donor PDMS to the acceptor PDMS for the determination of KPDMSw. The desorption rate constants (k) for the five test chemicals were obtained using Equation (2). Except for data for UV531 (R 2 = 0.81) and UV328 (R 2 = 0.76), the experimental data fit well (R 2 > 0.9). The KPDMSw values were then calculated using Equation (3), and log KPDMSw ranged from 6.14 to 7.22, and the uncertainties of the KPDMSw were calculated as the 95% upper and lower limits of the error propagation from the standard error of regression and the estimated uncertainties of the thickness of the ABL (Table 2). These values were used to estimate the Kow and Klipw values.   The values of K ow and K lipw were obtained from the slopes of the linear regression between the concentration in n-octanol or fish oil and that in PDMS (Figures 3 and 4). As shown in Figures 3 and 4, good linear relationships were observed for the five UV stabilizers within the concentration range investigated (R 2 > 0.97). The resulting values of log K ow and log K lipw , calculated using Equations (5) and (6), were in the ranges of 7.08-7.94 and 7.50-8.34, respectively ( Table 2). As shown in Table 2, log K lipw values were consistently greater than the log K ow values by 0.38-0.49 log units. Table 2 presents the literature values of log K ow submitted to the European Chemical Agency (ECHA) using the HPLC retention time method [29] and those estimated using EPI Suite version 4.11 [30] and the ALOGPS 2.1 program [31]. It should be noted that the K ow and K lipw values in this study were measured at 25 • C, the standard thermodynamic temperature for providing reference values. Further studies on the effects of environmental factors (e.g., temperature and electrolytes) on K ow and K lipw values would extend the applicability of those partition constants under various environmental conditions [19,32].

Discussion
The experimental values of K ow and K lipw can be obtain by using direct-measurement methods (e.g., slow-stirring, shake-flask, and generator column method) or indirect methods (e.g., reverse-phase high-performance liquid chromatography). However, shake-flask and generator column method and reverse-phase high-performance liquid chromatography might be not appropriate for determining of K ow and K lipw for highly hydrophobic compounds (log K ow > 6) [17,33]. The slow-stirring method has been reported in literature for measuring the K ow and K lipw for highly hydrophobic chemicals; however, the time required to reach to equilibrium generally takes from a few days up to a few weeks [17,34,35]. To overcome the difficulties in measuring the K ow and K lipw for highly hydrophobic chemicals (i.e., time-consuming, extremely low water solubility, degradation of parent compound during experiment period), the third-phase method, in which a polymer serve as a reference partitioning phase, was proposed to determine the K ow and K lipw for hydrophobic chemicals with log K ow > 6 [36,37]. As shown in Table 2, the log K ow values of the five tested UV stabilizers varied substantially depending on the experimental and estimation methods. The experimental log K ow data for the UV stabilizers have been as yet reported in literature using a HPLC-based method [29]. The experimental log K ow values in this study were consistently higher and could be more accurate than those reported in the registration dossier to ECHA [29], which reported only lower bounds estimated using the HPLC retention time ( Table 2). Because the HPLC retention times of chemicals with experimental log K ow values between 0 and 6 were used in the calibration of the method (OECD Guideline 117), greater uncertainties resulted when the log K ow values were above 6. However, for highly hydrophobic substances, log K ow values are often estimated using quantitative structure-property relationship (QSPR) methods (e.g., ALOGPS, KOWWIN) ( Table 2). Although the predicted log K ow values by KOWWIN or ALOGPS are lower than the experimental values in this study, the prediction for K ow by the QSPR methods agree quite well with experimental data in this case, especially by KOWWIN (the difference within 1 log unit, except for UV326). The predicted log K ow might not be accurate because the predicted K ow values from the commercial software packages are derived from the experimental K ow values of their nearest analogs [38,39]. An important additional reason here is that at least UV-328 can make an intramolecular hydrogen bond that is not captured in EPISuite [40]. Therefore, the lack of measured data for highly hydrophobic chemicals leads to the inevitable accuracy of the QSPR methods [39]. Although there are certain limitations, such as the long experimental time, high cost, and expensive instruments to detect trace concentrations of analyte to obtain reliable K ow values, these experimental methods are highly recommended for evaluation of UV stabilizers. Ab initio methods are also used as an alternative to predict partition constants for HOCs because they do not rely on the quality of experimental data in the training set. A commonly used program from this is the COSMOtherm program, which is based on quantum chemical descriptors [41,42]. Interestingly, the log K ow of UV327 and UV328 in this study agrees well with estimations made using COSMOtherm (log K ow of UV327 and UV328 are 7.91 and 8.5, respectively [40,43]. The difference between the predicted values by COSMOtherm and the experimental data for log K ow is within 0.6 log unit, supporting the usefulness of these ab initio methods for "difficult-to-test" substances. In this study, the K lipw values were generally higher than the K ow values, which can be explained by the fact that lipid storage has a highly organized structure compared to the less organized structure of n-octanol, and the driving forces for partitioning into saturated lipid storage can be differentiated from those for n-octanol [18]. There have been studies that experimentally determined lipid-water partitioning constants for a broad range of compounds, including ionic substances, strongly hydrophobic chemicals, and nanomaterials, via different methods, such as potentiometry, the equilibrium dialysis technique, and methods using solid-supported lipid membranes [18,44,45]. To the best of our knowledge, this is the first report of the K lipw values of the selected UV stabilizers. Although this study used a storage lipid (e.g., fish oil) as a representative lipid material, the lipid accumulation properties of chemicals in storage lipids might differ from those in membrane lipids (i.e., phospholipids). Previous studies have demonstrated that the K lipw values of hydrophobic chemicals toward storage lipids were similar to those of hydrophobic chemicals toward a model membrane (i.e., the difference was less than 1 log unit) [19,46]. In addition, storage lipids from different origins (i.e., olive oil, milk fat, fish oil, goose oil, and soybean oil) do not differ in their accumulation properties for various polar and nonpolar organic chemicals [19]. Thus, the K lipw values in this study are good indicators of the biopartitioning of UV stabilizers. For highly hydrophobic organic chemicals, the time required to attain a steady state is often longer than the lifespan of aquatic species, necessitating the consideration of other kinetic parameters, such as growth and metabolic transformation rates, to obtain the bioconcentration factor [47]. Bioaccumulation of UV stabilizers occurs primarily after their ingestion by organisms. There is evidence of the bioaccumulation of UV328 in fish, crustaceans, marine mammals, and algae [48]. Furthermore, biomagnification of UV531 was observed by Peng et al., and the lipid normalized concentration of UV531 in mantis shrimp was higher than that in its prey, sword and Kurumu prawns [49]. In another study, UV326, UV328, and UV329 were found at concentrations ranging from 1.34 to 45.6 ng g −1 (dry weight, dw) on Gran Canaria Island (Spain) [50]. Hasegawa et al. recently showed similar accumulation patterns between direct ingestion from the water column and indirect exposure via the trophic transfer pathway for UV327 in marine fish [51]. In a recent monitoring study, the sum of nine benzotriazole UV stabilizers (including the five target chemicals in this study) were found in river water (17.0-32.5 ng L −1 ) and sediment (2.0-22.6 ng g −1 , dw) in Korea [52]. Furthermore, the benzotriazole (UV326, UV327, UV328, and UV329) bioaccumulation in crucian carp muscle ranged from 0.345 to 5.94 (ng g −1 , wet weight), with UV329 being the dominant compound in all crucian carp. The field-derived log bioaccumulation factor (BAF) and biota-sediment accumulation factor (BSAF) for UV329 and UV326 were also calculated during 2018-2020 as 2.89, 2.67, −0.02, and −0.02, respectively [52].
These observations were consistent with the extremely high log K ow and log K lipw values obtained in this study. However, the log BAF and estimated partition coefficients between sediment and water were lower than those estimated from log K ow and log K lipw . As previously mentioned, competition between partitioning processes and other processes may explain these differences. In addition, there are other aspects to be considered for a better understanding of the environmental behavior of UV stabilizers because large amounts of UV stabilizers are used as plastic additives, and they are likely to leach very slowly from these plastics [53]. As microplastics have long environmental residence time [54], they may contribute to the long-range transport of UV stabilizers in the oceans. Because UV stabilizers do not readily biodegrade [43,55], the evaluation of their long-range transport and bioaccumulation potential is important to determine whether they should be listed as POPs. Therefore, future studies on their plastic-mediated, long-range transport as well as bioaccumulation in aquatic species are required. The experimental partitioning properties reported in this study are of considerable value for the evaluation of environmental flux by calculating fugacities in different media [56].

Conclusions
In this study, n-octanol-water partition (K ow ) and lipid-water partition (K lipw ) constants were experimentally measured to estimate the bioaccumulation of five UV stabilizers in aquatic environments. The partition constants between polydimethylsiloxane (PDMS) and water (K PDMSw ) of UV stabilizers were obtained using the dynamic permeation method. Consequently, the partition constants between octanol and water (K ow ) and lipid and water (K fishoil-water ) were determined using the third-phase equilibrium method. The results showed that the values of log K ow and log K lipw were in the ranges of 7.08-7.94 and 7.50-8.34, respectively, indicating the high potential bioaccumulation of the five UV stabilizers in aquatic environments. The experimental log K ow and log K lipw values also provide valuable information for risks assessment of UV stabilizers.