Assessing Aquatic Baseline Toxicity of Plastic-Associated Chemicals: Development and Validation of the Target Plastic Model

We developed a Target Plastic Model (TPM) to estimate the critical plastic burden of organic toxicants in five types of plastics, namely, polydimethylsiloxane (PDMS), polyoxymethylene (POM), polyacrylate (PA), low-density polyethylene (LDPE), and polyurethane ester (PU), following the Target Lipid Model (TLM) framework. By substituting the lipid–water partition coefficient in the TLM with plastic–water partition coefficients to create TPM, we demonstrated that the biomimetic nature of these plastic phases allows for the calculation of critical plastic burdens of toxicants, similar to the notion of critical lipid burdens in TLM. Following this approach, the critical plastic burdens of baseline (n = 115), less-inert (n = 73), and reactive (n = 75) toxicants ranged from 0.17 to 51.33, 0.04 to 26.62, and 1.00 × 10–6 to 6.78 × 10–4 mmol/kg of plastic, respectively. Our study showed that PDMS, PA, POM, PE, and PU are similar to biomembranes in mimicking the passive exchange of chemicals with the water phase. Using the TPM, median lethal concentration (LC50) values for fish exposed to baseline toxicants were predicted, and the results agreed with experimental values, with RMSE ranging from 0.311 to 0.538 log unit. Similarly, for the same data set of baseline toxicants, other widely used models, including the TLM (RMSE: 0.32–0.34), ECOSAR (RMSE: 0.35), and the Abraham Solvation Model (ASM; RMSE: 0.31), demonstrated comparable agreement between experimental and predicted values. For less inert chemicals, predictions were within a factor of 5 of experimental values. Comparatively, ASM and ECOSAR showed predictions within a factor of 2 and 3, respectively. The TLM based on phospholipid had predictions within a factor of 3 and octanol within a factor of 4, indicating that the TPM’s performance for less inert chemicals is comparable to these established models. Unlike these methods, the TPM requires only the knowledge of plastic bound concentration for a given plastic phase to calculate baseline toxic units, bypassing the need for extensive LC50 and plastic–water partition coefficient data, which are often limited for emerging chemicals. Taken together, the TPM can provide valuable insights into the toxicities of chemicals associated with environmental plastic phases, assisting in selecting the best polymeric phase for passive sampling and designing better passive dosing techniques for toxicity experiments.


Section S1: Evaluation of Target Model for Other Plastics
The critical plastic burdens of chemicals on additional plastic types, such as polypropylene (PP), polystyrene (PS), polyvinyl chloride (PVC), ultra-high-molecular-weight polyethylene (UHMWPE), and high-density polyethylene (HDPE), were estimated tentatively.Due to the limited experimental data available for these other plastics, it was not possible to evaluate them in separate sets based on distinct modes of toxic action.As a result, the evaluation set for these plastic phases includes chemicals known to have different modes of toxic action.The values for the critical plastic burden on these plastic types fell within the range of 0.01 -63.89 mmol/kg of plastic (Fig. 2d), with PS demonstrating the lowest and HDPE exhibiting the highest chemical critical burden.PVC demonstrated the highest variance in the distribution of critical burden, covering a range of more than seven orders of magnitude for 32 chemicals.This variance may be due to the fact that these chemicals belong to different chemical classes such as PCBs, PAHs, and S2 pharmaceuticals, which follow different modes of toxic actions.The critical burden of chemicals on HDPE (15.65 mmol/kg) was found to be similar to that on LDPE (24.95 mmol/kg), indicating that the density of polyethylene does not strongly affect the partitioning behavior of chemicals considered here.
It is important to note that the ASM estimated LC 50 values were used to compute the critical burdens of chemicals on these plastic types due to a lack of experimental values.Additionally, the sample size and structural diversity of the chemicals used to evaluate these additional plastic types were limited compared to the plastic types used previously to evaluate and validate the target plastic model.Hence, the results for these additional plastic types should be interpreted with caution as they were based on estimated LC 50 values due to the lack of experimental data and the limited sample size and structural diversity of chemicals for these plastic types.
The critical plastic burdens calculated for each plastic type were used to predict LC 50 values using TPM.
The predicted LC 50 values were then compared to the ASM estimated values instead of experimental values due to the limited data available.The results showed a good agreement between the predicted LC 50 values by the target PP model and by the ASM, with an RMSE of 0.45 log units.For the same set of chemicals (n=9), the target phospholipid and octanol models showed RMSEs of 0.11 and 0.17 log units, respectively.ECOSAR's predictions also matched well with ASM predictions, with an RMSE of 0.28 log units.However, the BL showed an RMSE of 0.61 log units with respect to the ASM predicted values for these chemicals.In this comparison, PCB 187 and 128 showed the highest residuals for the target PP model, which may be attributed to the poor data quality of measured partition coefficients of these strongly hydrophobic chemicals between PP and the water phase.
LC 50 values predicted by the target PS model did not match favorably with the ASM predicted LC 50 values based on 8 chemicals, showing an RMSE of 1.52 log units.In particular, nonylphenol showed a deviation of more than 2.75 log units from the ASM predicted value.Interestingly, the ECOSAR class for this chemical S3 is phenols, which generally exhibit excess toxicity compared to the baseline toxicity of neutral organics.
Additionally, significant deviations between the predictions of the two models were observed for very hydrophobic PCB congeners such as PCB 171, 200, and 206.Although other models worked well for these chemicals, the deviation for the TPM may be attributed to the poor quality of the reported experimental plastic-water partition coefficients or to factors other than partitioning responsible for the exchange of chemicals between the polystyrene and water phases.
The performance of the target PVC model and ASM in predicting LC 50 values for compounds from various chemical families was evaluated.Overall, the comparison of the predictions showed that the agreement between the models was not favorable, with an RMSE of 1.67 log units.However, this comparison included strongly hydrophobic chemicals such as PCB congeners and DEHP (di-2-ethylhexyl phthalate), which showed the highest deviations, not only for the target PVC model but also for other models used in this study.In contrast, the target PVC model's performance was satisfactory for moderately hydrophobic PCB congeners.For DEHP, even the ASM failed, as the difference between the experimental and ASMpredicted LC 50 values was more than five orders of magnitude.For PCB 209, with a log   of 8.27, the deviations for the PVC model were also more than five orders of magnitude.For most of the pharmaceutical drugs, there was good agreement between the predicted LC 50 values of the PVC model and ASM.However, there were exceptions, such as flunitrazepam and chlorpromazine, which showed significantly higher residuals among the pharmaceutical drugs.Huge deviations between the predictions of two models for certain chemicals may be rationalized by considering specific toxic nature of the chemicals and/or the uncertain data quality stemming from the experimental challenges of measuring such chemicals in partitioning and toxicity experiments.For example, the ECOSAR class identified for DEHP is esters, implying it might be following a mode of toxic action other than baseline toxicity, and with a log   value of 7.6, it also belongs to the strongly hydrophobic category.Measuring physicochemical properties of strongly hydrophobic chemicals free from experimental artifacts is a challenging task.

S4
Similarly, flunitrazepam and chlorpromazine follow a mode of toxic action other than baseline, as indicated by their ECOSAR classes as amides and aliphatic amines, respectively.Such specific toxic modes of actions are difficult to account for by partitioning processes alone.
The performance of the target HDPE model in predicting LC 50 values for chemicals was also evaluated.The comparison of the predictions revealed that the predicted LC 50 by target HDPE model compared favorably with the ASM-predicted LC 50 , as shown by its RMSE of 0.38 log units.Other models also performed well for this chemical set.However, the dataset used for this evaluation comprised only hydrocarbons with moderate hydrophobicities (log   ranging from 2.73-5.81).These chemicals are known to follow the baseline toxic mode of action, and their measured partition coefficient values between the plastic and water phases are expected to not suffer too much from experimental artifacts.Therefore, the better performance of HDPE for such chemicals is not surprising.It is worth noting that the dataset used to evaluate the UHMWPE was not quite meaningful, as it only comprised three chemicals, with two belonging to the strongly hydrophobic category and one belonging to a mode of toxic action other than baseline toxicity.