Leveraging Multiple Data Streams for Prioritization of Mixtures for Hazard Characterization

Abstract

There is a growing need to establish alternative approaches for mixture safety assessment of polycyclic aromatic hydrocarbons (PAHs). Due to limitations with current component-based approaches, and the lack of established methods for using whole mixtures, a promising alternative is to use sufficiently similar mixtures; although, an established framework is lacking. In this study, several approaches are explored to form sufficiently similar mixtures. Multiple data streams including environmental concentrations and empirically and predicted toxicity data for cancer and non-cancer endpoints were used to prioritize chemical components for mixture formations. Air samplers were analyzed for unsubstituted and alkylated PAHs. A synthetic mixture of identified PAHs was created (Creosote-Fire Mix). Existing toxicity values and chemical concentrations were incorporated to identify hazardous components in the Creosote-Fire Mix. Sufficiently similar mixtures of the Creosote-Fire Mix were formed based on (1) relative abundance; (2) toxicity values; and (3) a combination approach incorporating toxicity and abundance. Hazard characterization of these mixtures was performed using high-throughput screening in primary normal human bronchial epithelium (NHBE) and zebrafish. Differences in chemical composition and potency were observed between mixture formation approaches. The toxicity-based approach (Tox Mix) was the most potent mixture in both models. The combination approach (Weighted-Tox Mix) was determined to be the ideal approach due its ability to prioritize chemicals with high exposure and hazard potential.

Keywords: chemical prioritization; mixtures; mixtures safety assessment; polycyclic aromatic hydrocarbons; sufficiently similar mixtures.

Figure 1

Chemical Composition of Mixtures. Chemicals greater than 3% v/v in the mixture are shown in colored sections. Colors and area of color correspond to a given chemical and the %v/v of that chemical in its respective mixture. Anything less than 3% v/v is listed below its respective mixture figure. (A). 30 chemicals identified above detection limits out of 63 PAHs air samples were analyzed for. (B). Chemicals prioritized strictly based on abundance, top seven PAHs that made up over 97% v/v in Creosote-Fire Mixture were selected. (C). Chemicals prioritized using empirically derived and predicted toxicity metrics. (D). Chemicals prioritized by weighting empirical and predicted toxicity metrics with chemical concentrations. ¹ Full list of chemical composition in each mixture can be found in Tables S9 and S10. * 2-ethylnaphthalene was not included in Abundance Mix due to unavailability of the standard during synthesis of this mixture.

Figure 2

Comparison of Empirically Derived vs. QSAR Predicted Toxicity Values. Predicted RfD and RfC values were within the same magnitude as empirically derived values. IUR predictions were also within the same magnitude as empirical values. OSF had the largest difference between empirical and predicted values, which seemed to be associated with increasing molecular weight. *—empirically derived values. Reference Dose (RfD); Reference Concentration (RfC); Oral Slope Factor (OSF); Inhalation Unit Risk (IUR).

Figure 3

Correlations of Toxicity Metrics. (A) Correlation matrix of rankings for toxicity metrics used for Toxicity Mix. (B) Correlation matrix of toxicity metric rankings for Weighted-Toxicity Mix. Each square is labeled with correlation coefficient. Anything with an X was not significant. Significance cut-off was p < 0.05 using Pearson correlation coefficient. Correlations were conducted for each chemical only for complete observations.

Figure 4

Concentration-Response Curves for Mitochondrial Membrane Potential in NHBE. Curves are in order of decreasing potency based on predicted EC₅₀ values. (A) Toxicity Mix was the most potent mixture with an EC₅₀ of 50.5 µM. (B) Weighted-Toxicity Mix was the second most potent mixture with an EC₅₀ of 572 µM. (C) Abundance Mix was the least potent mixture with EC₅₀ of 1402 µM. (D) Creosote-Fire Mix did not elicit significant bioactivity compared to control, no EC₅₀ was calculated. Concentrations significantly different from control are denoted with an asterisk (*). p < 0.05 *; p < 0.01 **; p < 0.001 ***.

Figure 5

Concentration-Response Curves for Cell Viability in NHBE. Curves are in order from high to low potency based on predicted EC₅₀ values. (A) Toxicity Mix was the most potent mixture with an EC₅₀ of 31.9 µM. (B) Weighted-Toxicity Mix was the second most potent mixture with an EC₅₀ of 753 µM. (C) Abundance Mix was the least potent mixture with EC₅₀ of 1920 µM. (D) Creosote-Fire Mix had significant bioactivity for two highest doses, however, no EC₅₀ was calculated. Concentrations significantly different from control are denoted with an asterisk (*). p < 0.05 *; p < 0.01 **; p < 0.001 ***.

Figure 6

Concentration-Response Curves for Mortality and Any Effect in Zebrafish. (A) Percent incidence of Zebrafish mortality and predicted LC₅₀ and (B) Percent incidence of any effect and predicted EC₅₀ value. Toxicity Mix was the only mixture with significant bioactivity in zebrafish. Any concentration at or above binomial significance threshold are denoted with an asterisk (*).