Systems toxicology assessment of a representative e-liquid formulation using human primary bronchial epithelial cells

Abstract

The development of reduced-risk products aims to provide alternatives to cigarettes that present less risk of harm for adult smokers. Responsible use of flavoring substances in these products may fulfill an important role in product acceptance. While most flavoring substances used in such products are also used by the food industry and are considered safe when ingested, their impact when inhaled may require further assessment. To aid in such an assessment, a three-step approach combining real-time cellular analysis, phenotypic high-content screening assays, and gene expression analysis was developed and tested in normal human bronchial epithelial cells with 28 flavoring substances commonly used in e-liquid formulations, dissolved individually or as a mixture in a base solution composed of propylene glycol, vegetable glycerin, and 0.6% nicotine. By employing this approach, we identified individual flavoring substances that potentially contribute greatly to the overall mixture effect (citronellol and alpha-pinene). By assessing modified mixtures, we showed that, although cytotoxic effects were found when assessed individually, alpha-pinene did not contribute to the overall mixture cytotoxicity. Most of the cytotoxic effect appeared to be attributable to citronellol, with the remaining substances contributing due to synergistic effects. We developed and used different scoring methods (Tox-Score, Phenotypic Score, and Biological Impact Factor/Network Perturbation Amplitude), ultimately enabling a ranking based on cytotoxicity, phenotypic outcome, and molecular network perturbations. This case study highlights the benefits of testing both individual flavoring substances and mixtures for e-liquid flavor assessment and emphasized the importance of data sharing for the benefit of consumer safety.

Keywords: Bronchial epithelial cells; Electronic cigarettes; Flavoring substances; High-content screening; Systems toxicology.

Graphical abstract

Fig. 1

The Flavor Toolbox: a three-step workflow to assess the toxicity of flavored solutions in NHBE cells. The Flavor Toolbox is a complementary approach to standard toxicity assays (e.g., Ames assay, mouse lymphoma assay, etc.). It is designed to screen a large number of e-liquids for potential toxicity prior to performing whole aerosol assessment on human organotypic tissue cultures. The Flavor Toolbox workflow comprises the following three steps: (i) STEP1, which quantifies the toxicity of the exposure using a real-time impedance-based measurement, expressed as Tox-Score; (ii) STEP 2, which measures and investigates the phenotypic impact of the exposure using HCS image analysis; and (iii) STEP 3, which combines transcriptomic data and computable biological networks in a systems toxicology approach. For each STEP, a computationally derived score (Tox-Score, Phenotypic Score, and NPA/BIF score) was derived to quantify the e-liquids’ exposure effects.

Fig. 2

(A) RTCA-based cell viability dose response curves for the 28-flavor mixture (red dots) and the corresponding base solution (blue dots). Green line corresponds to 50% of cell viability and dotted lines indicate extrapolated EC₅₀ values. (B, C) Tox-Scores (y-axis) are shown as a function of their corresponding p-values computed for (B) each individual flavoring substance present in the 28-flavor mixture and (C) the flavor mixtures without citronellol, alpha-pinene, or both after a 24-hour exposure. Red dot in b and c indicate flavoring substances and mixtures selected for subsequent HCS investigation. The vertical line indicates a p-value of 0.05. Each dot corresponds to one flavor solution, with those selected for subsequent HCS-based investigation shown by white dots. Abbreviation: SEM, standard error of the mean.

Fig. 3

Circular bar plots of HCS endpoint MEC ratios for (A) the 28-flavor mixture, (B) various individual flavoring substances with different Tox-Scores, and (C) flavor mixture without citronellol and/or alpha-pinene. Each MEC ratio (reported next to each segment of the circular chart for each HCS endpoint) was computed by dividing the mean base solution MEC (from n = 3 replicates) by the mean flavor mix MEC. A unilateral t-test was computed with null hypothesis: The base solution MEC mean is higher than the flavor mix MEC mean. The t-test p-values are reported as follows: *** <0.001, ** <0.01, and * <0.05. The “- “sign on top of an MEC ratio denotes an imputed MEC value. Red circles correspond to an MEC of 1. Abbreviations: pH2AX, phosphorylated H2A histone family member X; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; ROS, reactive oxygen species.

Fig. 4

Bar plots (upper panel) represent relative BIF (RBIF) values (indicated above each bar) for NHBE cells exposed to the 28 flavor mixture or base solution using Inflammatory Processes, Cell Stress, Cell Proliferation, and Cell Fate networks. The percentages indicate the relative biological impact derived from the cumulated network perturbations caused by the treatment relative to the reference (REF, defined as the treatment comparison showing the highest perturbation; REF = 100%). For each treatment comparison, the δ value (−1 to 1) indicates how similar the underlying network perturbations are with respect to the REF. A δ value of 0 indicates no similarity of the perturbed networks, a δ value of 1 indicates identical network perturbations, and a δ value of −1 indicates a completely opposite network responses. The star plots (lower panel) represent the contributions of each network family (Inflammatory Processes, Cell Stress, Cell Proliferation, and Cell Fate) to the RBIF for each treatment. The sum of the segment is equal to the BIF score for the treatment. Percentage values in black indicate the dilution tested.

Fig. 5

NPA heatmap of the subnetworks impacted by the base solution alone or by the flavor mixture at two dilutions (0.25% and 0.50% v/v) at the 4- and 24 -h time points. A network is considered perturbed if, in addition to the significance of the NPA score with respect to the experimental variation, the two companion statistics (O and K) that report the specificity of the NPA score with respect to the biology described in the network, are also significant (as indicated by an asterisk). The darker the color the stronger the perturbations. Abbreviation: IPN: Inflammatory Processes Network.

Fig. 6

Flavoring substances assessment workflow. This diagram shows how the flavor mixture assessment was performed in the present study (A) and how it will be readapted based on the key learnings (B). Abbreviations: RTCA, real-time cell analysis; HCS, high-content screening; GEX, gene expression.