Profiling of the Tox21 chemical collection for mitochondrial function to identify compounds that acutely decrease mitochondrial membrane potential

Abstract

Background: Mitochondrial dysfunction has been implicated in the pathogenesis of a variety of disorders including cancer, diabetes, and neurodegenerative and cardiovascular diseases. Understanding whether different environmental chemicals and druglike molecules impact mitochondrial function represents an initial step in predicting exposure-related toxicity and defining a possible role for such compounds in the onset of various diseases.

Objectives: We sought to identify individual chemicals and general structural features associated with changes in mitochondrial membrane potential (MMP).

Methods: We used a multiplexed [two end points in one screen; MMP and adenosine triphosphate (ATP) content] quantitative high throughput screening (qHTS) approach combined with informatics tools to screen the Tox21 library of 10,000 compounds (~ 8,300 unique chemicals) at 15 concentrations each in triplicate to identify chemicals and structural features that are associated with changes in MMP in HepG2 cells.

Results: Approximately 11% of the compounds (913 unique compounds) decreased MMP after 1 hr of treatment without affecting cell viability (ATP content). In addition, 309 compounds decreased MMP over a concentration range that also produced measurable cytotoxicity [half maximal inhibitory concentration (IC50) in MMP assay/IC50 in viability assay ≤ 3; p < 0.05]. More than 11% of the structural clusters that constitute the Tox21 library (76 of 651 clusters) were significantly enriched for compounds that decreased the MMP.

Conclusions: Our multiplexed qHTS approach allowed us to generate a robust and reliable data set to evaluate the ability of thousands of drugs and environmental compounds to decrease MMP. The use of structure-based clustering analysis allowed us to identify molecular features that are likely responsible for the observed activity.

Figure 1

qHTS concentration response data binned into curve classes 1–4. Concentration response [ratio (590 nm/535 nm)] curves for the FCCP control titrations and > 10,000 substances tested, including all the replicates. FCCP is the positive control, and curve class 4 represents the inactive compounds.

Figure 2

Screening statistics. (A) Stability analysis showing the reproducibility of each curve class, calculated using the ratio readout. High-quality curves (i.e., classes 1.1, 1.2, 2.1, 2.2) were more reproducible than other curve types for antagonist activity (inhibitory curve classes); a similar pattern was observed with lower reproducibility for the agonist curves. Class 4 inactive curves were fairly reproducible. (B) Distribution of the results (percent of the total library) based on the combined call using the MMP assay (ratio and independent channels), cell viability, and compound autofluorescence assays.

Figure 3

Structure–activity relationship (SAR) of compounds that decreased MMP. All compounds with associated structures present in the library were clustered, based on structural similarity using the self-organizing map (SOM) algorithm (Kohonen 2006). Each cluster was then evaluated for enrichment for active antagonists (compared with the library average) using Fisher’s exact test. Enriched clusters are shown in red and deficient clusters in blue; scale values represent the log p-value for each cluster. Representative scaffolds are shown for some of the more enriched clusters. The red boxes define regions of molecules that share a common substructure.

Figure 4

Activity of three hydroxybenzophenone isomers (A) and three parabens (B). (A) Changes in the ability of hydroxybenzophenone isomers to decrease MMP; 3-hydroxybenzophenone was the more potent isomer, and 2-hydroxybenzophenone was inactive in this assay. (B) Decreases in MMP for different paraben molecules; activity increased with increasing length of the side chain (and subsequent increase in hydrophobicity).