Off-target effects of psychoactive drugs revealed by genome-wide assays in yeast

Abstract

To better understand off-target effects of widely prescribed psychoactive drugs, we performed a comprehensive series of chemogenomic screens using the budding yeast Saccharomyces cerevisiae as a model system. Because the known human targets of these drugs do not exist in yeast, we could employ the yeast gene deletion collections and parallel fitness profiling to explore potential off-target effects in a genome-wide manner. Among 214 tested, documented psychoactive drugs, we identified 81 compounds that inhibited wild-type yeast growth and were thus selected for genome-wide fitness profiling. Many of these drugs had a propensity to affect multiple cellular functions. The sensitivity profiles of half of the analyzed drugs were enriched for core cellular processes such as secretion, protein folding, RNA processing, and chromatin structure. Interestingly, fluoxetine (Prozac) interfered with establishment of cell polarity, cyproheptadine (Periactin) targeted essential genes with chromatin-remodeling roles, while paroxetine (Paxil) interfered with essential RNA metabolism genes, suggesting potential secondary drug targets. We also found that the more recently developed atypical antipsychotic clozapine (Clozaril) had no fewer off-target effects in yeast than the typical antipsychotics haloperidol (Haldol) and pimozide (Orap). Our results suggest that model organism pharmacogenetic studies provide a rational foundation for understanding the off-target effects of clinically important psychoactive agents and suggest a rational means both for devising compound derivatives with fewer side effects and for tailoring drug treatment to individual patient genotypes.

Figure 1. Workflow.

Schematic overview of the chemical genetic screening process.

Figure 2. Serotonergic Drugs Showed Potency on Yeast.

(A) Number of drugs that did (black) or did not (white) inhibit wildtype yeast growth for each of the initial drug sets tested. (B) Titration of drug concentrations used in genome-wide screening. Wildtype yeast growth in serial dilutions of drug was recorded as optical density every 15 min over a 25 h period. (C) Number of deletion strains that were sensitive (r>2 and z>3, see Materials and Methods) to bioactive drug in genome-wide fitness profiles.

Figure 3. Hydrophobicity and Molecular Weight Discrimination for Non-Active and Active Compounds.

All compounds tested were plotted as a function of logP and molecular weight.

Figure 4. Global Landscape of Fitness Profiles.

Two-dimensional hierarchical clustering was used to group all log₂-fitness ratios obtained from the 81 drugs. Log₂-fitness ratios from 0 (no fitness defect) to 3.5 (severe phenotype) are color-coded according to the severity of the sensitivity (this paper focuses on sensitivities, see Materials and Methods). Only 0.1% of the log₂-fitness ratios were higher than 3.5 and became saturated in the figure. The separation of dopaminergic and serotonergic drugs (orange) from drugs in other categories (grey) is indicated. Groups of strains exhibiting highly similar fitness profiles across the psychoactive drugs are extracted from the global clustergram, and the deletion strains included in each group are listed in the order determined by the hierarchical clustering algorithm. For each group of strains, the dominant function(s) of the deleted genes is indicated. Essential genes are underlined.

Figure 5. Enriched GO Processes.

Significant enrichment (p<0.0001) for GO Processes uniquely scored using sensitive (z>2) homozygous strains (blue) or heterozygous strains deleted for essential genes (red). GO Processes scored using both strain pools are indicated in purple (for details see Materials and Methods). Dopaminergic drugs are indicated in green, serotonergic in orange, and other drugs (from the initially analyzed diverse set) in grey. Drugs with affinity for both a dopaminergic and a serotonergic receptor are indicated according to which Tocris drug library they belong to. Closely related GO categories are collapsed for clarity (see Table S6).