Many psychoactive drugs are associated with adverse secondary effects in humans yet the mechanisms that underlie these off-target effects are poorly understood. To address mechanisms of drug action in a systematic manner, we profiled the genome-wide collection of budding yeast deletion strains for sensitivity to a broad spectrum of psychoactive compounds, of which dopaminergic and serotonergic drugs had a high bioactivity. Among 214 tested compounds, we uncovered 81 drugs that conferred a measurable growth defect on wildtype yeast. An appropriate dose of these active compounds was applied to the pooled heterozygous and homozygous yeast deletion sets to identify genes whose function is required for optimal growth in the presence of drug. Fifteen percent of all yeast strains (deleted for non-dubious ORFs) exhibited significant sensitivity (r>2, z>3) to these 81 psychoactive compounds and more than half of the drugs interacted with core cellular functions. Several clinically important drugs, such as fluoxetine, cyproheptadine, and clozapine were linked to diverse cellular processes. This observation may explain both the diversity of side effects observed in human patients and the therapeutic variability associated with these drugs. That is, polymorphisms in any of the conserved processes affected by a given drug are a likely source of the individual variation in response to drug. For instance, the response to the frequently prescribed antipsychotic clozapine is highly variable between individuals as the same dose can have markedly different efficacy and/or side effects in different patients 
. Genes functioning in vesicle transport, protein localization, telomere biology, and catabolic processes were required for clozapine resistance in yeast. In another example, fluoxetine is associated with side effects such as seizures, nausea, sleepiness, anxiety, and serious allergic reactions. This antidepressant affects numerous cellular processes including establishment of cell polarity, protein localization, and cytoskeleton organization and biogenesis. Given the limited number of FDA-approved drugs within the set of 81 compounds analyzed here and the overlapping side effects associated with these drugs, it is not yet possible to correlate any single side effect to a particular perturbed pathway.
The most frequently scored sensitivity for the 81 profiled antipsychotic drugs was due to loss of secretory pathway function, likely indicating the importance of vesicle transport (e.g.
to the vacuole) for drug detoxification. The lysosome (the mammalian vacuole equivalent) is known as the major site of degradation of both exogenous and endogenous molecules. For FDA-approved drugs, the requirement for vesicle transport genes was reflected in the frequent sensitivity of the neo1
deletion strain as the most sensitive strain in six FDA-approved drugs. Neo1 is an essential, highly conserved type 4 P-type ATPase involved in intracellular membrane- and protein-trafficking. Members of this family of P-type ATPases are implicated in the translocation of phospholipids from the outer to the inner leaflet of membrane bilayers. Our data suggested that interference with membrane structure and transport through inhibition of Neo1 is an additional, unwanted mechanism of action for clozapine, cyproheptadine, fluoxetine, paroxetine, sertraline and haloperidol, and their drug analogs. The importance in humans of functional 4 P-type ATPases is well documented as hereditary cholestasis, caused by defects in biliary epithelial transporters, has been directly linked to mutations in a 4 P-type ATPase gene 
In addition to the frequently observed requirement for uncompromised vesicle transport for drug detoxification, several drug sensitivity profiles were enriched for more specific processes. Within the FDA-approved drug group, the antidepressant paroxetine was unique in targeting RNA processing genes, pimozide interfered with membrane lipid metabolic processes, cyproheptadine preferentially targeted essential genes with chromatin remodelling functions, and fluoxetine interfered with establishment of cell polarity. Furthermore, seven dopaminergic compounds including the anti-Parkinson drug bromocriptine resulted in sensitivity of strains deleted in aromatic amino acid biosynthetic genes. This sensitivity may be a result of that dopaminergic drugs block aromatic amino acid uptake in yeast, requiring yeast to activate the corresponding biosynthetic pathways. Given the fact that aromatic amino acids are precursors to dopamine and serotonin, this was an interesting observation suggesting that the levels of intracellular precursors may be important in the response to certain psychoactive drugs.
Interestingly, interference with members of the chaperonin complex resulted in some of the most severe phenotypes. Seven of eight CCT-strains scored as significantly sensitive in several psychoactive drugs, among them CCT5
. The human homolog of this gene is associated with hereditary neuropathy 
. Although it is unclear how mutated CCT5
causes this disease, it has been postulated that its mutation leads to accumulation of misfolded cytoskeletal proteins, leading to defective assembly of actin into microfilaments resulting in neuronal apoptosis 
. In our yeast screens, CCT5
was needed for resistance to eight different compounds (cyproheptadine, paroxetine, fluoxetine, indatraline, MDL72222, CY208-243, 2-Chloro-11-(4-methylpiperazino)-dibenz[b,f]oxepin, N-Desmethyl-clozapine, and 3-alpha-[(4-Chlorophenyl)-phenylmethoxy]-tropane. We conclude that interference with tubulin and actin folding is an important, secondary mechanism of action of these compounds.
As an example of how the information from our yeast assays may lead to testable drug-gene interaction hypotheses in humans, we found that the levels of the yeast strain heterozygous for ACC1
was eleven-fold reduced in ritanserin as compared to the control, indicating that the acetyl-CoA carboxylase Acc1 may be a secondary target of ritanserin. Like its yeast counterpart, the human homolog ACACA
is required for de novo
biosynthesis of long-chain fatty acids and its activity drops during fasting 
. Because increased appetite is a reported side-effect during ritanserin treatment 
, it is tempting to speculate that biochemically mimicking fasting would increase appetite.
These studies raise several important issues for further consideration. Understanding the mechanisms that underlie adverse effects of clinically approved drugs is crucial for the development of next generation therapeutics with improved selectivity and efficacy. Moreover, knowledge of patient polymorphisms in off-target pathways may allow adverse effects of any given drug to be preempted by personalized pharmacogenomic strategies. It is also conceivable that some of the observed secondary drug effects are critical for therapeutic benefit.
In summary, a number of cellular processes were associated with sensitivity to the dopaminergic and serotonergic classes of psychoactive compounds. This points to additional, previously uncharacterized mechanisms of action for these drugs in humans and suggests follow-up experiments aimed at understanding a drug's mechanism of action on a genome-wide level. Our results suggest that model organism pharmacogenetics can be used as a comprehensive and unbiased tool in initial studies aiming at unraveling secondary effects and mechanisms of action for therapeutic compounds and their analogs. A more rigorous understanding of the complete mechanism of drug action in humans would be beneficial in the development of a new generation of better tolerated psychoactive drugs, and in personalized medicine.