High-throughput screening for fatty acid uptake inhibitors in humanized yeast identifies atypical antipsychotic drugs that cause dyslipidemias.

Fatty acids are implicated in the development of dyslipidemias, leading to type 2 diabetes and cardiovascular disease. We used a standardized small compound library to screen humanized yeast to identify compounds that inhibit fatty acid transport protein (FATP)-mediated fatty acid uptake into cells. This screening procedure used live yeast cells expressing human FATP2 to identify small compounds that reduced the import of a fluorescent fatty acid analog, 4,4-difluoro-5-methyl-4-bora-3a,4a-diaza-s-indacene-3-dodecanoic acid (C1-BODIPY-C12). The library used consisted of 2,080 compounds with known biological activities. Of these, ∼1.8% reduced cell-associated C1-BODIPY-C12 fluorescence and were selected as potential inhibitors of human FATP2-mediated fatty acid uptake. Based on secondary screens, 28 compounds were selected as potential fatty acid uptake inhibitors. Some compounds fell into four groups with similar structural features. The largest group was structurally related to a family of tricyclic, phenothiazine-derived drugs used to treat schizophrenia and related psychiatric disorders, which are also known to cause metabolic side effects, including hypertriglyceridemia. Potential hit compounds were studied for specificity of interaction with human FATP and efficacy in human Caco-2 cells. This study validates this screening system as useful to assess the impact of drugs in preclinical screening for fatty acid uptake.

During the past 20 years, obesity among adults and children has increased significantly in the United States. Being overweight or obese increases the risk for the development of adverse health conditions, including hyper-tension, dyslipidemia, diabetes, cardiovascular disease, and certain types of cancer, particularly colon and breast (1)(2)(3)(4)(5)(6). It is hypothesized that high circulating levels of free fatty acids, occurring as a result of various metabolic disturbances, including overeating and diabetes, lead to excessive fatty acid internalization. Cell death results when the accumulation of these free fatty acids, especially saturated fatty acids, exceeds the cell's capacity for triglyceride storage, resulting in high levels of fatty acid oxidation, oxidative stress, and ceramide production (7,8). This is termed lipotoxicity and is correlated with the development of obesity, insulin resistance, and cardiovascular disease (9)(10)(11)(12). Therefore, it is desirable to target factors that contribute to fatty acid internalization and accumulation to prevent lipotoxicity.
Several lines of evidence favor the involvement of proteins in the import of exogenous fatty acids. First, fatty acid transport into cells is regulated by insulin, and emerging evidence shows that key transcriptional regulators, including peroxisome proliferator-activated receptor a, sterol-regulatory element binding protein 1, and retinoid X receptor, participate in these regulatory processes (13)(14)(15)(16). Second, cells with high levels of fatty acid metabolism (either degradation or storage) transport fatty acids at higher rates compared with those with low levels of fatty acid metabolism (17,18). Third, fatty acid transport is inducible in several different cell types and is commensurate with the expression of specific sets of proteins that are likely to participate in this process (19)(20)(21). Among these are the fatty acid transport proteins (FATPs) and long-chain acyl-coenzyme A synthetases (ACSLs) (22)(23)(24). The involvement of both FATP and ACSL is consistent with the model of vectorial acylation, in which fatty acid uptake is coupled to esterification with CoA (25)(26)(27)(28). Our own work exploiting the genetically tractable yeast Saccharomyces cerevisiae has demonstrated that a membranebound FATP homolog, Fat1p, and an ACSL, either Faa1p or Faa4p, interact physically as well as functionally to catalyze import and activation (reviewed in Ref. 29). Fat1p and ACSL are both essential when cells are grown under hypoxic conditions, which render yeast auxotrophic for long-chain unsaturated fatty acids (30). Likewise, specific interactions have also been demonstrated between FATP1 and ACSL1 to potentiate fatty acid uptake in murine adipocytes (28). A number of studies have shown that fatty acid transport, particularly for protonated fatty acids, may also occur by diffusion or flip-flop between the two faces of the membrane (31)(32)(33)(34)(35)(36)(37). Using adipocytes, it was demonstrated that the association and flip-flop of oleate across the plasma membrane occurred very rapidly (33). A recent review of the diffusional model of fatty acid transport suggests that flip-flop is tied to the metabolic conversion to acyl-CoA, consistent with our model of vectorial acylation (32). Together, these reports provide compelling evidence for targeting fatty acid transport to limit the cellular uptake of fatty acids and thus lipotoxicity.
Previously, we developed a live cell high-throughput screening (HTS) assay using a fat1Dfaa1D yeast strain to identify chemical compounds that will inhibit fatty acid uptake (P. N. Black and C. C. DiRusso, United States Patent 7,070,944) (38). In this system, transport is a function of mouse FATP2 expressed from a yeast promoter and longchain fatty acid activation is provided by the endogenous ACSL Faa4p. We demonstrated that this HTS assay is rapid to execute, inexpensive to implement, and has appropriate sensitivity for HTS. In the present work, we conducted a pilot study using a 2,080 compound library from MicroSource Discovery Systems, Inc. (Gaylordsville, CT), called SpectrumPlus: to improve and validate this live cell HTS method. We also devised secondary screens to eliminate false-positives in the yeast system and further evaluated the potential hit compounds using the human Caco-2 cell line. The members of the largest family of structurally related compounds identified in this screen are derivatives of a tricyclic phenothiazine core from which atypical antipsychotic drugs have been developed. Among these are chlorpromazine and the related compound clozapine. Many of the drugs in this class cause adverse side effects, including weight gain, hypertriglyceridemia, hyperglycemia, and ketoacidosis, which in severe cases have led to patient death (reviewed in Ref. 39). Consequently, several of these drugs have received black box warning labels. The mechanism by which these drugs cause these clinical symptoms is unknown. Our identification of these compounds as inhibitors of fatty acid uptake leads us to suggest that the metabolic dysregulation associated with the administration of these drugs is caused in whole or in part by reduced fatty acid uptake, resulting in the disruption of normal cellular lipid trafficking.

Chemical compound library
The SpectrumPlus compound library (2,080 compounds) was obtained from MicroSource Discovery Systems, Inc. The library includes five subsets of compounds: 1) Genesis Plus, composed of 960 compounds that represent new and classical therapeutic agents as well as established experimental inhibitors and receptor agonists; 2) Pure Natural Products Collection, a unique collection of 720 diversified pure natural products and their derivatives, including simple and complex oxygen heterocycles, alkaloids, sequiterpenes, diterpenes, pentercyclic triterpenes, sterols, and many other diverse representatives; 3) Agro Plate, containing 80 compounds representing classical and experimental pesticides, herbicides, and purported endocrine disruptors; 4) Cancer Plate, including 80 cytotoxic agents, antiproliferative agents, immune suppressants, and other experimental and therapeutic agents; and 5) Spectrum Plus Plate, containing 240 biologically active and structurally diverse compounds. The compounds are supplied as 10 mM solutions in DMSO and were diluted in PBS to a final concentration of 80 mM for screening in yeast using a Caliper RapidPlate 96/384 Dispenser (Caliper Life Sciences, Hopkinton, MA).

Reagents and cell culture
Yeast extract, yeast peptone, and yeast nitrogen base without amino acids and dextrose were obtained from Difco (Detroit, MI). Complete amino acid supplement and single amino acids were from Sigma (St. Louis, MO). Fatty acid-free bovine serum albumin (FAF BSA) and other chemical reagents were also from Sigma, unless stated otherwise. Corning: white with clear flatbottom 96-well and 384-well microplates were used for screening of yeast. For Caco-2 cells, tissue culture-treated 96-well transparent with clear flat-bottom polystyrene microplates were from Fisher Scientific (Pittsburgh, PA). The fluorescent long-chain fatty acid analog, 4,4-difluoro-5-methyl-4-bora-3a,4a-diaza-s-indacene-3dodecanoic acid (C 1 -BODIPY-C 12 ), was obtained from the Molecular Probes Division of Invitrogen.
Caco-2 cells were maintained in Earl's minimal essential medium with 20% FBS in a 95% air/5% CO 2 atmosphere at 37jC, as described previously (40). For growth and differentiation, the BD Biosciences Intestinal Epithelium Differentiation Media Pack was used. Cells were plated in basal seeding medium at a density of 2.5 3 10 5 cells/cm 2 on a collagen-coated blackclear 96-well plate (BD Biosciences). Entero-STIM medium was added to each well after removal of the basal seeding medium 24 h later. Both media contained mito-serum extender. After another 24 h, cells were washed once with complete balanced Hanks' solution containing Ca 21 and Mg 21 without phenol red before the C 1 -BODIPY-C 12 uptake assay.

Cloning and expression of human FATP2 in yeast
For expression of the human FATP2 in yeast, the cDNA encoding the protein was amplified using PCR from a cDNA within NIH Image Clone I.D. 30348317 using forward primer 5 ¶-gcatagccctcgagatgctttccgccatctacaca-3 ¶ and reverse primer 5 ¶-cgcagaccaagctttcagagtttcagggtttttagc-3 ¶. The amplified DNA was cloned into the yeast expression vector pDB121 to give pDB126. In this construct, the expression of targeted human FATP2 is under the control of the GAL10 promoter (41). The human FATP2 plasmid and pRS416GAL4-ER-VP16 encoding a synthetic transcriptional activator, which is a protein fusion between the Gal4 DNA binding domain, a b-estradiol-responsive regulatory domain, and a VP16 RNA polymerase activation domain (42), were cotransformed into strain LS2086 for HTS assays. Expression of human FATP was induced by the growth of yeast cells overnight in selective medium containing 10 nM b-estradiol. The level of human FATP2 protein was routinely measured using an antibody directed toward a T7 epitope fused to the C terminus of human FATP2 in Western blots of cell extracts.

HTS
For screening of the 2,080 compound MicroSource library, a Biomek: FX laboratory automated workstation (Beckman Coulter, Inc., Fullerton, CA) was used. LS2086 yeast cells expressing human FATP2 (pDB126/fat1Dfaa1D) or transformed with the empty vector pDB121 (negative control cells) were pregrown in YNBD without leucine and uracil (YNBD 2leu 2ura); the cells were subsequently subcultured to absorbance at 600 nm 5 0.02 in the same medium containing 10 nM b-estradiol to induce expression and incubated overnight at 30jC with shaking. Mid-log-phase yeast cells [0.8-1.2 optical density at 600 nm (OD 600 )] were harvested and resuspended in PBS at a cell density of 6 3 10 7 /ml before dispensing to a 384-well assay plate (22.5 ml/well). Wells in the first two rows of each plate received the negative control cells, and all other wells received cells expressing human FATP2. Compounds (2.5 ml) were then added to a final concentration of 80 mM. After 2 h of incubation at 30jC, the C 1 -BODIPY-C 12 uptake assay was performed. To each well, a mixture of C 1 -BODIPY-C 12 (final concentration, 1.25 mM), FAF BSA (final concentration, 0.75 mM), and trypan blue (final concentration, 2.1 mM) was added to give a final volume of 100 ml. Trypan blue quenched the extracellular fluorescence, and only cell-associated fluorescence was measured after 30 min on a Bio-Tek Synergy HT multidetection microplate reader (Bio-Tek Instruments, Inc., Winooski, VT) with filter sets of 485 6 20 nm excitation and 528 6 20 nm emission. On each plate, two rows each of negative control cells and positive control cells were incubated with 0.8% DMSO in PBS and assayed simultaneously to calculate a Z ¶ factor for each plate. The results presented are the values obtained for two screening experiments (see Tables 1, 3-7 below).

Determination of inhibition constant
Ligand competition curves were fit by nonlinear least-squares regression using a one-site competition model and Prism software (GraphPad Software, Inc., San Diego, CA) to determine the concentration of added competitor that reduced C 1 -BODIPY-C 12 fluorescence readout by 50% (IC 50 ). Inhibition constant (K i ) values were calculated from the IC 50 by the equation of Cheng and Prusoff (43): where radioligand is the concentration of C 1 -BODIPY-C 12 and K d is the apparent affinity of C 1 -BODIPY-C 12 for the yeast cells expressing human FATP2 or Caco-2 cells. The apparent K d was determined by measuring uptake activity in the presence of different C 1 -BODIPY-C 12 concentrations.

Toxicity assays
Wild-type yeast strain YB332 was used to evaluate the compound toxicity. Cells were grown in YPDA overnight and sub-cultured into the same medium at 0.1 OD 600 in the presence of different concentrations of compound as indicated. Optical density was measured every hour during 8 h of growth. To test viability after treatment, cells were grown and treated with compound for 2 h as usual to test uptake inhibition but were harvested by centrifugation, washed three times with PBS, serially diluted, and plated on YPDA. After growth for 2 days at 30jC, colony numbers were counted and scored to give the number of colony-forming units per milliliter.
To evaluate potential toxic effects on Caco-2 cells, we used the MTT [for 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay (44). For these experiments, cells were treated with compound exactly as in the uptake assay using a 96-well format. At the end of the incubation period with test compound, the medium was removed and cells were treated with 100 ml of MEM containing 1.2 mM MTT for 3 h at 37jC. The reaction was terminated and cells were lysed by the addition of 5% SDS and 0.005 N HCl with incubation at 37jC for 1 h. The final absorbance was read at 570 nm to detect the formation of the purple formazan product. Cell viability was determined relative to a standard curve prepared using untreated cells.

Lipid analyses
To identify lipids labeled with C 1 -BODIPY-C 12 , LS2086 yeast cells (pDB126/fat1Dfaa1D) were grown in 50 ml of YNBD (2leu, 2ura) containing 10 nM b-estradiol to 1.0 OD 600 at 30jC. The culture was then concentrated by centrifugation and resuspended in PBS to a cell density of 6 3 10 7 /ml. To half of the cell suspension, perphenazine (in PBS) was added to a final concentration of 40 mM; a control flask received PBS alone. Incubation with the compound continued for 2 h, then C 1 -BODIPY-C 12 (final concentration, 1.25 mM) in FAF BSA (final concentration, 0.75 mM) was added for 30 min to evaluate uptake. Metabolism was stopped by the addition of 0.1 volume of 6.6 M perchloric acid. Lipids were extracted essentially as detailed previously (45). After extraction, the dried lipids were resuspended in 50 ml of chloroform, and aliquots (5 ml) were spotted to Whatman Partisil LK5 TLC plates (250 mm, 150 Å silica). To resolve phospholipids, the solvent was a mixture of 25 ml of methyl acetate, 25 ml of 2-propanol, 28 ml of chloroform, and 7 ml of 0.25% KCl (solvent I); to resolve neutral lipids, the solvent consisted of three parts cyclohexane to two parts ethyl acetate (solvent II). All solvents were of HPLC grade and were obtained from Fisher Scientific. To detect C 1 -BODIPY-C 12 , the plates were scanned on a Bio-Rad Molecular Image FX scanner equipped with an external laser using filter sets of 488 nm excitation and 530 nm emission. Positions of the labeled lipids were compared with those of verified lipid standards (Sigma) resolved on the same TLC plates.

Visualization of C 1 -BODIPY-C 12 uptake into Caco-2 cells
Caco-2 cells were cultured as detailed above on four-well tissue culture-treated microscope slides (BD Falcon; BD Biosciences); two wells were used as controls and two wells were used for compound treatment. Treatment of cells with 80 mM perphenazine for 1 h and uptake of C 1 -BODIPY-C 12 were essentially as detailed for HTS with the following modifications. After treatment with compound or MEM (control), medium was removed and cells were washed with MEM. C 1 -BODIPY-C 12 (final concentration, 5 mM in 5 mM FAF BSA) was added for 3 min without trypan blue. Cells were then immediately rinsed with MEM, and Elvanol mounting medium was applied to each well. Cells were imaged using a Zeiss LSM510 confocal laser scanning microscope (Carl Zeiss MicroImaging, Inc., Thornwood, NY) equipped with a 633 objective using fluorescence modes. The instrument settings for brightness and contrast were optimized to ensure that the confocal microscope was set for its full dynamic range relative to untreated cells on the same slide; the same settings were used for all subsequent image collections. An argon laser source was used for imaging, with excitation at 488 nm and emission at 505 nm.

Data analysis
Values in arbitrary units of the fluorescent signals from each HTS plate were exported to Excel (Microsoft Corp., Redmond, WA) spreadsheet templates, and the assay quality-control Z ¶ factor was calculated (46). ChemTree software (Golden Helix, Inc., Bozeman, MT) was used for additional statistical analysis and to study structure-activity relationships.

Library screening
In previous work, we developed a live cell, highthroughput method for the selection of small chemical modulators of fatty acid uptake dependent upon a mu-rine FATP expressed in a S. cerevisiae strain deficient in the long-chain fatty acid transporter Fat1p and the major ACSL Faa1p (fat1Dfaa1D) (38). In this study, this strain was used to express the human FATP2 under the control of the GAL10 promoter. Fatty acid uptake was measured as an increase in cell-associated fluorescence attributable to the accretion of the fluorescent fatty acid analog C 1 -BODIPY-C 12 through a human FATP2-dependent process. External fluorescence was quenched by the vital dye trypan blue. We used this system to screen a 2,080 chemical compound library, SpectrumPlus (MicroSource), to select for compounds that altered fatty acid uptake. This library consists of compounds that are structurally diverse, including known drugs, experimental actives, and pure natural products. The collection is advantageous in that the compounds included have been chosen for diversified structures containing unique structural scaffolds. Additionally, most compounds included in the set have known pharmacological and toxicological profiles that can be used to determine whether the hit compound or structur- ally similar compounds can be safely and effectively prescribed for additional medicinal uses.
In our HTS, library compounds were tested at 80 mM on 384-well plates with yeast cells expressing the human FATP2. The fatty acid uptake activity of the cells was determined by comparing the cell-associated fluorescence of cells expressing human FATP2 with that of cells carrying the vector (pDB121) after incubation with the fatty acid analog C 1 -BODIPY-C 12 (38). The Z ¶ factors were calculated from values obtained from positive and negative controls (two lanes of each per plate) and ranged from 0.61 to 0.80, indicating a robust assay with good discrimination between the positive and negative controls (47). Two screening trials were performed, and there was excellent consistency between the two experiments. Potential inhibitors were identified using a cutoff value of 3 SD from the mean of the positive and negative controls ( Table 1). As expected, most compounds (98%) had no effect on C 1 -BODIPY-C 12 uptake (Fig. 1). Four compounds enhanced the fluorescence signal, and 39 compounds were identified as potential inhibitors, as they reduced the cell-associated fluorescence compared with that of the cells incubated with vehicle alone. Of the 39 compounds, 22 (1.06%) reduced fluorescence by 80-100% and 17 (0.82%) gave intermediate levels (40-79% compared with positive controls).

Secondary screens eliminate autofluorescent and quenching compounds
This primary screening does not distinguish between effectors that alter apparent fluorescence readings by nonspecific mechanisms, apart from those that interact directly with FATP. Nonspecific effectors might include those that increase apparent fluorescence as a result of intrinsic properties specific to the compound or those compounds that act as quenching agents. Additionally, some compounds may disrupt membrane integrity and facilitate the internalization of trypan blue, which in turn quenches the imported fluorescent ligand. Therefore, all compounds identified as hits in the primary screen were subsequently tested for autofluoresence, the ability to quench C 1 -BODIPY-C 12 fluorescence, and the ability to disrupt membrane integrity ( Table 2).
Two compounds identified in the primary screen, acriflavinium hydrochloride and calcein, yielded high intrinsic fluorescence that was not quenched by trypan blue, suggesting that these compounds did not enhance fatty acid uptake but increased fluorescence by nonspecific mechanisms (Fig. 1, Table 2). Sixteen of the 39 compounds that were hits from the primary screening were colored. These colored compounds might cause an inner filter effect, reducing the fluorescence signal of C 1 -BODIPY-C 12 as a result of overlap of their absorption spectrum with the emission spectrum of the BODIPY fluorophore. To test this concern, we determined the absorp- Fig. 1. High-throughput screening for fatty acid transport inhibitors. Cellular accumulation of 4,4-difluoro-5-methyl-4-bora-3a,4adiaza-s-indacene-3-dodecanoic acid (C 1 -BODIPY-C 12 ) was measured in arbitrary fluorescence units (AFU) for yeast cells expressing human fatty acid transport protein 2 (FATP2) after incubation with compounds (diamonds) or with DMSO (triangles; positive control). Accumulation of C 1 -BODIPY-C 12 in cells with empty vector was used as the negative control (squares). Compounds (80 mM in 0.8% DMSO) or DMSO alone were incubated with yeast cells expressing human FATP2 for 2 h at 30jC before C 1 -BODIPY-C 12 uptake assay. Data points are from one trial.  tion spectrum of these compounds from 300 to 800 nm and found two structurally related compounds, gentian violet (purple) and pararosaniline pamoate (red), that have absorption maxima at 590 and 550 nm, respectively. Therefore, they most likely quench C 1 -BODIPY-C 12 fluorescence (emission at 510-625 nm). As expected, when these compounds were added directly to 5 mM C 1 -BODIPY-C 12 in the absence of cells, the fluorescence was reduced significantly (Table 2). These results showed that both compounds acted as quenching agents to cause the apparent reduction in cell-associated C 1 -BODIPY-C 12 fluorescence. Another compound, celastrol (orange), also modestly quenched the BODIPY fluorescence. Compounds that act as membrane permeants will decrease cell-associated fluorescence because they will allow the vital dye trypan blue to enter the cells. Among these, the common detergent digitonin (48) is found in the SpectrumPlus library and was identified by its ability to reduce fluorescence to background levels. To eliminate compounds that disrupt membrane integrity, cells were preloaded with C 1 -BODIPY-C 12 in the presence of trypan blue for 30 min; each compound was added and cells were incubated for 2 h, then we revaluated cell-associated fluorescence. A reduction in fluorescence intensity after the addition of compound compared with fluorescence intensity before the addition of compound was interpreted as indicative of a loss of membrane integrity. Among the 39 compounds, only the cationic detergent digitonin acted as a strong membrane-permeabilizing agent using this test (Tables 1, 2). The SpectrumPlus library also contains at least one other cationic detergent, polymixin B sulfate. This compound weakly inhibited fatty acid uptake (?30% inhibition compared with positive controls) and was not included in the selected list of hits from our primary screen. Other compounds with known detergent-like properties, including centrimonium bromide (described below), did not affect fluorescence readings of cells preloaded with C 1 -BODIPY-C 12 .

Structural analysis of hit compounds in yeast
After eliminating the autofluorescent compounds and quenching and permeabilizing agents, we compared the structures of the remaining compounds to identify similar structural cores using ChemTree software. Using a multiple tree clustering algorithm, ChemTree automatically groups compounds that act according to similar quantitative structure-activity relationship mechanisms. Four groups of compounds that inhibited C 1 -BODIPY-C 12 uptake were thus identified as having similar structural scaffolds. Members of the first set are structurally related to the natural ligands of FATP, fatty acids, in that each has a long aliphatic hydrocarbon chain, including centrimonium bromide and benzalkonium chloride ( Table 3). However, the hydrocarbon chain alone was not sufficient for the inhi- bition of C 1 -BODIPY-C 12 uptake, because at least five other compounds with this structural motif in the SpectrumPlus collection were inactive: batyl alcohol, roccellic acid, vocadyne acetate, avocadyne acetate, and caperatic acid. One compound, embelin, had intermediate activity. The inactive compounds have the common characteristic of being highly charged, with multiple hydroxyl groups on the terminal carbon of the chain that may preclude interaction with the transporter. The largest number of hits that fell into one structural class was related to the phenazothiazines ( Table 4). Five of these were highly active in preventing uptake,   Table 3 for details.
To further assess the effects of this class of compounds on fatty acid uptake and metabolism, we compared C 1 -BODIPY-C 12 incorporation into lipids of yeast cells expressing human FATP2 treated either with PBS (control) or with perphenazine (Fig. 2). As expected, C 1 -BODIPY-C 12 was incorporated into both phospholipids and neutral lipids of control cells, and treatment with perphenazine substantially reduced the labeling of essentially all lipid classes. A third group of related hit compounds contained a tricyclic core motif differing from the phenazothiazines in that the central ring has seven members rather than six ( Table 5). The structures of these compounds are very similar, and subtle variations contribute to substantial differences in activity. Three compounds strongly inhibited uptake at 80 mM: clomipramine, methiothepin maleate, and cyclobenzaprine. Two were slightly active (clozapine and imipramine), whereas amitriptyline and loratadine were inactive.
The final set of compounds, although more structurally diverse within the group, were related in that all were heterocyclic and highly oxygenated ( Table 6). These varied in the number of rings, from the simplest two-ring compound juglone to the five-ring structure of aklavin.

Toxic compounds are eliminated by evaluating yeast growth
We previously demonstrated that heat-killed cells become permeable to the fatty acid analog and the quenching agent trypan blue, resulting in low fluorescence readings using the C 1 -BODIPY-C 12 assay (38). This led us to question whether the mechanism of apparent inhibition of transport was merely attributable to cell death and associated membrane dysfunction. To test compound toxicity, wild-type yeast cells (YB332) were grown in YPDA to mid-log phase and then diluted to 0.1 OD 600 . Serial dilutions of a test compound were added to medium, and growth of the yeast cells was monitored relative to cells incubated without the addition of compound. Figure 3 shows the growth curves of YB332 in the presence of PBS (control) or 100 mM centrimonium bromide, anthralin, or perphenazine. At 100 mM, perphenazine seemed to inhibit the growth of yeast cells. Therefore, we further tested the viability of yeast cells in the presence of this putative inhibitory compound by plating cells on solid medium after treatment with compound for 2 h to mimic the conditions used in our HTS experiments. Using this method, .95% of the cells were viable. Therefore, we do not believe that inhibition by perphenazine was merely the result of cell death.

Evaluation of hit compounds in Caco-2 cells
Caco-2 cells are derived from a human colon adenocarcinoma (49). In culture, these cells spontaneously differentiate into polarized, columnar cells that are representative of the small intestine (50). When grown on plastic, they form domes typical of normal and transporting epithelium (51,52). This cell line has been used extensively as a model for transport studies (50,53). Previous work has shown that fatty acid uptake in Caco-2 cells is saturable, decreased by trypsin treatment of cells, and can be competed by natural fatty acids (54). These cells express higher levels of both FATP2 and FATP4 than other FATP isoforms (P. N. Black, P. Fraisl, and C. C. DiRusso, unpublished data). To further evaluate hit compounds identified from our screening of yeast cells, we performed C 1 -BODIPY-C 12 uptake inhibition experiments in 96-well format using Caco-2 cells. We initially tested the hit compounds in Caco-2 cells at 100 and 10 mM ( Table 7). At 100 mM, most of these compounds inhibited Caco-2 uptake of C 1 -BODIPY-C 12 to varying extents. Seven compounds reduced uptake to essentially control levels, whereas most had intermediate activity. When the concentration of compound was de- Fig. 2. Perphenazine inhibits the incorporation of the fluorescent fatty acid analog into various lipids. Lipids were isolated from yeast cells expressing human FATP2 after treatment with PBS (2) or perphenazine (1) and C 1 -BODIPY-C 12 as detailed in the text. A: Solvent system I was used to distinguish phospholipids from neutral lipids. B: Solvent system II was used to separate neutral lipids. Images shown are representative of two experiments. TLC plates were scanned using a phosphorimaging system. Positions of commercial standards are labeled as follows: NL, neutral lipids; PE, phosphatidylethanolamine; PS, phosphatidylserine; PC, phosphatidylcholine; TLN, triolean; CE, cholesteryl ester; C, cholesterol. creased to 10 mM, only three compounds inhibited uptake by 25-40% of control levels ( Table 7).
We next evaluated the concentration-dependent inhibition of centrimonium bromide, perphenazine, and anthralin on Caco-2 cells by comparison with yeast (Fig. 4). The results shown in Fig. 4B indicate that the K i values of centrimonium bromide and perphenazine for Caco-2 cells was higher than that of yeast cells (Fig. 4A). For anthralin, we were unable to determine the K i in Caco-2 cells, because at 200 mM the uptake was still ?70% of the control value. As was the case in yeast, C 1 -BODIPY-C 12 was incorporated into Caco-2 cells, labeled various cellular lipids, and did not show preferential compartmentalization (Fig. 5). Treatment with a representative compound, perphenazine, substantially reduced incorporation without causing cell death; at 80 mM perphenazine, there was 94 6 9% cell survival, as estimated using the MTT cytotoxicity assay (five experiments).   Table 3 for details.

Validation of the screening method
In this study, we tested a HTS system for use in identifying fatty acid uptake inhibitors. The target of this HTS, the human fatty acid uptake system, is likely to be involved in the development and progression of many disease states that result from excessive intracellular accumulation of free fatty acids. We chose to use S. cerevisiae cells expressing the human FATPs for our primary screening because  Table 3 for details.
1) the biochemistry and molecular genetics of fatty acid uptake in yeast are well understood and analogous to those of mammalian cell systems; 2) phenotypic tests for fatty acid uptake and metabolism are readily available; 3) yeast cells are small in size, grow rapidly, and are resistant to cell lysis upon automated manipulation; and 4) the yeast HTS system is simple and inexpensive to implement. Our strategy was to carry out the HTS and initial screening for lead compounds using the humanized yeast system and then to analyze these compounds for their inhibition of fatty acid uptake in human cells. We validated this approach by screening a 2,080 compound library and identifying groups of compounds with similar chemical structures that inhibited fatty acid uptake in both yeast cells and human Caco-2 cells.
Hit compounds similar in structure to long-chain fatty acids BODIPY-labeled fatty acids have been used to monitor fatty acid uptake (22,26,27) and have been shown to be biosynthetically incorporated into lipids of cells (55; this work) and cultured organs (56). The length of C 1 -BODIPY-C 12 is approximately equivalent to that of an 18carbon fatty acid. Our laboratory also demonstrated that the uptake of this fluorescent fatty acid analog is effectively competed by the natural fatty acid oleate in yeast (38). In the present study, of 2,080 compounds screened, four groups of structurally related compounds were identified as hits as a result of inhibition of C 1 -BODIPY-C 12 uptake. Among these were two ionic detergents, centrimonium bro-  mide and benzalkonium chloride, which are similar to fatty acids in having a long hydrocarbon chain. These compounds may act as nonspecific detergents or may mimic fatty acids by interacting with FATP specifically, for example, by occupying the fatty acid binding site on the protein. We noted that these two compounds have critical micelle concentrations in the same range as cholic acid and deoxylcholic acid (1-4 mM), which did not reduce apparent  fatty acid uptake, so nonspecific detergent effects do not adequately explain the inhibition of uptake by centrimonium bromide and benzalkonium chloride. Filipin, nystatin, and b-escin, three other compounds that disrupt cell wall and/or membrane architecture, also had no effect on uptake in our primary screening assay. Therefore, at the concentration used in these experiments (80 mM), it does not appear that these compounds disrupt FATP2-mediated fatty acid transport by nonspecific actions.

Some atypical antipsychotic drugs also inhibit uptake
The other two groups of compounds all have a three-ring structure similar to the BODIPY moiety of C 1 -BODIPY-C 12 (Tables 4, 5). One concern is that these compounds are merely competitive inhibitors of the fluorescent fatty acid analog and not the natural ligands. However, this is most likely not the case, because C 1 -BODIPY-C 12 is taken up in a FATP-specific manner and can be metabolized upon import. Furthermore, the native ligand (oleate) effectively competes with C 1 -BODIPY-C 12 (38). Thus, it is presumed the fluorescent ligand binds to the transporter in a manner analogous to the natural ligand. Compounds identified in this screen within these two structural classes fall into a family with well-known pharmacological properties that include antipsychotic drugs, which act by inhibition of neuroreceptors, including the dopamine receptor (39). Hints at a role for these compounds in fatty acid and lipid metabolism come from studies showing that these compounds are associated with hypertriglyceridemia, cholestatic liver, and insulin resistance (57,58). A recent gene expression array study demonstrated a correlation between alterations in gene expression resulting from treatment of cells in culture with phenothiazine drugs and expression changes induced by treatment of cells with arachidonic acid (59). Together, these data strongly suggest that these drugs may affect lipid homeostasis in whole or in part through the inhibition of fatty acid uptake into cells. On the other hand, although these adverse side effects might seem to eliminate this class of compounds from consideration as potential fatty acid uptake inhibitors, they may prove useful in other ways. First, they will be valuable to understand the mechanisms causing the unwanted side effects of the drugs now in clinical use. Second, they potentially could be modified/rationally designed to interact more specifically with neuroreceptors rather than the FATP proteins and visa versa. Third, they are expected to be informative regarding mechanisms of FATP activity.
Few antifungal drugs in the library altered transport activity Several compounds identified as hits were the antifungal agents tomatine, juglone, and sapinoside A. These compounds might act by killing the cell or inhibiting metabolic processes, including uptake. Because our primary screening assay requires only a 2 h incubation period with the compound, and because the addition of compound before C 1 -BODIPY-C 12 uptake did not result in reduced cell-associated fluorescence, we do not believe that these compounds are merely killing the cells. Cell death also results in increased membrane permeability. Metabolic poisoning might have the same effect in reducing membrane function, but this mechanism is more difficult to disprove with the types of assays available. We also note that the SpectrumPlus collection contains numerous antifungal agents that had no effect on C 1 -BODIPY-C 12 uptake by our method. Among these were fluconazole, griseofluvan, and clotrimazole.
In summary, we have designed multiple systems to identify and evaluate fatty acid uptake inhibitors. Primary screening and initial specificity evaluation used yeast cells that express a human FATP family member. Secondary screening has been developed in human cells (Caco-2) to verify and further evaluate the specificity of the lead compounds for inhibition of fatty acid uptake. This uptake assay has some aspects in common with another assay recently reported by Liao et al. (60). However, our method differs in that any cell type (from yeast to human cells) can be used with a high degree of sensitivity and reproducibility using an inexpensive quenching agent. In the future, we hope to use this method to identify compounds that discriminate between different classes of fatty acids or different FATP family members. Apart from identifying fatty acid uptake inhibitors, this system will also be of value in identifying fatty acid uptake activators. Additionally, although we have targeted human FATPmediated fatty acid uptake in genetically defined yeast, modifications could be made in this system to target other participants in fatty acid transport, metabolism, and trafficking operative in living cells. Another anticipated use of this screening system is in determining whether or not lead drugs interfere with fatty acid uptake and metabolism before they enter whole animal or human trials. In this regard, this inexpensive, rapid, live cell assay should prove an adaptable secondary screen for this purpose.