Identification of clinically approved small molecules that inhibit growth and affect transcript levels of developmentally regulated genes in the African trypanosome

Trypanosoma brucei are unicellular parasites endemic to Sub-Saharan Africa that cause fatal disease in humans and animals. Infection with these parasites is caused by the bite of the tsetse fly vector, and parasites living extracellularly in the blood of infected animals evade the host immune system through antigenic variation. Existing drugs for Human and Animal African Trypanosomiasis are difficult to administer and can have serious side effects. Resistance to some drugs is also increasing, creating an urgent need for alternative trypanosomiasis therapeutics. We screened a library of 1,585 U.S. or foreign-approved drugs and identified 154 compounds that inhibit trypanosome growth. As all of these compounds have already undergone testing for human toxicity, they represent good candidates for repurposing as trypanosome therapeutics. In addition to identifying drugs that inhibit trypanosome growth, we wished to identify small molecules that can induce bloodstream form parasites to differentiate into forms adapted for the insect vector. These insect stage parasites lack the immune evasion mechanisms prevalent in bloodstream forms, making them vulnerable to the host immune system. To identify drugs that increase transcript levels of an invariant, insect-stage specific surface protein called procyclin, we engineered bloodstream reporter parasites that express Green Fluorescent Protein (GFP) following induction or stabilization of the procyclin transcript. Using these bloodstream reporter strains in combination with automated flow cytometry, we identified eflornithine, spironolactone, and phenothiazine as small molecules that increase abundance of procyclin transcript. Both eflornithine and spironolactone also affect transcript levels for a subset of differentiation associated genes. While we failed to identify compounds that increase levels of procyclin protein on the cell surface, this study is proof of principle that these fluorescent reporter parasites represent a useful tool for future small molecule or genetic screens aimed at identifying molecules or processes that initiate remodeling of the parasite surface during life cycle stage transitions.

Introduction Trypanosoma brucei is a unicellular protozoan parasite that causes both Human and Animal African Trypanosomiasis (HAT and AAT, also known as sleeping sickness in humans and nagana in livestock, respectively). These diseases affect both humans and ungulates, causing a severe human and economic burden in regions of Sub-Saharan Africa where they are endemic [1]. Because trypanosomiasis primarily affects the rural poor, there is a paucity of efficacious drugs available. Moreover, all drugs in current use have severe side effects, some drugs used to treat late stage trypanosomiasis are highly toxic, and resistance to existing drugs is rising [2]. Thus, new strategies to treat trypanosomiasis are urgently needed.
We reasoned that existing clinically approved drugs might represent good candidates for alternate trypanosomiasis therapeutics, as they do not have to go through the same rigorous toxicity screening as newly developed drugs. With that in mind, we performed a small molecule flow cytometry-based screen using the Johns Hopkins University Clinical Compound Library and identified 154 small molecules that inhibit parasite growth, many of which have not previously been reported. We verified that several of the compounds from our screen inhibit growth with IC 50 s in the low micromolar range. We hope that identification of these compounds will be a resource for researchers interested in repurposing existing compounds for treatment of neglected disease.
At the same time, we wondered whether we might be able to simultaneously screen for compounds that make trypanosomes poorly adapted to living in the bloodstream of their mammalian host. Specifically, identifying compounds that make the bloodstream parasite vulnerable to the host immune system could produce alternative trypanosomiasis therapeutics. Trypanosomes live extracellularly within the mammalian host and untreated infections are almost invariably fatal. This is because T. brucei has evolved a number of mechanisms that are effective at evading the mammalian host antibody response. Many of these depend on the expression and recycling of the Variant Surface Glycoprotein (VSG) coat that densely covers the surface of bloodstream trypanosomes [3]. There are thousands of antigenically distinct genetic variants of VSG in the genome [4], and by periodically switching the variant expressed on the parasite surface, the host antibody response can be effectively thwarted. This immune evasion allows the parasite to live long enough within the mammalian bloodstream to be transferred to the tsetse fly insect vector following a bloodmeal, where it transitions to the insectspecific procyclic form. The environment of the fly is quite different from that of the mammalian bloodstream, and the parasite adapts itself accordingly by changing its morphology and metabolism [5]. Indeed, it has been shown that thousands of genes show altered transcript levels between bloodstream and procyclic forms [6][7][8]. A key event marking the transition from bloodstream to procyclic forms is the remodeling of parasite surface proteins to replace VSG with an invariant surface protein called procyclin, of which there are only four isoforms [9].
If bloodstream parasites were induced to express this invariant procyclin protein on the surface with a small molecule compound, this could provide a 'handle' for the mammalian immune system to eliminate parasites whose surface is usually a moving target due to the evershifting repertoire of VSGs. As an example, we previously showed that the small molecule bromodomain inhibitor I-BET151 has this effect in bloodstream form cells, resulting in the expression of procyclin on the surface and decreased virulence in a mouse model of infection [10]. In addition, others have attempted to identify compounds that induce expression of procyclin in bloodstream parasites using a GUS reporter assay [11,12].
To identify these compounds, we developed a fluorescent reporter system to allow us to easily detect transcript levels of the gene EP1, which codes for one isoform of the invariant insectstage specific procyclin protein. Here we present proof of principle that our EP1 fluorescent reporter can be used for flow cytometry-based screening using an autosampler. This allows for automated 96 well plate sampling without the need for a human operator, resulting in less labor-intensive flow cytometry-based drug library screens that can simultaneously screen for growth inhibition and a fluorescence-based phenotype. First, we showed that fluorescence levels in the reporter line faithfully mimic expected levels of EP1 transcript in bloodstream, procyclic, and differentiating parasites. Second, we identified several compounds that increase EP1 transcript levels in bloodstream form parasites.
While none of the compounds that increase EP1 transcript levels are ideal as trypanosomiasis therapeutics, we think our reporter system holds promise both as a screening tool for other small molecule libraries and as a system to learn more about unidentified factors that stabilize procyclin protein after initiation of differentiation. Much progress has been made in elucidating the quorum sensing signaling pathways that are necessary to trigger parasite differentiation from the bloodstream form to the insect form via the stumpy intermediate [13][14][15]. It is also clear that RNA Binding Proteins (RBPs) play an important role in maintaining life cycle stagespecific mRNA levels and triggering differentiation events [16][17][18][19]. However, the gene regulatory mechanisms by which VSG transcription is halted and procyclin transcription is initiated and/or stabilized remain poorly characterized. Our long-term goal is to take advantage of this reporter system to perform genetic screens, allowing us to identify the key players responsible for stabilizing procyclin transcripts during differentiation from the bloodstream form to the procyclic form. This will allow us to unravel the missing link between environmental signals for differentiation that are received and transduced at the parasite surface and the downstream effect on transcript levels that ultimately allow the parasite to adapt from the bloodstream environment to the insect environment.

Culture growth and strain details
We used the Lister 427 L224 strain of bloodstream T. brucei parasites with a dual BES marked line that expresses VSG3 from BES7 (with Neo R downstream of the promoter) and a Puro R gene downstream of the BES1 promoter (expressing VSG2) [20]. Bloodstream parasites were cultured in HMI-9 at 37˚C with 5% CO 2 . Procyclic PF427 parasites were grown at 27˚C in SDM79.

Generation of the EP1/GFP reporter plasmid
DNA encoding GFP flanked by 314 bp immediately upstream of the EP1 ORF and 279 bp immediately downstream of the EP1 ORF was synthesized by Life Technology. Both upstream and downstream EP1 fragments contain the entire 5' and 3' UTRs of EP1. The 5'UTR/GFP/3'UTR fragment was amplified with PvuII and HindIII sites and cloned into the pyrFEKO Hygro R plasmid using the same sites. We then amplified a 289 bp homology region downstream of the EP1 3'UTR fragment with SpeI and XhoI, which was cloned downstream of the Hygro R gene using the same sites. The 5'UTR/GFP/3'UTR fusion fragment was amplified using Forward ggccatcagctgagtcaatagtgcattttg and Reverse ggccataagcttgaatgaaaaaaaatagaagtgaaa primers. The downstream homology region was amplified using Forward ggccatactagtctttgaatttggatcttaaaattattattg and Reverse ggccatctcgagcaacttcagctgcggggc primers. For transfection into our Hygro resistant procyclic strain, the construct was modified to replace the Hygro R gene with a Puro R gene.

Strain construction
Transfection of bloodstream form parasites was performed using an AMAXA Nucleofector. Parasites were resuspended in 100μl T cell solution with 10μg linearized DNA and electroporated on the X001 setting. Following transfection, cells were grown in 5μg/ml Hygromycin (Invivogen). PF cells were transfected using the same procedure, but following transfection the cells were plated into conditioned media in 0.1μg/ml Puromycin (Invivogen). Correct integration of the construct in transfected cells was tested using PCR with a forward primer upstream of the 5'UTR of EP1 (gtccgataggtatctcttattagtatag) and a reverse primer within the GFP gene (agaagtcgtgctgcttcatgtggt). Successful isolation of genomic DNA was confirmed using control primers that amplify the downstream EP1 homology region: For ggccatactagtctttgaatttggatcttaaaattattattg and Rev ggccatctcgagcaacttcagctgcggggc.

Differentiation
Bloodstream parasites were spun down and resuspended in differentiation media [21] supplemented with 6 mM cis-aconitate (Sigma-Aldrich A3412) at 27˚C for three days.

Sytox Orange staining
Cells were resuspended in 5μM Sytox Orange (Fisher Scientific S11368) and incubated for 15 minutes at 37˚C prior to flow cytometric analysis.
The data are fitted with a sigmoidal curve 50 ÞÞ^HillÞ where GR inf = GR(count>inf), reflecting asymptotic drug efficacy. GEC 50 , the drug concentration at half-maximal effect, Hill, the Hill coefficient of the fitted GR curve reflecting how steep the dose-response curve is.
Cell cycle analysis. Cells were fixed using 1% formaldehyde in PBS for 5 minutes. Cells were then washed three times in PBS and then resuspended in a solution of 0.2 mg/ml RNAse and 0.05 mg/ml propidium iodide in PBS. Following a 2-hour incubation at 37˚C, cells were analyzed by flow cytometry.
Quantitative PCR analysis. For quantitative PCR to quantify mRNA levels, RNA was extracted from parasites treated with drug or vehicle control for 3 days using RNA Stat-60 (Tel-Test) following the manufacturer's protocol and quantified on a NanoDrop2000c. 2.5μg of RNA was used to generate cDNA using the SuperScript IV VILO Master Mix (Fisher Scientific 11756050) according to the manufacturer's protocol. cDNA was amplified using 2X Sybr green master mix (Life Technologies 4309155) and primers and quantified on an Eppendorf Realplex2 instrument. Primers used were as follows: Tb427. 10 Mammalian cell growth assay. RAW264.7 mouse macrophage or C32 human melanoma cells were maintained in DMEM (Sigma-Aldrich) supplemented with 10% fetal bovine serum, as well as 100 μg/mL penicillin and 100 μg/mL streptomycin. 10,000 RAW264.7 cells per well were seeded in 96-well plates in 100 μl/well 24 hours before the assay. On the day of the assay, 0.75 or 1.5 μl of 3.3 mM JHCCL drugs were added to 150 μl of cell-containing media to achieve 16 or 33 μM of each compound per well, respectively. Cells were treated with drugs for 6 hours at 37˚C and 5% CO 2 . C32 cells were treated with drugs for 24 hours. Determination of cell viability by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay was performed as described [23]. Cell viability is shown as the percentage of surviving cells obtained relative to cells treated with DMSO as vehicle control (100%

Identification of in vitro inhibitors for T. brucei growth
To identify existing compounds that might be repurposed as trypanosomiasis therapeutics, we obtained the Johns Hopkins University Clinical Compound Library of 1,585 small molecule drugs. 1082 of these compounds are FDA approved while the remaining are foreign approved drugs. We plated 1000 bloodstream parasites in 100μl of media in a 96 well plate format. We then added 1μl of each drug to the well for a final concentration of 33μM. We chose this concentration because it had been used previously with this same library to successfully identify compounds that cross-inhibit pathogenic agents and the host proteins they exploit [24]. We reasoned that follow-up experiments could be used to determine whether specific compounds could still inhibit growth in a more desirable nM or low μM range. Because our screen used a flow cytometer that can be programmed to analyze a fixed volume of cells, we were easily able to identify drugs that slowed trypanosome growth after 3 days of treatment by simply counting the number of analyzed cells that fell within a live cell gate. The entire assay was performed in duplicate for each drug in the library. We binned each drug into four categories: highly inhibitory, inhibitory, slightly inhibitory, and not inhibitory (  Table 1). For the most highly inhibited category of drugs, we searched the literature to see if they had previously been characterized as anti-trypanosomal. Berberine, ipecac syrup, disulfiram, methylene blue hydrate, promethazine hydrochloride, perphenazine, promazine hydrochloride, and triflupromazine have all previously been reported to inhibit growth in T. brucei and our results agree with those previously published reports [25][26][27][28][29] (S1 Table, S2 Table). Several other drugs, including nitrofurazone, aminacrine, clemastine fumarate, bepridil hydrochloride, amiodarone hydrochloride, prochlorperazine dimaleate, protriptyline hydrochloride, fluoxetine hydrochloride, apomorphine hydrochloride, and proadifen hydrochloride have been shown to inhibit growth in T. cruzi [27,[30][31][32][33][34][35][36][37][38]. Finally, cyproheptadine hydrochloride was shown to inhibit growth in T. evansi (S1 Table) [39]. We also identified a number of drugs that inhibit T. brucei growth in vitro that, to our knowledge, have not previously been studied in this context (S1 Table), including antimalarials, antibacterials, antiseptics, anesthetics, antihypertensives, antiemetics, antidepressants, vitamins, and a vasodilator. In addition to these highly inhibitory drugs, a further 119 drugs were shown to have a 10-100 fold inhibitory effect on parasite growth, including homidium bromide (used to treat cattle [40]) and pentamidine, a well-established trypanosomiasis drug [41] (S3 Table, S4 Table).
We additionally compared the results from this study to two other recent NTD-focused drug screens. One of these used the Medicines for Malaria Venture Pathogen box dataset (https://www.mmv.org/mmv-open/pathogen-box/about-pathogen-box#composition_), which includes 70 anti-kinetoplastid compounds. We cross-referenced all 400 compounds in the NTD Pathogen box with the compounds in our library and found that five of the compounds that were identified as inhibiting growth in our study were also included in the Pathogen Box. These include three compounds that were used as reference compounds in the Pathogen Box data: the antimalarial drug primaquine, the antibacterial drug clofazimine, and the antiprotozoal drug pentamidine. Both primaquine and clofazimine have previously been reported

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei to be active against T. cruzi [42,43], while pentamidine is used against T. brucei infections. Similar to our study, fluoxetine was reported to be effective against kinetoplastids in the Pathogen Box data. Finally, clemastine fumarate was reported to be effective against trichuriasis caused by the human whipworm in the Pathogen Box data, but was not mentioned in the context of kinetoplastid infections. We further compared results from our study to a set of compounds found to be inhibitory to T. brucei growth in a recent screen from GSK (https://www. ebi.ac.uk/chembl/g/#browse/activities/filter/document_chembl_id%3ACHEMBL3430912); however, none of the compounds identified in this screen overlapped with the compounds in our library. In addition to assaying these drugs for their effect on trypanosome growth, we also measured percent survival after treatment for 24h with 16μM drug for two different cell lines: murine Raw264.7 macrophages and human C32 melanoma lines (S2-S4 Tables). Additional toxicity information gathered from pubchem is provided in S5 Table. These additional data can be used to help prioritize the drugs with the highest mammalian cell line survival rate and the lowest toxicity. We hope that this list can be used as a resource for those in the trypanosomiasis drug development field.

Confirmation that drugs identified in the screen inhibit trypanosome growth
To confirm that the drugs identified in the screen did indeed inhibit parasite growth, we ordered fresh stocks of six drugs from Sigma-Aldrich and measured growth inhibition. We chose three drugs from our most highly inhibitory category: flunarizine hydrochloride, aprepitant, and clemastine fumarate. Clemastine fumarate is an antihistaminic drug that has been shown to inhibit growth in T. cruzi [31]. Aprepitant is an antiemetic and flunarizine hydrochloride is a vasodilator. To our knowledge neither of these latter two have been studied for their effects on kinetoplastids. In addition to these three highly inhibitory compounds, we chose one drug from our inhibitory category (triprolidine) and two drugs from our slightly inhibitory category (phenothiaizine and bufexamac). Neither triprolidine nor bufexamac have been shown to inhibit trypanosome growth, while phenothiazine has previously been shown to be strongly inhibitory [44,45]. To assess the effect of each of these six drugs on trypanosome growth, we adopted a flow cytometry assay using the nucleic acid stain Sytox Orange, which positively stains dying and dead cells with compromised membranes. Parasites were plated at 100,000 cells/ml, and an aliquot of parasites was stained with Sytox Orange after 24h of growth. A fixed volume of cells was then analyzed by flow cytometry and the number of live cells was determined by gating on Sytox Orange negative populations whose size (measured by forward scatter) is consistent with healthy parasites (Fig 1A). Whereas healthy cells do not stain positively with Sytox Orange (Fig 1A, left), those treated with an inhibitory drug do show populations of positively staining Sytox Orange parasites as well as populations of smaller particles that represent dead cells and debris (Fig 1A, right). Based on the number of cells in the live gate, the remaining unstained parasites were diluted back down to 100,000 parasites/ml. The growth curves in Fig 1B show that parasites treated with five of the six drugs we tested had a severe growth inhibited phenotype, while treatment with triprolidine had no appreciable effect on growth (S8 Table). Survival data for mammalian cell lines treated for 24h at 33μM for all verified growth inhibition drugs is provided in S6 Table. We went on to estimate the IC 50 values for each of the six drugs by growing parasites for 48h in a range of concentrations. Five of the six drugs showed IC 50 values between 3.7 and 12.3μM (Fig 1C, see insets). Consistent with our observations on cell growth, the IC 50 value measured for triprolidine was very high (S1 Fig inset) and the r 2 value for this drug indicates a Error bars indicate standard deviation of 3 biological replicates. � , �� , ��� indicates p < 0.05, p < 0.01, and p < 0.001 respectively as measured by an unpaired two-sided Student's T test. (C) Percent growth inhibition over a range of concentrations for the indicated drug. Data were fitted and indicated values calculated using the GRMetrics R package. GR 50 , the concentration at which the effect reaches a growth rate (GR) value of 0.5 based on interpolation of the fitted curve (dashed lines). GIC 50 , the drug concentration at half-maximal effect for calculated growth rate. r 2 GR, the coefficient of determination for how well the GR curve fits to the data points. IC 50 , the concentration at which relative cell relatively poor fit for the data. Therefore, it should be noted that the screen can produce some false positives, and promising candidates need to be verified with fresh stocks prior to further study. Interestingly, we did not observe consistently higher IC 50 values for drugs in the slightly inhibitory (phenothiazine, bufexamac) category when compared to those in the highly inhibitory category (aprepitant, clemastine fumarate, flunarizine). We conclude that compounds with only a 15-25-fold reduction in cell counts as measured in our screen can still have highly inhibitory effects on growth.

Construction and validation of the EP1/GFP reporter cell line
We reasoned that a flow cytometric based screen could allow for simultaneous screening for compounds that inhibit trypanosome growth and for compounds that increase transcript levels of the invariant insect-stage specific procyclin gene EP1. In order to facilitate flow cytometric screening for compounds that increase transcript levels of the EP1 gene in bloodstream parasites, we integrated a construct containing GFP flanked by native EP1 5' and 3' UTRs at the endogenous EP1 locus (Fig 2A). It should be noted that an increase in GFP expression in parasites harboring this construct could be from higher levels of transcript being produced, increased stability of the transcript, or both. The construct was integrated in both bloodstream and procyclic parasites, and the second EP1 allele was left intact. A PCR assay confirmed the correct integration of the reporter construct in both bloodstream and procyclic parasites (Fig  2A). We expect that if this reporter construct faithfully recapitulates native EP1 expression, GFP levels should be relatively low in bloodstream form parasites and high in procyclic parasites. Fig 2B shows that GFP levels in EP1/GFP reporter cells are low in bloodstream parasites and high in procyclic parasites, as expected. We also expect that GFP expression in bloodstream EP1/GFP reporter parasites should increase when parasites are induced to differentiate to the procyclic form using cis-aconitate and incubation at 27˚C, and we found that this was indeed the case ( Fig 2C). Finally, we tested the effect of the bromodomain inhibitor I-BET151, which was previously shown to induce EP1 expression in bloodstream parasites [10]. Treatment of bloodstream EP1/GFP reporter cells with I-BET151 resulted in an increase in GFP expression after 3 days, and the effect was more pronounced after 5 days (Fig 2D). We conclude that GFP expression mirrors the expected expression of EP1 in the EP1/GFP reporter line by all the metrics tested.
A proof of principle screen shows that the reporter strain can be used to identify small molecules that increase transcript abundance of the procyclin gene EP1 Our EP1/GFP reporter parasites were used for the flow cytometric screen of the Johns Hopkins University Clinical Compounds used to identify inhibitory compounds described above. After 3 days of growth, we subjected each well to analysis by flow cytometry to measure the number of parasites in the live gate and the level of EP1/GFP expression. Each plate also contained parasites growing in DMSO to serve as a solvent control and I-BET151-treated parasites as a positive control. This assay was performed in duplicate for each of the drugs. After the two initial assays, the median GFP fluorescence intensity was calculated for DMSO-treated parasites and drug treated parasites. Any drug that caused a 1.5-fold increase in average median GFP fluorescence intensity in drug-treated parasites compared to DMSO-treated control parasites was count = 0.5. r 2 IC 50 , the coefficient of determination for how well the traditional curve fits to the data points. See methods for more details. https://doi.org/10.1371/journal.pntd.0007790.g001

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei moved forward to the third round of screening. During this third round, in addition to the 3-day assay, drug-treated parasites were immediately subjected to analysis by flow cytometry following the addition of the drug to eliminate any drugs that were themselves fluorescent. After this third round, the median GFP fluorescence intensities for drug-treated parasites and solvent control-treated parasites were calculated again, and a Student's t-test was performed to identify drugs that caused a significant increase in median GFP fluorescence intensity. For those drugs that induced a significant shift, we imposed a 2-fold median GFP fluorescence intensity cutoff (S7 Table). This left us with 9 candidate drugs, for which we ordered fresh stocks from Sigma-Aldrich. Parasites were again subjected to the 3-day flow cytometry assay, except this time they were grown in 1ml cultures. We were able to replicate the increase in EP1/GFP expression for Spirinolactone, Phenothiazine, and Eflornithine (Fig 3). Notably, all of our false positives for EP1/GFP expression inhibited cell growth (S7 Table). We hypothesize that the low cell counts produced noise in the data and that dying cells may have produced autofluorescence in the GFP channel. During follow-up experiments, the higher number of cells assayed by flow cytometry may have alleviated one or both of these problems.
Of the drugs we identified as increasing EP1/GFP expression, eflornithine treated bloodstream parasites have previously been reported to show some similarities to stumpy cells, although these stumpy-like forms are not competent to differentiate to procyclic forms [46]. As stumpy forms have been shown to transcriptionally preadapt themselves to insect stage procyclic cells by upregulating procyclic specific transcripts [47], the increase in EP1/GFP expression in eflornithine treated parasites is consistent with these parasites sharing some characteristics of stumpy form parasites. As mentioned above, phenothiazine and its derivatives have previously been reported as anti-protozoal, strongly inhibiting cell growth in T. brucei [44,45]. Lastly, to our knowledge, spironolactone has not been studied in the context of T. brucei infection, although it has been tested for its effects on myocardial protection in a model of Chagas disease [48]. Spironolactone is used for the treatment of hirsutism, as is eflornithine, which is one of the most important anti-trypanosomal drugs in current use [49].

Phenothiazine causes severe growth defects and cell cycle abnormalities
We wished to determine if the observed increase in EP1/GFP expression for each candidate drug was a non-specific result of cell cycle arrest, so we tested the drugs that caused a measurable increase in EP1/GFP expression for their effect on parasite growth. While neither spironolactone nor eflornithine induced a profound effect on parasite growth at 33μM, phenothiazine caused a growth arrest in treated cells by 24h ( Fig 4A). As expected, eflornithine did have effects on parasite growth at higher concentrations of 120μM (Fig 4A). To test the effect of each of these drugs on cell cycle, we performed propidium iodide staining and flow cytometry on parasites treated with each of the drugs for three days. In keeping with our observations of

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei cell growth, phenothiazine treated parasites exhibited severe cell cycle abnormalities after 3 days of treatment (Fig 4B). We further tested phenothiazine treatment of parasites at earlier time points and found significant cell cycle abnormalities within 24h of treatment (Fig 4C). At 1 day the most pronounced effects were the accumulation of parasites in the G2 phase of the cell cycle, a decrease in S phase cells, and the appearance of multinucleated cells. At 2 days an accumulation of parasites lacking a nucleus (zoids) was apparent, and this effect was even more pronounced at 3 days (Fig 4D). The percent of parasites in each of the cell cycle stages was significantly different at all time points tested when compared to the DMSO treated controls. These results are in keeping with previously published reports that show that phenothiazine treated parasites exhibit severe growth defects through inhibition of trypanothione reductase [44,50]. We conjecture that the effect of phenothiazine on EP1/GFP expression might be a non-specific stress response, while this does not appear to be the case for either spironolactone or eflornithine. Survival data for mammalian cell lines treated for 24h at 33μM for all verified EP1/GFP drugs is provided in S6 Table.

Surface protein expression of procyclin is not observed following treatment with drugs observed to increase EP1/GFP transcript levels
We tested whether the increase in EP1/GFP transcript levels corresponded to an increase in the corresponding procyclin protein on the parasite surface. Because our antibody to procyclin fluoresces in the same channel as GFP, we treated the parental strain for our reporter strain for 5 days with 33μM eflornithine, phenothiazine, and spironolactone, stained the parasites with anti-procyclin, and assayed procyclin protein expression by flow cytometry. I-BET151 was used as a positive control as it has previously been shown to increase surface protein

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei PLOS NEGLECTED TROPICAL DISEASES expression of procyclin [10]. Previous experiments showed that the highest level of procyclin expression for I-BET151 was at 5 days, so we used a similar treatment for our new drug candidates and replenished the drug every two days. None of the drugs produced measurable procyclin protein on the parasite surface after treatment (Fig 5A), and we could not detect procyclin protein in whole cell lysates of drug-treated parasites by western blot (S2 Fig). An increase in EP1 transcript abundance without a concomitant increase in protein expression has been observed before in the context of DNA replication mutants [51]. While eflornithine is already in use as a trypanosome therapeutic and phenothiazine may have potential as a therapeutic based on its effect on cell growth, we conclude that spironolactone is not a promising drug candidate for trypanosomiasis since the effect on EP1 transcript levels is not recapitulated at the level of protein (Fig 5A). However, future studies beyond the scope of this work might aim to identify the protein target of spironolactone to try and ascertain whether this protein maintains low levels of EP1 transcript in bloodstream cells.
A previous study seeking to identify compounds that increase EP1 transcript levels reported that a number of compounds that increased EP1 transcript levels also had an effect on the level of VSG protein on the surface, as measured by flow cytometry [11,12]. The general effect observed was an increase in VSG median fluorescence intensity (MFI), and the authors speculated that this was due to loss of the VSG coat from the surface, which resulted in more epitopes exposed for antibody binding [12]. Our group also reported an increase in VSG MFI prior to complete loss of VSG protein from the surface following treatment with the bromodomain inhibitor I-BET151 [10]. Therefore, we performed flow cytometry using anti-VSG antibodies on parasites treated for 2 or 3 days with phenothiazine, spironolactone, and eflornithine. No effect on MFI for VSG was detected for spironolactone or eflornithine, indicating that an increase in EP1/GFP transcript levels is not sufficient to alter the VSG MFI ( Fig  5B). Both 2 and 3 days of phenothiazine treatment produced one population of parasites with higher VSG MFI compared to DMSO-treated controls and a second population of phenothiazine-treated parasites with lower VSG MFI compared to DMSO controls (Fig 5B). Because phenothiazine is such a potent growth inhibitor, we tested whether lower concentrations of the drug might produce the effect on VSG in healthier parasites. We found that a 3d treatment with 16μM phenothiazine produced an effect on VSG MFI but that this was not the case at lower concentrations. We observed almost no effect on EP1/GFP transcript levels at these lower concentrations (Fig 5C and S9 Table). In addition, treatment with another growth inhibitor, clemastine fumarate, produced altered VSG MFI with very little effect on EP1/GFP expression ( Fig 5C). All of the previously reported compounds that increased EP1 transcript levels and altered VSG MFI also induced growth inhibition [12], so it is difficult to disentangle whether the altered VSG MFI is a non-specific effect in sick parasites or whether it is in some cases a consequence of altered EP1 transcript levels.

The expression for a number of transcripts associated with differentiation is altered following treatment with spironolactone and eflornithine
Treatment with either eflornithine and spironolactone increased EP1/GFP expression in our bloodstream reporter cells, and this made us wonder whether other transcripts associated with differentiation show altered expression levels in drug treated parasites. While phenothiazine also increased EP1/GFP expression, we worried that this effect might be a non-specific stress response, and accordingly did not test phenothiazine's effect on other differentiation-related transcripts. We treated parasites with either eflornithine or spironolactone for three days, isolated RNA, and performed qPCR to ask whether expression levels were altered for a subset of genes whose expression changes during differentiation to procyclic forms [52]. In addition to an increase in EP1 and GFP transcript levels in drug treated cells, we also observed an increase in transcript levels for Procyclic Associated Gene 2 (PAG2, Tb927.10.10220), which is found just downstream of EP1 in the same Polycistronic Transcription Unit (PTU) and is known to have increased expression during differentiation (Fig 6). Another surface protein expressed in early procyclic parasites (GPEET, Tb927.6.510) showed an increase in eflornithine treated parasites, but this increase was not statistically significant. Because differentiating parasites must alter their metabolism to adapt to the insect midgut, we also checked expression levels for a number of genes associated with these changes in metabolism. Both hexokinase (HK1, Tb927. 10.2010) and glycosomal membrane protein (GIM5B, Tb927.9.11600) show an initial dip in expression levels following initiation of differentiation and then a slow increase in expression, while glycosomal fructose biphosphate aldolase (ALD, Tb927. 10.5620) and pyruvate kinase (PYK1, Tb927. 10.14140) show a monotonic decrease in expression following initiation of differentiation [52]. We observed a subtle but statistically significant decrease in PYK1, GIM5B, and HK1 following treatment with eflornithine and a decrease in HK1 and ALD following treatment with spironolactone. No changes were observed in splicing factor 1 (SF1, Tb927.10.9400), which we used as a negative control as it has not previously been associated with changes in expression for differentiating parasites (Fig 6). It should be emphasized that changes in expression levels for differentiation-associated genes in our drug treated parasites are much lower than those observed when we induced differentiation in our EP1/GFP reporter parasites using a three day treatment with 6mM cis-aconitate and incubation at 27˚C (S3A Fig) and for what has been reported in the literature [52]. We note that we did not observe a decrease in transcript levels for either GIM5B or HK1 in our differentiation-induced reporter parasites at 3 days but we did detect this decrease at earlier time points (S3B Fig) This is consistent with the previously reported transient dip in expression levels for these genes observed at early time points prior to 72 hours in the literature [52]. Although the expression changes are muted compared to those in differentiating cells, it is interesting that small changes in EP1 expression in eflornithine and spironolactone treated parasites are accompanied by other small changes in expression levels for a subset of genes known to be associated with differentiation. We conclude that the effect of both eflornithine and spironolactone is not solely confined to genes associated with changes in the proteins that make up the parasite surface, and may additionally initiate multiple components of the differentiation program.

Discussion
The flow cytometry screen described here allowed us to identify compounds that inhibited trypanosome growth, which can be used as a starting point for alternate trypanosomiasis therapeutics. We purposely used U.S. and foreign approved compounds because their toxicity is better characterized than those of new compounds, which might streamline drug development should promising candidates be found. However, there is no reason that other drug libraries could not be screened using the same methods described here. Our library consisted of~20 96-well plates, each of which took~1h to screen. We typically screened 6 plates a day, allowing us to complete a full screen of the 1,585 compounds in duplicate within 2 weeks. While we conducted a medium-throughput screen using 96 well plates, flow cytometry systems can now

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei accommodate screening using 1,536 well plates, allowing researchers to screen 50,000 wells per day [53,54]. These technologies would allow for very rapid high-throughput screens.
Our screen uncovered 154 compounds that were slightly to highly inhibitory for trypanosome parasite growth. We were able to confirm a small subset of these by ordering fresh drugs from Sigma-Aldrich and performing growth and IC 50 assays. 8 of the most highly inhibitory drugs had previously been shown to have effects on T. brucei growth, 1 had previously been shown to affect growth in T. evansi, and 10 had been shown to inhibit growth in T. cruzi. To our knowledge, the remaining 16 compounds that were in the highest category for growth inhibition have not previously been reported to affect T. brucei growth. We hope that our toxicity data in conjunction with already existing toxicity data can help to prioritize which of these drugs might be most promising as novel therapeutics. It is worth noting that pentamidine, a trypanosomiasis drug currently in active use, was sorted into the moderately inhibitory category, implying that the 102 compounds in this category could have potential as trypanosomiasis therapeutics.
We were able to confirm the growth inhibitory effects of a small subset of drugs across the inhibitory categories by ordering fresh drugs from Sigma-Aldrich and performing growth and IC 50 assays. One of the compounds we tested (triprolidine) was a false positive, for which we don't have a good explanation. It is possible that evaporation of the drug plate resulted in some compounds being tested at higher than the intended concentrations, which would be consistent with the fact that we did find growth inhibitory effects of triprolidine at high concentrations in the IC 50 assay. That said, we were able to confirm that the majority of compounds we tested had growth inhibitory effects consistent with the drug screen data. It is also worth noting that our screen likely contained some false negatives. Among other reasons, the hygromycin resistance marker used to select for the EP1/GFP reporter cells may have inactivated some of the drugs in the library due to phosphorylation.

PLOS NEGLECTED TROPICAL DISEASES
Screen for small molecules that inhibit growth and promote differentiation in T. brucei We also showed proof of concept for a reporter strain that can be used to track the molecular events that initiate surface remodeling during trypanosome differentiation from the bloodstream form to the procyclic form, an event that is critical to life cycle progression. We first showed that expression of EP1/GFP in our reporter line behaves as expected in bloodstream, procyclic and differentiating cells. One weakness of our reporter is that it was used in a monomorph strain rather than a pleomorph strain. While pleomorphs can be treated in vitro to closely mimic developmental transitions from bloodstream to procyclic parasites via the intermediate stumpy stage that is observed in the wild, monomorph lines are incapable of stumpy formation, making them a less physiological model. With this in mind, we created a pleomorphic EP1/GFP reporter strain and piloted it for use in flow cytometric screens. Unfortunately, EP1/GFP expression in these pleomorphic lines is much more variable than in the monomorph lines, and tends to increase with increasing parasite density (S10 Table and S4 Fig), which is consistent with preadaptation of stumpy parasites for transition to the insect vector [47]. Therefore, we decided to use monomorph lines for screening purposes, and for any drugs that produced procyclin protein on the surface, these drugs could be separately tested in pleomorph lines using anti-procyclin antibodies. Monomorph lines have been successfully used by others to identify factors important for differentiation when coupled with follow-up studies in pleomorph parasites [55][56][57].
This study demonstrates that our developmental reporter line can be used to conduct flowcytometry based small molecule drug screens. Importantly, our screening pipeline was developed such that growth inhibition can be tracked at the same time as the transcript abundance for our gene of interest. This allowed us to quickly identify drugs that might cause a non-specific increase in EP1/GFP expression, perhaps due to a stress response induced by the drug. Phenothiazine's effect on EP1/GFP transcript levels may well be non-specific, as the drug is a potent growth inhibitor, although many other drugs that inhibited growth did not have this effect. Our screen is similar in concept to a previously reported screen that tracked EP1 expression levels using a GUS reporter system [11,12]. This reporter was also used in a mediumthroughput screen of 7,500 compounds in a 96 well plate format using drug concentrations between 1 and 20μg/ml. For comparison, our screening concentration of 33μM corresponds to a 6-13μg/ml treatment for our EP1/GFP hits. GFP expression was found to be~26 times higher when we treated parasites with cis-aconitate than when we treated them with eflornithine (Fig 6 and S3 Fig). This phenomenon was also observed in the GUS screen; compound-induced GUS signals were up to 5000-fold lower than when parasites were treated with cis-aconitate [12]. All of the hits from the GUS screen were shown to inhibit parasite growth, and the top hits also had an effect on surface VSG MFI [12]. Similar to the hits from the previous study, phenothiazine-treated parasites showed inhibited growth, higher transcript levels of EP1, and altered VSG MFI (Fig 5B and 5C). However, while EP1 transcript levels were higher in eflornithine and spironolactone-treated parasites, neither of these compounds showed effects on parasite growth or VSG MFI. To our knowledge, the level of procyclin protein expressed on the parasite surface was not tested in parasites treated with compounds that induced higher GUS signals in the previous study. In our study, we failed to identify any compounds that produced procyclin protein on the surface of parasites, underscoring the already established paradigm that additional developmental steps may be required to remodel the parasite surface following transcription of the relevant RNAs. In addition, defects in glycosylation of the procyclin protein and/or defects in the addition of the GPI anchor could also explain the lack of procyclin protein observed on the surface of the parasites [65][66][67][68]. Given that the bromodomain inhibitor I-BET151 has previously been shown to increase both RNA and protein levels of procyclin [10], a more targeted screen using a library of epigenetic inhibitors might uncover more compounds that lead to this phenotype in the EP1/GFP reporter line.
While the GUS reporter system has been put to excellent use to identify compounds that increase transcript levels of EP1, a fluorescence-based reporter might have some advantages when performing genetic screens for native factors that are important for this process during developmental transitions. For example, parasites could be transfected with either an RNAi library or an overexpression library in bulk, and fluorescence positive parasites could be physically separated from the population using fluorescence activated cell sorting. This would allow for easy identification of gene products that increase the transcription and/or stability of procyclin, which could provide insights into molecular mechanisms for surface remodeling that occurs during developmental transitions.
Treatment with both spironolactone and eflornithine did not affect expression of EP1 alone, as we found effects on several other differentiation-associated genes. This phenotype was also previously observed for I-BET151, although its effects on differentiation associated genes were of greater magnitude [10]. In future, it would be interesting to test whether artificially increasing EP1 levels via overexpression of the gene is sufficient to induce expression changes in other differentiation-associated genes. It should be noted that we have no data to support a direct effect for these compounds on EP1/GFP expression, and the changes in transcript abundance for both EP1 and the other differentiation-associated genes tested may well be due to secondary effects.
While eflornithine had previously been reported to induce a stumpy-like, but developmentally incompetent, transition in monomorphic cells [46], the effect of spironolactone on T. brucei physiology has not been studied. In mammalian systems, spironolactone is an antiandrogen drug that non-selectively antagonizes mineralcorticoid receptors, including both progesterone and androgen receptors. It is used as a hypertensive agent to decrease morbidity and mortality of heart failure [58]. To our knowledge, no proteins with similarity to mineralocorticoid receptors have been identified in T. brucei. We used BLAST to interrogate various kinetoplastid genomes with both the DNA-binding domain of the human glucocorticoid receptor and the ligand-binding domain of the mineralocorticoid receptor. While no hits were uncovered in T. brucei, we did obtain a few hits for genes in T. cruzi and Leishmania panamensis using the DNA-binding domain. All of the identified genes in T. cruzi (Tc00.1047053504039.110, Tc00.1047053506769.55, and Tc00.1047053510555.21) with an E value of less than 0.01 were for hypothetical proteins and ranged from 26%-33% identities. All three of them were also shown to be transcriptionally downregulated after infection in amastigotes and upregulated in epimastigotes [59]. For BLAST with the ligand-binding domain of the mineralocorticoid receptor, we only obtained two hits with an E value of less than 0.05, one in Perkinsela (XU18_4875) and the other in Leishmania Mexicana (LMXM_29_2680). Both genes code for hypothetical proteins and had a 33-34% identity with the query. Future studies could potentially uncover the target of spironolactone in T. brucei by overexpressing a library of T. brucei ORFs and assaying for a decrease in GFP expression in spironolactone treated EP1/GFP reporter cells. These types of overexpression suppressor screens have previously been used successfully to identify drug targets in yeast and other organisms [60]. With the gene target identified, genetic studies could uncover the role of this target (if any) in the initiation of transcription or stabilization of EP1 transcript following the appropriate developmental signals in bloodstream form cells.
While we focused on generating alternate trypanosomiasis therapeutics by inducing life cycle progression in T. brucei, this strategy could likewise be applied to other causative agents for neglected disease. Indeed, this strategy is actively being pursued for treatment of malaria, toxoplasmosis, and schistosomiasis [61][62][63]. We think our reporter strain (and other developmental reporter strains such as the Pad1-GFP strain that tracks stumpy formation [64]) offer a powerful tool to perform genetic screens for developmental phenotypes. With the increasing availability of knockdown and overexpression libraries in T. brucei, there is great potential to uncover new life cycle-specific biological roles for the numerous proteins in T. brucei that are annotated as hypothetical. The combination of fluorescent reporter strains with high-throughput flow cytometry assays facilitates screening strategies that are both less labor-intensive and time-intensive than previous methods.
Supporting information S1 Fig. Triprolidine does not affect T. brucei growth at low concentrations. Percent growth inhibition over a range of concentrations for the indicated drug. Data were fitted and indicated values calculated using the GRMetrics R package. GR 50 , the concentration at which the effect reaches a growth rate (GR) value of 0.5 based on interpolation of the fitted curve (dashed lines). GIC 50 , the drug concentration at half-maximal effect for calculated growth rate. r 2 GR, the coefficient of determination for how well the GR curve fits to the data points. IC 50 , the concentration at which relative cell count = 0.5. r 2 IC 50 , the coefficient of determination for how well the traditional curve fits to the data points. (TIF)

S2 Fig.
A) Anti-procyclin western blot for parasites isolated after 3d of treatment with the indicated drug. Procyclic parasites were used as a positive control. Anti-H3 was used as a loading control. Note that the samples ran high toward the right-hand side of the blot. B) Anti-procyclin western blot for parasites isolated after 2d or 3d of treatment with the indicated drug. Procyclic parasites were used as a positive control. Anti-H3 was used as a loading control. Because phenothiazine-treated parasites were so sick it was difficult to get sufficient numbers of cells for the assay; consequently, these samples and their controls are loaded with less protein.