Identifying anti-prion chemical compounds using a newly established yeast high-throughput screening system

SUMMARY

Prion-like protein aggregation underlies the pathology of a group of fatal neurodegenerative diseases in humans, including Alzheimer’s disease (AD), Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS), and transmissible spongiform encephalopathy (TSE). Currently, few high-throughput screening (HTS) systems are available for anti-prion small molecule identification. Here we describe an innovative phenotypic HTS system in yeast that allows for efficient identification of chemical compounds that eliminate the yeast prion [SWI⁺]. We show that some identified anti-[SWI⁺] compounds can destabilize other non-[SWI⁺] prions, and their antagonizing effects can be prion- and/or variant-specific. Intriguingly, amongst the identified hits are several previously identified anti-PrP^Sc compounds and a couple of FDA-approved drugs for AD treatment, validating the efficacy of this HTS system. Moreover, a few hits can reduce proteotoxicity induced by expression of several pathogenic mammalian proteins. Thus, we have established a useful HTS system for identifying compounds that can potentially antagonize prionization and human proteinopathies.

eTOC Blurb

To develop effective anti-prion drugs is an unmet medical need. In this issue of Cell Chemical Biology, Du et al. have established a phenotypic assay for high-throughput screening of anti-prion small molecules. Their results demonstrate that this system has the potential to identify compounds antagonizing prions and prion-like neurodegenerative diseases.

Graphical Abstract

INTRODUCTION

Prions are self-perpetuating protein conformations that are often linked to protein misfolding, aggregation, and amyloidogenesis. While PrP^Sc, the causative agent of TSEs, was the first identified prion (Prusiner, 1982), recent evidence suggests that numerous additional amyloidogenic proteins that are associated with neuropathologies, such as tau, α-synuclein (α-syn), and some mutant forms of superoxide dismutase 1 (SOD1) (Ayers et al., 2016; Luk et al., 2012; Sanders et al., 2014), are also prion-like - forming pathogenic conformations that can spread from cell to cell and thus allow the progression of the disease from one tissue to another (also see review of (Soto, 2012)). In addition, many prion and prion-like proteins with diverse cellular functions have been identified in organisms across the kingdoms of life - mammals (Cai et al., 2014), Aplysia (Si et al., 2003), Drosophila (Majumdar et al., 2012), Arabidopsis (Chakrabortee et al., 2016), bacteria (Yuan and Hochschild, 2017) and fungi (Saupe et al., 2016; Wickner, 2016). Remarkably, there are several amyloid prions, including [PSI⁺], [URE3], [PIN⁺], [SWI⁺], [MOT3⁺], and [MOD⁺] identified in the budding yeast Saccharomyces cerevisiae (Crow and Li, 2011; Liebman and Chernoff, 2012; Wickner, 2016). The prion phenomenon appears to be more ubiquitous than we once thought and importantly, prions can play roles in both normal cellular function and etiology of disease. Thus, discovery of anti-prion small molecules may lead to development of useful chemical probes for studying prion biology and effective therapeutic drugs for treating prion and prion-like diseases.

In this regard, several systems with distinct approaches have been developed to screen for anti-prion compounds, including animal and cell culture models, cell-free in vitro conversion assays, in silico strategies, and immunotherapeutic tactics (Bolognesi and Legname, 2015; Cordeiro and Ferreira, 2015; Pagadala et al., 2017). These screens have resulted in identification of several chemical compounds with anti-PrP^Sc activity in vitro, ex vivo, and in vivo – whether through direct targeting of PrP or improving protein quality control by increasing molecular chaperone levels or reducing ER stresses (Boyce et al., 2005; Imberdis et al., 2016; Kamatari et al., 2013; Karapetyan et al., 2013; Sim, 2012). However, none of the identified compounds prevent or effectively treat prion diseases. Similarly, despite decades of research, no effective therapeutic drugs have been developed to cure other prion-like misfolding diseases. Thus, there is an urgent need for developing effective high throughput screening (HTS) platforms to identify anti-prion small molecules.

Although specialized neuronal properties are absent in yeast cells, basic eukaryotic cellular features are conserved from yeast to human. There are at least 200 genes in yeast that are homologous to human genes involved in diseases (Hartwell, 2004). The similarity of protein-homeostasis (proteostasis) machinery, vesicular trafficking, and autophagy have led to the establishment of a number of yeast models for mammalian protein-misfolding diseases (Fushimi et al., 2011; Ju et al., 2011; Krobitsch and Lindquist, 2000; Kryndushkin et al., 2011), and several novel risk factors in mammals that are relevant to synucleinopathy (Auluck et al., 2010; Basso et al., 2013), AD (Park et al., 2011; Treusch et al., 2011), and ALS (Couthouis et al., 2011; Kim et al., 2013), have been identified by taking the advantage of the simplicity and amenability of the yeast system. Yeast prions have also provided powerful platforms to study prionogenesis and transmission mechanisms (Crow and Li, 2011; Liebman and Chernoff, 2012). Indeed, despite the lack of homologies in sequence or function, yeast prions possess some structural and transmission features that are substantially similar to that of PrP^Sc, such as forming amyloids under physiological conditions (Alberti et al., 2009; Du et al., 2010; Glover et al., 1997; King et al., 1997; Patel and Liebman, 2007; Suzuki et al., 2012), and existing as multiple distinct transmissible conformational variants (Bradley and Liebman, 2003; Derkatch et al., 1997; Schlumpberger et al., 2001; Stein and True, 2014). Yeast prion research has thus provided valuable information regarding the structural basis of prions and their transmission mechanism (Sigurdson et al., 2005). The findings that some small molecules that antagonize the yeast prion [PSI⁺] also inhibit PrP^Sc accumulation in cell lines, and conversely compounds inhibitory to mammalian PrP^Sc are able to effectively inhibit the transmission of [PSI⁺] in yeast (Bach et al., 2003; Tribouillard et al., 2006) suggest that prionization mechanisms might be conserved in yeast and mammals. Thus, yeast prions may serve as platforms to identify anti-prion compounds. In this regard, a [PSI⁺] prion-based small molecule screening for inhibitor compounds was reported (Bach et al., 2003). Albeit a useful system, this [PSI⁺]-based screening methodology requires a rather long testing period and involves manually transferring prion cells onto antibiogram filters prior to a color-based visual test (Bach et al., 2003; Tribouillard et al., 2006). More recently, a liquid culture-based assay was reported for identifying [PSI⁺] inhibitors (Jennings et al., 2018). Although this more simplified assay identified several additional anti-[PSI⁺] small molecules with novel structures, it was rather low throughput. Alternative yeast prion-based methods that are more effective, easier to assay, and more amenable to automated high-throughput screening systems are thus desirable.

The yeast prion [SWI⁺] was discovered in our laboratory (Du et al., 2008). The protein determinant of [SWI⁺] is Swi1 – a subunit of an evolutionarily conserved, ATP-dependent chromatin-remodeling complex known as SWI/SNF (Cairns et al., 1994). The SWI/SNF complex regulates ~ 6% of total gene expression in yeast (Sudarsanam et al., 2000). Swi1 contains an asparagine (N)-rich prion domain (PrD) at its NH3-terminus (Crow et al., 2011; Du et al., 2010; Valtierra et al., 2017). PrDs are defined as regions within a prion protein that are essential for prion de novo formation and transmission, and they are also often the region required for protein aggregation and amyloidogenesis (Alberti et al., 2009; Du, 2011; Michelitsch and Weissman, 2000; Ross et al., 2005). Like other yeast prions, [SWI⁺] is dominantly and cytoplasmically transmitted, and thus is “infectious” (Du et al., 2008). Amyloid architecture was shown to be the structural basis of [SWI⁺] (Du et al., 2010). [SWI⁺] interacts with heterologous prions (Du et al., 2017; Du and Li, 2014; Nizhnikov et al., 2016), and relies on a delicate molecular-chaperone network for its de novo formation and propagation (Du et al., 2008; Hines et al., 2011). Interestingly, depending on the specific traits that are examined, [SWI⁺] may exhibit partial or complete loss-of-function phenotypes. For example, [SWI⁺] cells grow slowly in raffinose media, showing a partial loss-of-function phenotype on raffinose usage (Du et al., 2008), but they exhibit a complete loss-of-function phenotype of exhibiting multicellular features (Du et al., 2015). Like swi1Δ cells, the transcription of FLO1, one of the flocculin (FLO) genes conferring multicellular features (Guo et al., 2000) is completely inactivated in [SWI⁺] cells (Du et al., 2015). Based on this discovery, we established a [SWI⁺] reporter by replacing the FLO1 ORF with the URA3 ORF at the FLO1 chromosomal locus and we showed that this reporter can allow for positive and negative selection of the [SWI⁺] prion (Du et al., 2015). Utilizing this reporter, we successfully analyzed the de novo formation and propagation events of [SWI⁺] (Du et al., 2017; Du et al., 2015). In this study, we further engineered this [SWI⁺] reporter to establish an automation compatible two-step growth assay to allow HTS for chemical compounds that reactivate FLO1pr-URA3, indicative of the loss of [SWI⁺]. Using this platform, we have identified a number of hits that can indeed eliminate or inhibit [SWI⁺]. Further analysis showed that some of these hits can also inhibit other non-[SWI⁺] yeast prions and/or suppress toxicities caused by expression of several human pathogenic proteins linked to neurodegenerative diseases.

RESULTS

Establishing a [SWI⁺]-based HTS System for Identifying Anti-[SWI⁺] Compounds

We previously constructed an Ura3-based reporter system that can positively select for and negatively select against the [SWI⁺] prion (Du et al., 2017; Du et al., 2015). By integrating a functional FLO8 at the HIS3 locus of the lab strain S288C to compensate for the harbored flo8 non-sense mutation at codon 142, and by replacing its chromosomal FLO1 ORF with the URA3 ORF so the expression of URA3 would be regulated by the FLO1 promoter (FLO1pr-URA3), we showed that the FLO1 promoter is active in S288C [swi⁻] (non-prion) cells but inactive in [SWI⁺] cells due to Swi1 aggregation and sequestration of several other transcriptional activators required for FLO1 expression (Du et al., 2015). As shown in Figure 1A and 1B, [swi⁻] cells are able to grow in synthetic complete (SC) media without uracil (SC-ura) but are unable to grow in 5-Fluoroorotic Acid (5-FOA) because Ura3 can convert 5-FOA to a toxic compound, whereas [SWI⁺] cells cannot grow on SC-ura but are able to grow on SC media containing 5-FOA. As [SWI⁺] loss will convert cells from Ura⁻ to Ura⁺, which can be positively selected using SC-ura medium, we speculated that this system would allow us to positively identify small molecules that cure [SWI⁺], and perhaps we could develop it into a HTS platform to screen for anti-[SWI⁺] compounds. To do so, we further engineered the FLO1pr-URA3 reporter strain of S288C by deleting its ERG6 gene in order to enhance cell permeability. ERG6 encodes an enzyme involved in ergosterol metabolism and it has been shown that erg6Δ results in altered lipid composition of the plasma membrane and increased cell permeability (Bach et al., 2003; Gaber et al., 1989). We thus replaced the ERG6 gene with the LEU2 auxotrophic marker (Figure 1C and 1D). Deletion of ERG6 indeed increased the sensitivity of [SWI⁺] curing by 5 mM guanidine hydrochloride (GdnHCl), an inhibitor of the molecular chaperone Hsp104 required for [SWI⁺] propagation (Du et al., 2008) and the only known chemical compound that inhibits [SWI⁺] (Figure 1E). However, GdnHCl with a concentration of 0.5 mM or lower showed no curing effect on [SWI⁺] (Figure 1F). The curing media used in the primary screening in this study was supplemented with 0.2 mM GdnHCl.

Figure 1. Development of a [SWI⁺]-based Assay for HTS of Anti-prion Compounds.

(A) A single-step gene replacement to swap the chromosomal FLO1 ORF with the URA3 ORF to generate a [SWI⁺] reporter - FLO1 promoter-controlled URA3 (FLO1pr-URA3). (B) Late-log phase cultures of the FLO8-restored FLO1pr-URA3 [SWI⁺] and [swi⁻] cells were spotted after serial dilution onto a synthetic complete (SC) plate without uracil (-uracil) or with 5-FOA (+ 5-FOA) and a SC plate containing 2% raffinose (raffinose) or 2% glucose ((SC) glucose)) as the sole carbon source, and images were taken after 3 days of incubation. (C) A diagram showing the strategy of ERG6 gene deletion. The chromosomal ERG6-ORF is replaced with the LEU2 ORF. (D) Verification of ERG6 deletion by PCR. The presence of the 743 bp band confirms the gene replacement. (E) The ERG6 gene deletion increases the sensitivity of [SWI⁺] to curing by guanidine hydrochloride (GdnHCl). Log-phase cultures of the indicated strains were grown in selective SC medium containing 5 mM GdnHCl for the indicated time before spotting onto SC-ura plate. Water was used as a negative control. (F) GdnHCl with a concentration less than 0.5 mM did not cure [SWI⁺]. For growth assay in liquid culture (top), [SWI⁺] FLO8 FLO1pr-URA3 erg6Δ cells were grown for 24 h in SC media containing the indicated amount of GdnHCl, and then diluted into SC-ura to a density of 5×10³ cells/mL. Growth was monitored by measuring the optical density at 600 nm (OD 600 nm) using a plate reader for ~32 h. Similarly, cells grown in SC media containing the indicated amount of GdnHCl for 24 h and spotted directly on to SC-ura plate to monitor the prion loss. (G) The effect of DMSO on the assay. Isogenic strains (FLO8 FLO1pr-URA3 erg6Δ) of [SWI⁺] and [swi⁻] were grown in SC medium supplemented with the indicated amount of DMSO and GdnHCl (water as control) for 24 h, and then were diluted to inoculate SC-ura medium in a 96-well plate. The growth was monitored by OD 600 nm measurement. In all cases, three independent experiments were done and a representative is shown.

This condition was also used for identifying antagonizing compounds against the yeast prion [PSI⁺] in a previously reported study, as this minimal GdnHCl concentration may increase the sensitivity of the assay (Bach et al., 2003). Since DMSO was used as a vehicle for most available chemical libraries, its effect on our assay was also examined. We tested the [SWI⁺] curing efficiency of GdnHCl (0.2 or 5 mM) in the curing media of various concentrations of DMSO. As shown in Figure 1G, DMSO with a concentration ≤ 2% did not significantly alter the curing effect of GdnHCl. However, DMSO with a concentration of 5% or higher was toxic to both [SWI⁺] and [swi⁻] cells (Figure 1G and data not shown). Thus, in subsequent experiments throughout this study, 5 mM GdnHCl and DMSO (<2%) were used as positive and negative controls, respectively. We also observed that a 24-hour period is the minimal length of time required for efficient curing of [SWI⁺] (data not shown). Based on these tested conditions, we set up a protocol for high-throughput analysis using 384-well microplates (Figure 2A, also see STAR Methods). Following this protocol, we achieved an average Z’ score of ~0.68 and a P value of 2.5E-171 (Figure 2B), using 5 mM GdnHCl and 2% DMSO. Z’ is a measure of statistical effect size of an assay and a Z’ score equal or higher than 0.5 indicates the excellence of the examined assay. Thus, our assay appeared to be robust and suitable for high-throughput screening.

Figure 2. Primary Assay Protocol, Assay Quality Assessment, and Screens.

(A) A diagram illustrating the HTS primary assay protocol for identifying anti-[SWI⁺] molecules. (B) Evaluation of the quality of the growth-based HTS assay using only negative (2% DMSO) and positive (5 mM GdnHCl) controls following the protocol illustrated in panel A. Shown is OD 600 nm of cultures from detection medium (SC-his-ura). Growth in SC-his-ura indicates prion loss or inhibition. All even number of wells contain GdnHCl and all odd number of wells contain DMSO. Indicated are also average Z’ score and P value (T-test) calculated with methods described in STAR methods. (C) Summary of anti-[SWI⁺] screening results. n, replication number of positive (GdnHCl) and negative controls used in calculating average Z’ factor and P value for each library. See details in STAR Methods, also see Figure S1 for raw data of primary screening and Table S1 for detailed features of all verified hit compounds.

Screening for Anti-[SWI⁺] Compounds

Using this established high-throughput system, we screened five chemical libraries that were available in the Northwestern University High Throughput Analysis Laboratory – an NIH Clinical Collection (NCC) with 450 compounds that are almost entirely drugs that have been in phase I-III clinical trials and have not been represented in other arrayed collections; a MicroSource Spectrum Library of 2,000 compounds encompassing a wide range of biological activities and structural diversity; a collection of 3,200 compounds from the National Cancer Institute/Developmental Therapeutic Program (NCI/DTP); an ASDI library of 6800 compounds with diverse chemical structure that meet the Lipinski guidelines for drug-likeness; and the Chembrige DIVERset-CL with compounds of enhanced potential for therapeutic development, ~ 4,800). We screened the NCC, Spectrum, and DIVERset-CL libraries with duplicates of individual compounds and obtained reproducible results (Figure S1). For the NCI/DTP and ASDI libraries, we used a single copy of each compound for our primary screens. The values of Z’ of these primary screens were calculated to be from 0.51 to 0.71 and the P values of the difference between the positive and negative controls in these screenings range from 4.8E-24 to 1.9E-53 (T-test) (Figure 2C and Figure S1), suggesting that our screening experiments were valid and the results were trustworthy. We obtained 115 total hits from the five libraries (Figure 2C). Cherry picking assays verified that at least 48 hits (47 unique compounds because tacrine hydrochloride was obtained from both the NCC and Spectrum) were able to convert the [SWI⁺] cells from Ura⁻ to Ura⁺, indicating that the [SWI⁺] prion may have been lost upon treatment. The structural information of the top 22 hits is shown in Figure 3. Additional information about the chemical and biological features of all verified hits is shown in Table S1.

Figure 3. The Structures and EC50^G Values of the Top Verified Hits.

Hits are displayed in an order based on their potencies ranked by EC50^G. Calculation of EC50^G is described in STAR Methods. Hits in red were previously reported to have activities against yeast prion(s), mammalian prion, AD or PD. Names/synonyms are also marked. Key atoms are highlighted with distinct colors in each structure (Wire-Frame model) as illustrated.

We observed diverse effectiveness of our verified hits in [SWI⁺]-curing, with an estimated EC50^G (defined as 50% of anti-[SWI⁺] activity of that of 5 mM GdnHCl) ranging from 0.13 μM to 746.8 μM (Table S1). These hit compounds also exhibit diverse structures (Figure 3). Among the 22 top-ranked hits, 7 were previously shown to have activities antagonizing yeast prions, mammalian prion, AD or PD (Figure 3. See Table S1 for details), demonstrating the efficacy of our assay for identifying compounds against prions and amyloid-mediated neurodegenerative diseases. Besides the above seven compounds, novel structures were observed that we decided to characterize further. Thus, 12 commercially available compounds (Table S1, highlighted in blue) from the NCC, MicroSource Spectrum and Chembrige DIVERset-CL libraries were purchased and examined for their effectiveness against [SWI⁺]. This was done by treating the [SWI⁺] erg6Δ::LEU2 FLO1pr-URA3 FLO8::HIS3 cells with compounds at various concentrations ranging from 0 μM to 1000 μM for 24 hours, and after proper dilution, spreading them onto a SC-his plate to estimate the total number of spread cells and a SC-his-ura plate to quantify the number of Ura⁺ cells. The effectiveness of these compounds to eliminate [SWI⁺] was estimated by quantifying the percentage of Ura⁺ cells, which had presumably lost [SWI⁺] and thus converted from Ura⁻ to Ura⁺, over the total cells analyzed. The conditions of 5 mM GdnHCl and 2% DMSO were used as a positive and a negative control, respectively. As shown in Figure 4, the treatment of [SWI⁺] cells with GdnHCl generated ~70–80% Ura⁺ cells (the green dot in each panel), while no Ura⁺ cells were observed for DMSO-only treatment (the red dot in each panel). Similar results were obtained when the test was done in liquid cultures supplemented with the tested compound. Two compounds, nimustine and quinacrine, were not effective in the given concentration range examined whereas the other 10 compounds were effective. Amongst the 10 effective compounds, tacrine hydrochloride and aminacrine were the most effective in eliminating [SWI⁺]. The lowest concentration examined to have an inhibitory effect on [SWI⁺] is about 1 μM for both compounds (Figure 3). Dose response curves of the rest of 8 compounds were shown in Figure 4. No reports on the effect of phloretin and pilocarpine against yeast prions or PrP have been published, therefore, a more in-depth analysis of these two compounds was performed. The EC₅₀, a concentration that gives half-maximal prion curing activity or converts approximately 50% of the [SWI⁺] cells from Ura− to Ura⁺ in this study, is ~2.5 μM for phloretin and ~300 μM for pilocarpine (Figure 4). For hits obtained from the Chembridge DIVERset-CL library, cherry-picking results confirmed that six compounds (cpd7–12) were effective in eliminating [SWI⁺]. To our knowledge, none of the six compounds have been shown to have anti-prion activities. Therefore, we further examined the ability of the compounds to eliminate [SWI⁺] at a concentration range from 0 μM to 500 μM for Cpd7–11, and from 0 μM to 1000 μM for Cpd12. As shown in Figure 4, EC₅₀ values for these compounds are in the range of 100 – 250 μM.

Figure 4. Dose Response Assessment of the Effectiveness of Hit Compounds against [SWI⁺].

The [SWI⁺] FLO8::HIS3 erg6Δ::LEU2 FLO1pr-URA3 strain was grown in curing medium (SC-his containing the tested compound with the indicated concentration) for 24 h before spreading onto plates of SC-his (for total growth) and SC-his-ura (for cured cells). Colony numbers were quantified after 3 days of growth. Shown is the percentage of Ura⁺ colonies out of the total counted colonies on SC-his. Curing medium was supplemented with 0.2 mM GdnHCl. Treatments of 2% DMSO (red dot) and 5 mM GdnHCl (green dot) treatments are also included as controls. Nonlinear regression was used for curve fitting and EC50 calculation. The loss of [SWI⁺] for selected Ura⁺ isolates was confirmed with secondary assays (see Figure S2).

Secondary Assays to Verify the Reliability of the Pilot Screening Results

Next, we examined if the conversion from Ura⁻ to Ura⁺ upon treatment was indeed a result of [SWI⁺] loss rather than mutations or other factor(s) interfering with our FLO1pr-URA3 reporter system. We previously demonstrated that Swi1 forms prion aggregates that can be only observed in [SWI⁺] but not in [swi⁻] cells, and [SWI⁺] cells exhibit a compromised ability to use non-glucose carbon sources such as raffinose and thus show a partial growth phenotype (Raf^±), while [swi⁻] cells can efficiently use raffinose (Raf⁺) (Du et al., 2008). Therefore, Swi1 aggregation and growth in raffinose media were used as secondary assays to examine the [SWI⁺] status of cells treated with hit compounds. We used 2% DMSO and 5 mM GdnHCl as negative and positive controls, respectively.

For raffinose assays, after treating the [SWI⁺] erg6Δ::LEU2 FLO1pr-URA3 FLO8::HIS3 cells with a hit compound, we spread the cell mixtures onto SC-his plates and then replica-plated onto SC-his-ura plates. The resulting Ura⁻ and Ura⁺ isolates were then analyzed for their ability to grow on SC-raffinose plates. As shown in Figure S2, all Ura⁺ isolates obtained after treatment with pilocarpine or tacrine were 5-FOA⁻ / Raf⁺, whereas all tested Ura⁻ isolates were 5-FOA⁺ / Raf^±. Similar results were observed for other tested compounds (phloretin, and Cpds 7–12), which were comparable to 1 mM GdnHCl treatment (Figure S2). Thus, all Ura⁺ isolates are uniformly Raf⁺, indicative of the prion loss. As expected, the DMSO treatment of [SWI⁺] erg6Δ::LEU2 FLO1pr-URA3 FLO8::HIS3 cells only yielded Ura⁻ isolates that were 5-FOA⁺ / Raf^±.

For aggregation assays, a [SWI⁺] FLO1pr-URA3 FLO8::MET15 strain was transformed with a Swi1-NQYFP-expression plasmid that is controlled by either a TEF1 or a GAL1 promoter, and after compound treatment, Ura⁻ and Ura⁺ isolates were randomly picked for aggregation assay either in SC-glucose media for the TEF1 promoter or in SC-galactose media for the GAL1 promoter. Without treatment, Swi1 NQ-YFP aggregated in [SWI⁺] cells but diffused in [swi⁻] cells (Figure S2 controls). After a treatment with DMSO, as expected [SWI⁺] cells remained Ura⁻, and showed aggregated Swi1 NQ-YFP signals (Figure S2 bottom right). After the treatment of the [SWI⁺] strain with pilocarpine and tacrine hydrochloride, the obtained Ura− isolates always showed Swi1 NQ-YFP aggregation, while all Ura⁺ isolates did not (Figure S2). Similarly, the switch from Ura⁻ and Ura⁺ also correlated well with the loss of [SWI⁺] when we examined another 10 selected hit compounds (data not shown). Collectively, these data indicated that the Ura⁺ isolates acquired after the chemical treatment indeed had lost [SWI⁺], and thus these examined compounds possess anti-[SWI⁺] activity.

The Effect of Anti-[SWI⁺] Compounds on Other Yeast Prions

We next examined if the identified anti-[SWI⁺] compounds could also antagonize other yeast prions, in particular [PSI⁺], [URE3], and [MOT3⁺]. For [PSI⁺], strains containing the ade1–14 allele with a premature stop codon in the ORF of ADE1 were used (Figure 5A and (Chernoff et al., 1993)). Sup35, the protein determinant of [PSI⁺], is a translational terminator, and its prion form compromises its function and causes read-through of a premature stop codon; thus, [psi⁻] 74-D694 cells cannot grow in media without adenine and develop into red colonies on YPD due to a pigment accumulation, whereas [PSI⁺] cells can grow in media without adenine and develop into white colonies on YPD due to the presence of functional Ade1 (Derkatch et al., 1996; Uptain et al., 2001). This phenotype has been used as a convenient and reliable reporter for analyzing [PSI⁺]. The protein determinant of [URE3] is Ure2, a transcriptional repressor of the DAL5 gene (Brachmann et al., 2006). [URE3] can be easily assessed in a strain containing a DAL5pr-ADE2 (DAL5 promoter controlled ADE2) reporter, which is on in [URE3] cells but off in [ure-o] (non-prion) cells (Figure 5B and (Brachmann et al., 2006)). Thus, like [PSI⁺] and [psi⁻], [URE3] and [ure-o] cells can be distinguished easily as they form white and red colonies on YPD, respectively. The protein determinant of [MOT3⁺] is Mot3, a transcriptional repressor of the DAN1 gene (Alberti et al., 2009; Holmes et al., 2013). Based on this repressive activity of Mot3, a DAN1pr-URA3 reporter reported by Alberti et al allows for positive selection for [MOT3⁺] cells on SC-ura as the prion form Mot3 had lost its repressive activity on DAN1 expression (Alberti et al., 2009).

Figure 5. The Effects of Anti-[SWI⁺] Molecules on Other Non-[SWI⁺] Yeast Prions.

(A-C) Illustration of ADE-based yeast prion reporter systems. (A) [PSI⁺] prion reporter; (B) [URE3] prion reporter; and (C) [MOT3⁺] prion reporter. In all three reporter systems, prion cells exhibit Ade⁺ phenotype and appear as pink/white colonies on YPD plates, while non-prion cells exhibit as Ade⁻ phenotype and appear red on YPD. (D) After being treated with 500 μM of the specified compounds for 48 h in YPD liquid medium (supplemented with 0.2 mM GdnHCl), the indicated cells were spread onto YPD. The appearance of red colonies is indicative of prion loss. Shown are representative results. Note, both [PSI⁺]^S and [PSI⁺]^W shown in this panel were created by co-overproduction of Swi1 and Sup35 NM-GFP (Du and Li, 2014). (E) Prion loss in D was quantified. Shown are percentages of non-prion isolates (red colonies) from at least three independent experiments. (F-G) Two [PSI⁺][PIN⁺] strains created by overproduction of Sup35 NM-GFP from a [PIN⁺] strain (Du and Li, 2014) were treated with pilocarpine nitrate at the indicated concentrations for 24 hours in YPD supplemented with 0.2 mM GdnHCl. Five mM GdnHCl and water were included as controls. After treatment, cells were spread onto YPD for colony color visualization. Shown are representative results of colony images (F) and quantitative plots (G) from at least three independent experiments. Note, sectored colonies stand for partial loss of [PSI⁺].

To establish a convenient visual assay of [MOT3⁺], we replaced the endogenous ADE1 promoter with the DAN1 promoter (DAN1pr-ADE1). This allows [MOT3⁺] cells to develop into white colonies on YPD, whereas the [mot3⁻] (non-prion) cells develop into red colonies under identical conditions because the prion form of Mot3 is inactive in repressing the DAN1 promoter, (Figure 5C). Upon the establishment of the DAN1pr-ADE1 reporter, the same Ade⁺-based assay can be used for all prion elements to be examined, providing a convenient and objective means to allow us to conduct side-by-side comparison of these prions in response to treatments of the hit compounds. Under this system, all prion cells will exhibit an Ade⁺ phenotype and develop into pink or white colonies on YPD whereas non-prion cells will be Ade⁻ and form red colonies on YPD.

In our initial tests, we used 500 μM of each compound in the curing media to treat against the following prions: [PSI⁺] ([PSI⁺]^W for a weak variant, and [PSI⁺]^S for a strong variant), [URE3], and [MOT3⁺]. We tested the confirmed anti-[SWI⁺] hits – pilocarpine nitrate, tacrine hydrochloride, aminacrine, phloretin, and Cpd7–12 (Figure 5D–5E and data not shown). As expected, the positive control, 5 mM GdnHCl, eliminated all tested prions (data not shown), but the negative control, DMSO-only, had no effect (Figure 5D). Phloretin and Cpd7 did not cure any of these prions under all assay conditions (data not shown). Other tested compounds exhibited diverse curing activities in a prion-specific manner (Figure 5D and 5E). Overall, [MOT3⁺] was the most sensitive prion among those tested. About 80%−90% of [MOT3⁺] cells lost the prion upon treatment with tacrine, aminacrine, Cpd10, and Cpd12, with Cpd12 being the most effective. In comparison, [URE3] was the prion most resistant to the treatment. It could only be destabilized by tacrine, aminacrine, and Cpd10, with a curing efficiency of about 25%, 15%, and 45%, respectively (Figure 5D and 5E). Further tests showed that [URE3] could not be eliminated by pilocarpine at a concentration as high as 6 mM (data not shown). Cpd12 was also the most effective compound in eliminating the two [PSI⁺] variants with an efficiency of ~80% but was not effective for [URE3] curing (with an efficiency of <10%). In comparison, upon Cpd9 treatment, only ~1% and ~8% cells lost [URE3] and [MOT3⁺], respectively. The two previously reported anti-[PSI⁺] compounds, tacrine and aminacrine can eliminate [PSI⁺] with a curing efficiency of about 35%−60% and 45%−60%, respectively, under our examined conditions. Only subtle differences were observed among the two variants of [PSI⁺] upon the treatment of tacrine and aminacrine. Pilocarpine could eliminate [MOT3⁺] and [PSI⁺]^W with curing efficiencies of about 35% and 25%, respectively, but had minimal effects on other prions. Cpd8 could cure [MOT3⁺], [PSI⁺]^W, and [URE3] with an efficiency of ~75%, 10%, and 5%, respectively, but had no detectable effect on [PSI⁺]^S. Cpd10 was effective at curing [MOT3⁺] (~90%) and [URE3] (~45%), but not effective in curing the [PSI⁺] prion. Cpd11 could destabilize [MOT3⁺], with approximately 18% cells losing the prion, but not effective at curing others. Importantly, when red colonies that appeared upon the treatment of the three examined prions were restreaked on YPD, they remained red, indicating that the color change was indicative of prion loss rather than phenotypic inhibition.

No significant destabilizing effect was observed for [PSI⁺] or [URE3] without the supplementation of 0.2 mM GdnHCl in the curing media (data not shown). However, [MOT3⁺] could be destabilized by all 8 compounds in the absence of 0.2 mM GdnHCl, with tacrine being the most effective compound - approximately 70% cells lost [MOT3⁺] upon the treatment.

Interestingly, pilocarpine exhibited different effectiveness in antagonizing different [PSI⁺] variants. The [PSI⁺]^S and [PSI⁺]^W strains shown in Figure 5D and 5E were generated by co-overproduction of Swi1 and Sup35-NMGFP from a [pin⁻] strain, and were relatively tolerant to the 0.5 mM pilocarpine treatment; while another two [PSI⁺] variants-[PSI⁺]^S [PIN⁺] and [PSI⁺]^W [PIN⁺] made by overproduction of Sup35-NMGFP from a [PIN⁺] strain were largely cured at the same concentration (Figure 5F and 5G). To elucidate if these differences are attributable to the conformation of [PSI⁺] variants being created by distinct approaches or to the presence of different prion states of Rnq1, and to determine if pilocarpine treatment can destabilize [PIN⁺], we conducted additional experiments. First, a collection of 23 [PSI⁺] 74-D694 strains with different strengths of [PSI⁺] phenotypes and obtained by several different induction conditions were compared for their sensitivities to pilocarpine curing. In agreement with results shown in Figure 5E, pilocarpine effectively eliminated [MOT3⁺] but not [URE3] in the presence of 0.2 mM GdnHCl (Figure 6A). It seems that the weaker the [PSI⁺] variant was, the easier it was cured. However, there are exceptions. For instance, among the 15 [PSI⁺] variants obtained by Sup35 and Swi1 overproduction in the absence of [PIN⁺], #1, 3, 4, 8, 13, 14, and 22 were more resistant to the curing than the rest. Apparently, variants with similar [PSI⁺] strengths can have different tolerances towards an identical curing condition; for variants induced by Sup35-NMGFP fibril transformation in a [PIN⁺] strain, #11 appeared to be a much weaker variant than #2, but the former was significantly harder to get cured (Figure 6A). These results again indicate that the curing of [PSI⁺] by pilocarpine is variant-specific. Although the redness of colony color on YPD is conventionally thought to be linked to the prion strength of [PSI⁺], our results indicate that [PSI⁺] variants showing similar colony color on YPD can have totally different responses to pilocarpine treatment and thus they may have distinct prion conformations. For the four [PIN⁺] variants shown in Figure 6B, we reproduced the published data showing that a stronger [PIN⁺] variant showed a higher [PSI⁺]-inducing capacity (Bradley et al., 2002) and the Rnq1 protein was aggregated in cells containing all [PIN⁺] variants (Figure 6C). Among them, except for cells containing the very strong [PIN⁺] variant that showed mainly multiple foci, cells containing other variants mostly showed single-dot foci (Figure 6C). Interestingly, pilocarpine did destabilize the tested [PIN⁺] variants except high [PIN⁺] (Figure 6D).

Figure 6. Variant-specific Effects of Pilocarpine on [PSI⁺] and [PIN⁺].

(A) Log-phase cultures of various [PSI⁺] variants and control strains generated by different methods were spotted onto the indicated plates supplemented with the indicated concentrations of pilocarpine nitrate (pilo) and/or GdnHCl (Gd), and images were taken after 3–5 days of growth. (B-D) The effect of pilocarpine on [PIN⁺] variants. (B) After 24 h of overproduction of Sup35-NMGFP (CUP1-promoter) in the indicated strains, cultures were spotted onto SC-ade and YPD plates to assay the [PSI⁺]-inducibility of the tested [PIN⁺] variants. (C) To analyze the prion state of Rnq1 in the indicated strains, Rnq1-GFP was expressed from pCUP1-RNQ1GFP and its aggregation was assessed by a fluorescence microscopy assay. (D) The four indicated [PIN⁺] strains were grown on YPD plates supplemented with different amount of pilocarpine nitrate and/or GdnHCl and then spread onto SC-uracil plates containing 50 μM CuSO₄. Individual colonies were analyzed for the aggregation status of Rnq1 by a fluorescence microscopic assay. L, low [PIN⁺]; M, medium [PIN⁺]; H, high [PIN⁺]; and vH, very high [PIN⁺]. Data in this figure was collected from at least three independent experiments, representative (B and C) or averages of combined results (D) are shown.

Taken together, these tested anti-[SWI⁺] hit compounds have exhibited distinct antagonizing effects on [PSI⁺], [URE3], and [MOT3⁺]. Their effects can be not only prion-specific, but also variant-specific, as shown in the case of the pilocarpine curing on [PSI⁺] and [PIN⁺].

Examining the Ability of Anti-[SWI⁺] Small Molecules to Antagonize Proteotoxicity Linked to Neurodegenerative Diseases

Next, we tested if any of the anti-[SWI⁺] hit compounds could suppress the toxicity caused by FUS or FUS/TLS (fused in sarcoma/translocated in liposarcoma), α-syn, and TDP-43 (TAR DNA-binding protein 43), three human amyloidogenic proteins that are linked to a range of neurodegenerative diseases. This premise was based on our finding that the two anti-[SWI⁺] hits identified from the NCC library were tacrine and amiridine (Figure 7A), which were previously shown to have anti-[PSI⁺] and anti-PrP^Sc activities, and were once used as clinical drugs approved by FDA for AD treatment (Figure 7A and (Levy, 1989; Shevtsov et al., 2014; Tribouillard-Tanvier et al., 2008)).

Figure 7. Suppression of Toxicity Associated with Expression of FUS, α-syn, and TDP-43.

(A) The two anti-[SWI⁺] hits identified from the NCC library, tacrine and amiridine, are FDA-approved clinic drugs previously used for treating AD. (B) Phloretin reduced the toxicity caused by FUS expression. Yeast cells were grown in 2% galactose for overexpression of FUS and its two indicated mutants, and OD 600 nm was recorded for 72 h in SC-ura medium. Shown is a representative result of at least 3 experiments. (C-E) Proteotoxicity associated with expression of FUS (wt), α-syn, and TDP-43 can be reduced by phloretin (C), Cpd9 (D) and Cpd11 (E). Growth were monitored similarly as described for panel B. Representative results are shown for FUS (wt), α-syn, and TDP-43 with a galactose concentration of 0.2%, 0.1%, and 1%, respectively. OD 600 nm was recorded for 48 h for α-syn but 72 h for the other two proteins in experiments shown in panel C-E. (F) Aggregation assays of the three indicated proteins after 24 h of expression. Percentage of cells containing aggregation is quantified upon treatment with the three compounds. Note, toxicity assays were performed for all 12 purchased compounds, but only results of the three effective compounds are shown.

We first examined the effect of the 12 purchased compounds on aggregation of the C-terminal GFP fusions of the wild-type FUS and its two ALS-associated mutants, P525S and R524L (Kapeli et al., 2017). The expression of FUS and mutants was driven by the GAL1 promoter in a 2μ plasmid as described by Fushimi et al (Fushimi et al., 2011). As reported previously, these proteins aggregated and were toxic upon induction with 2% galactose (Figure 7B). Phloretin, which effectively cured [SWI⁺], but not other yeast prions, could suppress the toxicity associated with the expression of all three FUS alleles at a minimal effective concentration of 200 μM (Figure 7B and data not shown). We then extended our assays to include α-syn and TDP-43, which were also tagged by GFP at their C-termini and were controlled by the GAL1 promoter, and both were previously used to study proteotoxicity in yeast (Auluck et al., 2010; Caraveo et al., 2014). We tested various concentrations of galactose to obtain different expression levels as they would likely result in different levels of toxicity. We found that galactose concentrations of 0.2%, 0.1% and 1% were adequate to induce detectable toxicities in our assays for FUS, α-syn, and TDP-43, respectively (data not shown). We observed that phloretin is also effective in suppressing the toxicity associated with the expression of α-syn and TDP-43 in yeast (Figure 7C). In addition, Cpd9, which can cure [SWI⁺] and destabilize [MOT3⁺], also reduced the toxicity caused by the expression of all three pathogenic proteins (Figure 7D). However, Cpd11, which is able to cure [SWI⁺] with an EC₅₀ of 100 μM and to significantly destabilize [MOT3⁺] at a concentration of 500 μM, only slightly reduced the toxicity of FUS and had no effect on α-syn and TDP-43 (Figure 7E). The other nine compounds did not have any detectable effect on suppressing the proteotoxicity associated with the expression of FUS, α-syn or TDP-43 (data not shown).

We next examined if phloretin, Cpd9, or Cpd11 could affect the aggregation of the three amyloidogenic proteins. At the concentrations that these compounds relieved the indicated proteotoxicity, we found that phloretin and Cpd9 only slightly reduced the aggregation of FUS and α-syn and had no detectable effect on the aggregation of TDP-43 after 24 h induction in the indicated galactose concentration as shown in Figure 7F. Under identical conditions, however, Cpd11 showed a strong inhibition of FUS aggregation, but showed no detectable effect on the aggregation of α-syn or TDP-43 (Figure 7F).

DISCUSSION

Based on the yeast prion [SWI⁺] discovered in our laboratory, we previously established a simple growth assay that can faithfully report the prion status of the Swi1 protein (Du et al., 2015). In this study, we have amended this assay and developed it into a HTS platform for identifying small molecules that antagonize the [SWI⁺] prion. Using this yeast-based HTS platform we have successfully conducted several screens and identified a number of promising anti-prion small molecule compounds, some of which have novel structures and possess high curing efficiency (Figure 3). Importantly, these hit compounds have been validated by our multiple orthogonal secondary assays (Figure S2), indicating that our primary assay is able to identify true hits that eliminate [SWI⁺]. In addition, this assay is simple, robust, and cost-effective. It only requires the use of microtiter plates, media, pipet tips, and a few other inexpensive consumables. As shown in Figure 2A, the complete curing process of our assay can be finished within a 48-hour period. Importantly, to our knowledge, this is the first HTS-assay that allows for positive selection of anti-prion compounds and with high effectiveness.

The effective anti-[SWI⁺] compounds identified in this study can be used as chemical probes to study the [SWI⁺] prion. Before this study, GdnHCl was the only known chemical inhibitor of [SWI⁺] with effective concentration in the mM range. Some of the hit compounds identified in this study, e.g. phloretin has ~μM potency. Furthermore, the fact that 8 out of the 12 purchased anti-[SWI⁺] compounds are also effective in antagonizing one or more of the three non-[SWI⁺] prions that we tested - [PSI⁺], [URE3], and [MOT3⁺] (Figure 5) indicates that our HTS system can target both prion specific pathways and common pathways shared by multiple prions. Interestingly, a few anti-[SWI⁺] hit compounds obtained in our pilot screens - such as quinacrine, guanabenz, tacrine, and amiridine or their derivatives - have been previously shown to antagonize [PSI⁺] and PrP^Sc. Moreover, tacrine (Cognex) and amiridine were previously approved by the FDA as clinical drugs for AD treatment (Bach et al., 2003; Tribouillard-Tanvier et al., 2008). These results suggest that our HTS system is useful for identifying chemical modulators of the protein folding and amyloidgenesis pathways that are shared by prion and aggregation-prone proteins.

In addition to the moderate antagonizing effects of tacrine hydrochloride and aminacrine on [PSI⁺], [URE3], and [SWI⁺], we showed that these compounds can also eliminate [MOT3⁺] with a high effectiveness (Figure 5D and 5E). Indeed, when the four prions were treated with a panel of anti-[SWI⁺] hit compounds identified in this study, [MOT3⁺] is overall the most sensitive prion, whereas [URE3] is the least (Figure 5). Although it remains to be tested if these compounds have distinct effects on different variants of [URE3] and [MOT3⁺], our results demonstrate that these identified anti-prion compounds are prion-specific. Similarly, compound-specific inhibitions of proteotoxicity linked to expression of FUS, α-syn, and TDP-43 are also evident (Figure 7). Anti-prion compounds may act in cis, for example, through targeting amyloid or the expression of a prion gene, or in trans, through targeting cellular pathways involved in prion transmission. As a phenotypic assay, our system may identify either type of antagonizing compounds with a diverse range of effectiveness. The existence of prion variants contributes to an additional layer of complexity in terms of prion-curing. For example, pilocarpine shows distinct curing efficiency against different [PSI⁺] variants used in this study (Figure 6A). Different prions and their variants may exhibit distinct curing results upon a compound treatment as a result of their distinct amyloid structure and their interaction network inside the cell. Notably, none of the examined 12 compounds can efficiently destabilize all four prions tested – [SWI⁺], [PSI⁺], [URE3], and [MOT3⁺] (Figure 5).

To fully understand the curing mechanisms of the hit compounds, follow-up experiments are needed to identify their precise target(s). Since our screens were based on a phenotypic assay, potential targets can be in diverse pathways required for prion propagation – from protein synthesis, folding, amyloidogenesis to degradation. The fact that that some obtained hits possess known anti-prion activity provides some insight into the targeted pathways. Among them is Guanabenz (GA), an agonist of α2-adrenergic receptors, a drug routinely used for treating human hypertensive disorders (Holmes et al., 1983). Previously, it was shown that GA could antagonize the yeast prion [PSI⁺] and [URE3], and an ovine PrP^Sc -infected MovS6 cells (Tribouillard-Tanvier et al., 2008). However, the α2-adrenergic receptor activity and the anti-prion activity of GA seem to not be linked as derivatives of GA that no longer have the α2-adrenergic receptor activity remained effective in clearing PrP^Sc accumulation in MovS6 (Tribouillard-Tanvier et al., 2008). GA does not directly act on the conversion process of PrP^C to PrP^Sc or bind to protein aggregates/amyloids. Instead, it is believed that GA inhibits prion propagation by binding to ribosomes to modulate the protein folding activity on the ribosome during translation (Blondel et al., 2016). We suspect that a similar mechanism is also responsible for the [SWI⁺]-curing effect by GA. Another potent anti-prion compound from our screen is tacrine, a clinical drug that was used to treat AD for more than a decade prior to its discontinuation. Besides antagonizing Aβ, tacrine was also shown to have an anti-prion effect against both yeast prions ([PSI⁺] and [URE3]), and mammalian PrP^Sc accumulation (Tribouillard-Tanvier et al., 2008). It is thought that the anti-prion effect of tacrine is due to its acetylcholinesterase (AChE) inhibitor activity as AChE can bind to misfolded Aβ and prion proteins through hydrophobic interactions to form stable complexes to promote their aggregation (Galdeano et al., 2012; Inestrosa et al., 1996; Pera et al., 2006). Interestingly, five other compounds - amiridine (hit1), aminacrine (hit9), quinacrine dihydrochloride (hit8), N-(7-methoxy-1,2,3,4-tetrahydroacridin-9-yl) benzamide (hit28) and hydroxytacrine maleate (hit11) share similar structures with tacrine. They also exhibit potent anti-[SWI⁺] activity as tacrine does. Except for hit28, the others have been reported for their activities against prions, AD or PD (see references in Table S1). The prevalence of the tacrine-like hits in this and other studies suggest the “three-ring” structure targeting acetylcholinesterase has broad applicability in antagonizing prions and prion-related disorders. Unfortunately, tacrine is no longer in use for AD treatment due to its vast side effects. As a phenotypic screening system, our HTS system has a potential to capture compounds with diverse structures and with various action mechanisms. As shown in Table S1 and Figure 3, a large portion of the obtained hits have structures different from the known anti-AD or anti-PrP^Sc compounds. Significantly, some of them, for instance, hits 10, 13, 25, 34 and 36 share similar structural scaffolds, and the structures of Cpd9 (hit40) and Cpd11 (hit42) are also similar (Figure 3 and Table S1). Further analysis and chemical modifications of these hits may lead to derivatives with novel targets and action mechanisms. New therapeutic strategies against prions and other prion-like pathogenic proteins may be developed.

Given that all yeast prions included in this study require the function of Hsp104 for propagation (Crow and Li, 2011; Ferreira et al., 2001; Jung and Masison, 2001) and that none of the hit compounds we examined in this study can cure all of the tested prions, we conclude that Hsp104 may not be a major target of our studied hits. Considering that the curing efficiency of the examined hits strictly relies on the addition of 0.2 mM GdnHCl, which by itself does not cure any of the tested prions, a slightly compromised Hsp104 function may be a prerequisite to curing. It is possible that these hits do not directly target Hsp104 but affect pathways that are functionally connected to Hsp104. In addition, since the ERG6 gene deletion can only slightly increase the permeability of the yeast cell wall (Gaber et al., 1989), we anticipate that some compounds with a limited curing capacity in yeast may be more effective against prion or prion-like proteins in mammalian cells. Taken together, our study suggests that this newly established yeast-based HTS system provides a useful tool for identifying compounds with diverse inhibitory mechanisms against prion transmission and other prion-like proteopathies.

SIGNIFICANCE

A cure for fatal prion diseases and prion-like protein-folding disorders such as AD, PD, and ALS is an urgent medical need. Identification of small molecules that antagonize prion-like pathogenic proteins through HTS is an effective approach that can lead to discovery of novel small molecules as chemical probes for studying prion biology and development of new therapeutic drugs for treating these devastating diseases. However, few effective HTS platforms are currently available for this purpose. Taking advantage of the [SWI⁺] prion in the budding yeast Saccharomyces cerevisiae, we have established a robust HTS assay that can efficiently identify anti-prion small molecules. Unlike previously reported assays, our system allows for positive selection of non-prion cells, and thus, cells that have lost the prion upon compound treatment can be easily identified. Our growth-based method is effective, simple, and robust. Using this primary screening assay, in conjunction with several orthogonal secondary assays, we have successfully carried out pilot screens accessing several diverse chemical libraries. We have identified a number of hits, including several known anti-PrP^Sc and anti-[PSI⁺] compounds as well as FDA-approved clinical drugs for AD treatment. These results confirmed the efficacy of this anti-prion HTS system. We have also identified anti-prion hits with unique structures and they exhibit antagonizing effects in prion- and/or variant- specific manners. In addition, some identified hits can also partially suppress the toxicity caused by expression of mammalian pathogenic proteins linked to neurodegenerative diseases. Taken together, these results demonstrate the utility of this anti-prion HTS system. We anticipate that this straightforward phenotypic screening platform will be useful for identifying chemical modulators that interact with various components of the cellular machinery governing protein folding, aggregation, amyloidogenesis, and protein-based transmission.

STAR METHODS

LEAD CONTACT AND MATERIALS AVAILABILITY

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Liming Li (limingli@northwestern.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Yeast is the only model organism used in this study and all strains are MATa type. A BY4741 strain (met15 leu2 FLO8-HIS3 flo1::FLO1pr-URA3 [SWI⁺],DY755) was previously constructed (Du et al., 2017) and used in primary screens after ERG6 deletion (see below). All [PSI⁺] strains used in this study are derivatives of 74-D694 (ade1–14 his3 leu2 trp1 ura3) previously created by using different [PSI⁺]-inducing approaches (Du and Li, 2014). [PIN⁺] variants of the 74-D694 background used in this study are a courtesy of the Liebman laboratory (Sharma and Liebman, 2013). A BY241 [URE3] strain (leu2 trp1 ura3 P_DAL5:ADE2 P_DAL5:CAN1 kar1) described previously (Brachmann et al., 2005) was obtained from the Wickner laboratory. To make a [MOT3⁺] strain carrying the DAN1pr-ADE1 reporter, the chromosomal ADE1 promoter of the BY4741 leu2 his3 FLO8-MET15 flo1::FLO1pr-URA3 strain was replaced by DAN1 promoter (unpublished data), and the resulting strain was used as the recipient for [MOT3⁺] transformation. The [MOT3⁺] prion was then transferred from a [MOT3⁺] strain (RHY1277, MATa/α; ura3Δ3/ura3–52; dan1::URA3-SpHIS5/DAN1), a gift from the Halfmann laboratory in Stowers Institute for Medical Research) by extract transformation using a previously described method (Du et al., 2010). The BY4741 strain used for toxicity study of FUS was purchased from ATCC. FUS expression plasmids were published previously (Fushimi et al., 2011). ERG6 replacement strains were created by transformation of a PCR amplified fragment containing a LEU2 gene flanked by the upstream and downstream sequence (30 bps each) of ERG6L-F (CATAATTTAAAAAAACAAGAATAAAATAATAATATAGTAGGCAGCATAAG-ATGTCTGCCCCTAAGAAG) and ERG6L-R (AAATAGGTATATATCGTGCGCTTTATTTGAATCTTATTGATCTAGTGAAT-TTAAGCAAGGATTTTCTT). Upon transformation, Leu⁺ isolates were selected and the ERG6 deletion was verified by PCR with primers of ERG6-F (GGTACCTCGTTCCCGTAC) and Leu2-R (CACCTGTAGCATCGATAG). The BY4741 FLO8-HIS3 FLO1pr-URA3 erg6Δ [SWI⁺] strain was then treated with 5 mM GdnHCl, resulting in the corresponding [swi⁻] strain. BY4741 FLO8-MET15 FLO1pr-URA3 [SWI⁺] (LY770) and [swi⁻] (LY773) strains were described previously (Du et al., 2015). Strains carrying GAL1-controlled ORFs of α-syn and TDP-43 were gifts of Dr. Caraveo (Northwestern University) and have been previously described (Auluck et al., 2010; Caraveo et al., 2014). All yeast strains used in this study were cultured in rich media (YPD) or synthetic complete (SC) media with defined amino acid dropout for selection. For raffinose phenotype assay, glucose in media was replaced with raffinose and supplemented with 0.5 μg/ml antimycin (Sigma-Aldrich, St. Louis, MO). The curing medium used in primary screening was SC-his supplemented with 0.2 mM GdnHCl, and the detection medium was SC-his-ura. For galactose-induced expression, the basic carbon source supporting the growth was raffinose (for FUS, α-syn and TDP-43), or sucrose (for [SWI⁺] strains). All yeast cells were grown at 30°C.

METHOD DETAILS

High-throughput Screening

Chemical libraries

Five chemical libraries were used for this study. 1) The ASDI library contains 6800 compounds of diverse chemical structures that meet the Lipinski guidelines for drug-likeness. 2) The NIH Clinical Collection (NCC) contains 480 molecules that were used in human clinical trials. 3) The collection from NCI/DTP Open Chemical Repository contains ~2,000 structurally diverse compounds, ~880 mechanistically diverse anti-cancer compounds, and 120 natural products. 4) The Spectrum collection is a library of ~2000 drugs that (i) have been introduced in the US, Europe and Japan and have known pharmacological profiles, (ii) are natural products with unknown biological properties, and (iii) are other bioactive compounds such as non-drug enzyme inhibitors, receptor blockers, membrane active compounds, and cellular toxins. 5) The ChemBridge DIVERSet-CL is a collection of 50000 small molecules with enhanced potential for therapeutic development. In this library, >90% of compounds pass a rigorous set of drug-like filters including Lipinski and Veber rules, and Pipeline Pilot SMARTS liability filtering. Approximately 4800 molecules were screened from the ChemBridge DIVERSet-CL library, and all compounds were screened for other four libraries in this study. All compounds used for primary screens study were solubilized in DMSO and stored at the Northwestern University High Throughput Analysis Lab (Evanston, IL). Compounds for subsequent experiments were solubilized in DMSO, water, or ethanol. Compounds used in experiments performed in the Li lab were purchased from Sigma-Aldrich, St. Louis, MO (quinacrine hydrochloride), Santa Cruz Biotechnology, Dallas, Texas (phloretin, nimustine hydrochloride, pilocarpine nitrate, tacrine hydrochloride and aminacrine), and Chembridge Corporation, San Diego, CA (Chembridge compound ID numbers: 91109249 (Cpd7), 74610971 (Cpd8), 71326175 (Cpd9), 66861843 (Cpd10), 47774380 (Cpd11), 63161364 (Cpd12)).

Primary screening procedure

The primary screening was conducted at Northwestern University’s High-throughput Analysis Lab with a protocol shown in Figure 2A. In brief, 0.5 mM (for NCC, NCI/DTP and ASDI) or 0.1 mM (for Spectrum) of small molecules were added to a 384-well microplate by using a Labcyte Echo 550 Acoustic Liquid Transfer System (5 mM GdnHCl as positive control, and 1% DMSO as negative control), with duplicates for the NCC and Spectrum libraries and singlet for the NCI/DTP and ASDI libraries. Then, 50 μL of the yeast cell (FLO8-HIS3 FLO1pr-URA3 erg6Δ [SWI⁺]) suspension in curing medium (SC-histidine, with 0.2 mM GdnHCl) was added using a Titertek Multidrop 384 system to a final cell density of 2.5×10⁴ cells/mL. Plates were sealed with a breathable membrane and incubated for 24 h at 30°C with shaking in Genemachines HiGro incubator shakers. After curing growth, cultures were pinned into the detection medium using Genetix QPix II XT colony picking system and grown in Genemachines HiGro incubator shakers. Finally, cell growth (optical density, OD 600 nm) was measured using a Biotek Synergy 4 Microplate Reader after 24h incubation. The plates were then taken out from the incubator and placed on bench for additional 24h in room temperature and measured again. Five mM GdnHCl and 1 % DMSO were used as positive and negative control, respectively. A valid screening assay has a Z’ score > 0.5 and a P value < 1.0E-23. Compounds with a curing activity (growth on detection medium) greater than 25% of that of 5 mM GdnHCl or 5-fold higher than 1% DMSO were selected as hit candidates for further analysis.

Quantitative Assay of Anti-[swi⁺] Hits

Cherry-picking and dosage effects of obtained hits were performed in the Northwestern University High-throughput Analysis Lab with protocols similar to those used in the primary screening. Based on the reads of OD 600 nm of cell cultures from the detection medium, anti-[SWI⁺] activity for each hit was calculated as percent activity relevant to that of 5 mM GdnHCl. Dose-response curves were plotted, and EC50^G values were estimated using GraphPad Prism. To further evaluate the [SWI⁺]-curing effectiveness, eight selected hits were manually assessed in liquid media and on plates (with GdnHCl and DMSO as controls). Assays in liquid media were done with a method similar to that used in the primary screen except that a 96-well plate was used, and a growth curve was created with a plate reader measuring OD 600 nm continuously for 24 or 48 hours. For the plate-based quantification, cultures were properly diluted and spread onto SC-his and SC-his-ura plates after the curing, and percent of Ura⁺ colonies were calculated afterwards.

Secondary Assays to Confirm [SWI⁺] Loss

The raffinose phenotype and aggregation state of NQ-YFP were used as secondary assays to confirm the [SWI⁺] status upon treatment with compounds as [SWI⁺] cells grow poorly in non-glucose sugars and Swi1 protein forms aggregates only in [SWI⁺] cells (Du et al., 2008). Individual Ura⁺ and Ura⁻ colonies were streaked onto raffinose plates and assessed as described previously (Du et al., 2008). The Swi1 NQ-YFP aggregation assay (driven by the TEF1 or GAL1 promoter) has been reported previously (Du et al., 2010; Du et al., 2017). For cells expressing Swi1 NQ-YFP under the TEF1 promoter, aggregation was assayed in transformants on the plates after transforming the Ura⁺ and Ura⁻ isolates with p413TEF-NQYFP (Du et al., 2017). For cells expressing Swi1 NQ-YFP under the GAL1 promoter, strains were first transformed with p413GAL1-NQYFP (Du et al., 2017), treated with compounds, and the resulting Ura⁺ and Ura⁻ isolates were examined by fluorescence microscopy after induction with 2% galactose for 4 hours.

Anti-[SWI⁺] Hits against Other Yeast Prions

In these analyses, 5 mM GdnHCl and DMSO (or water, ethanol) served as positive and negative controls, respectively. As all reporter strains carried an ADE reporter, cells harboring prions would give rise to pink / white colonies while non-prion cells would generate red colonies on YPD (Derkatch et al., 1996). The prion curing was either performed in YPD liquid medium (for experiments in Figure 5) or on YPD plates (for experiments in Figure 6). In liquid YPD, the curing time was 24 hours for [PSI⁺]^S [PIN⁺], [PSI⁺]^W [PIN⁺], but 48 hours for other strains used in the experiments of Figure 5 and Figure 6. After curing, cultures were properly diluted and spread onto YPD, and their colony colors were assessed and counted. The ratio of red colonies was quantified after 3–5 days of incubation. Sectored colonies were counted for [PSI⁺]^S [PIN⁺], and [PSI⁺]^W [PIN⁺] strains (shown in Figure 5G), but were counted as non-prion for other strains shown in Figure 5D–5E. To test the effects of pilocarpine nitrate on the 23 [PSI⁺] variants and the 3 control strains shown in Figure 6A, strains were first grown in liquid YPD to log-phase, and cultures (10⁶ cells / mL) were then spotted onto YPD plates after serial dilution. For [PIN⁺] strains, cells were spread onto SC-ura + 50 μM CuSO₄, and assayed for Rnq1 status by fluorescence microscopy after 3 days of growth. The [PSI⁺] inducibility of the different [PIN⁺] variants was tested by overproduction of Sup35 NM-GFP and their Rnq1 aggregation state was examined by expressing Rnq1-GFP from plasmid pCUP1RNQ1-GFP as described previously (Du and Li, 2014).

Anti-[SWI⁺] Hits against Pathogenic Proteins

For proteotoxicity assay, SC selective media (raffinose based) were inoculated by yeast strains carrying GAL1-driven ORFs of FUS (wild-type, P525, and R524), α-syn and TDP-434, or an empty vector, respectively, in a 96-well microplate. Log-phase cultures were supplemented with different amounts of galactose (from 0.025%−2%) and monitored for growth inhibition using a plate reader for 48 or 72 hours after addition of galactose. The optimized galactose concentration for each protein was determined in pre-test experiments. Growth assays were done in the presence and absence (DMSO, water, ethanol) of a testing compound. Cultures without galactose served as a no-toxicity control. The aggregation of these proteins was separately examined in the same growth conditions used for the toxicity assay, which was observed at 2, 24, and 48 hours after the galactose addition.

QUANTIFICATION AND STATISTICAL ANALYSIS

Z’ score and p-values (from T-test) are mean numbers from 96–544 samples and are shown in Figure S1 and Figure 2B and 2C. Both types of scores were calculated for individual plates in high throughput screening assays using 5 mM GdnHCl as positive control and DMSO or water as negative control. The Z’ score was calculated using a formula $\left(Z^{\prime }=1-\frac{3\left({\sigma }_p+{\sigma }_n\right)}{\left|{\mu }_p-{\mu }_n\right|}\right)$ to estimate the reliability of a high throughput screening assay. The T-test was performed with two-tailed distribution and two-sample equal variance to evaluate the significance of difference between the positive and negative controls – the rigor of a high throughput screening assay.

In this study, EC50^G (shown in Table S1 and Figure 3) and EC50 (Figure 4) was used to quantify the effectiveness of a chemical compound eliminating or inhibiting [SWI⁺] and were calculated with data of three repeated tests using GraphPad Prism 8 using nonlinear regression for curve fitting. In calculating EC50, the biological effect is anti-[SWI⁺] effectiveness based on the number of cells that had lost [SWI⁺] upon compound treatment. For EC50^G, the biological effect is a relative anti-[SWI⁺] effectiveness of a compound treatment to that of 5 mM GdnHCl treatment based on growth in detection medium (growth means [SWI⁺] loss). EC50^G was used for results from cherry picking experiments.

DATA AND CODE AVAILABILITY

The published article includes all datasets generated or analyzed during this study, except for part of primary screening data that are available upon request.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, Peptides, and Recombinant Proteins
Chemical library: ASDI	Northwestern university High Throughput Analysis Laboratory (NU-HTA)	ASDI
Chemical library: NIH Clinical Collection (NCC)	NU-HTA	NCC
Chemical library: NCI/DTP Open Chemical Repository	NU-HTA	NCI/DTP
Chemical library: Spectrum collection	NU-HTA	Spectrum
Chemical library: ChemBridge DIVERSet-CL	NU-HTA	ChemBridge DIVERSet-CL
quinacrine hydrochloride (Cpd5)	Sigma-Aldrich	Cat# Q3251 CID# 6239
Phloretin (Cpd4)	Santa Cruz Biotechnology	Cat# sc-3548 CID# 4788
nimustine hydrochloride (Cpd3)	Santa Cruz Biotechnology	Cat# sc-202732A CID# 39214
pilocarpine nitrate (Cpd1)	Santa Cruz Biotechnology	Cat# sc-250722 CID# 657349
tacrine hydrochloride (Cpd2)	Santa Cruz Biotechnology	Cat# sc-200172 CID# 2723754
Aminacrine (Cpd6)	Santa Cruz Biotechnology	Cat# sc-214431 CID #7019
N-(2-isobutoxybenzyl)-N-methylisonicotinamide (Cpd7)	Chembridge Corporation	ID# SC-91109249 CID# 50981921
N∼4∼, N∼4∼,5-trimethyl-N∼2∼-(3-pyridin-4-ylpropyl) pyrimidine-2,4-diamine (Cpd8)	Chembridge Corporation	ID# SC-74610971 CID# 50974888
N-methyl-3-(1-methyl-1H-pyrrol-2-yl)-N-[(3-pyridin-4-ylisoxazol-5-yl) methyl]-1H-pyrazole-5-carboxamide (Cpd9)	Chembridge Corporation	ID# SC-71326175 CID# 75593183
4-[2-(trifluoromethyl) morpholin-4-yl] thieno[3,2-d] pyrimidine (Cpd10)	Chembridge Corporation	ID# SC-66861843 CID# 56874379
3-isopropyl-N,1-dimethyl-N-[(3-pyridin-4-ylisoxazol-5-yl) methyl]-1H-pyrazole-5-carboxamide (Cpd11)	Chembridge Corporation	ID# SC-47774380 CID# 50963693
5-{[[4-(2,6-dimethylpyridin-3-yl) pyrimidin-2-yl] (methyl) amino] methyl}-2-methoxyphenol (Cpd12)	Chembridge Corporation	ID# SC-63161364 CID# 56873491
Experimental Models: Saccharomyces cerevisiae Strains
BY4741	ATCC	ATCC# 4040002
BY4741 FLO8-HIS3 FLO1pr-URA3 [SWI+]	(Du et al., 2015)	DY755
74-D694 derivatives of [PSI+]	(Du et al., 2014)	N/A
74-D694 erg6Δ [PSI+]S	This paper	N/A
74-D694 erg6Δ [PSI+]W	This paper	N/A
74-D694 low [PIN+]	The Liebman laboratory (Sharma and Liebman, 2013)	L1943
74-D694 moderate [PIN+]	The Liebman laboratory (Sharma and Liebman, 2013)	L1945
74-D694 high [PIN+]	The Liebman laboratory (Sharma and Liebman, 2013)	L1749
74-D694 very high [PIN+]	The Liebman laboratory (Sharma and Liebman, 2013)	L1953
BY241 [URE3]	The Wickner laboratory (Schlumpberger et al., 2001)	N/A
BY241 erg6Δ [URE3]	This paper	N/A
S288C/Σ 1278b hybrid strain of [MOT3+]	The Halfmann laboratory (Holmes et al., 2013)	RHY1277
BY4741 FLO8-MET15 FLO1pr-URA3 erg6Δ [MOT3+]	This paper	N/A
BY4741 FLO8-MET15 FLO1pr-URA3 [MOT3+]	This paper	N/A
BY4741 FLO8-HIS3 FLO1pr-URA3 erg6Δ [SWI+]	This paper	N/A
BY4741 FLO8-HIS3 FLO1pr-URA3 erg6Δ [swi−]	This paper	N/A
BY4741 FLO8-MET15 FLO1pr-URA3 [SWI+]	(Du et al., 2015)	LY770
BY4741 FLO8-MET15 FLO1pr-URA3 [swi−]	(Du et al., 2015)	LY773
Strain carrying GAL1-controlled ORF of α-syn	The Caraveo laboratary (Auluck et al., 2010)	N/A
Strain carrying GAL1-controlled ORF of TDP-43	The Caraveo laboratory (Caraveo et al., 2014)	N/A
Oligonucleotides
Primer: ERG6L-F: CATAATTTAAAAAAACAAGAATAAAATAATAATATAGTAGGCAGCATAAG-ATGTCTGCCCCTAAGAAG	This paper	N/A
Primer: ERG6L-R: AAATAGGTATATATCGTGCGCTTTATTTGAATCTTATTGATCTAGTGAAT-TTAAGCAAGGATTTTCTT	This paper	N/A
Primer: ERG6-F: GGTACCTCGTTCCCGTAC	This paper	N/A
Primer: Leu2-R: CACCTGTAGCATCGATAG)	This paper	N/A
Recombinant DNA
Plasmid: p42GAL1-wtFUSGFP	The Wu laboratory (Fushimi et al., 2011)	FUS
Plasmid: p42GAL1-P525LGFP	The Wu laboratory (Fushimi et al., 2011)	FUS P525L
Plasmid: F p42GAL1-R524SGFP	The Wu laboratory (Fushimi et al., 2011)	FUS R524S
Software and Algorithms
GraphPad Prism 8.0	GraphPad Software	https://www.graphpad.com/scientific-software/prism/

Highlights.

• Established a yeast growth-based HTS system for identifying anti-prion compounds

• Identified hit compounds possess prion and variant-specific curability

• Some hits are known for their anti-AD and/or anti-PrP^Sc activities

• Some hits reduce proteotoxicity and aggregation of mammalian pathogenic proteins