Compounds from an unbiased chemical screen reverse both ER-to-Golgi trafficking defects and mitochondrial dysfunction in Parkinson’s disease models

α-Synuclein (α-syn) dysfunction has a well-established association with several neurodegenerative diseases, including Parkinson’s disease (PD), dementia with Lewy bodies (DLB) and multiple systems atrophy (MSA). PD is characterized by the degeneration of dopaminergic (DA) neurons in the substantia nigra pars compacta (SNpc) and other brainstem and cortical nuclei (Braak et al., 2003); DLB by widespread cortical neuronal loss; and MSA by neuronal loss in the brainstem, cerebral cortex and cerebellum. Both PD and DLB exhibit the formation of Lewy bodies (LBs) and Lewy neurites (LNs), which are proteinaceous cytoplasmic aggregates composed of α-synuclein (α-syn) (Spillantini et al., 1997). In MSA, on the other hand, α-syn accumulates as morphologically distinct cytoplasmic inclusions in both neurons and glia (Spillantini et al., 1998). Both missense mutations and amplification of α-syn (Polymeropoulos et al., 1997; Kruger et al., 1998; Singleton et al., 2003; Ibanez et al., 2009) are linked to dominantly inherited forms of PD and DLB, and recent genome-wide analyses have linked α-syn to idiopathic PD, DLB and MSA (Farrer et al., 2001; Kay et al., 2008; Scholz et al., 2009). Thus, it is becoming increasingly clear that α-syn dysfunction is central to the pathogenesis of these diseases.

α-Syn is a small 14.5 kDa protein ubiquitously expressed in the brain that is capable of associating with membranes (Jo et al., 2000; Cole et al., 2002). Its function remains poorly understood. When not bound to lipids, α-syn is natively unfolded and is prone to forming toxic oligomers and higher order aggregates (Cole et al., 2002). These simple biophysical properties make α-syn an attractive subject for study in yeast, where the protein homeostasis machinery, vesicle-trafficking pathways and even lipid metabolism pathways are highly conserved (Outeiro and Lindquist, 2003; Cooper et al., 2006; Gitler et al., 2008).

Yeast has the best-characterized eukaryotic genome and is amenable to rapid genetic interrogation. To take advantage of yeast genetics in dissecting the biology of α-syn, we placed α-syn under the control of the galactose-inducible promoter. This allows synchronous induction of the transgene in all cells in the culture upon switching the carbon source from raffinose to galactose media. By varying the copy number and integration loci of α-syn, we found an unusually strong dose dependence between α-syn expression levels and toxicity (Cooper et al., 2006; Gitler et al., 2008); this is highly reminiscent of the extreme dosage sensitivity of α-syn observed with certain dominantly inherited forms of PD (Singleton et al., 2003; Ibanez et al., 2009). In this model, α-syn toxicity causes lipid droplet accumulation (Outeiro and Lindquist, 2003), impairs proteasome-mediated protein degradation (Chen et al., 2005; Sharma et al., 2006) and elicits vesicle trafficking defects (Outeiro and Lindquist, 2003; Soper et al., 2008) with specific ER-to-Golgi blockade (Cooper et al., 2006; Gitler et al., 2008). All of these phenotypes have been observed in neuronal models and implicated in the pathogenesis of PD. Specifically, proteasomal impairment results in increased levels of α-syn in patient-derived fibroblasts (Hoepken et al., 2008) and the formation of fibrillar α-syn inclusions in primary rat neurons (Rideout et al., 2004); lipid droplets can promote the oligomerization of α-syn (Cole et al., 2002) and are associated with LBs in PD brains (Gai et al., 2000); and deletion of α-syn results in depletion of presynaptic vesicle pools (Abeliovich et al., 2000; Murphy et al., 2000; Cabin et al., 2002).

The extreme dosage sensitivity of α-syn allowed us to construct a modestly toxic α-syn strain (IntTox). This enabled us to perform a high-throughput unbiased genetic screen to identify both enhancers and suppressors of α-syn toxicity (Cooper et al., 2006; Yeger-Lotem et al., 2009). This screen identified genes involved in ER-to-Golgi trafficking (Cooper et al., 2006; Gitler et al., 2008), nitrosative stress, oxidative stress and manganese transport (Yeger-Lotem et al., 2009). Human homologues of eight of these yeast genes (including PARK9/ATP13A2, a lysosomal ATPase; PLK2, a Polo-like kinase; and SYVN1, an E3 ubiquitin ligase) were tested in other models of PD and seven were confirmed to modify α-syn toxicity in neurons (Cooper et al., 2006; Gitler et al., 2008; Gitler et al., 2009; Yeger-Lotem et al., 2009). Thus, the yeast work revealed that α-syn is part of a diverse genetic interactome that is conserved from yeast to mammals. Furthermore, we identified YPK9, the yeast homolog of PARK9/ATP13A2, as a suppressor of α-syn toxicity in our genetic screen. Mutations in this gene are associated with autosomal-recessive Parkinsonism with dementia and pallidopyramidal degeneration (Kufor-Rakeb syndrome; OMIM 606693). Aside from this finding, there had previously been no link between this disease and α-syn. In yeast, YPK9 helped to alleviate manganese toxicity and restored α-syn localization to the membrane. The disease-associated mutations eliminated these functions when introduced into the Ypk9 protein. These results establish that the core biological processes affected by α-syn have been conserved throughout eukaryotic evolution and demonstrate the usefulness of yeast in investigating its biology and pathobiology.

Here, we investigate a yeast strain that expresses higher levels of α-syn (HiTox) than the strain used in our earlier genetic screens (IntTox) (Cooper et al., 2006; Gitler et al., 2009; Yeger-Lotem et al., 2009). In various experiments using ELISA assays, the expression level of α-syn in HiTox cells was 1.4- to 2.5-fold higher than the α-syn expression level in the IntTox cells. One striking difference between these two strains is that most of the genetic modifiers that strongly suppress α-syn toxicity in the IntTox strain do not effectively suppress toxicity in the HiTox strain on their own (Gitler et al., 2009; Kritzer et al., 2009). Instead, co-expression of two suppressors from diverse pathways is required to restore viability in the HiTox strain. This suggests that multiple pathways are being affected by α-syn in this strain. We turned to transcriptional profiling to elucidate these complex biological responses to α-syn and performed a high-throughput chemical screen to determine if it was possible to discover individual compounds that could rescue diverse aspects of α-syn toxicity.

We compared the transcriptional profile of HiTox cells with that of cells transformed with the vector alone (vector) and of cells with a non-toxic low-expression level of α-syn (NoTox) (Fig. 1A, supplementary material Tables S1 and S2). At non-toxic levels, α-syn expression elicited only a minimal transcriptional response that was slightly enriched for genes regulating carbon metabolism (Fig. 1B and supplementary material Table S1). The HiTox α-syn strain, however, elicited a marked transcriptional response as early as two hours after induction, before any dying cells can be detected (Fig. 1A). This response markedly intensified four hours following induction (Fig. 1A, supplementary material Table S2), when only a few cells in the culture were starting to die (supplementary material Fig. S1) (Cooper et al., 2006; Gitler et al., 2008).

Our analysis, using the gene ontology (GO) annotation of differentially regulated genes (Hong et al., 2008; Barrell et al., 2009), showed global downregulation of the expression of ribosomal genes [28% (percentage of all genes with transcriptional changes), p<10⁻³⁰] (Fig. 1B, supplementary material Table S2). This was one of the earliest changes detected. However, it is a common signature of just about any stressful condition in yeast (supplementary material Table S3) (Gasch et al., 2000) and is not particularly informative regarding α-syn toxicity.

Other transcriptional changes were more specific to α-syn toxicity. At the earliest time point, just two hours after induction, we found marked decreases in transcripts involved in the function and maintenance of mitochondria (Fig. 1B). After two additional hours of induction, the downregulation of mitochondrial and respiratory transcripts was much more pronounced (60%, p<10⁻⁴⁴) (Fig. 1B). This finding was particularly striking given that only two mitochondrial genes were identified as modifiers of the α-syn IntTox strain (Cooper et al., 2006; Yeger-Lotem et al., 2009). One was the suppressor Hap4, a transcriptional activator of respiratory genes (De Winde and Grivell, 1995). The other, Mks1, enhanced α-syn toxicity. Mks1 is a negative regulator of the retrograde signaling pathway. This pathway senses mitochondrial dysfunction and evokes a nuclear transcriptional response (Liu et al., 2003). Hence, in the IntTox strain, this pathway appears to be acting to alleviate mitochondrial dysfunction caused by α-syn. Presumably, in the HiTox strain, the capacity of this pathway to counteract the toxicity of α-syn is overwhelmed.

In addition to this alteration of mitochondrial transcripts, we found that transcripts of genes with oxidoreductase activities were increased two and four hours following induction (13%, p<10⁻⁹) (Fig. 1B). This set of oxidoreductase genes is known to increase in response to oxidative, osmotic or aldehyde stress (Attfield, 1997). Together with the mitochondrial signature, we hypothesize that mitochondrial dysfunction is associated with oxidative stress in the HiTox strain.

To investigate the implications of these transcriptional alterations, we turned to thin-section ultrastructural electron microscopy (EM) to examine mitochondrial morphology. As we have previously reported, the IntTox strain exhibited the striking accumulation and clustering of vesicles, at both the periphery and interior of the cells (Gitler et al., 2008), and the accumulation of lipid droplets in the cytoplasm (Outeiro and Lindquist, 2003). These changes also occurred in the HiTox strain. In addition, we observed hypertrophy of the ER, with invagination of the ER membranes and mislocalization from the periphery of the cell to the interior (Fig. 2Ac, inset).

Mitochondria in the HiTox α-syn cells were also markedly abnormal. Healthy mitochondria are typically compact, with stacked, hyperintense mitochondrial membranes (Fig. 2Aa,b) (Vander Heiden et al., 1997). Instead of retaining their typical ribbon-like architecture, the mitochondria of HiTox cells were swollen and circular, with isointense electron density with respect to the cytoplasm (Fig. 2Ac). These morphological changes correlate closely with loss of mitochondrial membrane potential and are indicative of dysfunctional mitochondria (Vander Heiden et al., 1997). Mitochondrial changes were only rarely observed in the IntTox strain (data not shown) (Gitler et al., 2008).

Given the relationship between mitochondrial dysfunction and reactive oxygen species (ROS) in PD (Foley and Riederer, 2000; Abou-Sleiman et al., 2006), we assayed for the production of ROS using 5-(and-6)-chloromethyl-2′7′-dichlorodihydrofluorescein diacetate (CM-H₂DCFDA), a cell-permeable compound that fluoresces green upon oxidation by specific ROS (peroxynitrite, OH radicals and H₂O₂). As expected, ROS was detected in fewer than 3% of vector, NoTox and IntTox α-syn cells (Fig. 2B and data not shown). HiTox α-syn cells, however, exhibited a marked amount of ROS, with over 30% of cells showing fluorescence when treated with CM-H₂DCFDA (Fig. 2B). Together with the ultrastructural mitochondrial changes, the generation of ROS suggests that high levels of α-syn elicit global mitochondrial dysfunction.

Recent studies have suggested that α-syn might localize to mitochondria (Cole et al., 2008; Devi et al., 2008; Parihar et al., 2008). In our previous studies using immunogold EM, we did not find localization of α-syn to mitochondria in yeast (Gitler et al., 2008). However, these studies had employed the IntTox strain, which has much less mitochondrial damage. Therefore, we revisited this issue, asking whether, in the HiTox yeast strain, α-syn elicited mitochondrial damage by mislocalizing to the mitochondria. We employed a more quantitative method, differential centrifugation, to isolate distinct cellular compartments. This technique enables the detection of even small quantities of α-syn. The purity of individual fractions was confirmed with compartment-specific markers (protein disulfide isomerase for ER, porin for mitochondria, alkaline phosphatase for vacuole and phosphoglycerate kinase for cytosol; Fig. 2C).

We then probed for α-syn in each compartment and found that less than 0.2% of total α-syn was present in the mitochondrial fraction (Fig. 2C); most of the α-syn was found in the soluble fraction, with a smaller amount separating with the ER and vacuole (Fig. 2C). The small amount of α-syn identified in the mitochondrial fraction probably represents minor contamination from other cellular compartments (note that, for example, a similar level of reactivity with antibodies against alkaline phosphatase, a vacuolar marker, was also observed; Fig. 2C). Importantly, on long exposure, there was no increase in the small amount of α-syn in the mitochondrial fraction of HiTox cells relative to cells expressing low, completely non-toxic levels of α-syn. Thus, mitochondrial dysfunction elicited by α-syn is not due to specific mislocalization to mitochondria through a cryptic signaling sequence. Rather, it is the perturbation of other cellular pathways impinged on by α-syn that leads to mitochondrial dysfunction and the generation of reactive free radicals.

We next used the HiTox α-syn strain in a high-throughput chemical screen to identify single agents that might be capable of rescuing the robust toxicity of this strain. We hoped that identifying such a compound would uncover deeply rooted aspects of α-syn biology linking the disparate pathological phenotypes associated with toxicity. The multiple toxic phenotypes of this strain also created such a high level of stringency that we anticipated recovering only a small number of compounds that could reproducibly suppress α-syn toxicity. To score positive in the screen, the compounds needed to be able to (1) enter cells, (2) avoid clearance by core metabolic pathways, (3) alleviate α-syn toxicity without causing toxicity on their own at the screening concentration (Fleming et al., 2008) and (4) protect against at least two different features of α-syn toxicity within the HiTox cells (Gitler et al., 2009).

A library of 115,000 small compounds was screened at a concentration of 17 μg/ml (~15 μM) for the ability to ameliorate HiTox α-syn toxicity. Our initial compound hits were subsequently tested at lower concentrations (0.1–10 μM) to identify the most potent molecules capable of rescuing the HiTox α-syn cells (data not shown). Although none of the compounds completely restored normal growth, a few restored cell growth to ~50% of the density of wild-type cells after 48 hours of growth in galactose media (data not shown). Four 1,2,3,4-tetrahydroquinolinones, denoted (1), (2), (3), and (4) (Fig. 3A,B), satisfied these stringent criteria: these compounds were effective at low micromolar concentrations; exhibited significant dose-dependent rescue of growth; and were not toxic to wild-type cells (supplementary material Fig. S2). By quantifying the growth rate between 5 and 10 hours, we determined that compound (3) was the most potent at the lowest concentrations, compounds (2) and (4) were equally potent at higher concentrations, and compound (1) exhibited only weak protective activity at all concentrations tested (Fig. 3C). These compounds also demonstrated robust protective activity in a flow cytometric assay for cell death (Fig. 3D), with the most effective compound [compound (3), Fig. 3C] reducing α-syn-induced cell death from ~25% to 10% after six hours of α-syn expression (Fig. 3D).

We next asked if the compounds simply reduced the expression of α-syn from the galactose promoter. By western blotting, there was no measurable reduction in the amount of α-syn (data not shown). However, the strong dosage sensitivity of α-syn raised the possibility that even a slight alteration in expression could potentially alter the toxic phenotype. To further exclude potential transcriptional or promoter effects, we examined the efficacy of the compounds against α-syn toxicity when gene expression was controlled using an estradiol-inducible transcriptional transactivator (Louvion et al., 1993). Under these conditions, the compounds were still protective against α-syn toxicity, thereby excluding the possibility that suppression was simply the result of reduced induction from the galactose promoter (supplementary material Fig. S3).

We also tested the compounds in a yeast model of Huntington’s disease (Duennwald et al., 2006) to determine whether these compounds are general modifiers of cellular toxicity or if they address specific pathological processes unique to α-syn. None of the compounds restored growth in yeast expressing a toxic huntingtin fragment bearing a polyglutamine expansion (supplementary material Fig. S4). Thus, the compounds are not simply generalized modulators of cell stress.

To address whether the compounds altered the oligomerization properties of α-syn, we examined the ability of the compounds to alter the fibrillization of α-syn using a thioflavin-T fluorescence assay. Monomeric wild-type α-syn was incubated at 25°C with slight agitation in the presence of thioflavin-T, which emits at 490 nm only when bound to amyloid fibrils. We found that the compounds had no effect on the rate of the conformational transition of α-syn from monomer to amyloid fibers (supplementary material Fig. S5), suggesting that they do not interact directly with α-syn.

We also identified two commercially available, structurally related 1,2,3,4-tetrahydroquinolinones that differ from the original compounds only by the functional groups appended to the phenyl ring [Fig. 3A; compounds (5) and (6)]. These related compounds did not rescue the growth of HiTox α-syn cells at any concentration tested (Fig. 3A,C,D and supplementary material Fig. S3).

Before pursuing detailed investigations characterizing the compounds, we sought to determine whether they were also effective in neuronal models of PD. First, we tested a Caenorhabditis elegans model expressing wild-type α-syn in DA neurons (Cao et al., 2005; Cooper et al., 2006). In this model, DA neurons remain in their proper neuroanatomical context. Because development is highly stereotyped in nematodes, wild-type animals always have exactly the same number of DA neurons. Directed expression of α-syn by the dopamine transporter promoter (P_dat-1) results in highly reproducible age-dependent loss of DA neurons in ~50% of animals within 172 hours (neurons were scored for both intact cell bodies and preservation of neuritic processes) (Cao et al., 2005).

As free-living soil organisms, nematodes have thick cuticles and strong drug efflux pumps; this necessitated that we first test their ability to tolerate 1% DMSO, the vehicle in which our compounds were administered. Under these conditions, neither the compounds nor 1% DMSO itself exerted any toxic effect on wild-type worms not expressing α-syn. Newly hatched embryos expressing α-syn were able to tolerate 1% DMSO for up to 24 hours and two-day-old adult worms tolerated 1% DMSO for 8 hours with no additional neuronal loss (data not shown). Remarkably, the exposure of newly hatched embryos to ~15 μM of compounds (1), (3), and (4) in 1% DMSO for just 24 hours achieved substantial rescue of α-syn toxicity. After aging worms for an additional 148 hours after removal of the compounds, the percentage of worms with a full complement of neurons had increased to 70% from 50% (Fig. 4A,C). Moreover, two-day-old adult worms that were treated with compound (1) for only 8 hours also retained more DA neurons and dendritic connections (Fig. 4B). Subtler but statistically significant neuroprotective effects were observed with compounds (2), (3) and (4), but not (5) and (6). It is remarkable that brief treatment of adult worms with these compounds was neuroprotective well after the onset of α-syn expression, revealing that it is possible to arrest cell death even after the onset of toxicity.

We next sought to test the efficacy of these compounds in a mammalian model of PD. Stage 17 embryos were harvested from pregnant rats and their midbrains were carefully dissected, gently dissociated and cultured in vitro for seven days. They were then transduced for 72 hours with a lentivirus encoding a mutant form of α-syn that is associated with autosomal-dominant PD (A53T). This well-established PD model (Liu et al., 2008a), although labor intensive, provides a robust and more reproducible measure of toxicity than we have been able to attain with any stable cell lines. This might be because the cultures recapitulate natural interactions between glia and neurons and/or because the cell lines have lost apoptotic mechanisms during immortalization and transformation. An additional benefit of this system is that lentiviral transfection is not restricted to a particular cell type and thus enables the assessment of the preferential toxicity and rescue of DA neurons relative to other neurons in the culture.

Transduction of rat midbrain cultures with lentivirus encoding α-syn-A53T elicited a reduction in the total number of neurons (MAP2-positive neurons) and GABAergic neurons (GABA-positive neurons) within 96 hours (Fig. 4D) (Liu et al., 2008a). Notably, these neurons exhibited marked retraction of their processes (Fig. 4D). Although few tyrosine hydroxylase (TH)-positive neurons are present in individual microscopic fields, quantification of many fields revealed that the reduction in the number of these neurons was even more dramatic (Fig. 4E) (Cooper et al., 2006; Liu et al., 2008a). Remarkably, treatment with (1), (2) and (3), but not with DMSO vehicle, yielded a robust increase in the relative number of TH-positive neurons (Fig. 4E) and substantially prevented the development of aberrant morphologies of the surviving neurons (Fig. 4D). Treated neurons, including TH-positive neurons, retained their complex, branched dendritic arborization (Fig. 4D). Intriguingly, although compound (4) was not effective, compound (6), which had been inactive in both yeast and C. elegans, exhibited a protective effect in this neuronal assay (Fig. 4E).

It is important to note that compounds (3), (4) and (6) are structurally very similar and differ only in the functional groups appended to the phenyl group (Fig. 3A). This structural similarity suggested that the compounds might interact with the same target. Indeed, when we performed competition assays in HiTox yeast cells, by mixing the most effective compound (3) with increasing concentrations of yeast-inactive compound (5) or (6), the latter strongly interfered with rescue by compound (3) (Fig. S2 and data not shown). Furthermore, transcriptional analysis of the compounds in yeast revealed that compounds (5) and (6) did have some, albeit limited, effect on reversing the transcriptional profile of HiTox α-syn (Fig. 1C, see below). Compounds (5) and (6) must therefore bind to the same target as compound (3), but simply do not affect its function appreciably. Thus, the bioactive compounds we identified in the yeast screen act on a conserved target(s) that exhibits only minor differences in protein structure over astonishing evolutionary distances.

Given the diversity of the pathological phenotypes present in the HiTox α-syn cells, we asked which were being reversed by these compounds. Transcriptional profiling of wild-type yeast cells revealed that the compounds elicited a negligible transcriptional response on their own, consistent with our finding that the compounds had no effect on the growth or viability of wild-type cells (supplementary material Table S4). In HiTox α-syn cells, the active compounds had a profound effect on the transcriptional changes due to α-syn toxicity. Compounds (1) and (2) markedly reversed many of the transcriptional changes induced by high levels of α-syn (Fig. 1C). Compound (4) showed a more modest but similar effect on the transcriptional changes in HiTox cells (Fig. 1C) [correlation between profiles of cells treated with (1) and (4), r=0.74; differential findings among several replicates limited the statistical analysis of transcripts from cells treated with (3)]. By contrast, yeast-inactive compounds (5) and (6) yielded only a partial reduction in the transcriptional changes induced by HiTox α-syn (Fig. 1C). Overall, treatment with (5) or (6) correlated most closely with untreated HiTox cells (r>0.86).

Notably, the reversal of α-syn-associated transcriptional changes was incomplete [r=0.59 for untreated versus (1); r=0.54 for untreated versus (2)]. Specifically, in HiTox cells treated with (1) or (2), the differentially regulated genes were still enriched for mitochondria-localized proteins (29%, p<10⁻¹¹); however, the magnitude by which the expression of these genes changed was reduced in comparison to untreated cells (Fig. 1B,C). Gene transcripts involved in transition metal ion homeostasis (6%, p<10⁻¹²) that increased in response to α-syn remained upregulated. Overall, the transcriptional profiles of HiTox cells treated with these compounds revealed broad reductions in ER, mitochondrial, membrane and oxidative stress pathways (Fig. 1C). Thus, the compounds appear to act on a central core pathway that is either upstream of these pathways or unites them.

Next, we examined the effects of the compounds on the generation of ROS, ER-to-Golgi vesicle trafficking and mitochondrial morphology. All of the active compounds caused a dramatic reduction in the generation of ROS in HiTox α-syn cells (Fig. 2B). The production of ROS in HiTox cells was completely restored to control levels by treatment with compound (3), but not with compound (5) (Fig. 2B). This finding raised the possibility that the compounds were cytoprotective by acting as scavengers of free radicals. We therefore tested several strong antioxidants for their ability to suppress α-syn toxicity. The antioxidants N-acetylcysteine, riboflavin and melatonin all protected against the toxicity of H₂O₂ in yeast cells (supplementary material Figs S6–S8). By contrast, compound (3) did not protect against H₂O₂-induced toxicity (supplementary material Fig. S6). None of the antioxidants protected against the cellular toxicity of α-syn in yeast (supplementary material Figs S6–S8). Similarly, these antioxidants did not protect against the toxicity of α-syn in the nematode or rat midbrain neuronal culture models (Liu et al., 2008b) (supplementary material Figs S9 and S10). Thus, rescue of toxicity by these compounds is not due to simple antioxidant activity.

ER-to-Golgi vesicle trafficking blockade is a significant component of α-syn toxicity in IntTox cells (Cooper et al., 2006) and is effectively rescued by the coexpression of Rab GTPases (Cooper et al., 2006; Gitler et al., 2008). This ER-to-Golgi block is associated with the mislocalization of α-syn from the plasma membrane to large foci within both IntTox and HiTox cells (Outeiro and Lindquist, 2003; Cooper et al., 2006), but not in NoTox cells. The Rab proteins are largely ineffectual in protecting HiTox cells on their own (Gitler et al., 2009). They do so only when combined with genetic suppressors of different functional classes (Gitler et al., 2009; Kritzer et al., 2009). We found that, unlike the Rab proteins, all four bioactive compounds restored α-syn localization to the plasma membrane on their own in HiTox cells (Fig. 5A).

To specifically and quantitatively assay ER-to-Golgi trafficking, we measured the compartment-specific modifications acquired by carboxypeptidase Y (CPY) as it traverses from the ER to the Golgi to the vacuole. In healthy wild-type cells, less than 5% of CPY is retained in the ER 20 minutes after pulse with radiolabeled CPY (Cooper et al., 2006). In untreated HiTox cells that had been induced for 4.5 hours before pulse labeling, 55% of CPY was retained in the ER after the 20 minute chase. This is similar to the ER-to-Golgi trafficking block we previously reported for the IntTox α-syn strain (Cooper et al., 2006). HiTox cells treated with vehicle or with inactive compounds (5) and (6) exhibited a similar decrease in CPY trafficking (Fig. 5B). Remarkably, treatment with compounds (1), (2), (3) or (4) restored CPY trafficking to wild-type levels, with only 5% retention in the ER (Fig. 5B).

Finally, to confirm the effects of the compounds on these diverse pathways, we examined HiTox cells by EM. HiTox cells treated with compounds (1–4) still contained some swollen and isointense mitochondria; however, the frequency of these mitochondrial abnormalities was reduced (Fig. 5C). In addition, the active compounds largely restored ER localization to the periphery of the cells and reduced the size and number of vesicle clusters within HiTox cells (Fig. 5Ca–d). By contrast, none of these abnormalities was reversed upon treatment with either (5) or (6) (Fig. 5Ce,f). The fact that these compounds restored the trafficking of CPY from the ER to the Golgi, in addition to reducing ROS production and restoring mitochondrial morphology, demonstrates that it is possible to ameliorate several diverse cellular pathologies associated with PD with a single chemical agent.

Because the compounds reversed several different pathological phenotypes mediated by α-syn in yeast, this suggested that they might also protect mammalian cells from other sources of PD-related damage. Rotenone, a potent mitochondrial complex I inhibitor, is commonly used to elicit mitochondrial dysfunction, oxidative stress and DA degeneration in mammalian systems. In cultured cells, it is considered a good model of the PD-related damage associated with mitochondrial toxins (Ayala et al., 2007).

As expected, exposure of primary rat midbrain cultures to 100 nM rotenone caused a robust decrease in the overall number of GABA-positive, MAP2-positive and TH-positive neurons. Rotenone also elicited marked morphological changes, including retracted neuronal processes and shriveled neuronal somata (Fig. 6). As observed with α-syn-A53T transfection, TH-positive neurons were lost to a greater degree than other neurons (Fig. 6). Treatment with compounds (1) or (4) greatly abrogated the toxicity of rotenone in a dose-dependent manner and restored normal neuronal architecture (Fig. 6 and supplementary material Fig. S11). Compounds (3) and (5) also showed a trend towards neuroprotection, though results with (3) did not reach statistical significance (supplementary material Fig. S11).

As previously reported, overexpression of Rab1 suppresses α-syn toxicity in yeast (the IntTox strain) by restoring ER-to-Golgi transport. It also rescues α-syn toxicity in C. elegans, Drosophila melanogaster and rat neuronal cultures (Cooper et al., 2006). However, Rab1 did not reverse the toxicity of rotenone (supplementary material Fig. S11). Thus, it appears that mitochondrial dysfunction is not a direct consequence of ER-to-Golgi trafficking defects alone, but rather of some other closely related pathway affected by α-syn. The compounds must be acting on a target proximal to both pathways to be able to protect against high levels of α-syn toxicity in yeast and both α-syn and rotenone toxicity in mammalian cells.

Our findings reveal deep links among distinct pathogenic aspects of α-syn toxicity. Our previous work demonstrated that ER-to-Golgi trafficking defects are a prominent feature of intermediate levels of α-syn toxicity in yeast (IntTox strain) and in higher organisms (Cooper et al., 2006). The IntTox strain also can be suppressed by single genetic modifiers (Yeger-Lotem et al., 2009). By contrast, the HiTox α-syn strain, which only expresses ~50% more α-syn, requires two genetic suppressors acting in separate pathways to rescue growth (Gitler et al., 2009; Kritzer et al., 2009). Transcriptional profiling revealed that, in this strain, in addition to ER-to-Golgi trafficking defects, mitochondrial dysfunction is an early pathological phenotype. Mitochondrial dysfunction, which is associated with the generation of ROS and marked mitochondrial morphological changes, was not a feature of the previously described IntTox strain. Thus, the HiTox strain presented an opportunity to search for other core cellular processes affected byα-syn toxicity.

The identification of mitochondrial dysfunction and ROS as a consequence of higher levels of α-syn toxicity in yeast cells is extremely compelling, given the wealth of evidence that links mitochondrial dysfunction and oxidative stress to PD [for a review, see (Mandemakers et al., 2007)]. Humans accidentally exposed to the mitochondrial toxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) develop Parkinsonism resulting from the immediate loss of DA neurons in the substantia nigra (Davis et al., 1979; Langston et al., 1983). Similarly, rotenone induces DA degeneration with the accumulation of α-syn into LB-like aggregates when administered chronically to rats (Betarbet et al., 2000; Sherer et al., 2003; Norris et al., 2007). In cell culture and in vivo, α-syn overexpression sensitizes both cells and mice to MPTP (Song et al., 2004; Smith et al., 2005). Moreover, aged transgenic mice overexpressing mutant α-syn exhibit profound mitochondrial defects within their neurons (Martin et al., 2006; Stichel et al., 2007), and mice lacking α-syn have increased resistance to mitochondrial toxins (Dauer et al., 2002; Drolet et al., 2004; Klivenyi et al., 2006) and alterations in mitochondrial lipid content (Ellis et al., 2005). Overall, despite a decade of intense investigation, a direct connection between α-syn toxicity and mitochondrial dysfunction has remained elusive.

Recent reports have suggested the existence of a cryptic mitochondrial targeting sequence within the amino terminus of α-syn (Cole et al., 2008; Devi et al., 2008; Parihar et al., 2008) that might be responsible for toxicity in this organelle. Although we cannot exclude the possibility that α-syn might interact with mitochondria in mammalian cells, we found that only a very tiny fraction of α-syn localizes to mitochondria in the HiTox yeast cells, which nevertheless exhibit profound mitochondrial defects (Fig. 2C). Furthermore, the same quantity of α-syn was detected in mitochondria in the NoTox strain, indicating that the localization of small quantities of α-syn to mitochondria is not responsible for toxicity. However, the mitochondrial toxicity we detect in yeast must be closely related to that caused by α-syn and rotenone in neurons, because compounds selected to rescue yeast from α-syn also rescue neurons from both α-syn and rotenone. Rather than affecting mitochondria through direct localization to that organelle, our data indicate that α-syn interferes with deeply rooted biological processes that impinge on pathways linking vesicle trafficking, the ER and mitochondria.

How might ER-to-Golgi trafficking defects and mitochondrial dysfunction be linked? Our earlier work has shown that the block in ER-to-Golgi trafficking is associated with impaired ER-derived secretory vesicles failing to fuse with the Golgi (Gitler et al., 2008). Increased α-syn expression could exacerbate ER homeostasis defects, which in turn could impact mitochondria in at least three different ways. The first derives from recent work demonstrating that mitochondria are focally tethered to the ER in both yeast (Kornmann et al., 2009) and mammalian cells (de Brito and Scorrano, 2008). This attachment allows mitochondria to sense and take up Ca²⁺ released by the ER into the cytosol (de Brito and Scorrano, 2008). In addition to serving as a buffer for excess Ca²⁺, Ca²⁺ sensing also allows mitochondria to localize to parts of the cell (i.e. the synapse) where the local energy requirements are high. Thus, decoupling connections to the ER would impair the ability of mitochondria to sense and respond to cellular energy demands.

Second, a blockade in ER trafficking might disrupt the autophagic destruction of abnormal mitochondria, so-called “mitophagy” (Van Laar and Berman, 2009). The accumulation of dysfunctional mitochondria would then exacerbate cellular toxicity through the generation of ROS and impaired metabolism. Indeed, in both mammalian neurons and yeast, ROS are a salient feature of α-syn toxicity. The resultant ROS might also potentiate neurotoxicity by promoting the oligomerization of α-syn (Norris et al., 2003), thus creating a vicious toxic circle.

Finally, the ER and mitochondria also act together to coordinate lipid synthesis (Daum and Vance, 1997; Kornmann et al., 2009). Specifically, after phosphatidylserine is synthesized in the ER, it is transferred to mitochondria for decarboxylation by Psd1, after which it is shuttled back to the ER for subsequent modification to form phosphatidylcholine (Kornmann et al., 2009). α-Syn may impede the flow of biosynthetic components between these organelles, just as it impedes the flow of vesicles from ER to Golgi. Strikingly, abnormalities in lipid metabolism have also been noted in our yeast model (Outeiro and Lindquist, 2003), as well as in human PD brains (Gai et al., 2000). The application of high-throughput genomic, proteomic and lipidomic assays to our yeast model offers an opportunity for elucidating the vitally important connections between α-syn, the ER and mitochondria.

Our results also suggest reasons why neurons might be particularly sensitive to α-syn dysfunction. Compared to most cell types, neurons have unusually strong demands for vesicle trafficking. Catecholaminergic neurons are particularly susceptible because catecholamines are synthesized in the cytoplasm, where they can exert toxicity unless they are immediately sequestered into secretory vesicles (Mosharov et al., 2009). Deficits in the production of secretory vesicles would therefore be particularly toxic to certain neuronal populations. Neurons also have intrinsically higher levels of respiration than most other cell types (Kandel, 2000) and are therefore vulnerable to the disturbances in mitochondrial activity that are caused by α-syn. Finally, the transcriptional response of yeast cells to α-syn also revealed disruption of metal ion homeostasis (Fig. 1B). This corroborates our previous finding that three metal ion transporters (for manganese, Ca²⁺ and iron) modify α-syn toxicity (Gitler et al., 2009). The brain is also particularly sensitive to perturbations in the balance of these ions. Thus, the defects caused by α-syn would logically create a special combination of vulnerabilities in certain neurons.

Our unbiased chemical screen of 115,000 compounds against the HiTox α-syn strain identified compounds capable of protecting against the multiple pathological processes elicited by α-syn toxicity. The ability of these compounds to rescue disparate toxicities in this strain and to rescue both α-syn and rotenone toxicity in mammalian cells is remarkable. The efficacy of the compounds in neuronal models reveals not only that the cellular processes affected by α-syn have been conserved through a billion years of evolution, but also so have the specific protein targets upon which the compounds act, with only modest differences in structure-activity relationships (SAR). These studies add weight to the hypothesis that PD is a disorder of basic cellular dysfunction that certain neurons are particularly sensitive to. Our unbiased approach successfully identified individual small molecules capable of tackling multiple aspects of disease pathophysiology. This highlights the power of using a simple model organism to elucidate complex biological processes and suggest new therapeutic strategies to treat human diseases.

The yeast strains used were in the w303 pump-deleted background (MAT α can1–100, his3–11,15, leu2–3,112, trp1-1, ura3-1, ade2-1, pdr1::kanMX, pdr3::kanMX). The vector control strain contained empty vector at the trp and ura loci (pRS304Gal, pRS306GAL). The NoTox α-syn strain contained α-syn fused to green fluorescent protein (α-syn-GFP) inserted at the trp locus and empty vector at the ura locus (pRS304Gal-α-syn-GFP, pRS306GAL). The HiTox α-syn strain contained α-syn-GFP inserted at the trp and ura loci (pRS304GAL-α-syn-GFP, pRS306GAL-α-syn-GFP).

The high-throughput screen was carried out at the Institute of Chemistry and Cell Biology at Harvard Medical School. Over 115,000 compounds from various collections, including commercial libraries, natural products and National Cancer Institute collections, were screened in a 384-well format. HiTox yeast cells were induced in galactose media (starting OD₆₀₀ of 0.001) and dispensed into individual wells (300 nl final) containing the test compound (17μg/ml final in DMSO). Individual plates were incubated at 30°C for two days. Each compound was tested in duplicate. Each plate contained two sets of controls: cells grown in glucose media (to suppress α-syn expression) and induced α-syn cells grown in the absence of any test compound. Individual compounds were assessed for their ability to suppress α-syn toxicity by a read-out in optical density (increased signal at OD₆₀₀) and by fluorescence microscopy (loss of aggregation).

Growth curves were monitored using Bioscreen (www.bioscreen.fi). Yeast strains were pre-grown in 2% raffinose medium and induced in 2% galactose medium (starting OD₆₀₀ of 0.1). 300 μl of induced cells were dispensed to individual wells, in the presence of either the compound or vehicle control (1% DMSO final). Each growth condition was analyzed in triplicate wells per run. Cells were grown at 30°C, with plates shaken every 30 seconds to ensure proper aeration, and OD₆₀₀ measurements were taken every half-hour over a two-day period. The resulting data (OD₆₀₀ versus time) were plotted using Kaleidagraph.

Imaging of live yeast cells to monitor the localization of GFP-tagged α-syn was carried out as previously described (Outeiro and Lindquist, 2003).

Yeast cells were subjected to flow cytometry for the measurement of ROS production and cell viability. Briefly, 500 μl of cells were treated with CM-H₂DCFDA (Invitrogen) and incubated with gentle shaking (750 rpm) at 30°C for 20 minutes using an Eppendorf thermomixer. These cells were briefly sonicated (to prevent clumping) before transferring to a 96-well plate, where they were examined using an EasyCyte flow cytometer (Guava Technologies) and analyzed with Flowjo (Treestar). To assay for viability, 500 μl of cells were briefly sonicated and treated with 10 μl of 1 mg/ml propidium iodide before proceeding with flow cytometry.

Yeast cells were grown overnight in 2% raffinose and induced in 2% galactose medium. Concurrent with galactose induction was the addition of compound (1 μM in 1% DMSO final) or DMSO (1% final). Cells were grown at 30°C for 4.5 and 7 hours, followed by radiolabeling and immunoprecipitation as previously described (Cooper et al., 2006).

NoTox and HiTox α-syn-expressing strains were cultured to log phase growth in 2% raffinose-containing media before induction with 2% galactose for four hours. Cells were harvested and mitochondria isolated according to Meisinger et al. (Meisinger et al., 2006). Protein from total, crude mitochondria and isolated mitochondria was normalized and 10 μg protein resolved by SDS-PAGE. Western blotting with antibodies against alkaline phosphatase, PDI1, PGK1, porin and α-syn-GFP was used to analyze enrichment of mitochondria.

All procedures were carried out at 25°C. Gravid P_dat-1::GFP + P_dat-1:: α-synuclein adults [strain UA44 (baIn11)] were synchronized and embryos were obtained as previously described (Cao et al., 2005). The embryos were left in M9 buffer (22 mM KH₂PO₄, 22 mM Na₂HPO₄, 85 mM NaCl, 1 mM MgSO₄) in glass conical tubes with appropriate concentrations of compound or vehicle control solutions (time 0), and incubated at 25°C for 24 hours to obtain a relatively synchronized L1 population, followed by three M9 buffer washes. The synchronized L1 worms were then spread onto regular nematode growth media (NGM) plates seeded with 100 μl of concentrated OP₅₀Escherichia coli. At 72 hours post-hatching, 120 worms were transferred to a fresh plate containing 0.4 mg/ml 5-fluoro-2′-deoxyuridine (FUDR) to inhibit progeny growth. The worms were analyzed at 172 hours post-hatching. Each experiment was done in triplicate and the average was calculated. The analyses of neurodegeneration were carried out according to the criteria previously described (Cao et al., 2005). For each analysis, 55–120 worms were analyzed. For worms treated at the two-day-old adult stage, embryos were incubated in M9 buffer after the synchronization. Two-day-old adult worms (94 hours post-hatching) were incubated with compound or vehicle control solutions for 8 hours and analyzed at 172 hours post-hatching.

E17 rat primary mesencephalic cultures were prepared as described previously (Cooper et al., 2006). The cells were plated on cover slips at a density of 1300 cells per mm². Four days after plating, the cells were treated with cytosine arabinofuranoside (20 μM) for 48 hours to restrict the growth of glial cells. Cultures were then incubated in fresh media for an additional 24 hours. The cells were transduced with α-syn-A53T lentivirus (MOI=5) as described previously (Cooper et al., 2006) and incubated with the compound (applied at a sublethal dosage; data not shown) or vehicle control (DMSO). After a 72-hour transduction period, the cells were treated with fresh media for an additional 24 hours before immunocytochemical analysis. Compounds (1–6), riboflavin and melatonin were applied concurrently with lentiviral transduction.

E17 rat primary mesencephalic cultures were prepared as described above. In experiments aimed at determining the effects of the compounds on rotenone-induced toxicity, cultures were treated with rotenone (final concentration 100 nM) and the compound or the vehicle control (DMSO) for 24–48 hours before immunocytochemical analysis. In experiments aimed at determining the effects of Rab1 on rotenone toxicity, primary midbrain cultures were transduced with lentivirus encoding Rab1 (MOI=2). After 72 hours, the cells were treated with fresh media containing rotenone (100 nM) for 24 hours. Control cells were incubated in the absence of virus and then treated with fresh media supplemented with vehicle (DMSO).

Immunocytochemistry was carried out as previously described (Cooper et al., 2006), using anti-MAP2 monoclonal IgG (1:500) and an anti-TH polyclonal antibody (1:500), and goat anti-rabbit IgG conjugated to AlexaFluor 488 (1:1000) and goat anti-mouse IgG conjugated to AlexaFluor 594 (1:1000). MAP2- and TH-immunoreactive primary neurons were counted in at least ten randomly chosen observation fields for each treatment using a Nikon TE2000-U inverted fluorescence microscope with a 20× objective. The investigator responsible for counting cells was blind to the identity of the microscope slides. The data were expressed as the ratio of TH-positive neurons to MAP2-positive neurons. Each experiment was repeated at least three times (except for Rab1, where N=2) using embryonic cultures isolated from independent pregnant rats. Statistical analyses were carried out using the program GraphPad Prism, version 4.0 (http://www.graphpad.com/prism/Prism.htm).

Microarray experiments were carried out using yeast expression arrays from the Whitehead Institute Center for Microarray Technology (WICMT). A starter culture of each sample (~5 ml) was grown from a single colony taken from a freshly streaked plate. To ensure synchronous induction, an aliquot of the starter culture was inoculated into fresh media and grown overnight to early stationary phase (OD 1.0). Yeast cells were induced to a starting OD of 0.3 in galactose media in the presence or absence of the compound (1 μM final), and harvested at zero, two, four and six hours post-induction (50 ml aliquot per time point). Three biological triplicates, harvested on different days, were used for each array comparison. Total RNA was obtained using hot phenol-chloroform extraction (Schmitt et al., 1990), checked for purity by gel electrophoresis and BioAnalyzer, and labeled according to standard protocol.

Yeast expression arrays were scanned and array data were extracted using GenePix. Data were analyzed using the limma package from Bioconductor. Spots were background corrected using a normal and exponential convolutional model, and an offset of 50 was added to reduce low-intensity noise. Expression log ratios were normalized by print tip using loess. Differential expression was assayed at each time point using a moderated t-test, correcting for false discovery rate, for three pairwise comparisons between 1× α-syn, HiTox and vector. The microarray analysis data have been deposited at the National Center for Biotechnology Information Gene Expression Omnibus (repository number GSE11633).

Thin-section EM was carried out as previously described (Gitler et al., 2008).

S.L. acknowledges support from the MGH/MIT Morris Udall Center of Excellence in Parkinson’s Disease Research (NS038372), the Michael J. Fox Foundation and a Collaborative Initiative Award from the Howard Hughes Medical Institute. C. elegans studies in the Caldwell laboratory were supported in part by grants from the Michael J. Fox Foundation and Howard Hughes Medical Institute. Primary midbrain culture studies in the Rochet laboratory were supported by grants from the National Institutes of Health (NS049221) and the American Parkinson’s Disease Association. Deposited in PMC for release after 6 months.