Harnessing the power of yeast to unravel the molecular basis of neurodegeneration


  • Sandra Tenreiro,

    1. Instituto de Medicina Molecular, Faculdade de Medicina da Universidade de Lisboa, Lisboa, Portugal
    Search for more papers by this author
    • These authors contributed equally to this study.
  • Matthias C. Munder,

    1. Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany
    Search for more papers by this author
    • These authors contributed equally to this study.
  • Simon Alberti,

    Corresponding author
    1. Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany
    • Address correspondence and reprint requests to Dr. Tiago Fleming Outeiro, Department of NeuroDegeneration and Restorative Research, Center for Nanoscale Microscopy and Molecular Physiology of the Brain, University Medical Center Goettingen, Waldweg 33, 37073 Goettingen, Germany. E-mail: tiago.outeiro@med.uni-goettingen.de or Dr. Simon Alberti, Max-Planck-Institute for Molecular Cell Biology and Genetics, Pfotenhauerstrasse 108, 01309 Dresden, Germany. E-mail: alberti@mpi-cbg.de

    Search for more papers by this author
  • Tiago F. Outeiro

    Corresponding author
    1. Instituto de Medicina Molecular, Faculdade de Medicina da Universidade de Lisboa, Lisboa, Portugal
    2. Instituto de Fisiologia, Faculdade de Medicina da Universidade de Lisboa, Lisboa, Portugal
    3. Department of NeuroDegeneration and Restorative Research, University Medizin Göttingen, Göttingen, Germany
    • Address correspondence and reprint requests to Dr. Tiago Fleming Outeiro, Department of NeuroDegeneration and Restorative Research, Center for Nanoscale Microscopy and Molecular Physiology of the Brain, University Medical Center Goettingen, Waldweg 33, 37073 Goettingen, Germany. E-mail: tiago.outeiro@med.uni-goettingen.de or Dr. Simon Alberti, Max-Planck-Institute for Molecular Cell Biology and Genetics, Pfotenhauerstrasse 108, 01309 Dresden, Germany. E-mail: alberti@mpi-cbg.de

    Search for more papers by this author


Several neurodegenerative diseases, such as Parkinson's disease (PD), Alzheimer's disease (AD), Huntington's disease (HD), amyotrophic lateral sclerosis (ALS), or prion diseases, are known for their intimate association with protein misfolding and aggregation. These disorders are characterized by the loss of specific neuronal populations in the brain and are highly associated with aging, suggesting a decline in proteostasis capacity may contribute to pathogenesis. Nevertheless, the precise molecular mechanisms that lead to the selective demise of neurons remain poorly understood. As a consequence, appropriate therapeutic approaches and effective treatments are largely lacking. The development of cellular and animal models that faithfully reproduce central aspects of neurodegeneration has been crucial for advancing our understanding of these diseases. Approaches involving the sequential use of different model systems, starting with simpler cellular models and ending with validation in more complex animal models, resulted in the discovery of promising therapeutic targets and small molecules with therapeutic potential. Within this framework, the simple and well-characterized eukaryote Saccharomyces cerevisiae, also known as budding yeast, is being increasingly used to study the molecular basis of several neurodegenerative disorders. Yeast provides an unprecedented toolbox for the dissection of complex biological processes and pathways. Here, we summarize how yeast models are adding to our current understanding of several neurodegenerative disorders.

Abbreviations used

Alzheimer's disease


amyotrophic lateral sclerosis


amyloid precursor protein

ß-amyloid peptide


eukaryotic translation initiation factor 4-gamma 1


frontotemporal lobar disease


fused in sarcoma


Huntington's disease


leucine-rich repeat kinase2


phosphatidylinositol binding clathrin assembly protein


Parkinson's disease




sodium dodecyl sulfate


superoxide dismutase 1


TAR DNA binding protein


vacuolar protein sorting 35 homolog

Protein misfolding and aggregation: a common feature among neurodegenerative diseases

Proteins need to fold and maintain, conformational flexibility to perform their biological functions. As a consequence, many proteins are only marginally stable in their native environments (Vendruscolo et al. 2011). This metastability, however, comes at a cost. It results in a constant stream of proteins transiently adopting potentially dangerous conformations. This situation is further exacerbated by the fact that protein synthesis is error-prone, producing a high number of proteins with unstable folds or excess unassembled subunits (Drummond and Wilke 2008). When the capacity of quality control mechanisms is exceeded, the accumulation of misfolded proteins becomes a liability to cellular homoeostasis. Thus, conditions that favor protein unfolding/misfolding and unspecific aggregation on a massive scale, such as environmental and nutritional stress, aging, or pathological changes, can have dramatic effects for cellular and organismal viability.

Two direct consequences are hypothesized to be associated with protein misfolding: the loss of the normal biological function of the protein, and a gain of toxic function. The toxicity is thought to be a direct consequence of the ability of misfolded proteins to form oligomers, protofibrils, fibrils and, ultimately, large aggregates (Fig. 1) (Dobson 2003; Soto 2003). Although the detailed mechanisms of gain of function toxicities are not well understood, recent data suggest misfolded proteins engage in inappropriate interactions with themselves and with other cellular components, exhausting the available pool of quality control components in the cell. As a consequence, imbalances in cellular homeostasis occur that can culminate in cellular dysfunction or death. Thus, protein misfolding is known to be a major cause of several human diseases, such as lysosomal storage disorders, cystic fibrosis, cancer, and a range of neurodegenerative disorders, such as Alzheimer's, Huntington's or Parkinson's disease. The chain of molecular events that lead to development of these neurodegenerative diseases, involves the accumulation of different misfolded protein species, ranging from oligomers to protein aggregates that accumulate intra- or extracellularly, depending on the particular disease (Soto 2003; Gorman 2008). Recently, the idea that several proteins associated with neurodegenerative disorders can spread in the brain, in a prion-like manner, has hit the spotlight (Marques and Outeiro 2012; Munch and Bertolotti 2012).

Figure 1.

Overview of protein misfolding and aggregation events and the involved cytoprotective mechanisms, which, upon failure, culminate in cell death. Genetic or environmental factors can cause the misfolding of proteins and their conversion into pathological oligomers that fibrillize and deposit into larger aggregates. In healthy cells, the cellular quality control systems (QCS) are able to counteract this cascade of events. The first steps of aggregation can be prevented or reversed by chaperones, while the later ones are counteracted by promoting the degradation of misfolded proteins, either by UPS or by autophagy. In parallel, spatial quality control (SQC) mechanisms are activated that organize small aggregates in specific locations in the cell: the aggresome is thought to be cytoprotective by sequestering toxic misfolded proteins and facilitating their removal by autophagy; it requires the microtubule network and dynein motors, and is associated with the centrosome; the juxtanuclear quality control compartment (JUNQ), in which soluble, misfolded, ubiquitylated proteins can be deposited in close proximity to proteasomes; the insoluble protein deposit (IPOD), in which insoluble, terminally aggregated proteins are accumulating that may be targeted for autophagy. However, in pathological conditions QCS and SQC can fail, resulting in the accumulation of the misfolded forms, which establish inappropriate interactions with other cellular components, ultimately resulting in cytotoxicity and cell death. The mechanisms of cytotoxicity are not fully understood and might be diverse. However, it is known that oligomeric species damage membranes and organelles (e.g., α-syn is thought to assemble into ion channels in membranes) and larger aggregates have been proposed to titrate essential cellular factors.

In general, different neurodegenerative diseases are associated with the misfolding and aggregation of specific proteins. For example, α-synuclein (α-syn) is the main component of intracellular inclusions known as Lewy bodies (Spillantini et al. 1997). Diseases associated with α-syn aggregation are referred to as synucleinopathies, and include Parkinson's disease (PD), dementia with Lewy bodies, and multiple systems atrophy, among others (Vekrellis et al. 2011; Lashuel et al. 2012).

Another group of protein misfolding diseases are the polyglutamine disorders, characterized by the intracellular accumulation of a disease-specific protein with aberrantly expanded polyglutamine (polyQ) stretches (Shao and Diamond 2007). Huntington's disease (HD) is among the most common of these disorders and results from CAG-expansions in the exon 1 of the IT-15 gene, which encodes the protein huntingtin (Huntington's Disease Collaborative Research Group's 1993).

Alzheimer's disease (AD), the most common neurodegenerative disease, is characterized by the accumulation amyloid β peptide in extracellular plaques and by the intracellular accumulation of hyperphosphorylated Tau protein in inclusions known as neurofibrillary tangles. These tangles are also a hallmark of other diseases, collectively known as tauopathies (Spires-Jones et al. 2009).

Amyotrophic lateral sclerosis (ALS) and frontotemporal lobar disease (FTLD) have been linked to mutations and misfolding of diverse proteins. One of them, superoxide dismutase 1 (SOD1), has been studied extensively in the context of familial ALS (fALS). Mutant SOD1 tends to aggregate and seems to cause disease via a toxic gain of function mechanism (Kryndushkin and Shewmaker 2011). Two other proteins, TAR DNA binding protein (TDP-43) and fused in sarcoma (FUS), are emerging as important disease factors in ALS and FTLD-U (Neumann 2009).

The main risk factor for all of these neurodegenerative diseases, however, is aging (Yankner et al. 2008), suggesting an age-associated decline in protein quality control function that may contribute to the misfolding and aggregation of proteins in particular populations of brain cells (Chen et al. 2011).

Yeast as a model for studying protein misfolding diseases

The budding yeast Saccharomyces cerevisiae features many technical advantages over other systems, such as a short generation time, facilitated genetic manipulations, inexpensive growth media, simple sterile techniques requirement, easy laboratory maintenance, and the possibility of long-term storage. For these reasons, yeast has been used by researchers as an eukaryotic model to uncover basic cellular mechanisms, such as cell division, replication, metabolism, protein folding, and intracellular transport (Fields and Johnston 2005).

Yeast was the first eukaryotic organism to have its genome fully sequenced, revealing a compact genome with about 6000 open reading frames (ORF). In contrast to higher eukaryotes, only 4% of the yeast genes contain introns (Goffeau et al. 1996). Since then, strong effort has been undertaken to functionally analyze uncharacterized yeast genes. Yeast also had a fundamental role in pioneering many high-throughput techniques, such as DNA and ChIP-chip microarrays, the creation of bar-coded systematic deletion sets, the large-scale identification of protein–protein and genetic interactions (interactome) and subcellular localizations (localisome), and transcriptome, proteome and metabolome analysis (Botstein and Fink 2011). In parallel, good databases and supporting bioinformatic tools were developed to store and share the data collected. In this respect, it is important to highlight the role of the Saccharomyces Genome Database (SGD, http://www.yeastgenome.org), an essential resource for the yeast community that compiles information about the yeast genome, its chromosomal features, their functions and interactions of their gene products. Through a permanent effort to update and develop new tools, SGD responds to the needs of storage, display, analysis, and integration of huge amount of data generated by large-scale approaches (Chan and Cherry 2012; Cherry et al. 2012). For all these reasons, yeast had a unique role in establishing entirely new research fields, such as ‘functional genomics’ and ‘systems biology’ (Petranovic et al. 2010; Botstein and Fink 2011). As a consequence, we now have at least a rudimentary understanding of the biological role of about 85% of the yeast genes, making yeast the currently best understood eukaryotic cell (Botstein and Fink 2011).

About one fifth of yeast genes are members of orthologous gene families associated with human diseases (Heinicke et al. 2007). Indeed, many processes and pathways that are also disease-associated are conserved between yeast and humans (Karathia et al. 2011), such as protein quality control (Brodsky and Skach 2011), vesicular trafficking (Bonifacino and Glick 2004), autophagic pathways (Klionsky et al. 2007), lipid metabolism (Petranovic et al. 2010), the unfolded protein response (Goeckeler and Brodsky 2010), mitochondrial biogenesis and metabolism (Rinaldi et al. 2010), and aging (Barros et al. 2010), among others. The fact that several pathways associated with neurodegeneration are conserved between yeast and humans was crucial to develop yeast as a model for neurodegenerative diseases, including mitochondrial quality control mechanisms (Braun 2012), apoptotic and necrotic cell death (Braun et al. 2010), protein folding, trafficking, and secretion (Bonifacino and Glick 2004).

Two different approaches have been used in the study of human diseases in yeast. When a homolog of the gene implicated in the disease is present in yeast genome, genetic ablation or over-expression experiments are possible. On the other hand, when the disease-associated gene does not have an obvious counterpart in yeast, the transgene can be heterologously expressed in yeast, and the resulting strains subjected to functional analyses. Moreover, even when the gene function is already known, yeast models have enabled us to gain important insight into the underlying molecular basis of pathology of disease-associated mutations.

Importantly, once validated, yeast models of human diseases are a very useful platform for high-throughput screens as a first-line approach in the discovery of new genes that might have potential as therapeutic targets, as well as in the identification of new drugs with therapeutic value (Outeiro and Giorgini 2006; Tenreiro and Outeiro 2010). Examples of these approaches and discoveries will be presented below.

Yeast as a model of PD

Almost two centuries after the first description of PD, there are only symptomatic treatments available for this pathology. Our ability to discover new PD treatments is limited for several reasons; one of them relates to the complex and multifactorial etiology of PD, which involves an intricate interplay of a large network of factors: genes, susceptibility alleles, environment and gene-environment interactions, and aging. Only in the last 15 years it was realized that 5–10% of the PD cases have a genetic root (Klein and Westenberger 2012).

The first gene to be identified as associated with familial cases of PD was SNCA, which encodes for α-synuclein (Polymeropoulos et al. 1997). In addition to its presence in Lewy bodies and Lewy neurites, as explained above, missense mutations in the SNCA gene (A30P, E46K, A53T), as well as duplication or triplication of the gene, are causative for PD (Polymeropoulos et al. 1997; Kruger et al. 1998; Singleton et al. 2003; Chartier-Harlin et al. 2004; Zarranz et al. 2004). Later, other mutations in different genes, including those encoding the leucine-rich repeat kinase2 (LRRK2), the E3 ubiquitin ligase parkin, the oxidative-stress-related chaperone DJ-1, the mitochondrial PTEN-induced putative kinase1 (PINK1), the lysosomal ATPase ATP13A2, the vacuolar protein sorting 35 homolog (VPS35), and eukaryotic translation initiation factor 4-gamma 1 (EIF4G1) were found to be associated with PD (Klein and Westenberger 2012). Apart from these genes that are unequivocally linked to heritable monogenic PD, numerous genetic risk factors have been identified in hundreds of genetic association studies. These are now available in the public online database ‘PDGene’ (http://www.pdgene.org) (Lill et al. 2012).

Because α-syn is a central player in synucleinopathies, much research effort is currently directed at the molecular properties of α-syn, in particular those regarding its propensity to aggregate, its structure and post-translational modifications, as well as its physiological function (Lashuel et al. 2012).

The current lack of effective therapies is mainly because of the absence of suitable animal models, in which all the PD features can be reproduced (Lee et al. 2012). To overcome this serious limitation, researchers are now adopting more rational approaches where different cellular or animal models are used for answering a given scientific question. Moreover, current approaches often combine different cellular and animal models, looking for a broader validation of the findings (Cooper et al. 2006; Su et al. 2010; Xiong et al. 2010).

The first yeast model of PD was described 10 years ago and consisted in the heterologous expression of the α-syn cDNA in yeast cells (Fig. 2). Interestingly, and consistent with the discovery that multiplications in the SNCA gene cause familial PD, it was found that α-syn induced dose-dependent toxicity. This toxicity was associated with subcellular redistribution of α-syn and resulted in the formation of intracellular inclusions in a dose-dependent manner (Outeiro and Lindquist 2003). Since then, dozens of studies were performed using yeast to model particular aspects of PD (Outeiro and Giorgini 2006; Miller-Fleming et al. 2008; Tenreiro and Outeiro 2010). At present, α-syn toxicity in yeast is known to involve several cellular pathways that were either (i) first described in yeast and then validated in higher eukaryotic models of PD or (ii) identified in other systems and then successfully recapitulated in yeast. Among the different pathways involved in α-syn toxicity in yeast, induction of apoptosis, lipid droplet accumulation (Outeiro and Lindquist 2003), mitochondrial dysfunction (Buttner et al. 2008; Su et al. 2010), proteasome impairment (Outeiro and Lindquist 2003; Chen et al. 2005; Sharma et al. 2006), oxidative stress (Sharma et al. 2006; Witt and Flower 2006), autophagy/mitophagy dysfunction (Petroi et al. 2012; Sampaio-Marques et al. 2012), vesicle trafficking defects (Outeiro and Lindquist 2003; Soper et al. 2008), and endoplasmic reticulum (ER)-to-Golgi trafficking impairment (Cooper et al. 2006), seem to be the most striking. These findings suggest that α-syn interferes with multiple pathways in the cell but this does not clearly define the order of events. The question is whether the defects arise as a cause or consequence of α-syn expression in yeast. In either case, these observations are also in line with the multiple hit hypothesis for neurodegeneration in PD, where a conjugation of factors needs to come together for disease to take place.

Figure 2.

α-Syn WT and familial mutants expression in yeast cells. (a) Yeast multicopy expression vectors were used to clone the human gene SNCA encoding α-syn WT or the familial mutants A30P, E46K or A53T with a GFP C-terminal fusion and under the regulation of the inducible GAL1 promoter. (b) These constructs were used to transform yeast cells and the α-syn expression levels were evaluated by western blot of protein total extracts, after 5 h of α-syn expression induction with galactose. (c) α-Syn cytotoxicity was also evaluated by spotting assay. Briefly, the cell suspensions were adjusted to OD 600 nm = 0.05 ± 0.005 and used to prepare 1/3 serial dilutions that were applied as spots (4 μL) onto the surface of plates containing YPD rich medium and incubated at 30°C for 2 days. (d) Fluorescence microscopy of yeast cells expressing α-syn fused with GFP (scale bar: 5 μm).

The first large-scale genetic screening using yeast as a model of synucleinopathies consisted in the study of a collection of deleted nonessential genes to identify enhancers of α-syn toxicity, an approach known as a synthetic lethal screen (Willingham et al. 2003). In a separate study using the same gene-deletion collection to identify modifiers of α-syn aggregation and sub-cellular localization, similar pathways were identified, including lipid metabolism, vesicular transport, and vacuolar degradation (Zabrocki et al. 2008).

Over-expression screens were also performed to identify modifiers of α-syn toxicity (both enhancers and suppressors). Proteins involved in vesicular trafficking, such as Ypt1p, were identified as suppressors of α-syn toxicity and, ultimately, resulted in the finding that the mammalian Ypt1p homolog, Rab1, was also a suppressor of dopaminergic cell loss in animal models of PD (Cooper et al. 2006). Another gene over-expression screen resulted in the identification of suppressors of the lethality of WT, A30P, and A53T α-syn in response to oxidative stress induced by hydrogen peroxide (Flower et al. 2007). Interestingly, the suppressors identified for the WT α-syn and the two familial mutations were considerable different, reinforcing the idea that mutant α-syn causes toxicity by distinct mechanisms (Flower et al. 2007).

Analysis of gene expression profiling experiments alongside with data from genetic screens using a new bioinformatic tool, named ResponseNet, led to the identification of new pathways involved in α-syn cytotoxicity in yeast, such as ergosterol biosynthesis and the target of rapamycin (TOR) pathway (Yeger-Lotem et al. 2009).

Screens of small molecules with therapeutic interest were also carried out using α-syn-based yeast models. Screening of large libraries of compounds lead to the identification of two cyclic peptides and four 1,2,3,4-tetrahydroquinolinones as suppressors of α-syn toxicity (Griffioen et al. 2006; Kritzer et al. 2009; Su et al. 2010). Moreover, some of these compounds were found to reduce the formation of α-syn inclusions, re-established ER-to-Golgi trafficking, and ameliorated α-syn-mediated damage to mitochondria (Su et al. 2010), the same pathways of toxicity identified in the genetic screens mentioned above.

More recently, other PD related genes have also been studied in yeast, thus expanding our current knowledge of the molecular mechanisms underlying the disease (Table 1). For example, yeast was used to identify interactors of LRRK2 using yeast 2-hybrid analysis (Shin et al. 2008; Zheng et al. 2008). Another study associated LRRK2 toxicity with defects in endocytic vesicular trafficking and autophagy, an observation that was validated in primary neuronal models and found to depend on its GTPase activity (Xiong et al. 2010).

Table 1. Genes involved in familial forms of PD that have been heterologously expressed in yeast
GeneYeast homologYeast modelReference
SNCA __Heterologous expression of human SNCAOuteiro and Lindquist (2003)
LRRK2 __Heterologous expression of human LRRK2Xiong et al. (2010)
ATP13A2 Ypk9Co-expression of YPK9 with human SNCA;expression of Ypk9 with PD-associated mutationsGitler et al. (2009), Schmidt et al. (2009), Chesi et al. (2012), Usenovic et al. (2012)

Yeast models also contributed to the functional analysis of the human lysosomal P(5B) ATPase ATP13A2 encoded by the gene PARK9 (Gitler et al. 2009; Schmidt et al. 2009; Chesi et al. 2012; Usenovic et al. 2012). The yeast ortholog, YPK9 (for Yeast PARK9), encodes a vacuolar membrane transporter that is also a suppressor of α-syn toxicity (Gitler et al. 2009). This protein plays a role in vesicular trafficking and antagonizes α-syn toxicity (Gitler et al. 2009; Usenovic et al. 2012). Ypk9p protective function depends on its vacuolar localization and ATPase activity, and is probably related to its role in manganese and other divalent heavy metal ions homeostasis. In agreement with this finding, metal ions constitute environmental risk factors for some forms of Parkinsonism (Schmidt et al. 2009; Chesi et al. 2012). Importantly, PD-associated mutations resulted in an aberrant localization of Ypk9p, thus supporting the notion that these mutations are causing a loss of function (Gitler et al. 2009).

Recently, the D620N mutation in VPS35 was found to be associated with familial PD (Zimprich et al. 2011; Sheerin et al. 2012). Vps35 is a highly conserved component of a heteropentameric complex that mediates retrograde transport of transmembrane cargo from endosomes back to the trans-Golgi network (Bonifacino and Rojas 2006). The role of pathogenic mutation in VPS35 is not fully understood but it has been hypothesized that it may act by disrupting recognition and binding to key cargo proteins (Zimprich et al. 2011). Importantly, VPS35 can associate directly with both soluble N-ethylmaleimide-sensitive factor attachment protein receptor (SNARE) proteins and the sortilin-related receptor SORL1 (an important player in AD) (Arighi et al. 2004; Nielsen et al. 2007; Bonifacino and Hurley 2008). Yeast has a functionally well-characterized ortholog of VPS35, suggesting it might be used as a model to investigate the molecular mechanisms involved in its pathogenesis (Liu et al. 2012).

Similarly, mutations in the translation initiation factor EIF4G1 were recently associated with familial PD cases but it is still not clear whether they constitute pathogenic mutations or rare benign variants (Klein and Westenberger 2012). Yeast harbors two eIF4G isoforms (1 and 2) that are ~ 50% identical in sequence and related to mammalian eIF4G (Goyer et al. 1993). Thus, yeast might also valuable to explore the pathological implications of EIF4G1 mutations.

Yeast as a model of HD

There are currently nine diseases known in humans that are associated with abnormally expanded stretches of polyQ in specific proteins. Expansion causes them to misfold, aggregate, and it induces cytotoxicity, including dysfunction and death of specific neuronal populations (Orr and Zoghbi 2007). Interestingly, the proteins involved share no homology outside the polyQ tract and have different cellular functions and localizations. As a consequence, the different polyQ diseases are associated with perturbations in various cellular pathways, which differentially affect specific neurons in the brain.

HD is an autosomal dominant movement disorder caused by expansion of a CAG repeat located in the first exon of the IT-15 gene, which codes for the protein huntingtin (HTT) (Huntington's Disease Collaborative Research Group's 1993). Longer repeats increase the propensity of htt to form intranuclear and cytoplasmic inclusions. Expansion above 39 repeats is fully penetrant for HD, while intermediate repeat lengths of 36–39 are incompletely penetrant (Orr and Zoghbi 2007). HD patients suffer from cognitive, psychiatric, and motor abnormalities, resulting from the loss and dysfunction of neurons in the striatum and deep layers of the cortex (Orr and Zoghbi 2007).

The contribution of yeast models to our current knowledge of the molecular basis of HD was extensively reviewed very recently (Mason and Giorgini 2011). Thus, here we will focus primarily on more recent studies.

The first yeast model of polyQ diseases involved the expression of exon 1 of huntingtin (htt) with different polyQ lengths fused to GFP (Krobitsch and Lindquist 2000; Muchowski et al. 2000). Although the Q25 htt variant (corresponding to a normal polyQ length) did not aggregate, insoluble inclusion formation increased with an increase in the polyQ length (Krobitsch and Lindquist 2000), recapitulating results obtained in cultured mammalian cells and animal models (Apostol et al. 2003, 2006; Woodman et al. 2007). The correlation between aggregation and toxicity of htt fragments in yeast was found to be dependent on the sequences flanking the polyglutamine stretches as well as on the existence of specific interacting proteins of the yeast strain expressing it, in particular the prion composition of the cell (Meriin et al. 2002; Duennwald et al. 2006a, b; Gong et al. 2012). Specifically, the htt exon 1 with expanded polyQ tracts was shown to impair protein homeostasis of the ER (Duennwald and Lindquist 2008) and endocytosis (Meriin et al. 2003, 2007), cause transcriptional deregulation (Hughes et al. 2001), increase ROS production by affecting mitochondrial function and morphology (Giorgini et al. 2005; Sokolov et al. 2006; Solans et al. 2006), induce activation of apoptotic pathways (Sokolov et al. 2006), and cause cell cycle dysfunction (Sokolov et al. 2006; Bocharova et al. 2008). Metabolic effects of polyQ expression in yeast were also identified, with alteration of the cellular concentration of several metabolites, such as alanine, glycerol, glutamine, and valine, which were proposed as promising biomarkers for HD (Joyner et al. 2010).

Combining new quantitative biochemical and microscopic approaches and yeast models, it was reported that, in contrast to polyQ tracts of 25 consecutive Qs, which never aggregate, and polyQ tracts of 103 Qs, which always aggregate, a mid-size polyQ tract of 47 Qs, forms aggregates in yeast cells as they age (Cohen et al. 2012). Aggregation is aggravated in the absence of silent information regulator 2 (Sir2), known to regulate metabolism and longevity (Schwer and Verdin 2008), while over-expression of heat shock factor 1 (Hsf1), known to regulate longevity and to be involved in maintaining proteostasis (Akerfelt et al. 2010), attenuates aggregation (Cohen et al. 2012). Accordingly, these two highly conserved proteins appear to affect aging-dependent aggregation of htt exon 1, revealing new pathways as potential therapeutic targets. Interestingly, induction of Sir2 was also found to reduce protein aggregation (Sorolla et al. 2011).

Once validated, yeast models of HD were used as platforms to unravel the molecular basis of the disease (Giorgini et al. 2005; Meriin et al. 2007). An important advance was the identification of the kynurenine pathway in a yeast screen for modifiers of polyQ toxicity (Giorgini et al. 2005). This pathway is involved in tryptophan degradation and is activated by mutant htt expression, resulting in higher levels of two neurotoxic metabolites, 3-hydroxykynurenine and quinolinic acid, consistent with observations in mammalian models and HD patients (Thevandavakkam et al. 2010). Inhibition of a key enzyme of this pathway, the kynurenine 3-monooxygenase (KMO), by chemical or genetic approaches, rescued htt toxicity in yeast, Drosophila, and mouse models of HD (Giorgini et al. 2005; Campesan et al. 2011; Zwilling et al. 2011), rendering this pathway as a very attractive therapeutic target in HD (Thevandavakkam et al. 2010).

Yeast models of HD were also used in drug screens and resulted in the identification of small molecules that showed potential as therapeutic tools to ameliorate polyQ toxicity in higher eukaryotes (Zhang et al. 2005; Bodner et al. 2006; Ehrnhoefer et al. 2006; Sarkar et al. 2007). In a recent study, a HD yeast model was also used to dissect the protective effect and mode of action of curcumin, a polyphenol present in the spice turmeric and known to have broad biological and medicinal effects, including efficient anti-oxidant, anti-inflammatory, and anti-proliferative activities (Verma et al. 2012). Curcumin was found to prevent polyQ aggregation by inhibiting the expression levels of the Vacuolar Protein Sorting protein 36 (Vps36), a component of the ESCRT-II complex, involved in trafficking soluble and integral membrane proteins from the trans-Golgi network to the perivacuolar region and, finally, to the vacuole. These results reinforce the relevance of protein trafficking pathways in aggregation of amyloid proteins previously demonstrated (Meriin et al. 2007).

Yeast as a model of Alzheimer's disease

AD, the most common neurodegenerative disease worldwide, is characterized by the aggregation of β-amyloid peptide (Aβ). Aβ peptide is generated by sequential proteolytic cleavage of the amyloid precursor protein (APP) (O'Brien and Wong 2011). Imbalance between generation and clearance of Aβ-peptide is thought to lead to the formation of cytotoxic Aβ oligomeric species and extracellular deposition of Aß amyloid aggregates, known as Aβ plaques.

APP is a transmembrane protein that exists in the plasma membrane and, as such, it travels through multiple subcellular compartments in neurons. This complicated the generation of a yeast model that faithfully mimicked Aβ toxicity. Therefore, early studies in yeast focused on either upstream (APP processing) or downstream (Aβ aggregation, tau phosphorylation) effects of Aβ pathology (Bharadwaj et al. 2010). However, a robust yeast model was eventually developed and resulted in the establishment of functional links between Aß toxicity, endocytic trafficking, and AD risk factors (Treusch et al. 2011). In this study, the most toxic Aβ fragment known, Aβ 1-42, was directed to the yeast secretory pathway by fusing an ER signal sequence to the N-terminus of the peptide (ssAβ 1-42). The expectation was that the peptide would get secreted and, because of restraints by the yeast cell wall, reintroduced into the cell by endocytosis, thereby recapitulating many of the potentially important aspects of Aβ toxicity in neurons. Expression of multiple copies of ssAβ 1-42 from a genomic locus led to cytotoxicity and accumulation of oligomeric species of Aβ 1-42 (Treusch et al. 2011).

Two ssAβ 1-42-expressing yeast model strains were subsequently used for a genome-wide over-expression screen that identified 23 suppressors and 17 enhancers of Aβ toxicity. The study focused on 12 genes that had clear human homologs and several of them could be directly linked to human AD risk factors. Importantly, the hits from the screen were validated in a C. elegans model of Aβ toxicity. One of the hits, YAP1802/phosphatidylinositol binding clathrin assembly protein (PICALM), which is strongly associated with AD, was more closely examined in cultured rat cortical neurons, and found to rescue Aβ-caused cell death in a dose-dependent manner (Treusch et al. 2011).

To further dissect the molecular mechanism of PICALM-mediated Aβ detoxification, the researchers returned to the yeast model. Aβ severely perturbed clathrin-mediated endocytic trafficking of the G protein-coupled receptor Ste3 from the plasma membrane to the vacuole. Co-expressing YAP1802/PICALM and two other hits from the screen linked to endocytosis (INP53 and SLA1) reversed the trafficking defects, thus confirming clathrin-mediated endocytosis as an important target of Aß toxicity (Treusch et al. 2011).

Yeast as a platform to study aggregation-prone proteins and prions

The classical view of prions is that of protein-only disease agents, involved in neurodegenerative diseases of mammals, like scrapie in sheep, bovine spongiform encephalopathy in cattle, and Creutzfeldt-Jakob disease in humans, to list but a few. All these diseases are based on catastrophic misfolding of the PrP protein into a self-templating amyloid state, which ultimately causes neuronal cell death. However, it is now clear that not all prions are detrimental, and there is increasing evidence for the existence of functionally relevant and even beneficial prion states of certain proteins. Examples are the Het-S prion in Podospora anserina (Coustou et al. 1997), the CPEB protein in Aplysia (Si et al. 2003), and the Sup35 prion found in wild yeast strains (Halfmann et al. 2012). In addition, a recent study of prion domains (PrDs) in yeast has broadened the prion concept and revealed a number of new potential prion candidates in the yeast genome (Alberti et al. 2009).

This study also demonstrated how yeast can be used as a living test tube to explore amyloid-based protein aggregation phenomena on a genome-wide scale (Lashuel and Pappu 2009). Alberti et al. used a hidden Markov Model (HMM) trained on the PrDs of four known yeast prions to screen the yeast genome for potential prion candidates and identified ~ 200 candidate prion domains (cPrDs). Astonishingly, although trained on yeast sequence, the HMM also successfully identified prion-like domains in other genomes (King et al. 2012; Malinovska et al. 2013). Clustering of the hits according to SMART protein domains revealed a highly significant enrichment of PrDs in RNA-binding proteins. Several disease candidates, such as TDP-43 and FUS, which are discussed later in this review, were identified among these hits.

The 100 best-scoring yeast cPrDs were rigorously tested for their aggregation and prion properties with a variety of biochemical and cell biological assays uniquely available for yeast. In the following paragraphs, we highlight some of these assays, because they are well suited for the study of aggregation-prone proteins from heterologous sources. In fact, they have already been applied to characterize several disease-relevant proteins in the context of yeast models of neurodegenerative diseases. An overview of the mentioned methods is given in Fig. 3. For a more comprehensive summary of methods that are currently available for the study of aggregation-prone proteins in yeast, we refer the reader to the following study (Alberti et al. 2010).

The availability of a suite of recombination-based Gateway vectors (Alberti et al. 2007) and the presence of several Gateway over-expression libraries from different organisms (Rual et al. 2004, 2005; Gelperin et al. 2005; Cooper et al. 2006; Lamesch et al. 2007) allow the rapid and efficient cloning of candidate ORFs into various yeast expression plasmids. For example, in the above-mentioned screen for yeast prions, candidate PrDs were fused to EYFP and expressed from the inducible GAL promoter. This led to the formation of fluorescent cytoplasmic foci for more than half of the tested cPrDs (Alberti et al. 2009). Over-expression of many cPrD-EYFP fusions also resulted in formation of sodium dodecyl sulfate (SDS)-resistant amyloid aggregates, as was shown by semi-denaturing detergent agarose gel electrophoresis (SDD-AGE). SDD-AGE can be applied in high-throughput and is a powerful tool to detect SDS-resistant high-molecular weight conformers of aggregation-prone proteins (Bagriantsev et al. 2006; Alberti et al. 2010) (Fig. 3a). In the same manner, disease-relevant proteins that have no homologs in the yeast genome have been over-expressed in yeast as fusions to fluorescent proteins, often as a first step to assess their aggregation properties (Outeiro and Lindquist 2003; Johnson et al. 2008, 2009; Kryndushkin and Shewmaker 2011; Kryndushkin et al. 2011, 2012).

Figure 3.

Yeast as a platform to study aggregation-prone proteins. The aggregation-prone protein domain of interest can be expressed fused to a fluorescence tag (a) or the Sup35 C-terminal domain (b) to perform the shown aggregation assays. (a) Fusion to a fluorescence tag allows direct visualization of protein aggregates by fluorescence microscopy (left panel). The same strains can be subjected to semi-denaturing detergent agarose gel electrophoresis (SDD-AGE, right panel). Sodium dodecyl sulfate (SDS) insoluble species of aggregated proteins migrate slowly in the agarose gel and appear as high-molecular weight smear. Immunodetection with antibodies directed against the tag circumvents the need for specific antibodies. (b) Fusing the domain of interest to the C-terminal domain of Sup35, a translation termination factor, allows monitoring of the aggregation properties of the examined protein domain through a phenotypic readout. In the absence of aggregation, Sup35 mediated translation termination fidelity is high (upper panel) and a premature stop codon present in the ade1-14 mRNA is recognized with high frequency. In this scenario, no Ade1 protein is synthesized and cells produce a red pigment from a metabolite of the adenine biosynthesis pathway. However, when Sup35 is sequestered and deactivated by aggregation (lower panel), translation termination fidelity is low, and the premature stop codon is ignored with high frequency, leading to synthesis of Ade1p and a white colony color.

Aggregation-prone proteins can also be investigated with a powerful phenotypic aggregation assay that is based on the translation termination factor and well-established yeast prion Sup35 (Fig. 3b). The Sup35 protein has two functionally distinct regions. The N terminal region contains the prion domain and the C-terminal region is responsible for translation termination activity. Sup35p can itself form a prion state, designated [PSI+], which is characterized by sequestration of Sup35p into non-functional, cytoplasmic amyloid aggregates and results in increased read-through of premature stop codons. The existence of one such premature stop codon in the ADE1 gene (ade1-14) of many yeast laboratory strains, causes a red colony phenotype and auxotrophy for adenine. However, when the cells are in the [PSI+] state, the ade1-14 stop codon is ignored with a high frequency, resulting in a white colony color and adenine prototrophy. As a consequence, prion-positive colonies can be identified by simple growth assays (growth in the absence of adenine) or by their white colony color on rich medium.

By replacing the Sup35p prion domain with the predicted cPrDs from the screen, all candidates were tested for their ability to enter into and stably maintain a prion phenotype (Alberti et al. 2009). Again, this method can be applied to aggregation-prone protein domains in general (Fig. 3b) and was, among others, used to study structural features of the mammalian prion PrP (Dong et al. 2007; Jossé et al. 2012), aggregation of Aß (von der Haar et al. 2007; Park et al. 2011), expanded versions of polyQ (Osherovich et al. 2004; Suzuki et al. 2012), and ALS-associated proteins such as FUS (Kryndushkin et al. 2011). Sup35-based prion assays were also used to identify drugs that interfere with prion formation (Bach et al. 2003, 2006; Saupe 2003; Tribouillard et al. 2006, 2007; Wang et al. 2008; Roberts et al. 2009; Duennwald and Shorter 2010; Shorter 2010; Duennwald et al. 2012). Many of these drugs were not only effective against [PSI+] but also interfered with aggregation of other proteins, such as PrP or polyglutamine. However, drug screens were not limited to WT Sup35p. For example, in one study, Aβ was fused to Sup35p and the resulting strain was used to phenotypically screen a compound library (Park et al. 2011). The study identified two compounds with anti-oligomer activity, suggesting that Sup35-based high-throughput drug screens in yeast are an efficient and cost-effective approach to identify new inhibitors of protein aggregation.

Sup35p offers a powerful toolbox for the study of aggregation-prone proteins. However, Sup35-based aggregation assays also have several weaknesses. One of them is that the candidate needs to enter into a rigid amyloid conformation (Fig. 3b, lower panel) to sufficiently inactivate the translation termination function of Sup35; the formation of less rigid amorphous aggregates was often found to be without effect (e.g., Kryndushkin et al. 2011). However, as exemplified by Aβ, oligomerization of an aggregation-prone protein can sometimes sufficiently inactivate Sup35p (Bagriantsev and Liebman 2006). Another complication is that the proteins that are fused to Sup35p need to be largely devoid of structure. In addition, they should not be larger than a few hundred amino acids, because large extensions at the N-terminus tend to interfere with the overall function of Sup35p. Despite these drawbacks, yeast prion-based aggregation assays will continue to make important contributions to our understanding of protein aggregation and will aid in the discovery of new drugs to treat neurodegenerative diseases.

Yeast as a model of ALS and FTLD-U

In addition to the disease models discussed so far, yeast is also arising as a strong model for ALS and FTLD-U (Kryndushkin and Shewmaker 2011). A number of different proteins, like FUS (Ju et al. 2011; Daigle et al. 2012), HNRNPA2B1 and HNRNPA1 (Kim et al. 2013), OPTN (Kryndushkin et al. 2012), SOD1 (Rabizadeh et al. 1995; Corson et al. 1998), and TDP-43 (Johnson et al. 2008, 2009; Armakola et al. 2012) have been studied in yeast in context of these disorders (see Table 2). Here, we will focus on TDP-43 and FUS, both of which are RNA-binding proteins with prion-like domains. These proteins are representative members of a group of proteins that are emerging as important players in neurodegenerative diseases (Lagier-Tourenne and Cleveland 2009; Couthouis et al. 2011; King et al. 2012).

Table 2. Genes involved in ALS and FTLD-U that have been studied in yeast models
GeneYeast homologYeast model (Reference)
FUS __Heterologous expression of FUS (Ju et al. 2011; Sun et al. 2011; Daigle et al. 2012) and screen for suppressors of cytotoxicity (Ju et al. 2011)
HNRNPA2B1 and HNRNPA1__Heterologous expression of HNRNPA2B1 and HNRNPA1 (Kim et al. 2013)
OPTN __Heterologous expression of OPTN (Kryndushkin et al. 2012)
SOD1 Sod1Heterologous expression of mutant SOD1 in sod1Δ yeast strains (Rabizadeh et al. 1995; Corson et al. 1998)
TDP43 __Heterologous expression of TDP-43 (Johnson et al. 2008, 2009) and screen for suppressors of cytotoxicity (Armakola et al. 2012)

A clinical study identified TDP-43 in ubiquitin positive inclusions in brain samples of ALS and FTLD-U patients, thus establishing TDP-43 as a possible link between ALS and FTLD-U (Neumann et al. 2006). Yeast turned out to be a very useful model system to unravel the underlying disease mechanisms of TDP-43. It was initially used to dissect the structural determinants of aggregation and toxicity (Johnson et al. 2008, 2009). Subsequent genome-wide screens identified suppressors and enhancers of cytotoxicity, two of which – Pbp1 and Dbr1 – were confirmed in more complex model systems (Elden et al. 2010; Armakola et al. 2011, 2012). Interestingly, Pbp1 is the yeast homolog of human ataxin 2, which is itself associated with another neurodegenerative disease (SCA2) and increases the risk of developing ALS, when its polyglutamine stretch is expanded to intermediate length (Elden et al. 2010).

TDP-43 normally localizes to the nucleus in mammalian cells. However, in disease-affected neurons it accumulates in cytoplasmic inclusions, which are a hallmark of ALS and FTLD-U. TDP-43 also localizes to the nucleus in yeast cells when expressed at normal levels. Only strong over-expression, which supposedly overwhelms the protein quality control machinery, leads to formation of cytoplasmic aggregates (Johnson et al. 2008). These aggregates are non-amyloid like and form independently of Hsp104, an established modulator of amyloid-like inclusions in yeast, such as those formed by Sup35p, or htt exon 1 with expanded polyQ stretches (Grimminger-Marquardt and Lashuel 2010; Winkler et al. 2012). Structural determinants of cytotoxicity and aggregation were determined by over-expressing a variety of truncated versions of TDP-43. It was found that the C-terminal region of the protein plus the complete second RNA-binding domain was necessary to cause cytotoxicity and aggregation (Johnson et al. 2008). Moreover, and maybe most importantly, ALS-linked mutations in the C-terminal region of TDP-43 increased aggregation in vitro and in vivo and led to elevated cytotoxicity (Johnson et al. 2009). This underscores the ability of yeast to faithfully recapitulate the effects of disease-relevant mutations.

RNA binding is an important functional property of TDP-43 (Johnson et al. 2008). A recent study revealed that the RNA-binding ability might be causally linked to the cytotoxicity of over-expressed TDP-43 (Armakola et al. 2012). Inhibition of the lariat debranching enzyme Dbr1 strongly suppressed TDP-43 cytotoxicity not only in yeast but also in neuroblastoma cells and rat primary cortical neurons. Dbr1 is involved in degrading RNA-lariats generated during splicing. These lariats accumulate in the absence of Dbr1 and were shown to bind and thereby detoxify TDP-43. The authors concluded that aggregating TDP-43 might bind to and sequester essential RNAs and RNA-binding proteins and that this might be a major cause for its cytotoxicity (Armakola et al. 2012).

FUS – like TDP43 – contributes to a wide range of neurodegenerative diseases (King et al. 2012). Very similar to TDP-43, the structural determinants of FUS aggregation and toxicity were determined in yeast. FUS also forms cytoplasmic, non-amyloid like aggregates, and RNA binding is essential for its cytotoxicity (Ju et al. 2011; Sun et al. 2011; Daigle et al. 2012). Deletion and over-expression screens resulted in a list of suppressors and enhancers of cytotoxicity (Ju et al. 2011; Sun et al. 2011), among which RNA-binding proteins and proteins involved in RNA metabolism were significantly enriched (Sun et al. 2011). However, the molecular mechanisms behind these genetic interactions have so far remained elusive.

FUS co-localizes with P body and stress granule components (Sun et al. 2011). Moreover, it co-aggregates and interacts with TDP-43 (Kryndushkin et al. 2011), suggesting that both proteins share very similar mechanisms of aggregation and toxicity. This hypothesis is supported by their structural and functional similarities (Lagier-Tourenne and Cleveland 2009) and by the fact that both proteins are components of cytoplasmic stress granules in mammalian cells (Colombrita et al. 2009; Bosco et al. 2010; Daigle et al. 2012). Furthermore, the finding that both proteins form non-amyloid like aggregates in yeast cells distinguishes them from other disease-relevant proteins, such as polyQ/htt, α-syn, and Aß, all of which form amyloids.

Concluding remarks and future perspectives

Understanding the fundamental mechanisms underlying neurodegeneration is essential for the development of effective disease-modifying therapies. Obtaining this knowledge has proven to be more challenging than initially anticipated.

Science evolves through incremental steps that, at times, enable large leaps into novel grounds. Many of these scientific breakthroughs were only possible because a few visionary scientists realized the importance of approaching complex biological problems with simple and accessible model systems. Yeast is a simple unicellular organism that has preserved many of the complex processes that occur in neuronal cells. Thus, yeast constitutes a powerful system to investigate the molecular underpinnings of pathologies, such as those associated with neurodegeneration; it will also allow us to identify novel therapeutic targets and drugs. Using state-of-the-art approaches that rely in gene-to-gene or genome-wide studies, yeast models strongly contributed to our current knowledge of the genes and pathways involved in the most prominent human neurodegenerative diseases, as demonstrated by their subsequent validation in higher eukaryotic models.

New sequencing techniques are allowing rapid and larger genome-wide association studies that will ultimately lead to the identification of new genetic risk factors. Yeast will be a very powerful tool to functionally analyze these newly identified variants.

Altogether, using budding yeast as a model for neurodegeneration will enable faster and more affordable discoveries that will, with time, result in effective therapeutic strategies.


This study was supported by Fundação para a Ciência e Tecnologia projects PTDC/SAU-NEU/105215/2008 and PTDC/BIA-BCM/117975/2010, and post-doctoral fellowship SFRH/BPD/35767/2007 (ST). TFO was supported by an EMBO Installation Grant and by the DFG Cluster of Excellence Center for Nanoscale Microscopy and Molecular Physiology of the Brain (CNMPB). SA is supported by the Max Planck Society and MCM by a fellowship from the Dresden International Graduate School for Biomedicine and Bioengineering (DIGS-BB). The authors have no conflicts of interest to declare.