Correspondence: Ian Macreadie, Sienna Cancer Diagnostics, c/- Bio21 Institute, University of Melbourne, 30 Flemington Road, Building 404, Vic. 3010, Australia. Tel.: +61 9347 0622; fax: +61 3 9347 4413; e-mail: firstname.lastname@example.org
Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by acute cognitive decline. The AD brain is featured by extracellular senile amyloid plaques, intraneuronal neurofibrillary tangles and extensive neuronal cell loss in specific regions of the brain associated with memory. The exact mechanism of neuronal cell dysfunction leading to the memory loss in AD is poorly understood. A number of studies have indicated that yeast is a suitable model system to decipher the molecular mechanisms involved in a variety of neurodegenerative disorders caused by pathological protein misfolding and deposition. Here, the knowledge from various studies that have utilized a yeast model to study the mechanism of pathways involved in AD pathogenesis is summarized.
It is estimated that over 24 million people worldwide suffer from dementia, with 4.6 million new diagnoses made every year, and it is estimated that by 2040, 80 million people will suffer from dementia. Alzheimer's disease (AD) is the most common progressive neurodegenerative disorder and is responsible for approximately 60% of dementia cases (Ferri et al., 2005). Short-term memory decline and visual-spatial disorder are one of the first clinical manifestations in AD. With disease progression, a decline in other cognitive domains manifests, including deficits in attention, language and spatial orientation, eventually resulting in complete inability to function independently in basic daily activities (Welsh et al., 1992; Perry & Hodges, 1999; Artero et al., 2003; Lambon Ralph et al., 2003). With the increasing age of the world population, the development of effective therapeutics and preventive strategies for AD is of high priority.
The onset of AD is age dependent including familial early-onset cases (5% of all AD cases) and late-onset cases. Genetic variability is a significant contributor to the risk of the disease, underlying an estimated heritability of around 70% (Avramopoulos, 2009). Using a combination of functional evidence, linkage analyses and sequence comparisons, the genes involved in familial, autosomal dominant AD (FAD) were discovered, leading to the identification of the three genes accounting for most FAD cases: APP (Goate et al., 1991), PSEN1 (Sherrington et al., 1995) and PSEN2 (Levy-Lahad et al., 1995). However, the majority (95%) of AD cases are late onset and do not follow Mendelian inheritance, in spite of showing significant heritability (Bergem et al., 1997; Gatz et al., 1997). APOE is the only significant genetic variant observed for late-onset AD (Strittmatter et al., 1993a,b). Apart from genetic variability, several other risk factors have been implicated for late-onset AD, but they have largely been inconsistent and fail to replicate in large representative populations.
APP/Aβ and tau
The AD brain is characterized by severe atrophy and by a profound loss of neurons and synapses, which is restricted to specific brain regions critical for learning and memory, including the temporal and parietal lobes, the cingulate gyrus and the frontal cortex. In addition to neuronal loss and atrophy, the AD brain has extracellular amyloid plaques associated with degenerating neuritis and inflammatory processes (Wisniewski & Terry, 1973; Price et al., 1991). The senile plaque is primarily composed of amyloid β (Aβ) peptide deposits, which form the plaque's nucleating core. Aβ is a natural cleavage product of amyloid precursor protein (APP) and is a key molecule associated with the disease as observed from gene association studies and functional evidences. Increased production of Aβ peptide is a feature of early-onset AD and can be caused by mutations observed in APP and γ-secretase complex (Jonghe et al., 2001). These mutations alter either the APP metabolism or the nature of secreted Aβ. Amyloid deposition and aggregation is primarily driven by increased local concentration of Aβ or due to the intrinsic aggregating property of the mutant Aβ form. Although no evidence of increased production of Aβ in late-onset AD has been reported till now, it has been suggested that reduced degradation/clearance of Aβ (Iwata et al., 2000; Miller et al., 2003; De Strooper, 2010) and reduced perivascular drainage (Weller et al., 1998) may lead to elevated levels of Aβ in the AD brain. However, the precise role of Aβ in neurodegeneration has been elusive, due to its highly complex and variable nature.
Neurofibrillary tangles (NFT)/inclusion bodies are a common feature in several neurodegenerative disorders characterized by dementia and/or movement dysfunction. In the AD brain, intracellular NFT is composed of protein tau assembled into paired–helical filaments, twisted ribbons or straight filaments (Lee et al., 2001). Protein tau is a microtubule-associated protein presumed to stabilize microtubules and to regulate axonal transport in brain. Phosphorylation of protein tau regulates its binding to microtubules (Goedert, 1996; Buee et al., 2000), and hyperphosphorylation is thought to contribute to the disease pathogenesis (Braak & Braak, 1991; Wang et al., 1995). Whether binding to microtubules and aggregation are different processes and precisely how they are modulated by phosphorylation and conformation remain to be determined.
Yeast: a versatile cellular model for neurodegenerative diseases
Yeast has been used as a model for several neurological disorders characterized by protein misfolding and aggregation (Winderickx et al., 2008; Braun et al., 2009). The main advantage of using yeast is its reduced complexity compared with the mammalian models. Although yeast may be devoid of a nervous system, most of the molecular signalling pathways and the proteins involved in human neurological diseases are homologous and functionally conserved (Bassett et al., 1996; Foury, 1997). Yeast also offers the advantage of powerful genetics and proteomics including a completely sequenced genome, comprehensive single gene deletion and overexpression libraries, determination of protein localization in vivo and tandem affinity purification-tagging approaches (Puig et al., 2001; Giaever, 2003; Suter et al., 2006). Yeast models have been extensively used to study downstream pathological events in AD (Table 1, Fig. 1). The utilization and benefits of yeast in studying APP/Aβ and tau proteins and developing high-throughput screens for AD chemopreventatives are discussed here.
Table 1. Yeast models developed for studying Alzheimer's disease pathology
The accumulation and aggregation of fragments generated from APP is central in AD pathogenesis. APP is a ubiquitously expressed transmembrane protein believed to be associated with neurotrophic signalling, cell adhesion and migration (Reinhard et al., 2005; Zheng & Koo, 2006). Cleavage by β-secretase generates a soluble N-terminal ectodomain (sAPPβ) and a membrane-associated fragment C99. Variable cleavages of C99 fragment by γ-secretase generate Aβ40/42 peptide fragments. Aβ42 is more amyloidogenic and prone to aggregation, and forms the nucleating core of the senile plaques observed in AD brains. The alternative non-amyloidogenic processing of APP by α-secretase cleaves in the middle of the Aβ region. However, the mechanism and regulation of APP processing by secretases is complex and poorly understood. Studies in yeast have focused on studying the role of secretases involved in the processing and generation of the Aβ peptide from human APP transfected in yeast.
In the first study, the expressed human APP in yeast resulted in the release of a soluble ectodomain into the media and the retention of a C-terminal APP fragment within the cell, which indicated the presence of a possible α-secretase activity similar to multicellular organisms (Zhang et al., 1994). Further analysis showed that APP processing by α-secretases occurred in sec1 and sec7 mutants, where transport to the vacuole or the cell surface was blocked, but not in sec17 or sec18 mutants, in which transport from the endoplasmic reticulum to the Golgi is blocked. The deletion of Yap3/Mkc7 or the transfection of an altered APP isoform (APP_K612Q) significantly reduced α-secretase processing, indicative of the possible involvement of Yap3p and Mkc7p in the processing of APP in the late Golgi (Zhang et al., 1997).
APP is subject to a variety of modifications and processing during its transport to the secretory pathway (Nunan & Small, 2000). O- and N-glycosylation of APP in the Golgi apparatus is known to be essential for its maturation and transport to the plasma membrane. However, human APP expressed in yeast was only O-mannosylated and not N-glycosylated. Further analysis using yeast pmt mutants incapable of mannosylation showed that Pmt4p was mainly responsible for APP mannosylation. Also, pmt4 deletion mutants showed decreased production of APP fragments, slowed maturation and aggregation of APP (Murakami-Sekimata et al., 2009).
Reconstituted human β- and γ-secretases in yeast
β-secretase is an aspartic acid protease involved in initial APP processing for the generation of the Aβ peptide by γ-secretase. Although no known mutations in the genes encoding the β-secretase complex are involved in AD, elevated levels of this enzyme are observed in sporadic AD cases (Li et al., 2004). In Saccharomyces cerevisiae, a cellular growth system was developed for studying the functional expression of β-secretase (Luthi et al., 2003). Human APP was fused to the C-terminus of the yeast enzyme invertase, essential for growth in sucrose media. The invertase–APP fusion protein was expressed as a type-I transmembrane protein in intracellular compartments of yeast cells lacking endogenous invertase. In these cells, co-expression of human BACE restored cell growth on selective plates upon cleavage of the invertase–APP fusion protein. This system provides the basis for a high-throughput screen for identifying β-secretase inhibitors. In order to establish a cell-based system for the positive selection of β-secretase inhibitors, the above model was reverse designed in order to confer on yeast cells the ability to grow only upon inhibition or reduction of β-secretase activity (Middendorp et al., 2004).
γ-secretase is a membrane protein complex that catalyses the regulated intramembrane cleavage of APP to release Aβ and the APP intracellular domain (Haass & Steiner, 2002). However, γ-secretase-induced APP proteolysis is poorly characterized. Saccharomyces cerevisiae cells lack endogenous γ-secretase activity. Therefore, cleavage of APP was studied by reconstituting human γ-secretase complex in yeast (Edbauer et al., 2003). An APP-based type I transmembrane reporter protein (C1-55-GAL4) was constructed as the substrate for reconstituted γ-secretase in yeast. The membrane-associated endoproteolysis was studied using the GAL4 reporter gene assay (Steiner et al., 1999). The reporter gene was co-expressed with the different components Presenilin (PS1wt), Nicastrin (Nct), anterior pharynx-1(Aph-1) and presenilin enhancer-2 (PEN2) of the γ-secretase complex. Also, a functionally inactive PS1D385A was co-expressed separately with the same set of proteins. γ-secretase activity was found to be strictly dependent on the co-expression of functionally active PS with Nct, PEN-2 and APH-1. Immunoprecipitation confirmed the formation of the γ-secretase complex in the membrane and the isolated complex was also found to be proteolytically active in vitro. Further work also showed that combined overexpression of PS1 and Nct, together with otherwise inactive Aph1-a G122D (Steiner et al., 2001), facilitated the rescue of γ-secretase activity (Edbauer et al., 2004). Similarly, Yagishita et al. (2008) reconstituted the γ-secretase complex using yeast microsomes, in order to obtain a tool for analysing this activity in vitro. In this study, a PEP4 knockout mutant was used to eliminate most endogenous protease activity and the microsomal extraction was performed by sucrose gradient centrifugation to prevent vacuolar contamination. A recent study (Futai et al., 2009) from the same group revealed 15 new PS1 mutants unrelated to familial AD, which are active in the absence of Nct. They contained an S438P mutation together with one missense mutation distributed through transmembrane and loop regions. The foremost advantage of using these yeast models to reconstitute secretase complexes for studying human APP processing is the ability to exclusively investigate and characterize the role of the different components involved, which would be very difficult to achieve in a mammalian system.
The accumulation and toxicity of APP/Aβ fragments in AD has been attributed to the dysfunction of the cellular proteasomal machinery. The production of fragment C99 by β-secretase cleavage is thought to form the rate-limiting step in the APP processing pathway leading to the production of the toxic Aβ peptide (Nunan et al., 2001). A recent study analysed processing of the C99 fragment in wild-type yeast and a mutant strain lacking proteasomal activity. Using Zip Tip immunocapture, a specific anti-Aβ antiserum (6E10), and matrix-assisted laser desorption ionization time-of-flight MS, one dozen APP-generated peptide fragments in wild-type yeast (PRE1 PRE2) and over three dozen unique fragments in proteasome mutant cells (pre1-1 pre2-1) expressing C99 were identified (Sparvero et al., 2007). In the pre1-1 pre2-1 cells, there was evidence of an intact Aβ peptide and a significant change in the population of fragments formed compared with the wild type. It was shown that defects in proteasome function allow C99 and C99-derived peptides to become substrates for other cellular enzymes. It is therefore implicated that this compensatory response leads to the generation of products that are more hydrophobic and aggregation prone and hence could be associated with disease progression in AD.
In contrast to S. cerevisiae cells, studies in Pichia pastoris transfected with human APP695 cDNA have shown evidence of possible β- and γ-secretase activities (Le Brocque et al., 1998). Notably, the β- and γ- secretase cleavage products of human APP, sAPPβ and a 4-kDa fragment with electrophoretic and immunoreactive properties identical to Aβ40/42, were detected in the media supernatant. However, the sites of γ-protease cleavage need to be further investigated.
In vivo Aβ oligomerization
Aβ is a hydrophobic protein cleavage product of APP. It is highly prone to aggregation under physiological conditions, which is determined by several interdependent environmental factors (Stine et al., 2003; Bharadwaj et al., 2009). Aβ aggregation and deposition is an important pathological feature of AD brains, but the mechanism of its oligomerization and implication to the disease is not completely understood. Yeast has been used to study the in vivo aggregation and dynamics of intracellularly expressed Aβ. The first study used a two-hybrid system in yeast to study the interaction of Aβ monomers by linking Aβ42 to the LexA DNA-binding domain (bait) and also the B42 transactivation domain (prey) (Hughes et al., 1996). Protein–protein interactions were measured by the expression of these fusion proteins in S. cerevisiae harbouring lacZ and LEU2 genes under the control of LexA-dependent operators. No interaction was observed when Aβ was replaced by Drosophila protein bicoid (LexA-bicoid) or when Aβ was modified at residues 19 and 20 (Aβ: F19T, F20T), indicating the specific nature of Aβ self-interaction. In a more recent work (Caine et al., 2007), natively expressed human Aβ42 using a copper-inducible system in yeast was undetectable: whether this is due to its rapid degradation still needs to be investigated. However, when the Aβ peptide was C- or N-terminally fused to a green fluorescent protein (GFP), it was readily detectable by immunoblotting. Fluorescence microscopy revealed that Aβ caused Aβ/GFP fusions to localize into punctuate patterns inside the cell (Fig. 2), also supported by immunoelectron microscopy. Also, immunodetection of cell extracts showed that Aβ/GFP fusions localized into the detergent-soluble membrane fraction in contrast to unfused GFP, which localizes in the soluble cytosolic fraction. Cells expressing GFP/Aβ fusions also showed reduced growth yield and an increased heat shock response, indicative of its toxic nature in yeast cells (Caine et al., 2007). This model has also been developed as an assay for screening compounds that affect green fluorescence in GFP/Aβ fusion-expressing yeast. Addition of folate was shown to increase the fluorescence of GFP/Aβ fusions in a dose-dependent manner in yeast mutants deficient in folate synthesis (Macreadie et al., 2008). Folate has been shown to have anti-oxidant properties (Nakano et al., 2001; Mayer et al., 2002; Coppola et al., 2005), and oxidative stress is an important pathological marker observed in AD brain (Butterfield et al., 2001). Also, dietary folate deficiency constitutes to be a risk factor for AD (Luchsinger et al., 2008; Kronenberg et al., 2009; Suchy et al., 2009). However, the underlying molecular mechanisms for the effects of folate on GFP/Aβ in yeast remain unknown.
Oligomerization of Aβ was also studied using Aβ/Sup35p fusions expressed in yeast (Bagriantsev & Liebman, 2006). The essential Sup35p contains three domains including the long N-terminal prion domain, a highly charged middle domain and the essential C-terminal RF domain (release factor) that performs termination of protein translation. The prion domain promotes the aggregation of Sup35p, thereby causing inactivation of its translation termination activity, and loss of the N-terminal prion domain restores the activity. Sup35p (NMRF) activity was assayed in vivo by examining the efficiency with which protein synthesis terminates at a premature stop codon (Chernoff et al., 1995). Strains carrying the ade1-14 nonsense allele and bearing fully active NMRF produce only a truncated (inactive) version of Ade1p, and are unable to grow on a synthetic medium lacking adenine. However, when Aβ42 was fused to MRF (Sup35p lacking N-terminal domain), the translational termination efficiency at the premature stop codon of ade1-14 was compromised and cells grew on adenine-deficient media. When a previously known oligomerizing-deficient Aβ (Hilbich et al., 1992) modified at positions 19, 20 and 31 (Aβm1MRF, Aβm2MRF) was fused to Sup35p, the cells failed to grow on media lacking adenine, indicating the restoration of translational termination. Using immunoblots, AβMRF was found to form sodium dodecyl sulphate-stable oligomers; however, Aβm1MRF and Aβm2MRF were largely monomeric. In yeast, Sup35p can form self-propagating infectious amyloid aggregates and guanidine is known to increase the size of the Sup35p aggregates (Kryndushkin et al., 2003) by inhibiting the ATPase activity of Hsp104. Notably, guanidine also increased the levels of AβMRF oligomers. However, this was shown not to be related to the inactivation of Hsp104. In addition, it was shown that Hsp104 prevents AβMRF from undergoing cellular degradation. In a different study, Aβ was replaced into the Sup35p prion domain (amino acids 5–112) to analyse its oligomerization (von der Haar et al., 2007). All these yeast systems have developed into novel methods to study Aβ oligomerization. The positions of amino acid residues important for intermolecular Aβ-interaction and aggregation have also been identified using these techniques.
Hence, these yeast systems are useful tools in the study of Aβ oligomerization, for screening compounds affecting the aggregation process and also to decipher the molecular pathways triggered by Aβ aggregation.
Extracellular Aβ toxicity
Increased levels of Aβ42 protein and deposition in the brain have been shown to correlate with the acute cognitive decline observed in AD (Lue et al., 1999; McLean et al., 1999). Studies have attributed Aβ42 aggregation and toxicity as one of the primary causes of AD-caused dementia (Dahlgren et al., 2002; Shankar et al., 2007, 2008; Hung et al., 2008; Steinerman et al., 2008). However, the exact isoform and mechanism of Aβ42 causing neuronal dysfunction is still unknown. Chemically synthesized Aβ42 peptide has been extensively used for studying its pathological role in AD. However, its use in mammalian cell lines has several limitations. Firstly, Aβ is a normal physiological product of APP detected in a normal ageing brain (Kelly & Balch, 2003). Although the major proportion of the Aβ42 is secreted, low amounts are also detected intracellularly (LaFerla et al., 2007). It is therefore very difficult to exclusively investigate the effects of extracellular secreted Aβ42. Second, Aβ42 has a variable nature under physiological conditions (Bharadwaj et al., 2009), and the cellular toxicity of Aβ42 has been attributed to diverse mechanisms and different forms have been shown to possess varied toxic mechanisms (Dahlgren et al., 2002; Lesne et al., 2006; Shankar et al., 2008). Therefore, identifying a particular Aβ42 toxic species has been a major problem in mammalian cell culture studies as they require media and serum to maintain viability. Recently, Bharadwaj et al. (2008) developed a new method to study Aβ42 cellular toxicity in water, achieved through the use of yeast. In this method, Aβ42 was stably maintained in either an oligomeric or a fibrillar form, thereby enabling to study the effects of different isoforms of Aβ42 specifically. Using a survival plating cell viability assay, oligomeric Aβ42 was found to have a concentration-dependent toxicity and significantly more toxicity than fibrils. This observation is consistent with Aβ42 studies in mammalian cell lines, suggesting soluble oligomeric Aβ42 as the main determinant of neuronal dysfunction and cognitive decline in AD. It will therefore be very interesting to study its mechanism of toxicity in yeast compared with mammalian cell models. Because inhibition of Aβ42 toxicity has been a target of interest for the development of novel AD chemotherapeutics (Bastianetto et al., 2008; Amijee et al., 2009; Gibbs et al., 2010), the yeast toxicity assay can become a potential candidate for screening inhibitors of Aβ42 toxicity.
Only a few studies have investigated the expression of human tau in yeast. Protein tau is expressed as six isoforms derived from a single gene by alternative mRNA splicing. These isoforms differ by one or two additional insertions in the N-terminal domain (0N, 1N, 2N isoforms) and by the number of microtubule-binding repeats (3R and 4R isoforms). The C-terminal insert is known to influence microtubule binding; however, no function for the N-terminal insertion has been defined as yet. Human tau-3R and tau-4R isoforms expressed in yeast were found to aggregate and acquire pathological phosphoepitopes (Vandebroek et al., 2005). Notably, protein kinases Mds1 and Pho85 (orthologues of mammalian tau-kinases, i.e. GSK-3β and cdk5, respectively) modulate the phosphorylation and aggregation of tau in yeast. Functional assays of recombinant tau isolated from yeast also revealed an inverse correlation between phosphorylation status and microtubule binding and stability. This study also included a clinical mutant tau-P301L, which was shown to cause destabilization and bundling by aggregating on the microtubule (Vandebroek et al., 2006). Although human tau proteins expressed in yeast showed pathological characteristics, no significant growth defect was observed.
Humanized yeast models are significant tools to examine protein functions or cellular pathways that mediate the secretion, aggregation and subsequent toxicity of proteins associated with human neurodegenerative disorders. They have provided noteworthy contributions towards a better understanding of the cellular pathways involved in APP processing and Aβ oligomerization in AD. As indicated before, yeast offers numerous advantages, with its relatively less complex and well-characterized biology, compared with mammalian models. However, yeast has natural limitations: they are unicellular and not functionally linked to other cells. They are primitive compared with neurons as they lack morphological structures such as dendrites, axons and synapses and associated specialized functions. Nevertheless, yeast models remain a very valuable tool in investigating the cellular mechanisms involved in AD and also develop high-throughput screens for chemopreventatives.