Abstract
The yeast Saccharomyces cerevisiae is a widely used model organism for studying cell biology, metabolism, cell cycle and signal transduction. Many regulatory pathways are conserved between this yeast and humans, and it is therefore possible to study pathways that are involved in disease development in a model organism that is easy to manipulate and that allows for detailed molecular studies. Here, we briefly review pathways involved in lipid metabolism and its regulation, the regulatory network of general metabolic regulator Snf1 (and its human homologue AMPK) and the proteostasis network with its link to stress and cell death. All the mentioned pathways can be used as model systems for the study of homologous pathways in human cells and a failure in these pathways is directly linked to several human diseases such as the metabolic syndrome and neurodegeneration. We demonstrate how different yeast pathways are conserved in humans, and we discuss the possibilities of using the systems biology approach to study and compare the pathways of relevance with the objective to generate hypotheses and gain new insights.
Introduction
For many human diseases such as cancer, cardiovascular and neurodegenerative diseases, it remains challenging to find treatment strategies due to the complexity of the regulatory networks that underlie the disease onset and progression. For such complex diseases, there are often many different mechanisms that result in the same phenotype, and it is therefore difficult to identify the exact cause of the disease and design adequate and efficient treatment strategies. This calls for individualized treatment strategies, but a major obstacle to this is obtaining a diagnosis that allows the identification of the actual molecular events that are the root cause of the disease development in the individual patients. Even though there are many expectations for individualized medicines, for example for treatment of diseases associated with the metabolic syndrome, there is still a need for a far more detailed molecular understanding of complex regulatory networks involved in the development of complex diseases. In this context, model systems play an important role, as they allow for more detailed molecular studies than can be carried out either in the clinic or on isolated human cells. This holds particularly true for studies of regulatory biological networks, where it is not sufficient to study their elements separately; the systems as a whole need to be studied. Here, the systems biology approach provides concepts and tools for studies on entire networks and their regulations and properties, providing new insights. Clearly, reductionistic analysis of individual components (e.g. certain key kinases and/or transcription factors such as Snf1 or Tor1) has provided a wealth of information that is important for our understanding of cellular regulatory networks, but it is necessary to move to studies of complete networks to advance our understanding of whole pathways and their interactions (Vidal, 2009).
The yeast Saccharomyces cerevisiae has for decades been used as an industrially important cell factory (Nielsen & Jewett, 2008), and also as an excellent model organism for studying eukaryal cell biology. Although it is phylogenetically distant from human and mammalian cells, a number of key regulatory elements are highly conserved between yeast and human. Some straightforward advantages of this unicellular organism are that it is easy to cultivate fast, in large populations and in inexpensive media. It also comes both in haploid and diploid forms, which allows for both sexual crossing and clonal division (budding). Furthermore, there are a wide range of experimental and molecular tools that allow for genetic manipulations. This makes it easy to express heterologous genes either from an episomal plasmid or from a chromosomal integration, as well as making it fairly easy to insert, delete or mutate any sequence in the genome. The completion of the entire genome in 1996 represented the first available eukaryal genome (Goffeau, 1996), and a collection of single-deletion mutants is available for diploid cells and for nonessential genes as well as haploid cells (Winzeler, 1999). Furthermore, high-throughput data collected by functional genomic tools such as transcriptome analysis (Lashkari, 1997), proteome analysis (Zhu, 2001; Usaite, 2008), metabolome analysis (Villas-Boas, 2005; Jewett, 2006) and flux analysis (Sauer, 2006), interactome analysis (Uetz, 2000) (Lee, 2002; Harbison, 2004) and locasome analysis (Huh, 2003) are contributing to the set of valuable information, and yeast is probably the organism with the most comprehensive experimental dataset available. The large amount of data, both at the global level and at the molecular level, available for S. cerevisiae makes this yeast well-suited for a coordinated effort in systems biology (Mustacchi, 2006), where the objective is to obtain a quantitative description of cellular processes and, ultimately, global mapping of all key quantitative interactions within the cell.
The studies of human cardiovascular diseases have shown the involvement and the importance of energy metabolism, especially the metabolism of lipids and sugars. With the complexity of metabolism, where there is an interplay between a large number of biochemical reactions and Gibbs free energy stored in the form of ATP, redox potential stored in cofactors such as NADH and NADPH, and precursor metabolites used for the biosynthesis of both amino acids and lipids, it is valuable to have a simple model organism that allows for generation of generic models that can serve as a scaffold for further studies in humans or other mammalian systems. It is interesting to note how regulation of energy metabolism is highly conserved between yeast and human (Usaite, 2009; Zhang, 2010b), and there is increasing evidence that lipid metabolism is highly conserved between yeast and human (Nielsen, 2009).
Another area where yeast is a promising model organism is in studies of eukaryotic protein homeostasis or proteostasis. This involves controlling the concentration, three-dimensional structure, interactions and localization of individual proteins that make up a cell. Proteostasis is influenced by transcription, translation, post-translational modifications and the intrinsic chemistry of the protein, as well as by processes of folding/unfolding/refolding, protein damage, aggregation and proteolysis. All the mentioned processes are parts of numerous regulated networks of biological pathways that interact and compete. Some of the pathways that contribute to the proteostasis network are the heat shock response, oxidative stress response, unfolded protein response (UPR) and ubiquitin-proteasome system (UPS). Compromise of proteostasis can result in loss-of-function diseases, including the lysosomal storage diseases (Mu, 2008) and cystic fibrosis (Hutt, 2009), or gain-of-function degenerative diseases that involve protein aggregation, such as Huntington's disease (HD) (Williams & Paulson, 2008) and Alzheimer's disease (AD) (Jellinger, 2009).
Here, we review the prospect of using S. cerevisiae as a model organism for studying energy and lipid metabolism, how it is regulated by the conserved protein kinase Snf1 as well as regulation of proteostasis. These cellular processes interact through processes in the endoplasmic reticulum (ER), and hence there is a coupling between energy metabolism, lipid metabolism and protein metabolism, with the latter being further linked to apoptosis (Fig. 1). Through presentation of these different biological processes, we argue that there is a high degree of conservation in these pathways between yeast and human, and that yeast therefore has a large potential to serve as a model organism for the generation of generic eukaryal models for biological networks that can serve as a scaffold for more specialized studies in human cells. This could advance our understanding of the molecular mechanisms underlying complex diseases.
Overview of how energy metabolism, regulated by Snf1, lipid metabolism and protein metabolism, interact. Glucose is catabolized to acetyl-CoA (AcCoA) that can be oxidized to carbon dioxide through the TCA cycle and respiration. This results in production of Gibbs free energy in the form of ATP, but it also results in the formation of ROS. Glucose is also converted into amino acids that are polymerized into proteins, either in the cytosol or on the endoplasmic reticulum (ER), the latter for proteins that are to be secreted or sorted to membranes. Balancing of protein synthesis, protein folding and proteolysis is important to maintain proteostasis in the cell. Deregulation of proteostasis results in protein stress and damage that may cause cell death.
Lipid metabolism
Lipid metabolism is quite complex, involving a very large number of metabolic reactions, spanning different compartments in eukaryotic cells and resulting in the formation of a diverse group of chemical compounds. Lipids can be divided roughly into the following classes (Nielsen, 2009): (1) nonesterified fatty acids (NEFAs), or free fatty acids, which mainly serve as intermediates in lipid biosynthesis; (2) free sterols, which serve as structural components in membranes; (3) sterol esters that are formed from NEFAs and sterols and that serve as lipid storage compounds, mainly as lipid bodies; (4) triacylglycerides formed from glycerol and NEFAs, which serve as lipid storage, mainly stored in lipid bodies; (5) phospholipids formed from NEFAs, glycerol and an alcohol moiety, for example inositol, choline or ethanolamine, which serve as structural compounds in membranes; and (6) sphingolipids formed from palmitic acid, which are basically very long chain fatty acids that mainly act as regulatory compounds.
Despite the large chemical variety of lipids, they all have the same key carbon precursor, namely acetyl-CoA (see Fig. 2), and in both yeast and mammals, all initial steps of lipid biosynthesis occur in the cytosol (see Fig. 2). In mammals, acetyl-CoA in the cytosol is mainly derived from citrate through the reaction catalyzed by ATP:citrate-lyase. Saccharomyces cerevisiae does not possess an ATP:citrate lyase, but it relies on decarboxylation of pyruvate to acetaldehyde that is then converted further to acetate and acetyl-CoA (Fig. 2). Acetaldehyde is mainly converted to ethanol under fermentative conditions, but a sufficient amount of acetaldehyde is converted to acetyl-CoA, ensuring efficient lipid biosynthesis. Even though this pathway is central to fermentative metabolism in yeast, it also functions under fully respiratory conditions, providing a sufficient supply of acetyl-CoA for lipid biosynthesis.
A comparison of lipid biosynthesis and its regulation in mammals and yeast. AcCoA, acetyl-CoA; MalCoA, malonyl-CoA; AceAcCoA, acetoacetyl-CoA; FAs, fatty acids.
Lipid biosynthesis basically involves two branches from acetyl-CoA, one leading to sterols and the other to NEFAs serving as building blocks for biosynthesis of triacylglycerides, phospholipids, sterolesters and sphingolipids (Fig. 2). Figure 2 focuses on the similarities of the lipid metabolism in yeast and humans, but there are also some important differences: for example the carnitine shuttle plays an important role in mammals for the transport of acetyl-CoA between the cytosol and the mitochondria, whereas in yeast it is only active if carnitine is provided to the medium (Nielsen, 2009). Another important difference is that in mammals β-oxidation takes place in the mitochondria and is therefore closely related to respiration.
Transcriptional regulation of the sterol pathway is managed by transcriptional activators that are members of the sterol regulatory element-binding protein (SREBP) family. These transcription factors also upregulate the expression of several other genes encoding enzymes involved in the sterol pathway (Gibbons, 2003). In mammals, SREBP-2 is the transcription factor, and although yeast does not contain any homologues at the sequence level of SREBP-2, yeast Upc2 and Ecm22 were recently shown to have a function similar to SREBP-2 in mammals (Vik & Rine, 2001; Marie, 2008). SREBPs are bound to the ER membrane, but are activated and released from the ER membrane upon proteolytic cleavage by SREBP-cleavage-activating protein (SCAP; Gibbons, 2003). SCAP contains a multispanning ER membrane anchor that is referred to as a sterol-sensing domain (SSD), and increased cholesterol levels result in a conformational change of SCAP, which results in the binding of this protein to the so-called INSIG protein (SSD-interacting protein). Hence, SCAP is prevented from activating SREBP, resulting in a downregulation of genes encoding enzymes involved in sterol biosynthesis (Gibbons, 2003). Yeast has two homologues to INSIGs, Nsg1 and Nsg2, and these have been shown to have a functional role similar to the INSIG (Flury, 2005). Thus, even though the individual proteins do not share sequence homology, it seems that this regulatory system is conserved between yeast and human.
In the branch leading towards the NEFAs, acetyl-CoA is converted to malonyl-CoA by acetyl-CoA carboxylase (ACC and Acc1 in yeast). This is a committed step towards NEFAs and it is highly regulated. As discussed further below, Acc1 is inactivated by Snf1, and this regulation is important for controlling the flux into lipids. Malonyl-CoA serves as a precursor for the synthesis of NEFAs, which are formed by the large multifunctional fatty acid synthetases (FAS), two of which have been identified in yeast (Fas1 and Fas2). In mammals, the FAS have been shown to be regulated by SREBP, more specifically by SREBP-1, in a manner similar to enzymes of the sterol biosynthetic pathway. In yeast, the transcription factors Mga2 and Spt23 have been shown to regulate the expression of OLE1, which encodes for Δ9-desaturase, an enzyme that ensures insertion of a single double bond in unsaturated fatty acids. These transcription factors have been shown to be bound to the ER membrane and to be activated in a similar fashion as the SREBPs (Chellappa, 2001). Thus, in both yeast and mammals there seems to be a coordinated regulation of biosynthesis of NEFAs and sterols, which may be explained by the requirement for a coordinated biosynthesis of these two different types of lipids, both of which are needed for proper membrane function and for storage in the form of sterol esters. However, in the presence of high glucose levels, mammals have an additional regulatory system, namely insulin, which in the presence of high glucose concentration stimulates NEFA biosynthesis only, and hence allows for dedicated biosynthesis of storage lipids in the form of triacylglycerides (Gibbons, 2003). Based on transcriptional analysis of mouse hepatocytes that overexpressed SREBP-1 and SREBP-2, and a knockout of SCAP, all genes under SREBP regulation have been identified in mammals (Horton, 2003).
The integration of NEFAs into phospholipids and triacylglycerides and the degradation of NEFAs through β-oxidation are also conserved. In both yeast and mammals, two NEFAs are added to glycerol-3 phosphate to form phosphatidic acid (PA), which serves as a precursor for the biosynthesis of phospholipids by adding an alcohol to the phosphate group of PA. PA is also the precursor for triacylglycerides, as it can be converted to diacylglycerides, to which an additional NEFA is added to form triacylglycerides. The reactions and the regulation of β-oxidation are conserved between yeast and mammals. In mammals, oleic acid activates the peroxisome-proliferating activating receptor (PPAR), a transcription factor that leads to stimulation of peroxisome formation and hence increased β-oxidation in response to fatty diets. Yeast contains the transcription factor Pip2-Oaf1, which plays a similar role (Karpichev & Small, 1998), but these proteins share no homology with PPARs of mammals. In yeast, β-oxidation is also regulated by the transcription factor Adr1 (Hiltunen, 2003), which also regulates glycerol metabolism and ethanol oxidation. This transcription factor is probably conserved in the whole eukaryal kingdom as it is present in fungi (Salazar, 2009) and in humans (Das & Baez, 2008), but it is not known whether it serves a similar regulatory role in different eukaryotes.
Global regulation of metabolism by Snf1/AMPK
AMP-activated protein kinase (AMPK) in mammals and the Snf1 kinase in yeast share extensive conservation at the level of genes, structure of proteins and their downstream regulation, leading to the widespread use of S. cerevisiae as a model to understand the regulation and function of AMPK in mammals. Snf1/AMPK is a highly conserved serine/threonine kinase that signals low-energy status in the cell and invokes appropriate measures to inhibit energetically expensive processes such as biosynthesis of proteins and lipids while activating glucose catabolism (Hardie, 2007). Pharmacological activation of AMPK is used as a treatment of type-2 diabetes as it stimulates β-oxidation and inactivation of fatty acid biosynthesis (Hardie, 2007). Upon activation, AMPK also governs whole-body energy balance through cytokines such as leptin, ghrelin and adiponectin (Hardie, 2007).
AMPK was originally discovered as a kinase that is allosterically activated by AMP and inactivates lipid synthesis (Abdel-Halim & Porter, 1980). The yeast kinase, Snf1 (sucrose nonfermenting), was discovered in a screen for mutants that do not ferment sucrose (Celenza & Carlson, 1984). The sequencing of the AMPK genes led to the discovery that Snf1 is an orthologue of AMPK (Mitchelhill, 1994; Woods, 1994). Subsequent research on the activation of Snf1/AMPK revealed a structural similarity. Both Snf1 and AMPK function as heterotrimeric complexes with a catalytic subunit (α), regulatory subunit (γ) and a scaffold (β) that secures the complex. Both Snf1 and AMPK are activated by phosphorylation on a conserved threonine (T172 in mammals and T210 in yeast) on the catalytic subunit (Hawley, 1996; Stein, 2000; McCartney & Schmidt, 2001). The kinases that could phosphorylate Snf1 (Sak1, Elm1 and Tos3) were first discovered by whole-genome screening in S. cerevisiae and were demonstrated to phosphorylate AMPK in vitro (Hong, 2003). Although there are no clear orthologues of these three kinases in the human genome, the tumor suppressor kinase, LKB1 (Woods, 2003) and the Ca2+/calmodulin-dependent kinase kinase-β (Hawley, 2005) are the closest homologues and have been shown to activate AMPK in mammals. The evolutionary conservation between Snf1/AMPK is further evident from the cross-activation of AMPK and Snf1 by yeast and mammalian upstream kinases, respectively (Hong, 2003, 2005). Screening for mammalian kinases that could activate Snf1 revealed transforming growth factor-β as another candidate (Momcilovic, 2006).
The discovery of the upstream kinases was hoped to provide clear insight into the mechanism by which cellular energy status controls Snf1/AMPK activity. However, there are several lines of evidence that suggest that the upstream kinases are not regulated by energy status and therefore do not transmit the energy stress to Snf1 activation. For example, LKB1 is not specific to AMPK, but can activate several other kinases that respond to different stimuli (Lizcano, 2004). Another example is the absence of LKB1-mediated activation upon addition of AMP in vitro (Suter, 2006). Mutations in the regulatory subunit of Snf1/AMPK did not affect phosphorylation, but rather the dephosphorylation (Sanders, 2007). Finally, yeast cells lacking Elm1 kinase display elongated morphology. As there are no morphological differences between glucose-excess and glucose-limited conditions, Elm1 may not be sensitive to energy status. The current consensus hypothesis is that dephosphorylation is the likely regulator of Snf1/AMPK activity (Rubenstein, 2008). Whereas protein phosphatase 1 (Reg1-Glc7) deactivates Snf1 in yeast (Tu & Carlson, 1995; Sanz, 2000), protein phosphatase 2C deactivates AMPK in mammalian cells (Sanders, 2007). One potential deviation between the regulation of the activity of Snf1 and AMPK is in the mode of activation. Snf1 complex is not activated by AMP in vitro, although there is a clear correlation between the AMP: ATP ratio and its activation (Woods, 1994). Whatever the exact mechanism of activation, it is clear that high concentrations of ATP inhibit the activity of both kinases.
Upon activation, the downstream effects of Snf1/AMPK are geared to coordinate metabolism with developmental processes (Fig. 3). The mechanism of signaling involves harmonious regulation at the transcription and protein modification. Even though it was initially discovered as a regulator of glucose repression in yeast (Celenza & Carlson, 1984), the role of Snf1 as a master regulator of metabolism is evident from genome-scale studies (Usaite, 2009; Zaman, 2009). The genes (and proteins) involved in carbohydrate metabolism represent the largest group regulated by Snf1/AMPK. In yeast, this is mediated at the transcription level via the action of transcription activators such as Adr1, Cat8 and Sip4 (Hedges, 1995; Vincent & Carlson, 1998; Young, 2003), to induce ethanol uptake and gluconeogenesis and transcriptional repressors such as Mig1 (Treitel, 1998) to prevent the consumption of sugars other than glucose. In mammals, AMPK stimulates the expression of GLUT4 (primary glucose transporter) in skeletal muscle through the coordinated action of myocyte enhancer factor 2A and 2D, the activity of which is controlled by the peroxisome proliferator-activated receptor gamma co-activator 1 (PGC-1α) (Horman, 2006). GLUT4 is also activated by TBC1D1, in a manner similar to the activation of the hexose transporters by Grr1 (Ozcan & Johnston, 1995). In addition to extracellular sugars, storage carbohydrates, such as glycogen, also serve as an energy source. Glycogen production in yeast is controlled at the transcription level by Snf1 through Sip4 (Wiatrowski, 2004) and at the protein level by phosphorylating glycogen synthase on the conserved Ser7 (Hardy, 1994), whereas in mammals, only phosphorylation of glycogen synthase by AMPK is reported (Carling & Hardie, 1989).
Simplified representation of the role Snf1/AMPK plays in controlling various aspects of metabolism and cellular processes. The α, β and the γ subunits of Snf1/AMPK complex are shown in the center with the threonine (T172 in mammals or T210 in yeast on the α subunit) whose phosphorylation is required for the activity of the complex. The activation of Snf1/AMPK is shown in dotted arrows. The downstream effects of Snf1/AMPK described in the text are illustrated using solid arrows. The regulation that is prevalent only in mammals and not reported in yeast is indicated by pale arrows. The names of the proteins in mammals are shown in purple and their homologues in yeast, if present, are shown in orange. Green lines indicate activation and red lines indicate repression of the downstream function. The names of the different genes (proteins) are mentioned in the text.
The coordination of lipid metabolism with glucose availability and energy status is one of the most documented aspects of Snf1/AMPK regulation. As mentioned above, the first committed step in lipid synthesis is catalysis by ACC. Snf1/AMPK then phosphorylates this enzyme (Woods, 1994) to attain the delicate balance between fatty acid oxidation and synthesis, depending on the energy status of the cell. Fatty acid synthesis is also transcriptionally regulated by AMPK through the SREBP 1c, which controls the expression of many lipogenic genes. Analogous regulation in yeast is only reported for the Ino1 transcription factor (Shirra, 2001), although it is likely that many other transcription factors involved in fatty acid synthesis are also regulated by Snf1. Although there is no evidence for the regulation of sterol synthesis by Snf1 in yeast, cholesterol synthesis in mammalian cells is tightly regulated by AMPK phosphorylation of hydroxymethyl glutaryl CoA synthase (Clarke & Hardie, 1990). However, in a recent study on Snf1 it was found that Snf1 may regulate ergosterol biosynthesis through interaction with the transcription factors Upc2 or Ecm22 (see Fig. 1) (Zhang, 2010a).
Protein synthesis accounts for a large portion of energy consumption in the cell and it is therefore important to inhibit this synthesis during energy starvation. AMPK phosphorylates the ribosomal elongation factor (eEF2) to inhibit protein synthesis (Browne, 2004). Yeast translational elongation factors (Tef1 and Tef2) were identified as substrates of Snf1 by high-throughput analysis (Ptacek, 2005), although there is no report of molecular details on the regulation. An important link to controlling protein synthesis is through the target of rapamycin (TOR) kinases, which stimulate the initiation of protein synthesis by phosphorylation of multiple targets (Proud, 2004). Although it has been conclusively shown in mammals that AMPK inhibits mammalian TOR (Bolster, 2002), the hierarchy in the regulation between Snf1 and TOR in yeast is still a topic of intense debate (Orlova, 2006).
Many important targets for AMPK have been reported recently that demonstrate its global role in coordinating metabolism with developmental processes such as apoptosis, cell growth and proliferation. For example, AMPK induces the arrest of G1/S phase of the cell cycle with associated accumulation of tumor suppressor p53 (Okoshi, 2008). Pharmacological activation (using metformin) of AMPK also inhibited the growth of cancer cells (Imamura, 2001), clearly demonstrating the involvement of AMPK in tumor suppression. Recent evidence suggests that AMPK-mediated regulation involves NAD+ to control the activity of SIRT1 and PGC-1α (Canto, 2009), which are involved in aging and mitochondrial biogenesis, respectively.
The overlap in the regulation of developmental processes and metabolism between AMPK and Snf1 provides credibility that results obtained from studying the operation of Snf1 in yeast can be extrapolated to mammals and aid in the quest for new drugs and drug targets. Therefore, the yeast system can be used to address many unanswered, fundamental questions. A small sampling of these questions includes (1) how the kinase complex is regulated by nucleotide binding, (2) how nucleotide binding promotes phosphorylation, (3) how AMP inhibits dephosphorylation, (4) how ATP inhibits the complex when it binds to the complex in a similar manner as AMP, and (5) the role of the upstream kinases in relation to nutrient availability.
Proteostasis
Yeast contains a protein secretory pathway that is largely similar to the one functioning in human cells, allowing the study of the molecular basis for a range of human diseases, while maintaining all the advantages of a microorganism. The secretory pathway is responsible for folding and maturing all secreted and membrane-bound proteins, including receptors, nutrient transporters and vacuolar-resident proteins. In total, this represents a third of all protein produced (Tu & Weissman, 2004). The secretory pathway consists of the ER, Golgi and trans-Golgi network (TGN), and can include the cell membrane and vacuole/lysosome. It is notable that S. cerevisiae does not have the stacked cis-/medial-/trans-Golgi cisternae as do humans (Preuss, 1992). Instead, the three Golgi compartments can diffuse throughout the cell. However, other fungi, notably Pichia pastoris, do have similar stacked Golgi cisternae (Bevis, 2002). Through the ER, Golgi and TGN, many physical and chemical processes work in harmony to mature the protein. A malfunction at any step of this process can lead to chronic molecular stress and is responsible for several disease phenotypes observed in humans.
The ER is responsible for folding proteins, oxidization of the disulfide bonds, initial glycosylation and quality control of folding. Proteins that are not properly folded are retained in the ER, leading to an accumulation of protein that overwhelms the protein-folding capacity. These conditions activate the UPR, a transcriptional response conserved between yeast and humans (Patil & Walter, 2001). The UPR increases protein folding and oxidation capacity, alters the trans-Golgi vesicle trafficking and increases ER-associated degradation (ERAD), a translocation of misfolded proteins from the ER to the cytosol for degradation by the proteosome (Ron & Walter, 2007). In acute stress, the UPR can bring the secretory pathway back to equilibrium; however, under chronic conditions, the increased reactive oxidative species (ROS) and load on the proteosome result in apoptosis (Kaufman, 2002) (Fig. 4).
Improper protein folding in the ER leads to proteosome overload and ROS. ER stress is the molecular basis for many human diseases. Increased protein misfolding, which can be caused by a number of conditions in the cell, results in (a) increased load on the proteosome via ERAD and (b) increased ROS. Solid arrows represent the standard processing of ER-processed proteins. Grey arrows map an alternative route for misfolded protein that does not require the proteosome. ERAD, ER-associated degradation; ROS, reactive oxygen species; green circles, vesicles with properly folded protein cargo; red circles, vesicles with misfolded protein cargo.
A number of diseases result from ER protein-folding stress, leading to apoptosis in different tissues. β-Cell apoptosis, resulting in diabetic conditions, has been associated with chronic ER stress due to insulin mutants that do not fold properly (Oyadomari, 2002). Likewise, obesity-induced ER stress has been shown to decrease insulin sensitivity in a way that can be corrected by introducing chemical folding chaperones (Nakatani, 2005). Cystic fibrosis is a result of incorrect ER folding of the cystic fibrosis transmembrane conductance regulator, increasing the burden on the ERAD and proteosome (Gelman & Kopito, 2002). Prion-related neurodegeneration is also the result of a receptor that is not properly folded in the ER and is translocated to the cytosol, where the protein irreversibly aggregates (Hegde & Rane, 2003).
Efficient clearance or lack thereof, of terminally misfolded proteins is therefore significant in disease progression. The first characterization of the interplay between the UPR and ERAD was described in yeast (Friedlander, 2000; Travers, 2000). Friedlander. (2000) showed that knockouts of the ERAD pathway-associated ubiquitin-conjugating enzymes, Ubc1p and Ubc7p, resulted in upregulation of the UPR under nonstressed conditions. Likewise, under nonstressed conditions, basal ERAD activity was adequate to clear misfolded proteins in a UPR strain (Δire1). Travers. (2000) reached the same conclusion by monitoring the transcriptome under UPR-inducing conditions. As ERAD-associated proteosome saturation is at the root of several diseases, opportunities exist to modify the UPR to increase trafficking of misfolded proteins to the vacuole/lysosome or secretion outside the cell. In this way, ER stress could be alleviated without overburdening the proteosome. A yeast system would be excellent to study this mechanism.
ROS production is another cellular consequence of inefficient protein folding that can lead to apoptosis in different tissues. Disulfide bond formation results in the production of ROS by the transfer of electrons to the terminal acceptor by Ero1p. UPR activation in yeast has been shown to induce oxidative stress pathways (Kimata, 2006). As ER folding rate increases, as is necessary if proteins are lost to a terminally misfolded state, oxidation demand and ROS production also increase. ER-associated ROS production is likely to make up 25% of total ROS (Tu & Weissman, 2004) under unstressed conditions, and under ER-stressed conditions, considerably more ROS and oxidative damage could be caused. This ROS increase is known to activate apoptotic pathways but could be mitigated by preventing protein misfolding in the ER (Madeo, 1999).
Programmed cell death (PCD) is a natural mechanism by which multicellular organisms regulate tissue formation and homeostasis. Apoptosis is one of the pathways of PCD and its occurrence is extensive both in developing and adult animal tissues. Interestingly, in early development even the brain is extremely plastic: >50% of the neuron cells die in a regulated way in order to shape the neuronal networks in a correct way. In adult organisms, besides regulating the plasticity of many organs and tissues, apoptosis also serves as a quality control process: damaged, malformed, misplaced and infected cells, and cells that have been involved in the adaptive immune response, etc., are eliminated for the benefit of the whole organism (Jacobson, 1997). Sometimes apoptosis fails, leading, for example, to cancer and the promotion of infectious diseases such as HIV infection (Rathmell & Thompson, 2002), and sometimes it is triggered in cells that should rather have been kept alive, such as neurons and heart muscle cells.
Besides sculpting the developing brain, neuronal apoptosis has a potentially important role in neurodegenerative diseases. The principal molecular components of the apoptosis program in neurons include Apaf-1 (apoptotic protease-activating factor 1) and proteins of the Bcl-2 and caspase families and similar cell-death-signaling pathways might be activated in neurodegenerative diseases by abnormal protein structures (such as amyloid fibrils in AD). Even though physiological apoptosis in the developing brain and pathological apoptosis in the adult brain share similar molecular mechanisms in the effector phase, there are key differences in the mechanisms that trigger apoptosis (Yuan & Yankner, 2000). A common element in adult neurodegenerative disorders is the toxicity of abnormal protein structures or aggregates, such as in AD, Parkinson's disease (PD) and HD. Although these diseases have a different pathophysiology, the underlying common theme of abnormal protein structures groups them into so-called protein-misfolding disorders (Agorogiannis, 2004).
Many diseases appear to be caused by the misregulation of proteostasis, including both loss-of-function diseases (e.g. cystic fibrosis) and gain-of-toxic-function diseases (e.g. AD, HD and PD). Proteostasis is maintained by the proteostasis network, which comprises pathways that control protein synthesis, folding, trafficking, aggregation, disaggregation and degradation. Decreased ability of the proteostasis network to cope with misfolded proteins, due to aging and/or metabolic/environmental stress appears to trigger or promote proteostasis diseases (Powers, 2009). A causative link has been established between the formation of aggregates and neurodegeneration, but there are still numerous questions to be addressed, for example what exactly ‘pushes’ proteins to misfold, making them prone to aggregation? Another puzzle that remains is why protein aggregation, as such, harms the neuron and leads to apoptosis, giving the phenotype of disease. The cell has, of course, evolved protective mechanisms that most of the time ensure that the proteome is healthy and functional: chaperones ensure proper folding, and the UPS targets misfolded proteins (that fail to refold) for degradation thus preventing their aggregation. Protein aggregation and consequently neurodegeneration may occur because of age-related failure (or slow-down) of these surveillance systems, or as a result of increased production of misfolded proteins (due to stress or mutation) (Agorogiannis, 2004). In addition to the folding puzzle, more and more evidence shows that the traditional view on the cytotoxic role of fibrils, that are the end point of protein aggregation, should be revised, as soluble oligomers and protofibrils may also exert a pathogenic role. Surprisingly it even seems that fibrils may represent a nonfunctional state or inherent detoxifying mechanism that ‘packs’ potently harmful smaller particles into inert aggregates (Agorogiannis, 2004). Recently, more efforts have been put into interventions for modulating and repairing the proteome using proteostasis regulators (Balch, 2008). Interestingly, cells of multicellular organisms (mammalian, but also invertebrates) are not the only ones to undergo apoptosis. It has been shown that protists (Deponte, 2008) and unicellular organisms such as S. cerevisiae show apoptotic markers (Madeo, 2004, 2009). Not only does yeast show apoptotic markers, but in the last few years, it has become one of the preferred models for study of molecular mechanisms related to human health, such as the mechanisms of aging, apoptosis and protein folding (Willingham, 2003; Outeiro & Muchowski, 2004; Ocampo & Barrientos, 2008; Winderickx, 2008). The reason for the growing popularity of yeast as a model organism is based both on traditional aspects and on new discoveries. As mentioned previously, there are many benefits of working with an easy-to-handle model microorganism such as yeast but there are also other benefits of using yeast not only for medical and medicinal research (Mager & Winderickx, 2005; Petranovic & Nielsen, 2008) but also specifically for studies of protein folding and neurodegeneration.
So what makes yeast a good model for the study of complex human disorders such as misfolding diseases? Studies of basic cellular mechanisms such as DNA replication, recombination, transcription, translation, cell cycle progression and cell division, protein turnover, vesicular trafficking, signal transduction, the coordinated regulation of metabolic and cellular adaptations have made it clear that key cellular processes are well-conserved between yeast and higher eukaryotes, including humans. Indeed, approximately 31% of the yeast genes have a mammalian homologue and an additional 30% of yeast genes show domain similarity, and consequently about 30% of the genes known to be involved in human diseases may have a yeast orthologue (Botstein, 1997). Most initial experiments focused on complementation assays to elucidate the biological role of human proteins that have a yeast homologue, but lately more so-called humanized yeast systems have been used to study functional aspects of human proteins that do not have a yeast homologue, and several neurological disorders such as HD, PD and AD have been studied in yeast (Winderickx, 2008). It was also found that UPS malfunctioning is directly involved in some aspects of complex human diseases, such as neurodegeneration. The proteasome has been shown to serve an important role in S. cerevisiae as a ‘secondary antioxidant’ by clearing the cell of oxidized proteins (Chen, 2006). The major recognition motif for the 20S core proteasome are the hydrophobic surface patches that are formed by partial unfolding and exposure of hydrophobic amino acid residues during oxidation. Oxidized proteins appear to be relatively poor substrates for ubiquitination (Davies, 2001; Inai & Nishikimi, 2002; Grune, 2003; Shringarpure, 2003) and they appear first to aggregate and then to form covalent cross-links that make them highly resistant to proteolysis (Grune, 2003, 2004). The accumulation of extensively oxidized proteins may contribute to the formation of protein aggregates during diseases and the aging process.
The UPS plays a role in a variety of cellular functions, and deficiencies of that system can cause cellular dysfunction or death. Interestingly, ubiquitin itself can be a cause of UPS impairment: the aberrant ubiquitin UBB+1 mutant results from a dinucleotide deletion by molecular misreading, probably occurring during transcription (van Leeuwen, 2000), of the ubiquitin B gene UBB (van Leeuwen, 1998a, b). The deletion creates a frameshift resulting in the mutant ubiquitin called UBB+1. It has a 19-amino acid C-terminal extension, with the rest of the protein being identical to ubiquitin (van Leeuwen, 1998a). Due to the extension at the C-terminal, a glycine essential for ubiquitin's conjugation to substrates is missing. Thus, UBB+1 cannot be conjugated to targets, but is itself recognized by the UPS and ubiquitinated at lysines 29 and 48 and degraded by the proteasome (Lam, 2000; Lindsten, 2002). UBB+1 levels are very low under normal conditions and cells manage to degrade UBB+1 (Lindsten, 2002), but the protein tends to accumulate specifically in affected cells in several neurodegenerative diseases such as Alzheimer's and Huntington's (van Leeuwen, 1998a, b; de Pril, 2004), as well as in some non-neuronal pathologies characterized by involving misfolded proteins, such as liver pathologies (French, 2001; Wu, 2002) and sporadic inclusion body myositis muscle fibers (Fratta, 2004).
Although UBB+1 has been studied in several in vivo models, the exact molecular mechanism for UBB+1-mediated UPS dysfunction is still unclear and yeast models have been included in recent studies to study the molecular mechanisms in more detail (Tank & True, 2009). UBB+1 can accumulate to high levels in pathological states leading to induction of the heat shock response, resistance to oxidative stress, mitochondrial stress, cell cycle arrest and eventually apoptosis. PCD pathways (i.e. apoptosis, necrosis, pyroptosis and autophagic cell death) are themselves extensively studied topics that are of relevance not only for neurodegeneration but also in pathologies such as diabetes, myocardial infarction and cancer. Since the discovery of yeast apoptosis in 1997 (Madeo, 1997), yeast has gained momentum as a model for cell death studies (Carmona-Gutierrez, 2010). Studies of yeast apoptosis showing similarities to mammalian triggers and pathways have opened the way for a wider use of yeast in the study of apoptosis and how it is linked to protein homeostasis.
Conclusions and perspectives
It is clear that with the high degree of conservation between yeast and human cells, it is possible to use yeast as a model organism for studying these complex pathways, in particular their interactions. There is a close interaction between energy metabolism, lipid metabolism and protein metabolism, and it is therefore important to study these different pathways in the context of the whole system (Fig. 1). This is the essence of systems biology, where the functioning of complex pathways is studied quantitatively. Clearly, yeast as a model has some limitations, as not all key pathways on nutrient sensing and regulation are conserved, for example the glucose metabolism of yeast is not hormonally regulated, but it still offers a useful model, especially for studying the cross-talk between different processes that are normally studied separately.
Some of the contributions of systems biology to improving our understanding of the different pathways reviewed are as follows.
What are the targets of the protein kinases Snf1 and Tor1 and how do these two protein kinases (and their kinomes) interact? This question can be addressed through the use of phospho-proteomics combined, for example, with transcriptome analysis and integrative data analysis. This kind of analysis will make it possible to establish global interaction maps that can form the basis for more detailed kinetic modeling of how these two key kinases quantitatively interact. Furthermore, the dynamics of interaction will provide new insight into how different nutrient-sensing and regulation pathways are activated in response to changes in environmental conditions. This kind of information will be valuable for establishing similar models for human systems, where AMPK and mTOR are key targets for treatment of different types of diseases associated with metabolic disorders.
How is lipid metabolism regulated globally and how does it interact with the energy metabolism. Furthermore, how does lipid metabolism influence protein metabolism? The interaction of lipid and protein metabolism is quite complex, but it is possible to engineer yeast and perform very detailed molecular analysis, generating large experimental datasets that can be used for building a global model that might describe and predict the interactions of those two large pathways. Clearly, a fundamental understanding of the interface between lipid and protein metabolism has huge implications for identifying treatment strategies for diabetes type 2 and other metabolic disorders that cause changes in hormonal regulation.
How does the accumulation of protein aggregates affect cell aging and death? This can be studied through detailed modeling of protein synthesis, degradation and agglomeration. So far, yeast is the best model organism for collecting and obtaining sufficient experimental data to validate such detailed kinetic models. For this biological network as well, the complexity is great and new breakthroughs in our understanding can probably only be obtained through a combination of experimental work with different modeling approaches, for example construction of protein–protein interaction maps, Boolean type modeling to establish the signaling paths and detailed kinetic modeling of key steps in the pathway. Through model simulations it will be possible to evaluate different hypotheses, for example how the different components of certain pathways are connected and what the kinetics of their interaction is. It will also be possible to begin linking protein aggregation, cell aging and cell death to lipid and protein metabolism. This will obviously have large impacts on generating hypotheses for new treatment strategies for diabetes type 2, cancer and neurological diseases.
From these kinds of examples we are confident that the use of yeast systems biology will generate and test hypotheses and further advance our understanding of the molecular mechanisms underlying these important cellular processes. This understanding can serve as a scaffold for studies in mammalian and human cells that can lead to identification of new strategies for prevention of disease onset or progression and for development of treatments necessary for the development of individualized medicine.
Acknowledgements
Research work carried out by our research group in this field is sponsored by UNICELLSYS (http://www.unicellsys.eu), SYSINBIO (http://www.sysbio.se/sysinbio), Swedish Research Council, the Knut and Alice Wallenberg Foundation and the Chalmers Foundation.
References
Author notes
Editor: Claude Gaillardin
