Correspondence: Monique Bolotin-Fukuhara, Institut de Génétique et Microbiologie, Université Paris Sud/XI, Bâtiment 400, 91405 Orsay Cedex, France. Tel.: +33 1 69 156 201; fax: +33 1 69 157 296; e-mail: firstname.lastname@example.org
Nowadays, mitochondrial diseases are recognized and studied with much attention and they cannot be considered anymore as ‘rare diseases’. Yeast has been an instrumental organism to understand the genetic and molecular aspects of the many roles of mitochondria within the cells. Thanks to the general conservation of mitochondrial genes and pathways between human and yeast, it can also be used to model some diseases. In this review, we focus on the most recent topics, exemplifying those for which yeast models have been especially valuable.
In recent years, mitochondrial diseases have been recognized and studied much more attentively than in previous times. Attention to the patients and to the necessity of genetic counselling have considerably improved, but treatments are still lacking and even mechanisms of the different diseases are mostly far from being understood. In this article, we will try to review some latest results and the models that are now contributing to the understanding of this heterogeneous group of, mostly, neuromuscular diseases.
Mitochondria are at the heart of cell function. For many years, emphasis has been placed on their primary function, i.e. providing the majority of the cellular energy in the form of ATP. Besides their role in ATP synthesis, mitochondria are also involved in essential cellular processes: β-oxidation (Bartlett & Eaton, 2004) and maturation of Fe-S proteins; without the mitochondrial structure, the cell is not viable (Lill & Kispal, 2000). Mitochondria generate reactive oxygen species (ROS) during respiration, 1–2% of the oxygen consumed during respiration being not completely reduced to water. Under hypoxia, the mitochondrial respiratory chain (RC) also produces nitric oxide, which can generate other reactive nitrogen species (RNS). High levels of ROS or RNS produce oxidative and nitrosative stress and low levels function in cellular signaling; they also have important implications for several diseases such as inflammation and cancer (Poyton et al., 2009). Mitochondria also mediate apoptosis via the mitochondrial permeability transition pore (review in Cheng et al., 2006; Pradelli et al., 2010) but probably also via fission/fusion proteins conserved in evolution (Cheng et al., 2008). The mitochondria networks associated with the endoplasmic reticulum structure are pivotal to the control of Ca2+ signalling and processes that depend upon them such as apoptosis (Giorgi et al., 2009). All these pathways are tightly intertwined and as a consequence mitochondrial dysfunctions are associated with many multifactorial diseases such as diabetes, cardiac diseases (review in Baines, 2010) or cancer (Diaz-Ruiz et al., 2009; Mayevsky, 2009; Weinberg & Chandel, 2009). They are also associated with neurodegenerative diseases such as Parkinson (reviewed in Lin et al., 2009), Alzheimer (Yan et al., 2006) and Huntington diseases (Abou-Sleiman et al., 2006; Quintanilla & Johnson, 2009).
The precise relation between mitochondrial functions and such multifactorial diseases are not yet elucidated, and there are still much debates about any causal relationships (Johannsen & Ravussin, 2009). They are not the topics of this review and some will be treated in other chapters of this issue. Here, we will concentrate on diseases due to mutations that directly affect mitochondrial genes or gene expression in a monogenic manner. Because mitochondrial biogenesis requires two genetic compartments, the chromosomes and the mitochondrial genome localized within the mitochondrial compartment, one expects these mutations to be of nuclear and mitochondrial origins.
Yeast (Saccharomyces cerevisiae) is a simple eukaryotic organism, with a complete genome sequence and, more importantly, the best annotated one, thanks to a coordinated international effort. Many genetic tools that have been created during these years, including the complete collection of gene deletions and a considerable number of mechanisms and pathways existing in higher eukaryotes, have been first studied and described in yeast. Moreover, about 40% of human genes whose mutations lead to diseases have an orthologue in yeast (Bassett et al., 1996) and genomic screens have been extended to mitochondrial diseases (Steinmetz et al., 2002). It is the reason for which it has been widely used to decipher molecular mechanisms underlying diseases in general. However, the study of mitochondrial functions and dysfunction is of special interest in yeast because it is in this organism that mitochondrial genetics and recombination have been discovered (Bolotin et al., 1971) and that nucleomitochondrial interactions have been studied in depth. There are also specific reasons for choosing S. cerevisiae for mitochondrial studies. This organism is petite-positive i.e. it can lose its mitochondrial genome (assimilated to the rho factor) provided it is supplied with a fermentative substrate. Consequently, all mutations of the mitochondrial genome can be studied without cell lethality. The frequency of homologous recombination is very high (1% recombination is considered to correspond to about 100 bp in the mitochondrial genome). It is genetically easy to transfer mitochondria from one nuclear genetic background to another via karyogamy. Moreover mitochondria can be transformed making in vitro mutation analysis possible. The richness and ease of yeast molecular genetics opens big opportunities, and even the major difference existing between human and yeast mitochondrial genomes, i.e. the predominant heteroplasmy of human and the homoplasmy of yeast, can result in the easier definition of the pathogenic mutations.
To review mitochondrial diseases may be a very difficult task because the definition might include different kinds of metabolic disorders or degenerative syndromes. Moreover, some important aspects have been extensively reviewed and the reader might refer to very good recent articles by Di Mauro (2010) for historical aspects, by Wallace et al. (2010) for bioenergetics, by Spinazzola & Zeviani (2009) for nucleo–mitochondrial intergenomic cross-talk; a previous review by Schwimmer et al. (2006) has already given an important outline of yeast models of mitochondrial diseases. In recent years, it has become clearer how important the conservation is between human and yeast and hence the possible use of yeast models. In the present article, we try therefore to exemplify the newest subjects and those for which yeast models have been especially valuable.
Yeast models of mitochondrial diseases due to mutations in nuclear genes
Saccharomyces cerevisiae has played an important role as a model system to understand the biochemistry and molecular biology of mammalian cells. The genetic tools available have also made S. cerevisiae a powerful system to identify gene–disease relationship. Furthermore, the possibility to duplicate as haploid or diploid makes this organism a flexible tool for assessing the dominant or recessive nature of a mutation. Yeast offers invaluable guidance for approaching human diseases-associated gene functions particularly concerning mitochondrial ones due to the ability of yeast to survive without a functional mitochondrial RC, provided a fermentable carbon source is made available. When the concentration of glucose is reduced, respiration-deficient yeast mutants grow slowly, forming small (petite) colonies. Petite, OXPHOS-deficient, yeast mutants carry mtDNA abnormalities in the form of multiple rearrangements (rho-minus petites) or as mtDNA-less strains (rho-zero petites).
Now, mitochondrial functions are very diverse and so are the nuclear origins of pathologies. In most cases, the understanding of the defect in yeast was essential to clarify the human pathology. Mitochondrial diseases caused by mutations in nuclear genes can be grouped into several categories. Some disorders are due to mutations in RC subunits or in proteins involved in the corresponding assembly. Others are due to defects of ATP synthase or of other metabolic enzyme complexes, such as defects in biosynthetic enzymes for lipids or cofactors or in enzymes of the Krebs cycle. Several others are due to deficiencies in mitochondrial protein synthesis or in the corresponding proofreading apparatus.
In this first part of the present review, we will discuss some recent research advances where yeast models have made essential contribution to clarify the relationship between a nuclear mutation and the corresponding mitochondrial defect.
In the first section, we will discuss mitochondrial diseases associated with mtDNA instability, in the second section, we discuss the defects in metabolic assembly factors, while in the third section we will treat the very recently investigated cases of nuclear mutations resulting in diseases due to defects in mitochondrial morphology and interactions.
Defects associated with mtDNA instability
The maintenance of the mtDNA depends on a variety of nuclear-encoded proteins. While the most obvious link between the nuclear DNA and the mtDNA is represented by the genes coding for the enzymes involved in mtDNA synthesis, in some other interesting cases, the relationship between the mutation and the defect is not evident and yeast models have been very important to enhance knowledge.
The synthesis of mtDNA requires the activity of the mitochondrial polymerase POLG (Van Goethem et al., 2001) and helicase TWINKLE (Moraes, 2001; Spelbrink et al., 2001). In addition, a correct balance of the mitochondrial dNTP pool is essential for the maintenance of mtDNA copy number. In fact, mutations can lead to mtDNA depletion such as mutations in the human genes thymidine kinase 2 (TK2) (Saada et al., 2001) and deoxyguanosine kinase (dGK) (Mandel et al., 2001), encoding enzymes involved in mitochondrial dNTP recycling, in p53-regulated ribonucleotide reductase (Bourdon et al., 2007), involved in the de novo synthesis of dNTPs, and in thymidine phosphorylase (TP) (Nishino et al., 1999), involved in the catabolism of thymidine.
Defects associated with mtDNA instability – yeast models of POLG
The first observation that a disease associated with multiple deletions of mtDNA is caused by a mutation in a nuclear gene (Zeviani et al., 1989) dates back to 1989. In 2001, the first description of POLG mutations associated with mtDNA deletions (Van Goethem et al., 2001) appeared. So far, >150 mutations in POLG have been reported to be associated with various pathologies (for a complete list, see the database http://tools.niehs.nih.gov/polg/). Thanks to the conservation between the human POLG and the yeast orthologue MIP1 (Lecrenier & Foury, 2000; Foury & Kucej, 2001), it was possible to study in yeast the effect of POLG mutations found in patients. Yeast has proved a suitable model to validate the significance of new pathogenic POLG mutations. Generally speaking, most of MIP1 mutations equivalent to the POLG mutations led to an increased mutability of mtDNA, deletion, depletion, or point mutations. These results may provide relevant information on the molecular mechanisms of the replication defect underlying the disease. Most patients are compound heterozygous, or carry potentially relevant single-nucleotide polymorphisms (SNPs), which lead to the question as to whether these changes can determine additive or cooperative effects on the pathomechanisms, ultimately leading to the different clinical phenotypes. Yeast is a useful tool to answer these questions, as the effects of mutations either singly or in combination, both in cis and in trans, can be studied (Baruffini et al., 2010).
Some examples of studies in yeast that have contributed to the understanding of the effects of mutations and of the possible molecular mechanisms underlying the disease have been described here. The introduction of POLG mutations associated with progressive external ophthalmoplegia (PEO), L304R, 467T, G923D, R943H, Y955C, A957S, into yeast MIP1 caused increased mtDNA mutability, increased nuclear mutation rates and increased oxidative stress. Mutations in the polymerase domain caused the most severe phenotype, whereas the mutation in the exonuclease domain showed a less severe phenotype (Stuart et al., 2006). Yeast models have also been constructed of several human mutations alone or in combination, among which G848S-E1143G and H932Y-G1051R associated in trans, and A889T-E1143G associated in cis (Baruffini et al., 2007; Spinazzola et al., 2009; Stricker et al., 2009). The mutation A889T is slightly dominant in yeast, in agreement with the hypothesis that the gene penetrance is influenced by other genetic or environmental factors (Hisama et al., 2005). Furthermore, in vivo studies in yeast have shown that the mutation E1143G is not a neutral SNP (Baruffini et al., 2007), but rather acts as a cis-modulator of Mip1 activity, in agreement with the observation that in vitro E1143G was somewhat detrimental to protein stability (Chan et al., 2006). Experiments in yeast demonstrated that, in contrast to what was suggested previously (Graziewicz et al., 2004), mtDNA deletions are not related to accumulation of point mutations of mtDNA (Baruffini et al., 2006; Stumpf et al., 2010).
In yeast, the genetic and chemical rescue of mutant phenotype induced by POLG pathological mutations was also described. It was found that the mtDNA instability caused by mutations in MIP1 is reduced by increasing the concentration of the dNTP pool through an increase in RNR1 expression (Lecrenier & Foury, 1995; Zhao et al., 1998; Chabes et al., 2003; Baruffini et al., 2006; Stumpf et al., 2010) or deleting SML1, encoding Rnr1 repressor (Zhao et al., 1998; Baruffini et al., 2006). Whether the rescue due to an increased pool of dNTPs occurs preventing the stall of POLG or improving the efficiency of mtDNA repair, remains to be determined. Chemical rescue experiments demonstrated that the instability of mtDNA due to the POLG mutation Y955C depends, at least in part, on ROS damage (Baruffini et al., 2006). Finally, it is worth mentioning that there are examples where the effect of mutation in yeast suggests the existence of another mutation, in addition to that already identified, as a cause of disease (E. Baruffini, pers. commun.).
Defects associated with mtDNA instability – yeast models of ANT1
ANT1 is the gene encoding the muscle-heart-specific isoform of the mitochondrial adenine nucleotide translocator (Ant). ANT1 mutations are responsible for PEO. The ANT gene, primarily involved in ATP/ADP exchange across the inner mitochondrial membrane, is highly conserved in all eukaryotes, including S. cerevisiae. This allowed to study in yeast the effect of the mutations identified in patients. An A114P missense mutation in the human Ant1 protein was found to be associated with autosomal dominant PEO (adPEO) (Kaukonen et al., 2000). A128P mutation of the S. cerevisiae Aac2 protein, equivalent to A114P in human Ant1, causes a decrease in respiratory growth (Kaukonen et al., 2000). In a study aimed at evaluating the effect of this Ant mutant, it was found that it results in depolarization, structural swelling and disintegration of mitochondria, and it was hypothesized that the formation of an unregulated channel, rather than a defect in ATP/ADP exchange, was a direct pathogenic factor in human adPEO (Chen, 2002).
Four yeast models were constructed, each carrying a missense mutation identified in the ANT1 gene of adPEO patients (Fontanesi et al., 2004). All of them displayed a marked decrease in respiratory growth and a concurrent reduction of the amount of mitochondrial cytochromes, cytochrome oxidase activity and cellular respiration (Fontanesi et al., 2004). To evaluate whether the mutations were dominant in yeast, as in humans, the aac2 mutant alleles were also inserted in combination with the endogenous wild-type AAC2 gene. The mutations behaved as dominant for reduction in cytochrome content and increased mtDNA instability phenotypes indicating that S. cerevisiae is a suitable in vivo model to study the pathogenicity of the human ANT1 mutations. In yeast models, the efficiency of ATP and ADP transport was variably affected by the different AAC2 mutations. However, the observation that mutants retain the basic features of ANT/AAC proteins, together with a significant level of mitochondrial membrane potential (Fontanesi et al., 2004; Galassi et al., 2008), strongly argues against the hypothesis that the primary pathogenic role of ANT1 mutations associated with adPEO is to cause the opening of an unregulated channel followed by structural disintegration of mitochondria (Chen, 2002). It should also be noted that the uncoupling effects reported by Chen were observed in a strain overexpressing the aac2A128P mutant allele.
A yeast model carrying the mutation equivalent to the first (and so far the only) known recessive mutation in the ANT1 gene (A123D mutation) (Spinazzola & Zeviani, 2009), displayed a complete loss of transport activity, and a severe OXPHOS phenotype that was largely rescued by exposure to ROS scavengers, suggesting that increased redox stress is involved in the pathogenesis of the disease and that anti-ROS therapy may be beneficial to patients; the mutation was recessive in yeast as in human (Palmieri et al., 2005). It was reported elsewhere that by increasing the gene dosage of the A123D mutant allele, the mutation resulted in a dominant phenotype (Wang et al., 2008). However, this is neither a natural condition for yeast nor a representative of the human condition where the gene dosage is a constant.
To overcome the problem posed by a mutation mapping in domains not conserved between human and yeast, we took advantage of a yAAC2/hANT1 chimeric construction as a template to introduce pathogenic hANT1 mutation. Application to the case of the D104G mutation indicated that the chimeric construction could be a tool for studying pathogenic mutations in yeast (Lodi et al., 2006).
The relationship between ANT1 and mtDNA stability is not obvious. A possible explanation is that, in addition to its function on ATP/ADP translocation, the Aac2 protein plays a ‘structural’ role that contributes to maintain the integrity of respiratory complexes in the inner membrane of mitochondria.
mtDNA maintenance (MPV17-SYM1)
Another intriguing example of a gene necessary for the maintenance of mtDNA is human MPV17, mutation of which leads to a peculiar form of hepatocerebral mtDNA depletion syndrome (MDS). Even though Mpv17 mutations are one of the causes of MDS in humans (Poulton et al., 2009) and the discovery of this protein has been reported >20 years ago, its function is not yet understood. Originally considered as a peroxisomal membrane protein (Weiher et al., 1990; Zwacka et al., 1994), it was later demonstrated that Mpv17 is localized to the inner mitochondrial membrane (Spinazzola et al., 2006), as also previously demonstrated for the yeast orthologue Sym1 (Trott & Morano, 2004).
With the aim of clarifying the role of MPV17 in MDS, a mouse model Mpv17 −/− was studied (Viscomi et al., 2009). This mouse presents, like humans, severe mtDNA depletion in liver, but, unlike human, only a modest reduction of RC enzyme activities, probably due to an increased transcription of mtDNA. The yeast orthologue SYM1 was identified as a heat shock protein with a role in metabolism and/or tolerance to ethanol (Trott & Morano, 2004).
To validate the significance of pathogenic human mutations, these were introduced into a SYM1-defective yeast strain (sym1D). The mammalian gene, MPV17, itself can complement the phenotype of sym1D mutant, thanks to the high degree of conservation from yeast to human. The pathological mutations of human MPV17 are deleterious in yeast: they cause an OXPHOS-negative phenotype and result in an increase of mtDNA mutability (Spinazzola et al., 2006). Further studies in yeast have shed some light on the function of Sym1. The sym1 mutant mitochondria are morphologically abnormal, with flattened mitochondrial cristae and accumulation of electron-dense particles, as observed for Mpv17−/− mouse mitochondria (Viscomi et al., 2009), suggesting a role for Mpv17/Sym1 in the structural preservation of the inner mitochondrial membrane. This defect is not a consequence of the mtDNA instability because it has been observed under cultural conditions where no defect of mtDNA was observed, indicating that the morphogenetic effects of Sym1 are likely to precede and possibly determine its effects on mtDNA stability.
The phenotypes of double mutants (cit1 sym1, cit2 sym1) and the nature of multicopy suppressors (ODC1, YMC1) suggest for sym1 null mutant a defect in Krebs cycle confirmed by an enzymatic analysis that clearly indicates a heavy reduction of succinate dehydrogenase (SDH) activity. Accordingly, sym1D displays a significant reduction in the amount of glycogen that is dependent on gluconeogenesis, which is in turn regulated by the anaplerotic flux of tricarboxylic acid (TCA) intermediates from mitochondria to the cytosol (Dallabona et al., 2010). Interestingly, patients with Mpv17 mutations suffer from drastic, often fatal, hypoglycaemic crises, which are likely due to glycogen shortage in the liver (Spinazzola et al., 2006; Parini et al., 2009).
Blue-native gel electrophoresis immunovisualization clearly demonstrated that Sym1 is part of a high-molecular-weight complex (>650 kDa) (Dallabona et al., 2010). While further work is necessary to identify the primary role of Sym1, including the molecular dissection and characterization of the Sym1-containing protein complex, these results indicate that Sym1 is involved in the structural and functional stability of the inner mitochondrial membrane, thus controlling crucial mechanisms related to this compartment, including the activity of RC complexes, the morphology of mitochondria and the maintenance of mtDNA.
Defects in the biosynthesis of SDH assembly factors
SDH (or complex II or cII) is composed of four subunits (SDHA-D in humans, SDH1-4 in yeast), all encoded by nuclear genes. Despite the extensive knowledge on structural and catalytic properties of the complex, two assembly factors specific for the SDH have been only recently identified: SHDAF1 and SDHAF2 (YDR379C-A and SDH5 in yeast). The identification and characterization of SDH assembly factors is an example of how yeast can be used as a model system in ‘two ways’: from yeast to human and back.
The SDHAF1 gene has been identified in humans as linked to infantile leukoencephalopathy (Ghezzi et al., 2009). A yeast strain deleted in the SDHAF1 orthologue YDR379C-A (ydr379c-aΔ) was OXPHOS incompetent due to a severe and specific reduction of SDH activity. Transformation with YDR379C-A variants corresponding to the human mutant alleles did not restore OXPHOS growth of the ydr379c-aΔ strain, demonstrating that these mutations are really the cause of the disease. Experiments performed in human tissues and in yeast showed a marked reduction of cII holoenzyme and a Km value for succinate similar in wild type and in the null mutants, suggesting that defective SDH activity was caused by a reduced number of enzyme units rather than by qualitative alterations of complex II (Ghezzi et al., 2009). Although there are other examples of low cII content and activity associated with mutations in mitochondrial chaperonins such as yeast Tcm62 (Klanner et al., 2000), or proteins involved in Fe-S biosynthesis such as human and yeast frataxin or IscU (Rouault & Tong, 2008), SDHAF1 is the first protein identified with a specific role in cII assembly, as other Fe-S-dependent activities were normal in SDHAF1-defective organisms (Ghezzi et al., 2009).
The function of another assembly factor, SDH5 (EMI5/YOL071W) was discovered in yeast. The sdh5Δ mutant is characterized by defective oxidative growth, impaired respiratory activity, decreased chronological life-span, hypersensitivity to H2O2, lack of SDH activity and loss of complex II. Sdh5 interacts with Sdh1, but is not a stable component of complex II, and it is both necessary and sufficient for Sdh1 flavination and thus for SDH activity (Hao et al., 2009). After the identification of the yeast Sdh5 and the assignment of the function, it was demonstrated that mutations in the human orthologue gene C11 or f79 (renamed hSDH5) are responsible for paraganglioma. Thus, an uncharacterized mitochondrial protein in yeast was shown to play a crucial role in the biogenesis and function of complex II and its mutational inactivation was found to confer susceptibility to cancer in humans (Hao et al., 2009).
Nuclear mutations affecting the mitochondrial fusion and fission system
A new class of mitochondrial diseases (which has recently been the object of considerable attention) is caused by nuclear mutations affecting mitochondrial morphology (Detmer & Chan, 2007).
Actually, the refinement of microscopic techniques has revealed a completely new viewpoint on these organelles: their shape, their connections, their mobility in the cell and the corresponding defects.
The mitochondrial network is regulated by two dynamically opposed processes, fusion and fission of the mitochondrial membranes. Most proteins mediating yeast mitochondrial fusion and fission are conserved in flies, worms, plants, mice and humans, indicating that the fundamental mechanisms controlling mitochondrial morphology have been maintained during evolution. This conservation places S. cerevisiae as a good model for studying molecular mechanisms of mitochondrial dynamics.
Studies of yeast mutants, based on defects in mitochondrial shape, identify three major morphology pathways: fusion, fission and tubulation. When fusion is blocked, mitochondria fragment due to ongoing fission. When fission is blocked, mitochondria form interconnected nets due to ongoing fusion. When the tubulation pathway is disrupted, mitochondria are converted into large spheres. While the fusion and fission proteins have human homologues, proteins of the tubulation apparatus are not conserved in evolution. Indeed, the first proteins required for mitochondrial distribution and morphology (McConnell et al., 1990), and tubulation, (Sogo & Yaffe, 1994) were identified in the early 1990s by genetic screens in S. cerevisiae.
For many years, the fission/fusion equilibrium has been investigated mainly in yeast, but it was soon evident that the proteins involved in this key aspect of mitochondrial function are highly conserved in evolution. Accordingly, mammalian cells exhibit the same fission/fusion events observed in yeast and several human, mainly neurodegenerative, pathologies derive from alteration in this equilibrium.
Mitochondrial fission requires a family of conserved, dynamin-related GTPases, called Dnm1 in yeast (Bleazard et al., 1999) and Drp1 in humans (Smirnova et al., 2001). Dnm1 is a cytoplasmic protein and it can assemble into punctate structures on mitochondria in the sites of future mitochondrial fission. The evolutionarily conserved integral membrane proteins, Fis1 in yeast (Mozdy et al., 2000) and hFis1 in humans (James et al., 2003; Yoon et al., 2003), play an essential role in mitochondrial fission. Fis1 is a tail-anchored outer mitochondrial membrane protein with its N-terminal domain exposed to the cytoplasm. Fis1 recruits the Dnm1 to promote mitochondrial fission.
Figure 1 shows a cartoon of inner and outer mitochondrial membranes with localized yeast proteins involved in fusion and fission apparatuses.
Loss of fusion results in mitochondrial fragmentation due to ongoing fission events. Fragmented mitochondria eventually lose mtDNA by unknown mechanisms. Yeast cells defective in fusion cannot, therefore, grow on nonfermentable media, which require respiration for energy production (Chen et al., 2010). Mutations in the nuclear-encoded proteins involved in the above apparatuses result in important alterations in mitochondrial morphology and this has a relevant influence on the mitochondrial movements.
Fusion is important to protect mitochondrial function by enabling protein complementation, mtDNA repair and equal distribution of metabolites. Fission instead helps to isolate damaged mitochondria and promotes autophagy (Chen & Chan, 2009).
Mitochondrial mobility has an important function in yeast: suffice it to say that the presence of mitochondrial structures is necessary in the bud to ensure viability (García-Rodríguez et al., 2009; T. Rinaldi, pers. commun.).
On the other hand, in human neurons, the mitochondrial movement for long distances towards the synapses is essential for neural functions and, if defective, leads to neurodegeneration (Knott et al., 2008).
In Table 1, we report a list of human genes, with orthologues in yeast, involved in mitochondrial morphology, which produce, when mutated, neurodegenerative diseases. In a situation in which mitochondrial movements are essential for neuron physiology, mitochondria having an altered shape should be eliminated; therefore, a complex mechanism exists connecting mitochondrial dynamics, fusion/fission equilibrium, movement and mitophagy.
Table 1. Yeast genes involved in mitochondrial morphology and their diseases-associated human orthologues
CMT4A (Charcot-Marie-Tooth 4A) and CMT2K (a dominant form)
Autosomal recessive peripheral neuropathy
358 aa, 29 mutations
Several neurodegenerative diseases caused by mutations in the fission–fusion genes have been described (Liesa et al., 2009). Mutations in MFN2 gene cause Charcot–Marie–Tooth (CMT) subtype 2A, a peripheral neuropathy that is characterized by muscle weakness and axonal degradation of sensory and motor neurons (Kijima et al., 2005; Züchner et al., 2004). Mutations in OPA1 cause the most common form of optic atrophy: the autosomal dominant optic atrophy (ADOA). Patients with ADOA exhibit progressive loss of vision and degeneration of the optic nerve and retinal ganglion cells (Carelli et al., 2004). In addition, some mutations in the OPA1 GTPase domain cause ‘ADOA-plus’ phenotypes that are also characterized by deafness, sensory-motor neuropathy and muscle-movement disorders (Amati-Bonneau et al., 2008; Hudson et al., 2008).
Mutations in fission apparatus seem to be more severe. Recently, a dominant-negative mutation was reported in the human DRP1 gene (Waterham et al., 2007). The patient exhibited elongated mitochondria around the nucleus, a feature that is characteristic of impaired mitochondrial fission. Unfortunately, the patient died 37 days after birth. The neurological features of this patient consisted of severe neonatal hypotonia, abnormal brain development and abnormal gyral pattern. The early onset suggests that mutations in fission proteins are much more severe than those of the fusion mutations. Mutations in the human GDAP1 gene, with no homologue in yeast, cause CMT4A, another subtype of CMT syndrome (Baxter et al., 2002; Cuesta et al., 2002). The homozygous GDAP1 mutation causes early onset and more severe progression. GDAP1 protein localizes in the outer membrane and seems to participate in DRP1-dependent mitochondrial fission (Niemann et al., 2005).
To the list of diseases certainly deriving from mutations in proteins involved, at the level of mitochondrial membranes, in the fusion/fission equilibrium, we should probably add, at present, the Huntington disease, which heavily involves mitochondria, and is monogenic and not polygenic like Parkinson and Alzheimer diseases. This fatal autosomic dominant disease is related to the presence in the gene of an increased number of CAG triplets (n>35–40, which is the normal range). In these cases, the mutated Huntingtin, mtHtt, produces a fatal neuropathy and dementia in adulthood. The molecular mechanism of this disease is debated, but severe mitochondrial damage is present. If it is confirmed that Htt has a function in regulating the relationships among cell organelles, mitochondrial defects might be generated by malfunctioning of the fusion/fission equilibrium (Reddy et al., 2009).
This group of mitochondrial defects is only now starting to be studied and clarified in detail, but it is already clear that both inner and outer mitochondrial membranes are very dynamic compartments in which some proteins are stably localized, but others participate only when needed to determine the necessary structure. These phenomena are certainly present in all eukaryotic cells, but S. cerevisiae has been the model in which this complex dynamics has been studied (Khurana & Lindquist, 2010).
Human mitochondrial diseases due to mutations in the mitochondrial genome: a short overview
The human mitochondrial genome codes for 13 proteins of the RC: seven for complex I (ND1, ND2, ND3, NND4, ND4L, ND5 and ND6), cyt b from complex III, three for complex IV (CO1, COII, COIIII) and two (ATP6 and ATP8) for complex V. In addition, the human mitochondrial genome also encodes 22 tRNAs and two rRNAs, which are different from the cytoplasmic ones. Apart for complex I which is absent, all OXPHOS proteins as well as tRNAs and rRNAs have their equivalent in S. cerevisiae. Mutation rate in mitochondria is high (around 3 × 10−5 to 3 × 10−6 as compared with 2.5 × 10−8 for the nuclear genome; Nachman et al., 1996; Schriner et al., 2000). In theory, one would expect pathologic mutations to be randomly distributed in all the coding sequences, but this is not so. The list of potential (reported) and confirmed mutations is regularly updated on the MITOMAP website (http://www.mitomap.org). To this day, 27 confirmed tRNA mutations have been described, while 25 mutations in total have been identified in the other coding regions. This mutational bias in favour of tRNA genes, which represent only about 1/10th of the genome sequence, is also observed among all reported cases, and among tRNAs, a second bias is observed in favour of tRNALeu UUR. A possible explanation for this nonrandom distribution may be the more or less deleterious effects of some mutations that could be selected against in the germline because this phenomenon has been observed in mice in vivo (Fan et al., 2008; Stewart et al., 2008), but this is probably not the only explanation.
Very poorly diagnosed 20 years ago and largely underestimated, the presence of potentially pathogenic mtDNA mutations is now estimated to be 1 : 200 (Elliott et al., 2008) but result in mitochondrial diseases in only 1 : 5000 cases (Schaefer et al., 2008). This does not allow these diseases to be considered as rare anymore.
A large variety of syndromes is associated with these mutations (deletions, rearrangements and point mutations). They will not be detailed here but more information can be found on the MITOMAP website and in two very recent reviews (Wallace & Fan, 2009; Tuppen et al., 2010).
The specificities of mitochondrially encoded mutations
As compared with diseases derived from nuclear-encoded mutations, mitochondrial-encoded ones have their own specificities.
All mtDNA deletions and many mtDNA point mutations are heteroplasmic. Because a cell contains many mtDNA molecules, most cells contain in fact a mixture of mtDNA wild-type allele and mtDNA mutated allele. This proportion may vary in the progeny and in the different tissues. It is now clear that the severity of the disease is often associated with the relative proportion of mutated molecules, introducing the notion of ‘threshold’. It is also logical to expect more deleterious effects of the mutations in tissues that are strongly high-energy dependent, such as the heart, the brain or the muscles (see Tuppen et al., 2010 for a discussion of these points). However, these two basic facts do not account for all the observations that have been made. For example, many different diseases are associated with mt tRNA mutations, which cannot be related only to the molecular effect of the mutation: the same mutation in the same tRNA can be sometimes associated with a large variety of syndromes from mild to severe, and different tRNAs when mutated confer the same syndromes (reviewed in MITOMAP, 2009; Scaglia & Wong, 2008). It is probable that the nuclear genetic background of each individual plays a large role in the syndrome outcome. How could it be explained that a mother carrying a homoplasmic tRNAVal mutation (C1624T) is asymptomatic while her seven children from different partners were all suffering from extremely severe pathogenesis? A profound RC deficiency was observed in both the mother and her children (Mc Farland et al., 2002). The role of such ‘nuclear modifier gene’, necessary for the tRNA mutation to express the disease phenotype, has been reported in a tRNA Ile mutation (Davidson et al., 2009).
The complexity of the system where oogenesis and differentiation probably play an important role is far from being understood. The cellular programme and in particular the transcriptional regulation that presides to these critical steps as well as the mtDNA bottleneck that takes place during the oogenesis (reviewed in Tuppen et al., 2010) are certainly key factors to explain the variable penetrance and the phenotypic variability that are observed.
The role of the yeast model
Considering the complexity of mitochondrial diseases in general, animal models of such diseases are urgently needed. With the possibility to make transgenic mice at will, they have been largely developed for nuclear- encoded mutations. This is far from being the case for diseases due to mutations in the mitochondrial genome. Nucleic acids are naturally not imported into yeast mitochondria with the exception of a cytoplasmic tRNA Lys (Tarassov et al., 1995) and mammalian mitochondria cannot be transformed with an in vitro modified mitochondrial gene. Several factors could explain it: the very high number of mtDNA copy, the low efficiency of recombination and the absence of a powerful selective system. Even the incorporation of exogenous RNA into isolated mammalian mitochondria by electroporation has been unsuccessful because the RNA is not expressed (McGregor et al., 2001). Alternative strategies have been set up, based on somatic genetics, with the goal to construct genetically modified ES cells that can be later reintroduced into the mouse female germline. A few mice models carrying specific mutations have been produced (see reviews in Tyynismaa & Suomalainen, 2009; Wallace & Fan, 2009). Manipulation of the mitochondrial genome in the germline based on the selection of mutations resistant to mitochondrially targeted restriction enzyme have been elegantly applied to Drosophila (Xu et al., 2008) and this method could certainly also be applied to mice. However, these constructions are laborious and there are arguments (Fan et al., 2008) suggesting that a mouse model carrying a highly pathogenic mutation may not be possible. To circumvent these difficulties, researchers have tried to indirectly engineer the mtDNA by acting upon nuclear genes involved in mtDNA replication and maintenance. Particularly interesting are the ‘Mutator mice’ (Trifunovic et al., 2004). The in-depth analysis made by Foury et al. (2004) on the yeast mtDNA polymerase (coded by the MIP1gene) has shown that its proofreading domain can be mutated, leading in some cases to a strong mutator activity. Transposed to mouse, appropriate mutations also lead to a strong enhancement of mitochondrial mutations among which one can be lucky and find the ‘pathogenic mutation’ that is searched. This study is once more a very good example of how yeast can help in expertly guiding what could be done in mammals. For the moment construction of appropriate mice models is still very laborious and specific mutations very difficult to obtain. Despite these limitations, the various mice mutants that have been developed have been instrumental to mimic some syndromes and to show that mtDNA mutations can be the primary cause of neurodegenerative diseases. They also permitted to understand tissue specificity and the segregation of mtDNA molecules as well as to investigate the role of mtDNA in ageing and other degenerative phenotypes (Kujoth et al., 2005).
In contrast to the difficulties encountered to establish mitochondrially mutated mice, the yeast system offers some facilities to create at will the precise desired mutations.
While nobody pretends that the information obtained will provide knowledge on how mitochondrial diseases appear and are transmitted in a sophisticated differentiated organism, yeast ‘models’ of human mitochondrial diseases offer information on genetic and molecular aspects; they also allow to distinguish between the significant number of neutral mtDNA variants and pathogenic mutations, a question that in humans is difficult to resolve because of the high mutational rate of the mitochondrial genome and the presence of population-specific polymorphism (Tuppen et al., 2010). In addition, the power of yeast genetics allows to easily screen for compensatory mutations (see Exploiting these model yeast strains).
The key point, if only one, which amply justifies the utilization of yeast, is the possibility to transform mitochondria. Initially shown by Johnston et al. (1988) and Fox et al. (1988), the biolistic transformation (shooting at high velocity microprojectiles layered with manipulated DNA) has been improved enough to be used as a routine technique (Bonnefoy & Fox, 2007). Second, and opposite to what is observed in mice where many interesting mutations are eliminated after transmission to the female germline (see discussion in the previous paragraph), all mutations whatever their consequences can be kept in yeast. Yeast (S. cerevisiae) can survive on fermentative carbon substrates with completely nonfunctional mitochondria. One could add that it is very easy to change the nuclear background for the same mitochondrial genome, based on a karyogamy mutant (kar1-1; Conde & Fink, 1976). The kar 1-1 mutation considerably delays the karyogamy and allows to reassociate a given mitochondrial genotype with a new nuclear one. Finally, the fact that yeast becomes very rapidly homoplasmic when two mitochondrial populations are mixed could also be an advantage. This property does not allow to study the threshold effect (which can however be mimicked by the different nuclear background), but it leads to the simplification of a very complex system, which is a welcome first step to start any analysis.
Some examples of ‘humanized’ yeasts
Saccharomyces cerevisiae mitochondrial transformation is mostly based on two important characteristics (1) the possibility for yeast mitochondria to efficiently replicate many sequences – and in particular plasmids – which do not have the true mitochondrial replication origins (Fox et al., 1988) and (2) the high efficiency of homologous recombination. The fact that mitochondrial genetics has been developed for years in this organism offers a third essential tool: appropriate mutations and markers. Biolistic transformation is ‘inefficient’ in the sense that most cells are killed by the process. Selecting survivors and appropriate recombinants is based on a positive selection, supported by appropriate genetic markers. A detailed description of this process and associated protocols can be found in Bonnefoy & Fox (2007). Finally, we should mention the essential fact that the products of mitochondrial genes (proteins or RNA) are largely conserved between human and yeast.
Cyt b is certainly the protein that presents the largest spectrum of mutations; some of them were isolated by random mutagenesis many years ago, but the panel was largely extended via biolistic transformation (Meunier, 2001; Bratton et al., 2003; Fisher et al., 2004a, b; Blakely et al., 2005; Wenz et al., 2006). The positive screening is based on the ability to grow on respiratory medium (see Wenz et al., 2006 for details). Site-directed mutagenesis was also used to model selected regions of the mammalian Qo site in yeast cyt b in order to further understand the differential efficacy of these Qo-site inhibitor in the mammalian and pathogen bc1 complexes (Kessl et al., 2005). In these cases, selection was based on inhibitor resistance (Fisher & Meunier, 2005). These mutations were informative to define the biochemical defects of mutations and to explain their effects in relation with the structural three-dimensional (3D) scaffold. Suppressors were isolated (phenotypic reversion to the respiratory competent phenotype) and shown to be intragenic. They explain or confirm the 3D interpretation and its effect on the phenotype (see some examples in the review by Fisher & Meunier, 2001). A new screening method, based on the ARG8m mutation, has been devised to allow selection of any cyt b mutations whatever their phenotype (functional or nonfunctional). ARG8m is an allotopically expressed nuclear gene now expressed in the mitochondrial genome (Steele et al., 1996) and the screen is based on arginine auxotrophy/prototrophy (Ding et al., 2008).
The latter method has also been used to create mutations of ATP6. One of its advantages is also to eliminate deletions of mtDNA (rho−/rho°), which accumulate as a consequence of the primary mutation because arginine prototrophy requires a functional translation absent in rho− cells. Four mutations that mimic human pathological ones are described (Rak et al., 2007; Kucharczyk et al., 2009a, b, 2010). A thorough biochemical investigation has shown that these mutations have similar impact on the ATPase of human and yeast origin. They sometimes can clarify contradictory results or offer new informations not yet obtained in humans. Consequently, they open new directions for investigations. The last example of ‘humanized’ yeast strains is the case of tRNA mutations. Human mt tRNA mutations are over-represented and exhibit a very large diversity, structural (any domain of the cloverleaf structure), biochemical (processing, aminoacylation, interaction with ribosome, stability, etc.) and phenotypic (from severe syndrome such as MERFF or MELAS to milder ones such as diabetes or CPEO). A precise list and description of the human pathogenic mutations can be found in Scaglia & Wong (2008). It is also in a tRNA that the first ‘dominant’ mitochondrial mutation has been identified (C5545T in tRNATrpS, Sacconi et al., 2008), in contradiction with the ‘threshold rule’.
In yeast, the first mutations introduced into tRNALeu UUR gene were MELAS pathogenic mutations. Because this is a severe syndrome, one could expect that yeast cells bearing the same mutations would exhibit a strong phenotype. Based on this hypothesis and because it was impossible to screen with the ARG8m gene, which requires an active mitochondrial protein synthesis absent in this type of mutant, the authors relied on sequencing the tRNALeu UUR gene of only about 30 ‘petite’ colonies. Indeed the mutation was identified in two colonies, confirming the expected strong yeast phenotype. Such cells were not able to grow on respiratory substrates and moreover produced a high percentage of rho−cells. (Feuermann et al., 2003). Mutations for which the phenotype could be very weak or barely detectable would require one to sequence the gene in thousand or more colonies, considering the recombination frequency. To circumvent these difficulties, the authors later on turned to the use of an artificially created restriction site (ACRS)-PCR technique – an analysis that does not preclude any resulting phenotype – and introduced other mutations either in the tRNALeu UUR gene or in other tRNA genes. Molecular and structural consequences of those were analysed in detail (Montanari et al., 2008; De Luca et al., 2009) based on semi-denaturing Northern blots in which the amount, the size and the aminoacylation of the tRNAs could be examined. Two conclusions could be drawn from such studies: (1) a correlation between the severity of the in vivo phenotypes of yeast tRNA mutants and those obtained by in vitro studies of human tRNA mutants supports the view that yeast is a suitable model to study the cellular and molecular effects of tRNA mutations involved in human pathologies. This correlation was also observed for molecular effects. For example, the aminoacylation defect of the A3243G mutation (Sohm et al., 2003) is also observed in yeast (Montanari et al., 2008). The same similarity holds true for the thermosensitive aminoacylation defect of the yeast T3250C equivalent mutation, which parallels the moderate severity of the in vitro aminoacylation defect of the human mutation. (2) The phenotypic effects of the mutations is considerably dependent on the nuclear background of the wild-type strain in which the mutations have been introduced, but the relative strength of the phenotypes within one background is conserved in the others.
Exploiting these model yeast strains
As shown previously, once yeast mutations similar to human pathogenic mutations have been created, the first objective is to genetically and biochemically characterize them, a task that is made easier by the facilities to grow yeast cells (in terms of rapidity, yield and simplicity of manipulation) and the homoplasmic state. Moreover, one irreplaceable quality of yeast is the power of its genetics. Once a mutation is obtained, there is no limit to search for compensatory mutations (suppressors), which will alleviate the defective phenotype. Conversely, it is also possible to search for synthetic lethal mutations that, when associated with a mutation with a nondetectable or weak phenotype, will worsen it. Suppressors can be selected either as secondary mutation or as wild-type gene dosage modification (multicopy suppressors). These possibilities have indeed been exploited with the mutations described above.
The search for suppressors has already been described (see Some examples of ‘humanized’ yeasts) in the case of cyt b mutations; all of them turned out to be intragenic second-site mutations, and, when placed within the 3D structural model, they provided insights into the molecular interactions within the structure. Such knowledge had important consequences, in particular for drug selection (Fisher & Meunier, 2008). A different strategy was applied to identify suppressors for ATP6 and tRNA genes. The phenotypic correction was looked for with multicopy suppressor, i.e. a change in wild-type gene amount. The gene ODC1, a member of the mitochondrial carrier family that exchanges intermediates of the TCA cycle across the inner membrane is able, when overexpressed, to increase ATP production, thus correcting mutations with an impaired ATP synthesis (Schwimmer et al., 2005). It seems that under such conditions the ODC1 gene can improve the respiratory capacity of the yeast NARP T8993G mutation (cited in Kucharczyk et al., 2009c). In this case, the gene overexpression allows a metabolic correction, by producing alternative ATP synthesis.
Overexpression of the TUF1 gene, which encodes the mitochondrial translation elongation factor mtEF-Tu was initially identified in a screen to correct the defective phenotype of several MELAS mutations of the yeast mt tRNALeu UUR (Feuermann et al., 2003). Interestingly, it appeared later that the same gene can in fact correct all types of tRNA mutations. Cognate tRNA synthetases can also compensate the mutations (De Luca et al., 2006, 2009). The effect of tRNA synthetase could be understood because overexpression can overcome weaker affinity for the substrate, but the fact that mutated leucyl-tRNA synthetase with highly reduced catalytic activity maintains full suppressing effect rather suggests a chaperone-like and/or stabilizing function and this is probably also the case for the general suppression effect of mtEF-Tu (De Luca et al., 2009).
Saccharomyces cerevisiae has proven to be a performing organism to mimic human pathological mutations of mitochondrial origin but it has its own limitations. For example, complex I is absent and replaced by a nuclear-encoded NADH dehydrogenase. However, biolistic transformation is being developed in other yeasts. Recently, mitochondria of Candida glabrata were transformed with appearance of heteroplasmy. Deepening this observation may provide interesting information to control it (Zhou et al., 2010). Because Yarrowia lipolytica possesses the mitochondrial-encoded subunits of complex I, it may be used in the future to model complex I human mitochondrial mutations. At present, only Chlamydomonas reinhardii for which mitochondria can also be transformed are used to study this complex (Remacle et al., 2006).
The work in S. cerevisiae is only possible if the human and yeast gene products are highly similar, which means that all tRNA mutations cannot be analysed. To overcome this difficulty, efforts are made to have a complete human tRNA gene introduced into the yeast mitochondrial genome. Recently, Y. Zhou and M. Bolotin-Fukuhara (pers. commun.) have been able to show that the human tRNALeu UUR can be transcribed and properly matured, including the terminal CCA, into yeast mitochondria. It remains to be seen if it can be correctly aminoacylated and functional.
Based on exploiting the natural RNA import system, Kolesnikova et al. (2010) have developed a set of small RNA molecules opening the possibility of creating a new mitochondrial vector system able to target therapeutic oligoribonucleotides into deficient human mitochondria.
Alternatively, gene therapy based on mitochondrial gene versions expressed in the nucleus and artificially reimported into mitochondria or nuclear suppressors able to correct the defective mitochondrial gene can be foreseen in the future. Multicopy suppression is especially valuable for such approaches. The case of Ef-Tu, which is a general suppressor of all mutated tRNAs described up to now, is of special interest and all the more because it has been recently shown that it is also efficient to correct defective patient cell lines (Sasarman et al., 2008). Finally, one should not overlook the pharmaceutical approach. Yeast cells are easy to manipulate, fast to grow and such process can be robotized at low costs. They constitute therefore a method of choice for a rapid first screening of pharmaceutical products active in human diseases. Yeasts have been successfully used to find drugs active against mammalian prions (Bach et al., 2003; Tribouillard et al., 2006), and promising molecules active on yeast models of ATP synthase deficiencies have already been isolated (cited in Kucharczyk et al., 2009c).
In this review, we have discussed different aspects of mitochondrial diseases. Because, as noted in the Introduction, several good reviews already exist, some of them focused on yeast, we have chosen to focus on the more recently developed aspects of mitochondrial defects, in particular those concerning mtDNA maintenance, the importance of mitochondrial morphology and the mitochondrial transformation.
The first important remark to be made is concerned with the increasing complexity of the mitochondrial world that pathologies are disclosing to us. Starting from the second part of the last century, we first were fascinated by the presence of two genomes evolving in the same eukaryotic cell, but now we see in detail the dynamic relationship between the two genomes and the different organelles. The importance of mitochondria in complex pathologies and in the effects of ageing certainly adds to this complexity.
Here, we come to a central point: in all the described pathologies, the possibility to establish a yeast model has been an invaluable tool to control and verify basic effects. The possibility to control yeast ploidy, to shift mitochondria to different nuclear contexts and to immediately understand the effect of an mtDNA mutation in the homoplasmic yeast mitochondria have largely demonstrated the value of this model. The new knowledge on the fine-tuning of mitochondrial functions by proteins localized in inner and outer mitochondrial membranes, but also contacting inner and outer plasm, opens a new field in which the yeast model is indispensable and in which a new aspect, namely mitochondrial motility, becomes essential.
Finally, the successful use of yeast mitochondrial transformation can be an essential step towards a mitochondrial pharmacology.
Moreover, we would like to stress the very high conservation of these complex functions between humans and yeasts and hence the general applicability of the yeast models even when pathological characteristics should be considered.
M.B-F has been continuously supported by the ‘Association Française Contre les Myopathies (AFM) for the work on yeast mitochondrial tRNA models. L.F. has been supported by Telethon project GGP07164.