10.1 Introduction

The mitochondria present within our cells carry out important and diverse functions including oxidative phosphorylation and apoptotic cell death (Nunnari and Suomalainen 2012). Pathologies can arise from inhibition or imbalance in any of the cellular processes for which mitochondria are required. A healthy mitochondrion requires over a thousand proteins encoded by the nuclear DNA (nDNA) and mitochondrial DNA (mtDNA) (Gorman et al. 2016). This means that there are over 1000 mutational targets from which mitochondrial dysfunction can result. Oxidative phosphorylation relies on the formation and stability of 5 multi subunit complexes of nuclearly and mitochondrially encoded components (Sanz and Stefanatos 2008). Accordingly, this process is particularly sensitive to mutation and as such, most mitochondrial disease patients carry mutations in genes encoding OXPHOS proteins (Lightowlers et al. 2015).

The current standard of care for mitochondrial disease patients is palliative, with interventions that mainly target specific symptoms (Viscomi et al. 2015; Lightowlers et al. 2015; Gorman et al. 2016). To facilitate the design, discovery and validation of curative treatments, a deeper understanding of disease pathogenesis is required. Predictive models, which recapitulate the clinical and molecular pleiotropy exhibited by mitochondrial disease patients will allow examination of the molecular effects of mitochondrial dysfunction. The mitochondrial research community has accumulated a collection of complementary models with just this aim.

In this chapter, we will attempt to summarise the wide range of models which have been created and used to study the effects of disrupting a variety of mitochondrial proteins, with a focus on those genes which have been found to be mutated in patients (Table 10.1). This will include in vivo models such as mice, zebrafish, flies and worms. Finally, we will comment on the potential of these models in investigating the pathophysiology of mitochondrial disease and the development and validation of novel therapies.

Table 10.1 Models of Mitochondrial disease
Full size table

In vitro, cell culture models have been used to model the pathology of many patient derived mutations, using a variety of techniques. Most studies have used cytoplasmic hybrids (Cybrids). Cybrids allow the assessment of mtDNA alterations in a defined nuclear background. They have been used to study the bioenergetics of specific patient mutations and mtDNA heteroplasmy (Wilkins et al. 2014). Primary cells derived from patient biopsies have also been used to model mitochondrial diseases, however disease pathology is not always recapitulated in culture (Cameron et al. 2004). Finally, with the advent of induced pluripotency, the creation of stem cells from differentiated tissue also known as induced pluripotent stem cells (iPS), researchers can create patient specific iPS cells which can be differentiated into specific cell types, like neurons or muscle cells (Avior et al. 2016). This is particularly important for mitochondrial disease given the tissue specific effects of different mitochondrial syndromes. Despite the obvious limitations of an in vitro system, iPS cell models will be useful in validation of potential therapies. However, given the long standing and wide use of cell culture models we will focus this chapter on whole animal models of mitochondrial disease.

10.2 Mouse Models of Mitochondrial Disease

The mouse has been well established as the go to model in the study of several human diseases (Peters et al. 2007). Humans and mice share a large degree of genetic and physiological traits which make them an attractive model to study disease pathogenesis (Peters et al. 2007). In the last 20 years, technology for the genetic modification of mice continues to improve, making the process quicker and cheaper (Van Der Weyden et al. 2011). In that time mouse models corresponding to over 40 nuclearly and mitochondrially encoded mitochondrial proteins have been created. Broadly these can be divided into two categories, those genes involved directly in oxidative phosphorylation (OXPHOS) i.e. subunits and assembly factors of the mitochondrial respiratory complexes, and those proteins required for normal mitochondrial function such as those regulating mitochondrial DNA maintenance.

10.2.1 Complex I Deficiency

10.2.1.1 NDUFS4

The first mouse model created which targeted complex I directly was the NADH: Ubiquinone Oxidoreductase Subunit S4 (Ndufs4) knockout (KO) mouse created in the Palmiter Lab (Kruse et al. 2008). Ndufs4 whole body KO mice were created via Cre/loxP excision of exon 2 which resulted in undetectable levels of NDUFS4 protein (Kruse et al. 2008). Phenotypically Ndufs4−/− mice started to show signs of illness around 5 weeks of age, with symptoms of increasing lethargy, weight loss, progressive loss of motor ability and reduced survival at around 8 weeks, reminiscent of patients with isolated complex I (CI) deficiency or Leigh syndrome (Kruse et al. 2008; Dahl 1998). Leigh syndrome (LS) is a common infantile mitochondrial disorder which is characterized by progressive neurodegeneration (Dahl 1998). Ndufs4−/− mice had reduced levels of assembled CI and a corresponding decrease in CI activity as assayed by oxygen consumption (Kruse et al. 2008). Interestingly, when the same group conditionally removed Ndufs4 from neurons and glia, they were able to phenocopy whole body KO mice (Quintana et al. 2010). This highlights the importance of the neuronal component in the Ndufs4−/− phenotype and suggests it could underlie that of LS associated with CI deficiency. Moreover, neuronal pathology in both models correlated with disease onset and progression in humans (Kruse et al. 2008; Quintana et al. 2010; Rahman et al. 1996). Removal of NDUFS4 globally or specifically from the central nervous system results in a LS-like phenotype which overlaps highly with that of patients suffering from LS, and as such could be useful in the validation of therapies for LS.

The Larsson and Tian groups went on to study the effects of conditional ablation of NDUFS4 in mouse heart, describing a mild phenotype of cardiomyopathy (Karamanlidis et al. 2013; Li et al. 2000a). Given that NDUFS4 levels were undetectable upon ablation, insights into the tissue specific phenotypes observed via conditional knockdown came from a study by Calvaruso and colleagues who measured the residual CI activity in various tissues of the whole-body KO (Calvaruso et al. 2012). The study found that residual CI activity was present in all tissues but that this varied greatly, and in fact that the heart possessed the highest level of residual CI activity (Calvaruso et al. 2012).

Using random transposon insertion Leong and colleagues created another model of NDUFS4 loss, Ndufs4fky/fky, which in line with what had been observed in previous models displayed reduced CI activity in all tissues, progressive neuropathy and died prematurely at around 7 weeks of age (Leong et al. 2012).

Finally, the Pinkert lab created an NDUFS4 knock in (KI) mouse (Ndufs4 KI/KI) which carries a point mutation which results in the production of a truncated protein (Ingraham et al. 2009). This is a particularly interesting model, as most NDUFS4 mutations in humans lead to expression of truncated versions of the NDUFS4 protein (Budde et al. 2000; Iuso et al. 2006; Van Den Heuvel et al. 1998). Homozygous KI mice were embryonic lethal, while mice heterozygous for the KI allele were viable and displayed a mild phenotype, characterized by a 30% decrease in CI linked respiration and CI activity, and increased levels of lactate, commonly seen in patients (Ingraham et al. 2009).

This group of studies using several independent models of NDUFS4 dysfunction firmly establish the importance of this subunit in stability and regulation of respiratory complex I (Ingraham et al. 2009; Karamanlidis et al. 2013; Kruse et al. 2008; Leong et al. 2012; Quintana et al. 2010, 2012; Sterky et al. 2012). They also highlight how residual CI activity can vary greatly tissue to tissue, with implications not only on tissue specific pathology but also on organismal viability. As mutation of NDUFS4 is associated with LS due to CI deficiency (Ruhoy and Saneto 2014), understanding the effects of its depletion can serve only to aid the development and assessment of targeted therapies.

10.2.1.2 NDUFS6

NADH: Ubiquinone Oxidoreductase Subunit S6 (NDUFS6) is part of the enzymatic core of complex I that has been found to be mutated in patients suffering from isolated CI deficiency (Kirby et al. 2004; Spiegel et al. 2009). Like many patients suffering from CI deficiency, patients carrying mutations in NDUFS6 display a reduction in CI activity accompanied by acidosis, myopathy and encephalopathy. Several mutations of NDUFS6 have been described to result in very severe phenotypes, and death within weeks of birth (Spiegel et al. 2009).

There is currently one mouse model of NDUFS6 loss available, created via gene trap insertional mutagenesis (Forbes et al. 2013). The Ndufs6gt/gt mouse has decreased levels of assembled CI and CI activity across all tissues, however like in the case of Ndufs4, the degree of reduction varies from tissue to tissue, with the lowest residual activity found in cardiac tissue (Forbes et al. 2013). Interestingly, the authors describe a tissue specific splicing anomaly which may underlie this variability in Ndufs6 knockdown. Ndufs6gt/gt animals develop cardiomyopathy, have reduced ATP synthesis from complex I linked substrates and a deregulation of fatty acid metabolism with sex bias, as males were more severely affected than females. Additionally, both homozygous Ndufs6gt/gt and heterozygous Ndufs6gt/+ mice develop renal disease, with a greater severity in homozygous animals (Forbes et al. 2013).

Although this model displays classic phenotypes of isolated complex I deficiency many of the phenotypes associated with mutations in NDUFS6 are missing. In contrast to the short survival of affected patients, Ndufs6gt/gt can live up to 4 months (Forbes et al. 2013; Spiegel et al. 2009; Kirby et al. 2004).

10.2.1.3 Mt-ND6

MT-ND6 is a mitochondrially encoded core subunit of CI (Wirth et al. 2016). Mutations and allelic variants of MT-ND6 have been found in patients suffering from Leber hereditary optic atrophy (LHON), LS and mitochondrial encephalomyopathy lactic acidosis and stroke like episodes (MELAS) (Johns et al. 1992; Chinnery et al. 2001; Ravn et al. 2001). A severe frameshift mutation of MT-ND6 (13885insC) was shown to be actively removed from the oocyte in a mouse model of mitochondrial disease (Fan et al. 2008). Two further mouse models were created carrying the same missense mutation (G13997A) isolated from a lung carcinoma cell line, which corresponds to a mutation described in patients (G14600A) (Ishikawa et al. 2008). Homoplasmic mice created in the Hayashi lab had a very mild phenotype consisting of a 20–30% reduction in CI and CIII activity and a modest increase in serum lactate (Hashizume et al. 2012). Long term these mice failed to develop any neurological or ophthalmological pathologies (Hashizume et al. 2012).

The Wallace lab also introduced this mutation into a mouse model but in a different strain (Lin et al. 2012). They observed a clear recapitulation of LHON patient pathology (Lin et al. 2012). In addition to a ubiquitous reduction in CI activity, there were highly elevated levels of ROS in neuronal tissue of affected animals. Thus, the MT-ND6P25L model provides an opportunity to investigate the molecular pathogenesis of LHON and test potential therapies (Lin et al. 2012).

10.2.2 Complex II Deficiency

10.2.2.1 SDHD

Succinate Dehydrogenase Complex Subunit D (SDHD) has been shown to be mutated in several cases of isolated CII deficiency (Jackson et al. 2014; Alston et al. 2012, 2015). Two separate attempts were made to create a mouse model for the study of this gene (Piruat et al. 2004; Bayley et al. 2009). Sdhd−/− mice created in the Barneo lab were embryonic lethal while mice heterozygous for the Sdhd KO allele had no observable phenotype, despite a decrease in CII activity (Piruat et al. 2004). As mutations in Sdhd have also been shown to be involved in a form of hereditary paraganglioma (Schiavi et al. 2006), the Taschner lab created Sdhd−/− mice to assess the role of Sdhd as tumour supressor. These mice were also embryonic lethal and aged heterozygous mice did not develop tumours (Bayley et al. 2009). As such these models are not useful for the study of Sdhd in isolated CII deficiency or as a tumour suppressor.

10.2.3 Complex III Deficiency

10.2.3.1 BCS1L

BCS1 like Ubiquinol-Cytochrome C Reductase Complex Chaperone (BCS1L) has been shown to be involved in the final steps of CIII assembly (Fernandez-Vizarra et al. 2007). Mutations in this gene are associated with cases of LS due to CIII deficiency, GRACILE syndrome (growth restriction, aminoaciduria, cholestasis, iron overload, lactic acidosis and early death), Bjornstad syndrome and mitochondrial CIII deficiency (Visapaa et al. 2002; Hinson et al. 2007; De Lonlay et al. 2001; Blazquez et al. 2009). The Fellman laboratory created a mouse model of GRACILE syndrome by using gene targeting to introduce the Bcs1l 232A > G point mutation, Bcs1lg/g (Leveen et al. 2011). The pathology they observed in Bcs1lg/g mice displayed remarkable similarities to that observed in patients with fatal GRACILE syndrome (Leveen et al. 2011). This included, growth restriction, progressive liver disease and premature death. Importantly, the model recapitulated the tissue specificity observed in patients with pathology arising only in the kidney and liver. Metabolic comparison of pre-symptomatic and symptomatic mice revealed a deterioration of carbohydrate and fatty acid metabolism leading to a starvation like state which may underlie disease onset (Kotarsky et al. 2012). Further analysis of the molecular pathogenesis underlying the Bcs1lg/g phenotype could provide key insights into GRACILE syndrome.

10.2.4 Complex IV Deficiency

10.2.4.1 COX4I2

Cytochrome C Oxidase Subunit 4 Isoform 2 (COX4I2) is a nuclearly encoded subunit of CIV which has been shown to be expressed in lung and pancreatic acinar cells, with patients carrying mutations in Cox4I2 suffering from exocrine pancreatic insufficiency (Shteyer et al. 2009; Hüttemann et al. 2001). This subunit and that encoded by Cytochrome C Oxidase Subunit 4 Isoform 1 (COX4I1) is thought to play important roles the regulation and assembly of CIV. Cox4i2 KO mice develop a progressive lung pathology with reduced levels of ATP and CIV activity (Huttemann et al. 2012). However, no pancreatic pathology was observed in these mice, this discrepancy could also result from an absence of Cox4i2 expression in the pancreas of mice (Huttemann et al. 2012).

10.2.4.2 SURF1

SURF1 is an assembly factor involved in CIV biogenesis (Tiranti et al. 1998; Williams et al. 2001). Patients carrying mutations in SURF1 suffer from LS due to CIV deficiency (Zhu et al. 1998; Tiranti et al. 1998). Two separate Surf1 KO mice were created by the Zeviani lab which displayed an expected CIV deficiency but unexpectedly did not have any neuropathology, a hallmark of SURF1 patients (Agostino et al. 2003; Dell’agnello et al. 2007). Although these models will be useful in understanding the biological role of SURF1 they do not recapitulate the patient phenotype and therefore may not be suitable for the development and validation of novel therapies.

10.2.4.3 COX10

COX10 is a CIV assembly protein required for biosynthesis of the heme α prosthetic group required for functional CIV (Antonicka et al. 2003a). Mutations and allelic variants of COX10 have been identified in patients suffering from isolated CIV deficiency and LS due to CIV deficiency (Antonicka et al. 2003a). A mouse model carrying a conditional allele (Cox10cKO) was used to create model of mitochondrial myopathy, hepatopathy, neuropathy and encephalopathy (Diaz et al. 2008, 2005, 2012a). Conditional ablation of Cox10 in the liver of mice resulted in phenotypes which recapitulated clinical features of mitochondrial hepatopathies including enlargement of the liver and steatosis (Diaz et al. 2008). Ablation of COX10 in the peripheral nervous system resulted in a severe neuropathy characterized by demyelination, also observed in several patients carrying Cox10 mutations (Diaz et al. 2012a). Given the stark recapitulation of pathology seen in patients, these models will be important in understanding tissue specific pathogenesis and in the testing and validation of therapies which directly target the underlying mitochondrial dysfunction.

10.2.4.4 COX15

COX15 is a CIV assembly protein which plays a key role in the heme biosynthetic pathway (Petruzzella et al. 1998). Patients carrying mutations in this gene suffer from LS due to CIV deficiency and cardioencephalomyopathy (Antonicka et al. 2003b; Oquendo et al. 2004). COX15 is required for mouse embryonic development (Viscomi et al. 2011). Mice lacking COX15 specifically in muscle have acute CIV deficiency accompanied by a severe myopathy from 1 month of age (Viscomi et al. 2011).

10.2.4.5 SCO2

SCO2 is an assembly factor required for the assembly and function of CIV. Mutations in SCO2 can result in fatal infant cardioencephalomyopathy due to CIV deficiency (Papadopoulou et al. 1999). The E140K amino acid change, corresponding to the n.1541G > A missense mutation most commonly found in patients, was introduced into a mouse model (Baxevanis et al. 2000). This model had a clear CIV defect but unfortunately did not recapitulate human disease pathology.

10.2.5 Complex V Deficiency

10.2.5.1 ATPIF1

Generation of patient relevant mouse models of CV deficiency will allow testing of therapeutic interventions. There are currently no mouse models of CV deficiency based on patient mutations available. To model CV deficiency, a mouse overexpressing a modified form of human ATPase Inhibitory Factor 1 (ATPIF1) in neurons was created (Formentini et al. 2014). ATPIF1 is thought to play an inhibitory role in the hydrolysis of ATP by ATP synthase (CV) when mitochondrial membrane potential is decreased (Formentini et al. 2012). This mutant form of human ATPIF1 has increased affinity with the beta subunit of ATP synthase, inhibiting CV hydrolysis activity, which resulted in increased oxidative stress, reduced levels of respiration and CIV activation, as well as metabolic reprogramming that prevented cell death (Formentini et al. 2014). Generation of patient relevant mouse models of CV deficiency will allow a deeper investigation of molecular pathogenesis and testing of therapeutic interventions.

10.2.6 CoQ Deficiency

CoQ10 plays several roles in electron transport within mitochondria. It facilitates the flow of electrons from CI and CII to CIII, acts as an electron acceptor for several dehydrogenases, such as those involved in NADH shuttling from the cytosol to mitochondria (glycerol-3-phosphate dehydrogenase), in fatty acid beta-oxidation (ETF:QOR) and pyrimidine nucleotide synthesis (DHODH), and functions as a potent lipid soluble antioxidant in the plasma membrane and elsewhere (Lopez-Lluch et al. 2010).

Given the central role that CoQ plays in metabolism, it is unsurprising that mutations involving its biosynthesis lead to severe phenotypes and disorders. Human CoQ deficiency can result in several forms of mitochondrial disease including LS, myopathy, myoglobinuria, ataxia, cerebellar atrophy and severe infantile encephalomyopathy with renal failure (Fernandez-Ayala et al. 2014).

CoQ biosynthesis depends on a highly conserved multi-enzyme complex that involves at least eleven genes (named COQ1 to COQ10 plus PDSS2), although some of them have no catalytic activity (Bentinger et al. 2010). Below, we summarize three mouse models of CoQ deficiency.

COQ7 encodes a hydroxylase that takes part in the latter stages of CoQ biosynthesis. Tamoxifen-induced knockdown of Coq7 led to weight loss a few days after drug administration, with reduced respiratory chain electron transfer, mitochondrial respiration and survival (Wang et al. 2015). However ROS production was unchanged (Wang et al. 2015).

COQ9 encodes a lipid-binding protein that is necessary for incorporation of COQ7 into the Q biosynthetic complex (Lohman et al. 2014) and has been shown to be mutated in a patient suffering with primary CoQ10 deficiency (Duncan et al. 2009). This patient mutation was used to create a KI mouse (Coq9X/X) model of primary CoQ10 deficiency (Garcia-Corzo et al. 2013). Homozygous KI mice developed normally but developed neurological symptoms of paralysis and eventually death within 6 months (Garcia-Corzo et al. 2013). Accordingly, these mice had reduced levels of CoQ10 as well as decreased CI, mitochondrial respiration and ATP synthesis (Garcia-Corzo et al. 2013). As such this KI mouse model is an excellent model of encephalopathy due to primary CoQ10 deficiency.

PDSS2 encodes part of the decaprenyl diphosphate synthase involved in CoQ biosynthesis, which has been found to be mutated in a case of infantile COQ deficiency type 3 (COQ10D3) (Lopez et al. 2006; Salviati et al. 2005). Pdss2kd/kd mice carry a missense mutation which results in kidney failure and decreased levels of CoQ10 (Peng et al. 2008). This can be rescued by CoQ supplementation (Saiki et al. 2008). Conditional ablation within the liver did not result in pathology despite depletion of CoQ10 (Peng et al. 2008). Suggesting that there may tissue specific thresholds for sensitivity to decreases in CoQ10.

10.2.7 Mitochondrial DNA Maintenance

10.2.7.1 Twinkle

TWINKLE is the mtDNA helicase and plays a key role in mtDNA replication and maintenance (Spelbrink et al. 2001; Tyynismaa et al. 2004). Homozygous Twinkle KO mice were embryonic lethal but conditional ablation within skeletal and cardiac muscle resulted in depleted levels of mtDNA, OXPHOS defects and a reduced survival of only 19 weeks (Milenkovic et al. 2013). Mutation of Twinkle in humans can lead to several pathologies, including autosomal dominant progressive external ophthalmoplegia with mtDNA deletions (adPEOA) and mtDNA depletion syndromes (Spelbrink et al. 2001; Hakonen et al. 2008).

Mice carrying adPEOA associated alleles of Twinkle (TwinkleA360T, Twinkledup353–365) recapitulate many aspects the human pathology (Tyynismaa et al. 2005). TwinkleA360T and Twinkledup353–365 mice progressively accumulated mtDNA deletions, developing mitochondrial myopathy characterised by reduced CIV activity and increased mitochondrial proliferation (Tyynismaa et al. 2005). Interestingly, these “deletor” mice showed selective mitochondrial dysfunction in Purkinje cells, located within the cerebellum, and hippocampal CA2 pyramidal neurons which have been previously shown to be affected in adPEOA and other mtDNA depletion syndromes (Tyynismaa et al. 2005). Importantly, physiological analysis of muscle biopsies from these mice demonstrated that the accumulation of mtDNA deletions was associated with reduced OXPHOS function and induced a local and global starvation response (Tyynismaa et al. 2010). Using the Twinkle “deletor” mouse models, Nikkanen and colleagues outline a metabolic response to mitochondrial dysfunction which results in altered 1C metabolism and induction of serine directed glutathione synthesis as a possible mechanism to stabilise dNTP synthesis, which is also present in patients suffering from mitochondrial myopathy (Nikkanen et al. 2016). This validates the Twinkle “deletor” mouse as a disease model and provides several targets for therapy of mitochondrial myopathy.

10.2.7.2 POLG and POLG2

POLG encodes the catalytic subunit of the mitochondrial DNA polymerase γ, providing DNA polymerase and 3′-5′ exonuclease activity (Lamantea et al. 2002). Several hundred mutations of POLG have been described in patients which can result in several clinical presentations caused by mtDNA instability (Lamantea et al. 2002; Milone and Massie 2010). As such, accurately modelling disease pathogenesis of POLG mutation is challenging. Unsurprisingly, mtDNA stability is required during development, Polg KO mice are embryonic lethal and display severe mtDNA depletion (Hance et al. 2005).

Several mouse models have been created to address the pathology underlying POLG mutation (Lewis et al. 2007; Bensch et al. 2009; Kasahara et al. 2006; Zhang et al. 2000, 2005). The first Polg mouse model created by Zhang et al., used conditional expression of an exonuclease deficient Polg, PolgD181A within mouse cardiac muscle (Zhang et al. 2000). These mice developed cardiomyopathy associated with increased levels of mtDNA mutations (Zhang et al. 2000). Further analysis of this model revealed activation of a pro-survival response within the cardiac muscle, reasoned to be a response to activation of cell death pathways in the face of increased mtDNA instability (Zhang et al. 2005). This same mutation was also overexpressed in neurons and pancreatic β cells (Bensch et al. 2009). In neurons, expression of PolgD181A resulted in increased mtDNA instability in the forebrain leading to neurological dysfunction relating to behaviour and mood, in line with what has been observed in patients suffering from progressive external ophthalmoplegia (PEO) with mood disorder (Kasahara et al. 2006). In the pancreas, expression of PolgD181A within islets resulted in early onset diabetes due to β cell death and insulin insufficiency which associated with increased mtDNA mutations (Bensch et al. 2009).

To access the role of mtDNA mutations in aging two Knock IN mice for the PolgD257A mutation, which ablates proof reading of Polg, were created (Kujoth et al. 2005; Trifunovic et al. 2004). Although the role of mitochondrial dysfunction in aging is not the focus of this chapter, several relevant phenotypes where described in these mice (Kujoth et al. 2005, Trifunovic et al. 2004) . These mice, often referred to as “mutator” mice, accumulated mtDNA mutations and deletions which led to progressive OXPHOS deficiency (Kujoth et al. 2005; Trifunovic et al. 2005). Importantly, follow up analysis on this model revealed distinct tissue specific pathology due to this KI allele, PolgD257A.

Finally, the Y995C mutation associated with PEO was transgenically expressed in mouse cardiac muscle and resulted in increased oxidative stress, reduced levels of mtDNA and cardiomyopathy, as seen with the expression of PolgD181A, a proof reading deficient mutant, in mouse heart (Lewis et al. 2007).

POLG2 encodes the accessory subunit required for Polymerase γ function. Although rarer, mutations of POLG2 have been described in patients (Longley et al. 2006). Like Polg, homozygous loss of Polg2 results in embryonic lethality with severe mitochondrial defects, underlining the requirement of mtDNA stability for mammalian development (Humble et al. 2013). Importantly half the dose of either Polg or Polg2 is sufficient for mitochondrial function, development and lifespan.

Given the number of described pathogenic mutations the mouse may not be the most suitable model for assessment of the molecular pathology of each disease causing lesion. This is an example where complementary models such as Drosophila melanogaster or C. elegans could play a role.

10.2.7.3 TFAM

Mitochondrial transcription factor A (TFAM) is required for mtDNA transcription and packaging and mitochondrial biogenesis (Stiles et al. 2016). In humans, the singular described mutation of TFAM resulted in mtDNA depletion syndrome (Stiles et al. 2016).

Due likely to its role in maintaining mtDNA stability, homozygous loss of Tfam is embryonic lethal (Larsson et al. 1998). Tfam has been conditionally ablated in skeletal muscle, cardiac muscle, pancreatic β cells, adipose tissue, forebrain and dopaminergic neurons (Wredenberg et al. 2002; Wang et al. 1999; Silva et al. 2000; Sörensen et al. 2001; Ekstrand et al. 2007; Vernochet et al. 2012). In all models, Tfam deletion resulted in mtDNA depletion and progressive OXHOS deficiency and except for the adipose tissue, this led to tissue specific pathologies. In the adipose tissue, this dysregulation of mitochondrial biology led to an increased insulin resistance and obesity prevention on a high fat diet.

Given the penetrance of the phenotypes seen in mouse models of Tfam deletion it is not surprising that only 2 cases of patients with TFAM mutations have been described so far.

10.2.7.4 LRPPRC

Leucine-rich pentatricopeptide repeat containing protein (LRPPRC) works in complex with other mitochondrial proteins to regulate post transcriptional gene expression of mitochondrial encoded transcripts (Ruzzenente et al. 2012; Sasarman et al. 2010). The French-Canadian variant form of LS (FCLS) is caused by the c.1119C > T mutation of LRPPRC, which is characterised by a severe defect in CIV with neuro and hepatopathy (Debray et al. 2011).

Homozygous loss of Lrpprc results in developmental delay, CIV deficiency and embryonic lethality (Ruzzenente et al. 2012; Xu et al. 2012). Conditional ablation of Lrpprc in the skeletal or cardiac muscle of mice led to reduced lifespan, development of progressive cardiomyopathy and CIV deficiency (Ruzzenente et al. 2012). The sensitivity of CIV function to mutation in LRPPRC could be a result of binding of LRPPRC to mRNA of CIV core subunit COX1 (Xu et al. 2012). However, in a model of Lrpprc loss in the mouse heart, a defect in CV assembly was found to underlie the OXPHOS defect (Mourier et al. 2014). Our knowledge of LRPPRC function is modest and although those mouse models currently available recapitulate aspects of FCLS, we may need to build new (i.e. neuronal or liver specific ablation) or utilise complementary models to have a more disease relevant model for analysis of molecular pathology.

10.2.7.5 ANT1

Adenine nucleotide translocator isoform 1 (ANT1) is a member of a family of adenine nucleotides translocators (ANTs) which are responsible for the exchange of ADP and ATP across the inner mitochondrial membrane (Li et al. 1989). Mutations in ANT1 can be found in patients suffering from several types of mtDNA depletion syndromes with cardiomyopathy and myopathy, as well as adPEO (Kaukonen et al. 2000; Thompson et al. 2016). The Wallace laboratory created a Ant1 KO mouse (Ant1PGKneo) which faithfully recapitulated the features of a mitochondrial myopathy and cardiomyopathy (Graham et al. 1997). In line with patients, Ant1PGKneo mice had myopathy characterised by ragged red fibres (RRFs), COX negative fibres and dysfunctional mitochondria (Graham et al. 1997). These mice also displayed metabolic adaptations, including increased levels of lactate and Krebs cycle intermediates, as well as increased oxidative stress and accumulation of mtDNA deletions (Graham et al. 1997). Importantly and in contrast to patients, Ant1PGKneo mice do not display any neuronal or ophthalmological pathology which might be due to expression of another Ant family member (Lee et al. 2009; Phillips et al. 2010).

Despite the absence of a neuronal or ocular phenotype, the Ant1PGKneo mouse is a great model of mitochondrial myopathy and as such is an excellent model for molecular investigation of mitochondrial myopathy, as well as validation of targeted therapies.

10.2.7.6 TK2

The mitochondrial thymidine kinase is encoded by TK2 (Johansson and Karlsson 1997). It catalyses the first step of the nucleotide salvage pathway and has been shown to be mutated in patients suffering from mitochondrial depletion syndrome type 2 (MTDS2) and autosomal recessive PEO (PEOB3) (Boustany et al. 1983; Moraes et al. 1991; Tyynismaa et al. 2012). Homozygous Tk2−/− mice are viable but develop encephalopathy and die within the first 4 weeks of life (Zhou et al. 2008). This was mirrored in a KI model where a point mutation corresponding to a human pathogenic mutation, H121N, was introduced (Akman et al. 2008). Tk2KI mice also displayed encephalopathy, early mortality and reduced levels of mtDNA (Akman et al. 2008). Importantly and in contrast to patients, neither mouse model developed myopathy which is a key characteristic of MTDS2. However, a detailed analysis of the Tk2−/− mouse described an altered transcriptional signature and pre-clinical phenotypes in heart and skeletal muscle (Paredes et al. 2013). Conditional or inducible models of Tk2 loss specifically in the muscle may allow modelling of this myopathic mtDNA depletion syndrome.

10.2.7.7 TYMP

TYMP encodes thymidine phosphorylase which is essential for the nucleotide salvage pathway (Matsukawa et al. 1996). Mitochondrial neuro-gastrointestinal encephalopathy (MNGIE) is a mtDNA deletion syndrome characterised by gastrointestinal blockage, PEO and mitochondrial dysfunction (Hirano et al. 2012). Most MGNIE cases are a result of TYMP mutation (Hirano et al. 2012). The phenotype of a mouse model of TYMP deficiency (Tymp −/−, Up−/−) partially overlapped with that of MNGIE patients (Haraguchi et al. 2002). With increased levels of thymidine and deoxyuridine in several tissues and in the brain, progressive mtDNA depletion (Lopez et al. 2009).

10.2.7.8 RRM2B

RRM2B encodes the small subunit of the p53-controlled ribonucleotide reductase (Tanaka et al. 2000). Mutations in RRM2B have been associated with mtDNA depletion syndromes, PEO and MNGIE (Bourdon et al. 2007; Kollberg et al. 2009; Tyynismaa et al. 2009).

Rrm2b KO mice (Rrm2b −/−) developed normally, had retarded growth after weaning and died at 14 weeks of severe kidney failure (Kimura et al. 2003). In a comparative analysis of Rrm2b −/− mice and patients carrying mutation in RRM2B, severe depletion of mtDNA in the absence of a OXPHOS defect was described (Bourdon et al. 2007). In both studies, Rrm2b mutation resulted in a decrease in the dNTP pool suggesting that Rrm2b is required for replenishing the dNTP pool (Bourdon et al. 2007; Kimura et al. 2003). Rrm2b −/− mice mirror the patient phenotype and given the association of RRM2B with several mitochondrial diseases this and future models of Rrm2b mutation could be used to validate therapies for mtDNA exhaustion resulting from dNTP deficiency.

10.2.7.9 MPV17

MPV17 encodes a protein of the mitochondrial inner membrane whose function has not yet been delineated. Work in yeast has shown that Sym1 (yeast orthologue of MPV17) is required for maintenance of mitochondrial morphology and mtDNA maintenance in stress conditions (Dallabona et al. 2010). In humans, mutation of MPV17 leads to mtDNA depletion syndrome 6 (MTDPS6) which is an infantile syndrome characterised by liver failure. Unfortunately, Mpv17 KO mice (Mpv17−/−) did not develop liver failure despite a reduction in mtDNA levels leading to CI and CIV deficiency (Weiher et al. 1990). Liver failure was induced in Mpv17−/− mice fed a ketogenic diet (Bottani et al. 2014). Clues to the pathogenic mechanisms which underlie MTDPS6 may be found by focusing on understanding the function of MPV17.

10.2.8 Mitochondrial Dynamics

10.2.8.1 MFN1 and MFN2

The mitochondrial GTPases Mitofusin 1 and 2 (Mfn1/2) are essential for mouse development (Chen et al. 2003, 2007). However, they are partially redundant with overlapping and distinct roles. Loss of both MFN 1 and 2 in the skeletal muscle results in muscle atrophy, increased levels of lactate and premature death at 6–8 weeks of age (Chen et al. 2010). These mice have severely impaired muscle mitochondria, with a progressive reduction in mtDNA levels, despite a compensatory increase in mitochondrial biogenesis (Chen et al. 2010). Additionally, the remaining copies of mtDNA were found to have an increased number of point mutations and deletions. This work suggests that fusion is required to maintain mitochondrial integrity.

MFN2 has been conditionally deleted in several different tissues including the heart, kidney and liver (Papanicolaou et al. 2011; Gall et al. 2012; Sebastian et al. 2012). In all cases, loss of MFN2 resulted in dysfunction or tissue pathology, but did not mirror phenotypes observed in patients carrying pathological mutations of MFN2. To have a more patient relevant model, mice expressing mutant alleles of MFN2 associated with Type 2A Charcot Marie Tooth Disease (CMT2A), a peripheral polyneuropathy, were created (Detmer et al. 2008; Cartoni et al. 2010). As expected, in both cases animals displayed phenotypes consistent with those of CMT patients such as defective motor skills, axonopathy and incomplete disease penetrance (Detmer et al. 2008; Cartoni et al. 2010). However, the molecular pathogenesis may differ, as density and distribution of mitochondria in axons was not consistent between these models (Detmer et al. 2008, Cartoni et al. 2010). Nevertheless, these models have allowed examination of CMT2A pathogenesis and in future, could be utilised to further understand the interplay between mitochondrial distribution and neuronal function.

10.2.8.2 OPA1

Mutations in Optic atrophy 1 (OPA1) are associated with autosomal dominant optic atrophy (adOA) (Delettre et al. 2000). Mouse models homozygous for two mutant alleles which result in truncated forms of Opa1, die during early embryogenesis (Alavi et al. 2007; Davies et al. 2007). Heterozygous animals with around a 50% decrease in Opa1 displayed progressive vision loss as well as neurological symptoms, such as reduced locomotor activity. As seen in the case of MFN2, the phenotypes of these models did not completely overlap. However, further characterisation of both these models uncovered symptoms that were not restricted to retinal ganglion cells (RGCs), the cell type preferentially affected in adOA (Williams et al. 2012).

Mice homozygous for a common adOA deletion (Opa1delTTAG) died at around E10.5, while heterozygous mice recapitulated phenotypes seen in patients with the syndromic form of adOA (adOA plus) (Sarzi et al. 2012). This included deafness, encephalomyopathy, cardiomyopathy and defects of the peripheral nervous system (Sarzi et al. 2012). This was in addition to loss of RGCs and vision loss observed in previous mouse models and most adOA patients. At the molecular level these mice displayed reduced CIV activity and incorporation of CIV into stable super-complexes (Sarzi et al. 2012).

Taken together these models clearly demonstrate the requirement of OPA1 for normal mammalian development, where complete loss of OPA1 leads to a fragmented mitochondrial network unable to support life. Analysis of dominant Opa1 alleles highlights the pleotropic effects Opa1 mutations. Mouse models which show RGC specific phenotypes will be useful in further understanding the sensitivity of this tissue to Opa1 loss. The more severe syndromic model of adOA (Opa1delTTAG) not only mirrors adOA patients but also several clinical features seen across various forms of systemic mitochondrial disease. Validating this model for use in the identification of potential therapies.

10.2.9 Additional Mouse Models

10.2.9.1 AIFM1

Apoptosis Inducing Factor, Mitochondria Associated 1 (AIFM1) is a flavoprotein located in the mitochondrial intermembrane space, that acts both as a NADH:cytochrome c oxidoreductase in healthy cells and as an inducer of caspase independent apoptosis (Joza et al. 2009). Mutations in this gene can result in combined oxidative phosphorylation deficiency 6 (COXPD6) and Cowchock syndrome (Cowchock et al. 1985; Fischbeck et al. 1986; Ghezzi et al. 2010). A proviral insertion in the Aifm1 gene resulted in an 80% reduction in the harlequin mouse model (Hq) (Klein et al. 2002). Further analysis of this model of neurodegeneration showed an isolated effect on CI function. Mitochondria from retinal and neuronal tissue of Hq mice had reduced CI activity, abundance and expression of CI subunits (Vahsen et al. 2004). The relationship between levels of Aifm1 and CI was further underlined when conditional deletion within skeletal and cardiac muscle resulted in an 80% reduction in CI activity, atrophy of the skeletal muscles and cardiomyopathy (Joza et al. 2005). However, patients who carry mutations in AIFM1 do not seem to carry complementary defects in CI but instead in respiratory CIII and CIV (Ghezzi et al. 2010). Further analysis of patients and available models will help elucidate the link between these two entities.

10.2.9.2 FXN

FRATAXIN (FXN) is a mitochondrial protein which has been shown to play roles in several processes, including heme and iron-sulphur cluster biosynthesis and antioxidant defence (Ristow et al. 2000; Shoichet et al. 2002; Martelli et al. 2007). Friedreich’s ataxia (FDRA) is a mitochondrial disease which is caused by expansion of a GAA triplet repeat in an intron of FXN (Campuzano et al. 1996). Patients suffer from neurodegeneration, cardiomyopathy and diabetes, presenting with a distinct biochemical signature which includes a decrease in the levels of iron-sulphur cluster containing enzymes and an increase in markers of oxidative damage (Delatycki et al. 2000).

Homozygous Fxn null mice are early embryonic lethal, demonstrating the importance of Fxn in early embryogenesis (Cossée et al. 2000). Conditional KO of Fxn was performed in the heart, brain, liver and pancreatic β cells (Puccio et al. 2001; Ristow et al. 2003; Simon et al. 2004; Martelli et al. 2012). As in patients, tissues specific depletion of Fxn in the brain resulted in progressive neurodegeneration, in the heart in cardiomyopathy and in the pancreas, the onset of diabetes (Puccio et al. 2001, Ristow et al. 2003, Simon et al. 2004, Martelli et al. 2012). However, these phenotypes were accelerated and enhanced in mouse models when compared to patients but these mice failed to show any signs of oxidative damage (Puccio et al. 2001, Ristow et al. 2003, Simon et al. 2004, Martelli et al. 2012). Given what is known about other forms of repeat disease such as several forms of spinocerebral ataxia and myotonic dystrophy, the effects of expansion of the GAA triplet on the levels of FXN may be more nuanced than a complete KO of the gene. Accordingly a mouse model which overexpresses the human FXN gene with the GAA repeat mirrored the patient phenotype, developing a milder progressive form of ataxia with an accumulation of oxidative damage (Al-Mahdawi et al. 2006).

10.2.9.3 ETHE1

ETHE1 encodes a detoxification enzyme, present within mitochondria, preventing the accumulation of damaging hydrogen sulphide (H2S) (Tiranti et al. 2004). Ethylmalonic encephalopathy (EE) is a severe childhood mitochondrial disorder characterised by encephalopathy, petechiae and gastrointestinal disturbances. Patients also excrete large quantities of ethylmalonic acid in their urine (Tiranti et al. 2004).

To gain a better understanding of EE pathology Tiranti et al created Ethe1 KO (Ethe1−/−) mice and analysed them together with EE patients (Tiranti et al. 2009). They described a striking recapitulation of the patient phenotype, with Ethe1−/− mice showing symptoms of EE within the first two weeks of life and succumbing within the first 6 weeks (Tiranti et al. 2009). Ethe1−/− mice had arrested growth, pronounced CIV deficiency in muscle, brain and colon, increased levels of lactate, C4 and C5 acylcarnitines and increased levels of ethylmalonic acid and thiosulfate (Tiranti et al. 2009). Conditional ablation in muscle, brain and liver failed to phenocopy Ethe1−/− mice (Di Meo et al. 2011). While CIV deficiency was clear in targeted tissues, it was not sufficient to result in increased levels of ethylmalonic acid and thiosulfate in urine, observed in patients and Ethe1−/− mice (Di Meo et al. 2011).

The Ethe1−/− mouse model provides an excellent opportunity to develop targeted therapies for EE patients which can prevent or negate the accumulation of toxic H2S.

10.2.9.4 RNASEH1

RNASEH1 encodes a ribonuclease which is required in mitochondria for mtDNA replication (Reyes et al. 2015). RNASEH1 mutations have been associated with adult-onset encephalopathy with chronic PEO and mtDNA deletions (Reyes et al. 2015). RNaseH1 null mice are embryonic lethal due to depletion of mtDNA resulting in cell death (Cerritelli et al. 2003). Conditional and inducible deletion of RNaseH1 specifically in the liver also resulted in reduced levels of mtDNA, with clear defects in mitochondrial morphology. These mice also displayed progressive liver degeneration (Lima et al. 2016). These models have allowed the delineation and conformation of RNASEH1 function. The liver specific models could in future allow modelling of liver dysfunction (hepatopathy) due to mtDNA depletion.

10.3 Drosophila Models of Mitochondrial Disease

The fruit fly, Drosophila melanogaster, has long been used in the study of several human pathologies (Vidal and Cagan 2006; Chan and Bonini 2000; Stefanatos and Vidal 2011). Despite its’ relative simplicity, the fly genome shows remarkable conservation with that of humans, with the advantage of reduced genetic and functional redundancy. As a model organism to study mitochondrial disease, Drosophila present many advantages. There is a high degree of conservation of mitochondrial proteins and many patient phenotypes can be assessed in a whole animal model (Jacobs et al. 2004). The major advantage of the fly is the plasticity that the genetic tools available in Drosophila confer (Jennings 2011). While several mutations in mitochondrial proteins have been isolated in Drosophila, the true power of this model organism lies in the ability to utilise methods such as RNA interference (RNAi) in combination with the binary GAL4/UAS system for modulation of gene expression as well as the several methods of transgenesis available (Brand and Perrimon 1993; Lee and Luo 2001; St Johnston 2002; Dietzl et al. 2007; Perkins et al. 2015). Together these tools allow tissue specific or global, depletion or deletion of any gene of interest. Over the last 10 years the Drosophila research community has established several models of mitochondrial dysfunction. While we must keep in mind the differences between mammalian and insect physiology, the fly possesses similar if not the same cellular and physiological functions with a high degree of conservation. Importantly, flies with mutations in mitochondrial genes recapitulate several aspects of mitochondrial disease (Toivonen et al. 2001; Vartiainen et al. 2014).

10.3.1 Complex I

Li and colleagues have utilized P-element mediated excision of CG12079/ND-30, the fly orthologue of CI subunit Ndufs3, to create novel mutant alleles (Li 2013). They find that CG12079/ND-30 mutants have severely reduced CI activity and aberrant assembly (Li 2013). Most developing larvae do not make it to adulthood but characterization of rare escapers revealed increased oxidative stress as well as activation of compensatory increases in mitochondrial biogenesis (Li 2013). These flies also display seizure sensitivity and defective climbing ability, phenotypes which can be extrapolated to mitochondrial disease patients (Jacobs et al. 2004). RNA interference (RNAi) mediated depletion of Ndufs3 results in decreased ATP levels and membrane potential (Vos et al. 2012). Although this model was not further analysed it demonstrates the power of the availability of RNAi constructs against most of the genome to induce mitochondrial dysfunction (Dietzl et al. 2007, Perkins et al. 2015).

Mutants have also been isolated for two CI assembly factors Ndufaf1 (CG7598) and Ndufaf6 (CG15738) (Cho et al. 2012; Zhang et al. 2013). In the case of Ndufaf1, deletion resulted in developmental arrest with defects in mitochondrial structure and a loss of CI holoenzyme (Cho et al. 2012). RNAi mediated depletion of Ndufaf1 (CG7598) also resulted in loss of CI holoenzyme and defects in mitochondrial structure (Cho et al. 2012). Two separate mutations of Ndufaf6 isolated from an unbiased forward genetic screen resulted in decreased CI activity and increased ROS levels (Zhang et al. 2013).

Burman and colleagues took advantage of a mt-ND2 mutant created using a mitochondrially targeted restriction enzyme (Burman et al. 2014). In humans, mutations in MT-ND2 are associated with Leber Optic Atrophy (LHON) and CI deficiency (Johns and Berman 1991; Schwartz and Vissing 2002; Pulkes et al. 2005). ND2del1 flies, which carry a small deletion in ND2, show an increase in heat, mechanical stress, hypoxic and hypercarbic induced paralysis as well as a reduced lifespan (Burman et al. 2014). Although there were no signs of muscle pathology, aged mutants displayed progressive neurodegeneration (Burman et al. 2014). Biochemically ND2del1 flies have an isolated CI respiration defect, reduced levels of assembled CI, reduced membrane potential and ATP production. This model recapitulates many aspects of CI deficiency and provided insights into the role of mt-ND2. Importantly this model has also been recently used to parallel the therapeutic effects of rapamycin described in a mouse model of CI deficiency further validating its use in the investigation of mitochondrial disease pathogenesis (Johnson et al. 2013; Wang et al. 2016).

10.3.2 Complex II

SDHB is one of four subunits of CII. An insertional mutant for Sdhb (CG3283) in flies displayed a CII specific respiratory defect accompanied by a reduction in CII activity and an increase in mitochondrial ROS (Walker et al. 2006). Mutation of Sdhaf3 (CG14898) resulted in reduced CII activity, increased sensitivity to oxidative stress and motility defects (Na et al. 2014). Loss of Sdhaf4 (CG7224) also resulted in a severe decrease in CII activity and increased sensitivity to oxidative stress as well as neurodegeneration and reduced survival (Van Vranken et al. 2014).

10.3.3 Coenzyme Q

In flies, the first model of CoQ deficiency was isolated in the Gould lab using a mosaic screen to identify modifiers of neuroblast size (Grant et al. 2010). This technique in fruit flies allows analysis of wild type and mutant cells within the same tissue, as well as allowing the study of mutations that are lethal, at any stage in development (Lee and Luo 2001). They identified qless (CG31005), the Drosophila orthologue of PDSS1, involved in the first step of COQ biosynthesis (Grant et al. 2010). Investigation of qless in the developing CNS revealed a growth defect in qless mutant clones which were smaller than wild type clones (Grant et al. 2010). qless mutant clones undergo caspase dependant apoptosis and express the marker of mitochondrial stress, Hsp60A. These phenotypes are rescued by dietary supplementation with CoQ10 (Grant et al. 2010). As well as being the first to study the effects of PDSS1 loss in the fly, Grant and colleagues highlight how the power of mosaic analysis for the investigation of gene function.

COQ2 catalyses one of the final reactions in CoQ biosynthesis (Forsgren et al. 2004). Mutants of Coq2 (CG9613) created using P-element mediated excision display developmental arrest and decreased levels of ATP (Liu et al. 2011). Animals heterozygous for mutant CoQ2 had increased lifespan and inhibited insulin signalling which has been shown to have life extending effects (Liu et al. 2011). This model does not present a viable model of mitochondrial disease; however, mosaic analysis of this mutation may provide more insights into gene function.

Finally using RNAi technology, Fernandez-Ayala and colleagues have systematically depleted enzymes required for CoQ biosynthesis including, Pdss1 (CG31005), Coq2 (CG9613), Coq3 (CG9249), Coq5 (CG2453), Coq6 (CG2453), Coq7 (CG14437), Coq8 (CG32649), Coq9 (CG30493) and Coq10 (CG9410) ubiquitously. Apart from Coq6, depletion of these enzymes resulted in development lethality (Fernandez-Ayala et al. 2014). Coq6 knockdown flies had severe CoQ deficiency and the presence 5-dimethoxy ubiquinol (DMQ) a late stage CoQ biosynthesis intermediate was detected in COQ3, 6, and 9 knockdown animals (Fernandez-Ayala et al. 2014).

Mutation or RNAi mediated depletion of biosynthetic CoQ enzymes in flies results in primary CoQ deficiency. As such these models will allow study of the role of CoQ in development and mitochondrial biology in health and disease.

10.3.4 Complex III

The Zeviani Laboratory described mutations in tetratricopeptide repeat protein TTC19 which resulted in severe CIII deficiency accompanied by progressive encephalopathy (Ghezzi et al. 2011). A fly KO line for Ttc19 (CG15173) created by transposon insertion displayed a severe deficiency in CIII activity, reduced locomotor activity and an increased sensitivity to mechanical stress (Ghezzi et al. 2011). The parallels between the patient phenotype and that seen in the KO flies underlines the high degree of conservation of genes involved in mitochondrial respiration. Importantly, using RNAi technology to knockdown Ttc19 down by 70% did not result in the same phenotypes observed in KO flies, highlighting that for some disease genes more than one model may be required (Ghezzi et al. 2011).

10.3.5 Complex IV

The first model of CIV deficiency described in Drosophila was the levy1 mutation in the orthologue of COX6A1, CG11015 (Liu et al. 2007). Initially identified in a screen of temperature sensitive paralytic mutants, levy1 flies have a severe reduction in CIV activity, less ATP, increased sensitivity to mechanical stress and neurodegeneration as early as seven days old (Liu et al. 2007). Even though mutations in Cox6a1 have not been described in patients we have included this model as the severity of the phenotype and clear neurodegeneration recapitulates the main features of CIV deficiency in humans. This is very important given that knockdown of Surf1, that is often found mutated in patients with LS due to CIV deficiency, does not result in neurodegeneration in flies or mice (Da-Re et al. 2014; Dell’agnello et al. 2007).

Ubiquitous depletion of several nuclearly encoded subunits of CIV via RNAi, including Cox6b1,6c, and 7a, results in developmental lethality (Kemppainen et al. 2014). Tissue specific depletion in the CNS resulted in a variety of phenotypes including no effect i.e. indistinguishable from wild type and severe neurodegeneration depending the targeted subunit (Kemppainen et al. 2014). Again, highlighting that for some disease genes RNAi knockdown models may not be appropriate.

The SCO1 and SCO2 genes, represented by one member in flies (CG8885), are required for the biogenesis of the core subunit of CIV, MT-CO2. Patients carrying mutations in SCO1 suffer from CIV deficiency (Valnot et al. 2000). In flies, null mutations of CG8885 result in developmental lethality while milder alleles result in female sterility (Porcelli et al. 2010). All mutants display a reduction in CIV activity with a reduction in locomotor activity in mutant adults (Porcelli et al. 2010). Given these robust phenotypes flies carrying patient relevant mutations in this gene may provide a more tractable system to study disease pathogenesis and therapeutic validation.

Finally, van Bon and colleagues studied the role of Cep89 thought to modulate CIV activity in vitro in cell culture and in vivo in Drosophila, after isolating a mutation in a patient suffering from CIV deficiency (Van Bon et al. 2013). Ubiquitous RNAi depletion of the Drosophila orthologue of Cep89 (CG8214), resulted in reduced CIV activity and developmental lethality (Van Bon et al. 2013). Tissue specific loss of Cep89 in either the muscle or CNS also resulted in lethality with a few escapers who presented with severe motor defects. Closer inspection of viable knockdown flies revealed serious neuronal aberrations and learning defects. This study highlights how use of complementary models can allow molecular and pathological examination which can elucidate the role of novel mitochondrial disease genes identified in patients.

10.3.6 Complex V

Mutations of the mitochondrially encoded CV subunit MT-ATP6 have been shown to cause maternally inherited LS (MILS), Neuropathy, ataxia, retinitis pigmentosa (NARP) and familial bilateral striatal necrosis (FBSN). An endogenous missense mutation of mt-atp6 was isolated from Drosophila ANT1 mutants, sesb1 (Celotto et al. 2006; Palladino 2010). mt-atp6 mutant flies (ATP61) had abnormal mitochondrial morphology, no detectable CV activity but respiration was not affected. ATP61 flies had a reduced survival, progressive neuromuscular pathology and locomotor deficiency, mirroring the main features observed in MILS patients (Celotto et al. 2006, Palladino 2010).

10.3.7 Associated Mitochondrial Proteins

Excluding those genes which are directly implicated in OXPHOS, knockout and knockdown models of several associated mitochondrial proteins have been created in Drosophila. In humans, mutation of ANT1, the ADP/ATP transporter, results in Progressive External Ophthalmoplegia with Mitochondrial DNA deletions (adPEO) and mitochondrial DNA depletion syndrome (Kaukonen et al. 2000; Thompson et al. 2016). Flies carrying a point mutation in sesB (CG16944), the Drosophila ANT1 orthologue, are developmentally delayed, sensitive to mechanical stress and have compromised respiration (Zhang et al. 1999; Vartiainen et al. 2014). Analysis at the transcriptional and metabolic level revealed a glycolytic shift and increased mitochondrial stress (Vartiainen et al. 2014). Importantly, this phenotype overlapped with that of the fly model of mitochondrial disease, tko25t, orthologous to human MRPS12 (Royden 1987; Toivonen et al. 2001; Vartiainen et al. 2014). tko25t flies are developmentally delayed, sensitive to mechanical stress, have a combined OXPHOS defect and display metabolic remodelling (Fernandez-Ayala et al. 2010; Toivonen et al. 2001). This strongly suggests that some mitochondrial diseases could have common therapeutic targets.

Mutation of the DNA polymerase POLG2 (CG33650) in developing larvae leads to depletion of mtDNA, increased mitochondrial biogenesis and altered mitochondrial neuronal transport (Baqri et al. 2009). This study provided key insights into how altered mitochondrial trafficking in neurons populated with mtDNA deficient mitochondria may underlie pathogenesis of mtDNA depletion syndromes. The Larsson lab have recently used genetic engineering to create flies with an exonuclease (DmPOLγexo-) deficient form of POLG (CG8987) (Bratic et al. 2015). They found that DmPOLγexo- flies accumulated mtDNA mutations, although several fold less than the equivalent “deletor” mouse model despite a clear developmental arrest at the larval stage (Bratic et al. 2015). Using this model, the authors could assess the transmission of these deletions and the existence of a mtDNA mutation threshold after which pathogenic effects were observed.

The fluid state of the mitochondrial network is regulated by the action of Dynamin related protein 1 (DRP1), which promotes mitochondrial fission, and Optic Atrophy 1 (OPA1) and Mitofusin 1 and 2 (MFN1/2) which promote mitochondrial fusion. Homozygous mutation of Drp1 (CG3210), Opa1(CG8479) and Marf (CG3869) are developmentally lethal (Verstreken et al. 2005; Dorn et al. 2011; Debattisti and Scorrano 2013). Examination of Drp1 mutant larvae uncovered a role for DRP1 in maintaining cellular distribution of neuronal mitochondria and mitochondrial function as ATP levels were also depleted in Drp1 mutants (Verstreken et al. 2005). UAS/GAL4 mediated depletion of Opa1 or Marf, the single orthologue of MFN1 and 2, within cardiac muscle resulted in altered mitochondrial morphology and cardiomyopathy (Dorn et al. 2011). Confirming the functional redundancy between the human and fly genes, co-expression of MFN1 or MFN2 could rescue these phenotypes in Marf knockdown flies (Dorn et al. 2011). Opa1 heterozygous flies have a reduced lifespan, increased oxidative stress, altered mitochondrial morphology and reduced activity of CII and III. Mosaic analysis of Opa1 mutations demonstrated the requirement of Opa1 in the developing eye, where altered mitochondrial morphology and increased ROS levels resulted in a miss-patterned eye due to ectopic cell death (Yarosh et al. 2008).

10.4 Zebrafish Models of Mitochondrial Disease

Danio rerio, commonly referred to as the zebrafish, is a popular vertebrate model system used extensively in developmental studies (Eisen 1996; Fishman 1999). As tools for reverse genetic approaches in zebrafish have improved, their use to study the function of human disease genes has increased (Eisen 1996, Fishman 1999). There are several advantages to working with zebrafish including the availability of genetic tools for gene silencing and transgenesis, as well as high degree of genetic conservation (Wienholds et al. 2003; Soroldoni et al. 2009; Nasevicius and Ekker 2000).

To study mitochondrial disease using zebrafish, several tools and assays have been developed over the last ten years that allow real time visualization of mitochondrial networks and the assessment of physiological effects of mitochondrial dysfunction (Broughton et al. 2001; Kim et al. 2008; Plucinska et al. 2012). Below we highlight those cases where genes described to be mutated in patients have been modelled in zebrafish.

The first zebrafish model of mitochondrial dysfunction was described by Baden and colleagues (Baden et al. 2007). Using morpholinos against CIV subunit, COX5A (cox5aa and cox5ab) and CIV assembly factor, SURF1 (surf1), they could reduce CIV activity by around 50% (Baden et al. 2007). They also found a concomitant decrease in the levels of the mitochondrially encoded core subunit MT-CO1 (mt-co1). In all cases, morpholinos against cox5aa, cox5ab, or surf1 resulted in severe developmental phenotypes and ectopic apoptosis (Baden et al. 2007). Mutant embryos displayed delayed and malformed development of the nervous system, gut and heart. As such these embryos die in embryogenesis. Although this model reveals details around the effects of CIV deficiency during development which may be informative for infantile syndromes associated with CIV deficiency there are many patient features that cannot be extrapolated here. However, given the strong developmental effects which have been so carefully characterised here, this model could be combined with the developmental screens for small molecules which alleviate the phenotypes described.

More recently zebrafish were used to study the role of CIII subunit Ubiquinol-Cytochrome C Reductase Binding Protein (UQCRB) in vivo (Cho et al. 2013). UQCRB has been shown to be mutated in a patient suffering from CIII deficiency (Haut et al. 2003). The focus of the study was the role of CIII in angiogenesis, which was impaired in embryos treated with a small molecule inhibitor of uqcrb, terpestacin or injected with morpholinos against uqcrb (Cho et al. 2013). Importantly, there were also clear developmental effects of uqcrb knockdown suggesting that this model could also be used for developmental screening of effective therapies.

Developmental defects were also observed when two regulators of mitochondrial dynamics, MFN2 (mfn2) and OPA1 (opa1), were depleted independently using specific morpholinos (Vettori et al. 2011; Rahn et al. 2013). mfn2 knockdown resulted in decreased survival, with only 60% of injected embryos surviving (Vettori et al. 2011). Surviving animals had neuromuscular defects resulting in motility defects, reminiscent of patients carrying MFN2 mutations which cause CMT2A (Vettori et al. 2011). As such this is a useful model for the investigation of CMT2A while the characterization of the developmental defects will allow easy screening of effective small molecules. Depletion of opa1 also resulted in developmental defects with larvae dying at 7 days post fertilisation (dpf). opa1 knockdown larvae had a fragmented mitochondrial network, reduced respiration and compensatory upregulation of mitochondrial biogenesis and transcription (Rahn et al. 2013). Although not directly extrapolatable to human patients with OPA1 mutations, this like those zebrafish models described above could be used to identify active therapeutic small molecules.

10.5 Caenorhabditis elegans Models of Mitochondrial Disease

The transparent nematode C. elegans has been used extensively over the last 50 years to study diverse and complex biological processes (Kenyon 1988). Its short life cycle, defined cellular lineage and array of tools for genetic manipulation have made it a powerful force in the study of gene function (Kenyon 1988). The simplicity of its nervous system coupled with conservation of neuronal subtypes and communication make it an attractive model to study neurological disorders. Given the severe effects that mitochondrial dysfunction has on the nervous system of patients with mitochondrial disease, using C .elegans to study mutations identified in patients will allow investigation of the neuropathology of several mitochondrial disorders.

Several classical genetic mutants which correspond to mitochondrial genes found mutated in patients have been isolated in C. elegans (Kayser et al. 2001; Kayser et al. 2004; Ishii et al. 1998; Grad and Lemire 2004; Tsang et al. 2001). Mutants of the NDUFS2 orthologue, gas-1 have a reduce lifespan, retarded growth, increased sensitivity to temperature, oxidative stress and hyperoxia, as well as a reduced CI dependant respiration (Kayser et al. 2001). Mutation of mev-1, the orthologue of SDHC result in increased oxygen sensitivity and accelerated ageing. Analysis of the CII function demonstrated the requirement of this subunit for the participation of CII in electron transport (Ishii et al. 1998). Grad and Lemire created 3 models expressing distinct patient mutations of nuo-1, the orthologue of NDUFV1 (Grad and Lemire 2004). In all three cases, mutations of nuo-1 resulted in decreased lifespan, premature ageing, compromised respiration and lactic acidosis (Grad and Lemire 2004). clk-1, orthologue of COQ7, mutants are unable to produce CoQ9, accumulate the intermediate DMQ9 and display defects in CoQ dependant activities of CI and CIII, as has been described in other animal models of COQ7 loss (Kayser et al. 2004; Fernandez-Ayala et al. 2014).

These studies clearly demonstrate that mutation of genes found in patients with mitochondrial disease can induce many of same phenotypes in worms, validating their use as a model for the study of molecular pathogenesis and the effectiveness of available therapies.

10.6 Conclusion

As can be appreciated from the studies summarised, the extent of mitochondrial disease models currently available is impressive. Each with its advantages and limitations. The last 25 years of mitochondrial research has elucidated many important aspects of these organelles functions in health and disease. Despite this progress patients remain without treatments. As we move forward the focus remains on understanding the molecular pathology underlying mitochondrial disease but increasingly it will be on the development and validation of therapies. These aims both rely on a foundation of disease relevant and genetically defined models. Models of mitochondrial disease in flies, worms and fish display complex and pleiotropic phenotypes of mitochondrial syndromes which can be extrapolated to human patients. Although the mouse is an excellent model for human disease, the use of complementary models suitable for high throughput screening, where patient specific mutations can be introduced using state of the art genome editing will accelerate progress in both these areas.