The mitochondrial brain: From mitochondrial genome to neurodegeneration

Mitochondrial DNA mutations are an important cause of neurological disease. The clinical presentation is very varied in terms of age of onset and different neurological signs and symptoms. The clinical course varies considerably but in many patients there is a progressive decline, and in some evidence of marked neurodegeneration. Our understanding of the mechanisms involved is limited due in part to limited availability of animal models of disease. However, studies on human post-mortem brains, combined with clinical and radiological studies, are giving important insights into specific neuronal involvement.


Introduction
Mitochondrial DNA (mtDNA) mutations are increasingly becoming recognised as important causes of disease. They were first identified in 1988 [1,2] and were initially thought to be relatively rare, however, the incidence of known pathogenic mutations in the general population has recently been investigated. These studies have shown that approximately 1 in 500 individuals carry the m.3243A>G mutation [3], which can cause severe neurological disease, with a similar figure observed for the m.1555A>G mutation, which causes aminoglycoside-induced deafness [4,5]. Another approach to study the incidence of mtDNA disease is to document the number of clinically affected cases within a specific geographic region. This approach is limited because there is marked clinical variability which means many patients go unrecognised and because patients may not be referred to specialist centres. However, even these studies show there is a high disease burden with at least 1 in 10,000 of the adult population suffering from mtDNA disease [6].
In the twenty years since mtDNA mutations were first associated with human disease it has become apparent that the central nervous system is one of the predominant systems affected, especially in those patients with a severe phenotype. Despite this we only have limited understanding of the mechanisms of neurodegeneration. One thing that has become clear, however, is that certain mtDNA mutations have predilections for causing neurodegeneration in certain areas of the brain.
Patients with mtDNA disease present with a plethora of neurological manifestations, in particular, optic atrophy, ataxia, seizures, progressive weakness, dementia, stroke-like episodes and extrapyramidal features. The In addition to the increasing number of patients being recognised with mtDNA disease, mtDNA mutations have also been found to be present in high levels in several ageing tissues. For example, mtDNA mutations have been identified in the neurons of the substantia nigra in both elderly control subjects and in patients with Parkinson's disease [7,8]. There is a direct link between the presence of these mutations and a mitochondrial defect in these cells. Thus a greater understanding of disease mechanisms in patients with mtDNA disease may also give important insights into neuronal loss or dysfunction in ageing and age-related neurodegenerative diseases [9][10][11].
This review focuses entirely on the neurological and neuropathological features of patients with mtDNA disease. Mitochondrial disease can also be caused by nuclear gene disorders [12], many of which mimic the features seen in mtDNA disease, but these are not discussed here. We first review important aspects of mitochondrial genetics, crucial if we are to understand the pathogenesis of the changes seen. We then describe the clinical features seen in patients and the associated neuropathology. Unfortunately at present there is limited treatment available for patients with mtDNA mutations and we discuss various potential therapies.

Mitochondria -structure and function
Mitochondria are double membrane subcellular organelles which were originally primitive, autonomous, bacteria-like organisms. During evolution these organisms were engulfed by larger eukaryotic cells; enjoying protection in return for providing energy. The relationship was so successful that we are now dependent on mitochondria for over 90% of our cellular energy in the form of ATP generated by the process of oxidative phosphorylation. This endosymbiotic relationship is almost certainly the origin of the mitochondrial genome which rather surprisingly has been maintained throughout evolution [13].

ACCEPTED MANUSCRIPT
Mitochondrial function is closely linked to its unique double-membrane structure.
The outer mitochondrial membrane is porous to small molecular weight substances and encapsulates the mitochondria. The enzymes of the mitochondrial respiratory chain are found within the highly folded inner membrane which separates the matrix from the intermembrane space. The matrix contains the mitochondrial genome, as well as the enzymes involved in the tricarboxylic acid cycle and mitochondrial fatty acid oxidation. The cristae, or finger-like projections, of the inner mitochondrial membrane ensure the maximum surface area between the biochemical substrates in the matrix and the respiratory chain enzymes. The inner membrane is also highly impermeable allowing the free passage of only water, oxygen and carbon dioxide. Mitochondria are often considered as discrete organelles however this is not the case with mitochondria often forming networks within cells. There is constant fission and fusion of mitochondria and if this process is disrupted then mitochondrial disease develops and interestingly accumulation of mutations [14].
The fission and fusion process is also thought to be important in targeting mitochondria for turnover by the process of autophagy [15].

Mitochondrial Genetics and Biology
The mitochondrial genome Mitochondrial DNA was identified in 1963. It has subsequently been shown that multiple copies of this double-stranded, super-coiled, circular molecule are found within the matrix of mitochondria. There are between 2-10 mtDNA copies per mitochondrion, resulting in 1000-100000 copies per cell. This is the only extra chromosomal DNA in human cells. The mitochondrial genome consists of 16,569 A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
base pairs and the entire mtDNA sequence was first reported in 1981 [16] and is referred to as the Cambridge reference sequence. This has since been revised to eliminate errors caused by the use of bovine fragments in the original sequencing project [17].
The mitochondrial genome encodes 13 essential respiratory chain subunits as well as the genetic information for the coding of 2 ribosomal RNAs (rRNA) and 22 transfer RNAs (tRNA) required for the intramitochondrial sequence of these proteins. The majority of mitochondrial proteins are nuclear encoded and these proteins form the other subunits of the respiratory chain as well those required for mtDNA maintenance. Thus nuclear gene mutations of the proteins involved in mtDNA replication and/or repair will have profound affects on the mitochondrial genome.
The mitochondrial genome seems to be particularly vulnerable to damage. It has been proposed that the mutation rate is at least 10 times greater than that of the nuclear genome and there are a number of possible reasons for this. The mtDNA lacks protective histones and has limited repair mechanisms. It is also susceptible to nucleolytic attack from the free radicals produced by oxidative phosphorylation. Combined with this the mtDNA consists entirely of exons and consequently there is very little redundancy. As a result a point mutation or deletion can very quickly lead to a biochemical defect.

MtDNA replication, transcription and translation
The two strands of the mitochondrial genome differ in their distribution of bases C and G making one strand denser (called the heavy or H strand) than the other (light or L strand). There is a non-coding control region of about 1.1kb which is often referred to as the D-loop. This region contains essential sequences for initiation of replication and transcription.
MtDNA replication: Individual mtDNA can replicate independently of the cell cycle and is therefore described as relaxed. Clayton originally described mtDNA replication as an asynchronous mechanism originating from two points O H and O L [18]. An RNA primer generated from the light chain initiates replication of an mtDNA molecule at the origin of the heavy chain, O H, found in the D-loop. DNA polymerase gamma, encoded by a nuclear gene, then synthesises the DNA strand. The origin for the light chain, O L , is a small non-coding region surrounded A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
by tRNA genes. It is exposed when the heavy strand passes this region, about 2/3 of its way around the genome and L-strand replication is initiated in the opposite direction.
Recently a strand-coupled method has been proposed in which the two strands are synthesised simultaneously [19]. In this model of mtDNA replication, lagging L-strand synthesis starts shortly after the initiation of replication at O H and may involve extensive RNA synthesis prior to DNA synthesis. However, there remains debate as to which form of mtDNA replication predominates and even whether different forms of replication occur in different tissues.

MtDNA transcription:
There is less contention as regards mtDNA transcription and many of the proteins involved have been identified and an in vitro system established. Mitochondrial transcription is initiated from promoters on both H and L strand generating polycistronic transcripts, which are then processed to produce the individual mRNA, rRNA and tRNA molecules. This process requires a mitochondrial RNA polymerase, a transcription activator called TFAM, and either mitochondrial transcription factor B1 or B2 [20].

MtDNA translation:
Mitochondrial translation is still an area in which there are many uncertainties. It is a process controlled by nuclear encoded proteins [21] which include two specific mitochondrial initiation factors [22], three mitochondrial elongation factors [23] and at least one termination release factor [24].

Mitochondrial DNA mutations
The first pathogenic mtDNA mutations were described in 1988 and since then over 300 mtDNA mutations have been reported (Mitomap) [25]. These mutations show evidence of a biochemical defect, observed as a mosaic of cells lacking mitochondrial enzyme activity, often as cytochrome c oxidase deficiency [26]. This is due to a threshold effect with mitochondrial function only being affected if there are high levels of mutated mtDNA within individual cells [27]. It is thought that the critical factor is the amount of wild-type mtDNA present to complement the defective mtDNA. Typically between 70-90% mutated mtDNA is required for the clinical phenotype to develop [28], but this varies markedly for different mtDNA mutations. The threshold level also appears to vary for different cell types and this may explain why some tissues are more affected in the presence of some mtDNA mutations than others.
Mitotic segregation: in dividing cells not all daughter cells will contain the same level of mutated mtDNA. In cells undergoing mitotic division the cells with high levels of mtDNA mutation appear to be at a disadvantage and the mtDNA mutation is lost during life [29]. In post-mitotic cells such as muscle and neurons there cannot be loss of the mutated mtDNA by mitotic segregation and this may be a factor in their frequent involvement in mitochondrial disease. proposed that specialised processes within each cell, particularly within neuronal tissue, have a different energy requirement for example, for neurotransmitter release, ion pumping and electrical transmission [34]. In a similar way to tissuespecific dysfunction, cells may be functionally affected in different ways.

Maternal transmission and bottleneck
Evidence is accumulating for a role of apoptosis-induced cell loss, production of reactive oxygen species and altered calcium metabolism leading to neuronal dysfunction and neuronal degeneration although the relative importance of these three factors remains ambiguous. By identifying the underlying pathogenesis it may be possible to identify therapeutic solutions.

Oxidative stress
Oxidative damage has been implicated as a precursor to apoptosis or programmed cell death. Reactive oxygen species (ROS) are predominantly formed as a by-product of oxidative phosphorylation in the mitochondria [35] with at least 9 submitochondrial ROS-producing sites identified. It is estimated that a significant proportion of oxygen is converted to the superoxide anion during oxidative phosphorylation which is subsequently converted to hydrogen peroxide [36], an important mediator of cellular damage. The enzymes of the respiratory chain are in close proximity to the mtDNA which, in combination with the lack of protective histones, predisposes the genome to mutagenesis. These findings are consistent with studies into pathological mechanisms in specific mtDNA diseases whereby increased ROS production is seen in neuronal NT2 cybrid cells with the LHON mutation. In addition, exogenous administration of

ACCEPTED MANUSCRIPT
antioxidants can protect against ROS-mediated damage in cybrid cells containing the m.8993T>G NARP mutation [37].
There are unfortunately very few models of mtDNA disease in which to explore mechanisms [38]. This is because mtDNA mutations have limited capacity to pass through the germ-line. An alternative approach has been to develop transgenic mice with mutation in the proof reading domain of the mitochondrial polymerase gamma [39,40]. These mice accumulate mtDNA defects and interestingly there was no evidence of increased sensitivity to oxidative stress-induced cell death or of increased levels of ROS ( [40,41]. Intermittent periods of encephalopathy characterised by raised CSF and plasma lactate but not associated with seizures or stroke-like episodes are also seen in patients with m.3243A>G MELAS. A combination of recurrent encephalopathy and stroke-like episodes are thought to cause a slowly progressive neurodegeneration leading in some patients to cognitive decline and dementia [46]. Neuropathological studies: Neuropathological studies have identified infarctlike areas in the white matter of the cortex and subcortex, concentrating principally in the parietal, temporal and occipital lobes [50]. The cerebellum, thalamus and basal ganglia may also be affected [51,52] (Figure 1) . Infarcts are multiple, asymmetrical and, unlike vascular lesions, are not restricted to a particular vascular territory. Microscopically, however, they are similar to true infarcts in both acute and chronic stages. They may be associated with considerable neuron loss, astrogliosis and microvauolation [53]. Ventricular dilatation accompanying cortical atrophy due to extensive neuronal loss is also seen as MELAS progresses.

Apoptosis
Calcification of the vasculature of the basal ganglia is also a common neuropathological feature seen in patients with MELAS symdrome and can be identified on neuroimaging [51,54]. Despite this finding the neurones in the basal ganglia are relatively spared and patients rarely present with evidence of basal ganglia dysfunction. The pattern of calcification is similar to that seen in normal ageing and it is possible that the process is simply accelerated by the mitochondrial dysfunction in MELAS.

ACCEPTED MANUSCRIPT
Extensive damage to the cerebellum is also common in MELAS with loss of Purkinje cells and development of cerebellar ataxia clinically. In severely affected patients cactus formation is visible on Purkinje cells which are thought to be due to the accumulation of abnormal mitochondria within dendrites similar to that seen in ragged red fibres on muscle biopsy.

Mechanisms:
One theory for the generation of the necrotic lesions in MELAS is the vascular hypothesis which proposes that a deficit in oxidative phosphorylation in cerebral arteries causes an angiopathy which leads to alterations in vascular tone, ischaemia and infarction. This is suggested by the presence of enlarged, abnormal mitochondria in the endothelial and smooth muscle cells of these arteries [55]. Other studies suggest that the proliferation of mitochondria may be a consequence rather than a cause of cerebral ischaemia. Compelling evidence for the vascular hypothesis in MELAS comes from the post mortem examination of the blood vessels of two patients with MELAS. There was widespread respiratory chain deficiency in the blood vessels and high levels of the m.3243A>G mutation [53]. These results were compared to similar regions from controls which were found to have normal respiratory chain function. The presence of respiratory deficient blood vessels throughout the brain, however, implies that another mechanism must be responsible for dictating the cortical selectivity of brain lesions.
Electron-microscopy has demonstrated abnormal mitochondria within neurones as well as the brain vasculature [56] and data from MR spectroscopic data has shown impaired oxidative metabolism and increased concentrations of lactate in cortical lesions during acute episodes. Observations of impaired glucose uptake in the occipital and temporal regions with positron-emission tomography (PET) in a patient with MELAS may correspond to a higher metabolic rate of neurons in this region [57]. These findings have led to the formulation of the metabolic hypothesis of neurodegeneration in MELAS, which goes some way to explaining the focus of cortical lesion in the parieto-occipital lobes where it is argued there may be a higher metabolic demand on neurones.

ACCEPTED MANUSCRIPT
Clinical Features: MERRF is a severe neurodegenerative disorder, which often presents in childhood or early adulthood following normal development [47].

MtDNA mutation: MERRF is caused most commonly by a point mutation in the
MT-TK gene at position 8344 in the mitochondrial genome [58]. MT-TK encodes the tRNA Lys which is essential for protein synthesis.

Neuropathological features:
The dentate nucleus is particularly targeted in patients with the m.8433A>G mutation leading to severe neuronal loss accompanied by astrocytosis. Some studies have also demonstrated abnormally large mitochondria with inclusion bodies in this area [59]. An interesting microdissection study of a patient with m.8433A>G examined the proportion of mutant mtDNA in different cell types. This identified the vulnerability to this mutation of neurones in the dentate nucleus ( Figure 2) where 45% neuronal loss was noted compared with 7% of cerebellar Purkinje cells. This loss occurred despite the finding that the mutation load in Purkinje cells was 97.6±0.7% compared to 89.0±1.5% in the dentate neurones [60]. Purkinje cell loss is usually mild in MERRF.
The gracile and cuneate nucleus, Clarke's column of the spinal cord, the inferior medullary olives, the pons, and red nuclei of mid-brain have all been noted to suffer from some neuronal loss [51]. Although infarct-like areas are found in MERRF as in MELAS, loss of cerebral neurones is rare.

Mechanisms:
The molecular genetics of the m.8433A>G mutation have been explored in vitro with studies of cultured myotubes containing the mutation [61].

ACCEPTED MANUSCRIPT
In cells containing over 85% mutant impaired protein translation was demonstrated, especially in those proteins with a large number of lysine residues.
The authors postulate that the mutation was functionally recessive as a return to  increased susceptibility to apoptosis [66].
LHON is usually due to a homoplasmic mtDNA mutation; all copies are mutated mtDNA. As such, all maternal offspring will inherit the mutation, however, whilst 50% of males will be affected only 10% of females will develop visual loss. This incomplete penetrance implies a role for nuclear genetic and environmental factors in modulating the expression of the mutation, whilst the male preponderance suggests that there may be an X-linked susceptibility locus [68]. recognised that patients with a mutant load greater than 90% have a predominant CNS presentation which usually occurs in infancy [69]. This is known as Maternally Inherited Leigh Syndrome (MILS). As Leigh Syndrome progresses it is associated with stepwise developmental delay followed by developmental regression and death due to respiratory failure. Many other mutations, both mitochondrial DNA and nuclear, can cause Leigh's syndrome suggesting the developing brain is particularly vulnerable to disturbances of energy metabolism.

MtDNA mutation:
The m.8993T>G mutation was first described in a single family in which four members had a combination of peripheral neuropathy, ataxia and retinitis pigmentosa [70]. Following this original report, it has also been recognised that a mutation at the same base m.899T>C may also cause the NARP/MILS phenotype. Other mutations of the ATPase 6 gene may also cause the NARP phenotype [71]. Neuropathological studies: There have been few studies of extraocular muscles in this condition, but skeletal muscle biopsy typically reveals cytochrome c oxidase (COX) deficient fibres. Some of these fibres demonstrate characteristic sub-sarcolemmal accumulation of abnormal mitochondria, the classical raggedred fibre. Single muscle fibre analysis has revealed levels of pathogenic mtDNA deletions above a critical threshold level of > 80% mutant load in these COX deficient fibres.

Mechanisms:
The extraocular muscles (EOM) have a wide dynamic range making their structure, biochemistry and immunology distinct from skeletal muscle. This has been suggested as a reason for their selective involvement in certain mitochondrial disorders [75]. A study in 2006 looked at the specific findings in the EOM in patients with CPEO using high-resolution orbital MRI.
Unusual signal abnormalities were noted in the EOM with diminished function, particularly in the superior rectus and levator muscles. In these muscles atrophy was also seen, similar to that observed in neurogenic paralysis. MtDNA deletions are predominantly (~85%) flanked by short direct repeats.

Kearns-Sayre Syndrome
About one third of these patients have a 4977 bp deletion, known as the 'common deletion' which has a 13 base pair repeat sequence [76].

ACCEPTED MANUSCRIPT
Neuropathological studies: The neurodegeneration observed in KSS affects both white and grey matter in the brain (Figure 4) with widespread spongiform degeneration recognised as a key histological feature of the disease. White matter changes are noted in the cerebrum, cerebellum, thalamus, basal ganglia and spinal cord [51]. The appearance of the white matter depends upon the level of mitochondrial dysfunction and varies from mild atrophy to extensive vacuolation with a sieve-like appearance. Oligodendrocytes appear to be more dependent on optimal mitochondrial activity as they are preferentially affected in KSS. Examination of brain tissue under electron microscope has demonstrated that splits in the intraperiod line are responsible for the spongy appearance of myelin [77].
The grey matter changes and neuronal loss seen in KSS occur predominantly in the cerebellum, the cerebrum and the brainstem with the cortex and the dentate nucleus being relatively spared. The neuropathological findings described are in corroboration with various neuroimaging studies which routinely show diffuse cerebellar, cerebrum and brain stem atrophy. The relative absence of ATPdependent ion channels and water transporters in astrocytes of the grey matter is thought to be the underlying pathogenic consequence of mitochondrial dysfunction in these areas [78]. To delineate a mechanism for neuronal loss the COX activity of neurones throughout the brain of a patient with KSS were explored. Low levels of COX deficiency were found in the cerebellum,  [78]. It is thought that the mitochondrial dysfunction in these cells leads to an impairment of the blood-brain barrier. This may also play a key role in the pathogenesis of KSS whereby the active transport of folates, which are essential in DNA and RNA synthesis and serotonin metabolism as well as in synthesis of membrane phospholipids, is impaired. induced mitochondrial biogenesis and led to delayed onset of myopathy [84].

Potential Therapeutic Options
Whilst bezafibrate has been used in the treatment of patients with mitochondrial A C C E P T E D M A N U S C R I P T

ACCEPTED MANUSCRIPT
fatty acid defects [85], there have been no published studies in patients with mtDNA disease.
There have been a number of experimental molecular approaches based on either complementing the mtDNA defect or changing the balance of mutated to wildtype mtDNA. To complement mtDNA protein defects it is possible to express the wild-type protein allotopically from nuclear transfected constructs. In cell lines carrying the m.8993T>G NARP mutation, allotropic expression of the wild-type ATPase 6 protein partially rescued the biochemical phenotype [86]. It is also possible to complement mitochondrial tRNA defects. Yeast cytosolic RNA can surprisingly be imported into human mitochondria and it has been shown that the imported tRNA LysCUU can partly rescue the biochemical defect in cell lines carrying the m.8344A>G MERRF mutation [87]. Techniques to shift the mitochondrial genotype have also been investigated including the selective inhibition of mutated mtDNA replication [88] and the import of restriction endonucleases which specifically target the mutated mtDNA [89].
Another approach for the treatment of mitochondrial diseases is expression of single subunit alternative oxidases (AOX) which are found in many eukaryotes but not mammals. These oxidases have the potential to bypass the biochemical defect in oxidative phosphorylation. Recent studies have shown that transgene AOX expression does not produce a detrimental phenotype in Drosophilia and was able to rescue the phenotype two different mitochondrial disease models [90].
Although there is considerable effort to develop treatment for mtDNA diseases, most therapies are a long way from potential clinical use. An alternative approach is to use the unique maternal inheritance pattern of mtDNA diseases to try to prevent the transmission of mtDNA diseases from mother to offspring.
Preimplantation genetic testing is of value for some women with heteroplasmic mtDNA defects. Pronuclear transfer in early stage embryos may over a novel approach to preventing transmission of both heteroplasmic and homoplasmic mtDNA disease [91].