REVIEW

Neuroimaging in Mitochondrial Disorders

Andrea L. Gropman

Published online: 4 December 2012# The American Society for Experimental NeuroTherapeutics, Inc. 2012

Summary Mutations in either nuclear DNA or mitochon-drial DNA can result in disruption of oxidative phosphory-lation and lead to mitochondrial dysfunction. Mitochondrialdisease manifestations occur predominantly in the centralnervous system, peripheral nervous system, and/or involveseveral organ systems. The consequences range from man-ifestations of a single organ or tissues, such as musclefatigue, if confined only to muscle, seizures, intellectualdisabilities, dementia, and stroke (if to the central nervoussystem), leading to disability or even early death. The de-finitive diagnosis of a mitochondrial disorder can be diffi-cult to establish. Criteria and checklists have beenestablished and are more reflective of adult disease. How-ever, in children, when symptoms suggest a mitochondrialdisease, neuroimaging features may have more diagnosticimpact and additionally these can be used to follow thecourse, evolution, and recovery of the disease. This reviewwill demonstrate the common neuroimaging patterns inpatients with mitochondrial disorders and point out howvarious newer neuroimaging modalities may be exploitedto glean information as to the different aspects of mitochon-drial dysfunction or resulting neurological and cognitivedisruption, although reports in the literature using thesemethods remain sparse.

Keywords Brain injury . diffusion tensor imaging (DTI) .

functional MRI (fMRI) . inborn error of metabolism .

metabolism . magnetic resonance imaging (MRI) . magnetic

resonance spectroscopy (MRS) .metabolism .mitochondrialdisorder.

Introduction

Mitochondria are small organelles measuring 0.5–1.0microns that are found in the cytoplasm of all eukaryoticcells. Although they are small, they play a very importantrole in generating cellular energy, in the form of adenosinetriphosphate (ATP). Mitochondria produce ATP via oxida-tive phosphorylation, a process involving a series ofreduction-oxidation reactions during which electrons aretransferred through several multimeric complexes on theinner mitochondrial membrane [1]. Complex V, the ATPsynthase, is responsible for the synthesis of ATP from aden-osine diphosphate (ADP) and inorganic phosphate (seeSaneto and Sedensky, this supplement).

The mitochondria have a complex and compact structure,consisting of 4 compartments in which distinct biochemicalprocesses occur: 1) the inner membrane, 2) the outer mem-brane, 3) the intermembrane space, and 4) the matrix. Theenzymes of the electron transport chain involved in oxida-tive phosphorylation are located on the inner membrane.ATP is produced by generating a mitochondrial membranepotential secondary to a proton gradient between the mito-chondrial matrix and the inner mitochondrial membrane.

The term “mitochondrial cytopathy” is used clinicallyand refers to the group of inherited heterogeneous disordersdue to the dysfunctional mitochondrial respiratory chain,manifesting in the central nervous system (CNS) plus orminus other organs [2]. With an estimated prevalence of1:4:000 to 1:5000, they represent a frequent cause of meta-bolic disorders [3] in both pediatric and adult patients.

Mitochondrial disorders are unique in that there are 2genomes that are involved in the production of protein

Electronic supplementary material The online version of this article(doi:10.1007/s13311-012-0161-6) contains supplementary material,which is available to authorized users.

A. L. Gropman (*)Department of Pediatrics and Neurology, Children’s NationalMedical Center and the George Washington University of theHealth Sciences, Washington, DC 20010, USAe-mail: agropman@childrensnational.org

Neurotherapeutics (2013) 10:273–285DOI 10.1007/s13311-012-0161-6

subunits that comprise the respiratory chain complexes: thenuclear genome and the mitochondrion DNA (mtDNA). Mi-tochondrial disorders, therefore, can be due to defects in eithergenome or signaling and regulatory proteins that govern theinteraction between the products of the 2 genomes [1]. Unlikenuclear DNA, mtDNA may be present in multiple copies invarying amounts of wild-type and pathogenic genomes con-taining mutations, and the percentage of normal and mutatedgenomes may vary in the tissues. This concept is referred to asheteroplasmy. The degree of heteroplasmy is a factor in de-termining the severity and number of organ systems involvedin disease, but is not the only factor as the pathogenicity of themutation may play a role as well. For example, symptoms ofsevere mitochondrial disorders may not appear until adult-hood because many cell divisions are required for a cell toreceive enough mitochondria containing the mutant alleles tocause symptoms. Furthermore, most mitochondrial proteinsare encoded by the nuclear DNA, translated in the cytoplasm,and imported into the mitochondria. There is limited encodingcapacity of the mitochondrial genome (mtDNA), which onlyencodes 13 of 80 protein subunits of the oxidative phosphor-ylation system [4] (Table 1).

Mitochondria were first observed more than 100 yearsago by Altman, who called them “elemental organisms”because they were initially believed to be free living organ-isms within the cells. They were first linked to humandisease in 1962 when a patient with hypermetabolism wasfound to have loose coupling of electron transport and ATPsynthesis. In 1963, mitochondrial DNAwas discovered, butthe complete sequence was not available until 1981 [4]. Thegenetic basis of mitochondrial disorders remained obscureuntil 1988 when the first disease-causing mutations of mi-tochondrial DNAwere found [5]. For the last decade, owingto advancement in sequencing technologies, many genesresponsible for nuclear mitochondrial subunits have beenidentified, establishing their role in disease.

Many mitochondrial disorders present with irreversiblebrain injury [6, 7]. Mitochondrial-associated brain injury

patterns are often characterized by whether the processprimarily involves gray matter, white matter, or both, andbeyond that, whether subcortical or cortical gray matternuclei are involved [8]. Although expected to be globalinsults due to impaired energy metabolism, some mitochon-drial disorders result in selective injury to deep gray matternuclei or white matter, whereas others have patterns ofcerebellar atrophy or strokes in nonvascular territories [9].Selectivity for particular brain regions or even cell typesbased on morphology or neurotransmitter systems (astro-cytes/neuron: glutamatergic, Gamma-aminobutyric acid(GABA)-ergic) remains poorly understood, as is the casein other inborn errors of metabolism.

The presenting clinical neurological features of a mito-chondrial disorder may suggest gray or white matter in-volvement. Patients with cortical gray matter involvement,as expected, may present with seizures, encephalopathy, ordementia, whereas deep gray matter injury typically mani-fests with extrapyramidal findings of dystonia, chorea, athe-tosis, or other involuntary movement disorders. Whitematter disorders feature pyramidal signs (spasticity, hyper-reflexia) and visual findings. Involvement of the cerebellumor its connecting tracts may lead to ataxia. It is typical formore than 1 aspect of the brain to be affected in a singlemitochondrial disorder, such as basal ganglia and whitematter giving rise to complex neurological symptoms andsigns that often mimic more common conditions, such asmultiple sclerosis, stroke, or cerebellar ataxias [6].

The clinical presentation in any 1 patient with a mito-chondrial cytopathy is influenced by factors that include thespecific gene mutation, heteroplasmy within tissues, sec-ondary mutations or polymorphisms in both nuclear andmitochondrial genomes, as well as environmental and im-munological factors [10, 11] (Table 2).

Neurological features may reflect dysfunction of any partof the nervous system and encompass a broad range ofcommon neurological symptoms, including dementia, de-velopmental delay, stroke, epilepsy, psychiatric illness,

Table 1 Comparison betweenmtDNA and NuclearEncoded U nits

mRNA 0 messenger RNA;mtDNA 0 mitochondrion DNA;rRNA 0 ribosomal RNA;tRNA 0 transfer RNA

Mitochondrial Nuclear

Respiratory chain components 13 subunits >80 subunits

NADH dehydrogenase 7 >41

Succinate dehydrogenase 0 4

Ubiquinol cytochrome c reductase 1 10

Cytochrome c oxidase 3 10

ATP synthase 2 14

Other proteins 0 DNA/RNA polymerases,and so forth

Protein synthesis components 2 rRNAs All mitochondrial ribosomalproteins (>70)22 tRNAs

(13 mRNAs)

274 Gropman

neuropathy, and myopathy. Children with mitochondrialdisease may present with life-threatening illness in the new-born period; however, the majority of children come toclinical attention for nonspecific problems, including failureto thrive, developmental delay, ataxia, neurosensory loss(vision and/or hearing), seizures, hypotonia, strokes, andloss of developmental milestones.

Neuroimaging and Mitochondrial Disorders

Neuroimaging has already shown its potential for the inves-tigation and management of patients with mitochondrialdisorders allowing clinicians and researchers to study thetiming, extent, and potential of reversibility of neural injury[9], and in many cases providing noninvasive and repeatablebiomarker inquiry. The term “neuroimaging” now encom-passes many investigative modalities beyond routine T1 andT2 imaging, including volumetric analysis, diffusion tensorimaging (DTI), magnetic resonance spectroscopy (MRS),arterial spin labeling, and functional magnetic resonanceimaging (fMRI). Each modality can be combined to gaincomplementary information regarding the brain’s structural,functional, and metabolic dimensions, and how these maybe altered in pathologic states (Table 3). With newer

Table 2 Neurological Markers for Mitochondrial Disorders

Central nervous system Neuromuscular

Development delay/mental retardation Peripheral neuropathy

Hypotonia/floppy baby Exercise intolerance

Autistic features Muscle weakness

Dementia/encephalopathy Cramps after exercise

Headaches/migraine Easy fatigability

Stroke, ischemic episodes Cardiomyopathy

Ataxia Heart block

Episodic coma Arrhythmia

Siezures Ophthalmoparesis, CPEO

Myoclonus or myoclonic seizures Abnormal EMG/NCV

Perinatal insult Ptosis

Pyramidal signs

Hemiparesis

Intractable seizure, refractory

Spasticity

Dystonia

Chorea

CPEO 0 chronic progressive external ophthalmoplegia; AbnormalEMG/NCV 0 electromyogram/nerve conduction velocity

Table 3 Comparison of Imaging Modalities and Potential Role in Mitochondrial Disorders

Modality Principle Application

Structural magneticresonance imaging

Differences in proton spins between tissue typesallow this noninvasive modality to captured three-dimensional images of living tissue

In the clinic, MRI is used to detect microstructuralabnormalities in the human brain. Structural MRIcan be used to investigate volumetric differencesthroughout the brain between experimental andcontrol populations. Imaging parameters can beadjusted to collect different image contrast qualities,such as T1, T2, and FLAIR

Functional magneticresonance imaging

Magnetic differences between oxygenated anddeoxygenated hemoglobin allow active regionsof the brain to image noninvasively in an MRIscanner

fMRI has revolutionized the study of brain function,allowing researchers to administer cognitive tasksconcurrently with fMRI scanning to map cognitivefunctions to individual structures and networks ofstructures within the brain

Diffusion-weightedand diffusion-tensorimaging

These MR-based methods are used: water diffusionand the degree of ordered anisotropy that definewater movement along natural hydrophobic barrierswithin the brain, such as myelinated axons. DWIand DTI provide an index of brain structure

Diffusion-based imaging techniques are highly sensitiveto microstructural abnormalities invisible on traditionalMRI, particularly within the brain’s white matter usedin clinical research studies to identify differences inwhite matter integrity between clinical and healthypopulations

Magnetic resonancespectroscopy

The characteristic location and height of peaks alongthe MR spectrum are used to measure theconcentration of chemicals and metabolites.The three-dimensional structure and chemicalcomposition of organic molecules determine theirspectroscopic resonance frequency and the extentof their “downfield” chemical shift on massspectrum

Using a three-dimensional voxel, researchers andclinicians can use MRS to determine the chemicalprofile within a given region of the brain. This can beused as clinical tool, as in inborn error of metabolismthat alters the concentration of metabolites within thebrain, or by researchers seeking to identifyneurochemical abnormalities in disease populations

DTI 0 diffusion-tensor imaging; DWI 0 diffusion-weighted imaging; FLAIR 0 fluid attenuation inversion recovery; fMRI 0 functional magneticresonance imaging; MR 0 magnetic resonance; MRI 0 magnetic resonance imaging; MRS 0 magnetic resonance spectroscopy

Neuroimaging in Mitochondrial Disorders 275

modalities, such as DTI, MRS, and fMRI, mechanisms ofinjury are starting to be understood in mitochondrial andother metabolic disorders [12, 13].

What Imaging Tools are Available?

Anatomical Imaging: What can it Tell Us?

Routine magnetic resonance imaging (MRI), using T1- andT2-weighted MRI sequences along with fluid attenuationinversion recovery (FLAIR) and voxel-based morphometry(VBM), can be used to interrogate the macroscopic structurein mitochondrial disorders of the brain. However, muchmore information that can help inform clinical decisionscan be achieved as to the disease state using advancedtechnologies. Diffusion-weighted imaging (DWI) and DTI,for example, can be used to study microstructural variancein white matter fiber tracts [14, 15], whereas MRS is used tomeasure brain metabolism in static and dynamic models[16], namely lactate and N-acetylasparate levels in patientswith mitochondrial disorders. Functional MRI can be usedto study the neural nodes and networks underlying cognitiveoperations [17] that may be damaged in patients with mito-chondrial disorders and would be done best in the stablestate to assess degree of damage and can be repeated after anintervention or to track disease progression, mainly on aresearch basis. Using these various methods, 1 can probefocal, regional, and global neuropathological sequelae, andmonitor disease progression and response to therapies.

Here, we offer an introduction to the principles andapplications of each of these neuroimaging modalities, andwe review how they have added to our understanding ofneurological manifestations in mitochondrial disorders.

Nonspecific MRI Changes When Using Routine T1 and T2Imaging

Routine T1 and T2 images can be used to characterize graymatter and white matter microstructural and macrostructuralchanges. When abnormal brain pathology exists, 1 mayobserve signal abnormalities in T1- and T2-weighted scansthat reflect changes in brain tissue water. The limitations ofusing only routine imaging are that 1 is only able to detectdamage at a macroscopic level, typically when the patient isalready symptomatic; MRI findings may also lag behindclinical changes. This is particularly a concern in mitochon-drial disorders, in which, for example, recognition of theearliest stages of stroke in mitochondrial encephalopathy,lactic acidosis, and stroke syndrome (MELAS) will influ-ence treatment and management. In disorders, in whichthere is white matter damage, the use of FLAIR imaginghas added value in the detection of early white matter injury

[18, 19]. This is a particular concern in a patient withMELAS who may be having symptoms suggestive of animpending stroke, yet no stroke is seen on the routineimaging. In addition, routine MRI detection is limited tomacroscopic alterations in brain structure. It lacks the spatialresolution to provide information regarding microstructuralneuropathology, and does not capture dynamic processes inspace and time related to brain function and metabolism. Inaddition, many macrostructural neuropathological pheno-types lag behind the presentation of associated clinical man-ifestations. MRI changes may be nonspecific, such asshowing specific regions and general atrophy of cerebrumor cerebellum.

Anatomical FLAIR Imaging

FLAIR imaging is an MRI-based methodology that is sen-sitive to increases in interstitial water content in CNS dis-eases [20]; it plays a role in delineating white matterintegrity in brain tumors, multiple sclerosis, cerebralinfarcts, and metabolic white matter diseases. FLAIR essen-tially nulls signals from fluids with long longitudinal (T1)relaxation times. Lesions adjacent to cerebrospinal fluid areoften more clearly identified on FLAIR. In addition, FLAIRis useful in disorders of myelination (hypomyelinating anddemyelination), and as a tool for classifying normal baselinecharacteristics of myelin in the subcortical U fibers (thearcuate fibers at the cortical-white matter junction), deepwhite matter, and periventricular white matter. In children,myelination is incomplete at birth, and in healthy children itdevelops in a predictable, age-specific pattern [21]. Becausemany mitochondrial disorders present with white matterinjuries in the developing brain, FLAIR is expected to bean effective means of comparing subsequent myelinationpatterns in patients with early metabolic decompensationin mitochondrial disorders, as compared to development inhealthy individuals (Fig. 1).

Functional Magnetic Resonance Imaging

Functional MRI is mainly a research tool for imaging thetime course of activity associated with neurocognitive pro-cesses in the brain. This technology has wide-range appli-cations for both basic research into brain function, and forclinical research into the neurophysiology of neurologicaland psychiatric illness. Functional MRI investigation ofmitochondrial disorders may offer potential insight into theneurocognitive challenges experienced by mitochondrialpatients, and may help guide the development of education-al support structures. It has not been widely used in thisgroup of disorders.

First described by Roy and Sherrington [22] in 1890,neuronal metabolism and cerebral blood flow are intimately

276 Gropman

linked by neurovascular coupling. Functional MRI exploitsthis phenomenon to generate a blood oxygenation level-dependent signal in activated regions of the brain. Throughthe energy-expensive process of neurotransmission, active neu-rons consume glucose and oxygen, which are restored by localdilation of neural vasculature and subsequent increases in theflow of oxygenated hemoglobin. MRI pulse sequences aresensitive to the magnetic contrast between oxygenated anddeoxygenated hemoglobin, and can be used to map the hemo-dynamic response of local brain regions to the stimuli of aneurocognitive task presented to the subject in the scanner [23].

The principal drawback of fMRI, however, is that inter-subject variability in the spatial extent and precise locationof neurocognitive functions within the brain require rela-tively large numbers of patients to achieve an acceptablesignal-to-noise ratio, which is often impractical in rare dis-orders. This provides limitations in trying to explain theneurocognitive phenotype at the single subject level.

Voxel-Based Morphometry

VBM is a useful tool for the comparison of brain parametersin a clinical population relative to a group of healthy con-trols and allows 1 to quantitate volumetric differences inbrain structures. VBM relies on the representation of thebrain as a three-dimensional matrix of voxels, and eachvoxel in the VBM matrix provides an index of local brainvolume. Imaging is registered to a common three-dimensional template, as in fMRI. VBM templates are com-monly an average image comprised of scans from largenumbers of healthy subjects. The template image providesmorphological landmarks that allow the brain to be seg-mented into gray matter and white matter structures(Fig. 2). Analysis of VBM data across patient and controlgroups allows assessment of global differences in whole

brain, white matter, and grey matter volume throughoutthe brain, as well as fine differences within particularregions of interest at the voxel level [24].

Diffusion-Weighted and Diffusion Tensor MRI

For the past decade, DTI has emerged as a valuable meansof characterizing white matter structure and its alterations inan array of clinical contexts. Its sensitivity to microstructuralabnormalities not visible with traditional MRI, and its abil-ity to predict clinical correlates of white matter pathologyunderscore the value of this imaging platform for multimod-al imaging studies of disorders affecting the CNS.

Diffusion MRI encompasses DWI and DTI, which areboth voxel-based modalities that provide indices of axonalintegrity in vivo by measuring water movement in whitematter. DWI measures contrast between brain regions vary-ing with respect to water diffusivity. The diffusion of waterwithin a magnetic field gradient reduces the MRI signal, andas a result, regions in which diffusion is restricted evincelower signal loss and brighter appearance on DWI relative toregions of high diffusivity. This signal is called the apparentdiffusion coefficient (ADC) and is used to identify brainregions with abnormally high ADC or low ADC diffusivity,providing a means of identifying pathologic abnormalitiesin the white matter microstructure of the brain. DTI is alsosensitive to the ordered movement of water in the brain,whereas DWI scales each voxel to indicate the degree ofwater diffusion, and DTI indicates the direction of waterdiffusion in three-dimensional space. Acquisition of multi-ple pulse sequences with varying three-dimensional orien-tation generates a tensor, which indicates the orientation ofwater diffusion. Ordered water diffusion is said to be aniso-tropic, due to its lack of randomness, and DTI images rendervalues of fractional anisotropy (FA) to indicate the directionof water diffusion. Decreased FA in the white matter isgenerally interpreted as a signal of compromised axonalintegrity. Taken together, these measures provide a powerfuland sensitive basis for inferences about white matter micro-structure in healthy and diseased states [14, 15].

Fig. 1 Example of white matter lesions as seen on fluid attenuationinversion recovery imaging that may not be seen on T2 imaging

Fig. 2 Example of tissue parcellation for a volumetric study

Neuroimaging in Mitochondrial Disorders 277

Pathological variations in these measures serve as indices ofwhite matter injury, although the complexity of white mattermicroarchitecture underscores the need to interpret DTI findingsin the context of other disease-specific parameters, includingadditional imaging modalities and relevant clinical parameters.Depending on the nature of the pathology, either ADC or FAvalues may be increased, decreased, or show a mixed pattern.

Water diffusion in brain tissue is therefore affected by thepresence of barriers to its translational motion. The diffusionof water is restricted within a cell and between myelin sheets.The measured water ADC value is therefore frequently aniso-tropic and its value will change depending on the orientationof restricting barriers. DTI is best used in the imaging ofmitochondrial disorders when such information is combinedwith other MRI modalities to age, grade, and predict ischemicbrain injury for its entire temporal evolution.

Proton MRS

MRS is a powerful clinical tool for interrogating the neuro-chemical environment and characterizing the metabolic fea-tures of neurological disease. Although imaging coils havebeen tailored for sensitivity to a variety of isotopes, such as13C and 31P, proton 1H magnetic resonance spectroscopy(1H MRS) is most commonly used. 1H MRS produces aspectrum of peaks that contain information regarding thebiochemical concentrations in a region of interest (typically1–10 cm3) within the brain. These spectra contain informa-tion regarding brain metabolite concentrations relevant to avariety of neurological conditions, including brain tumors,traumatic brain injury, white matter disorders, epilepsy, andmetabolic disorders [25].

What Does MRS Tell Us About In Vivo Chemistry?

For an element to be studied by MRS there are severalrequirements. First, it should lend itself to excitation bymagnetic resonance. The chemical also has to be presentin a detectable concentration on the order of milligramsand should produce an acceptable signal-to-noise ratio.These conditions are met by 31P, 1H, 19F, 13C, 7Li, and23Na.

The ability to differentiate chemical compounds onthe magnetic resonance (MR) spectrum depends on thedifferences between compounds in the chemical config-uration of protons within the molecule, and how theseconfigurations differentially affect the electromagneticmicroenvironment in the region being sampled. Com-monly assayed metabolites include the following: N-acetyl aspartate (NAA), creatine, choline, myoinositol,and lactate. Other compounds, such as glutamate andglutamine benefit from short echo time and higher fieldstrength. The relative abundance of metabolites of inter-est can be determined by comparison of the peakheights between 2 compounds. Successful use of clinicalMR spectra requires a radiologist trained to recognizethe patterns in which neurochemical abnormalities elicitdeviations from normal MR spectra and clinical corre-lation to determine how these abnormalities may con-tribute to the diagnosis, course, and management ofdisease [26] (Fig. 3). Furthermore, MRS includes therequirement for stronger and more homogeneous mag-netic fields and is more sensitive to movement artifactsthan other modalities of MRI adding to expense in thecost and time.

Fig. 3 Example of proton 1Hmagnetic resonancespectroscopy showing 2subjects with peaks labeled

278 Gropman

What Do the Peak Heights of Various BiochemicalsIndicate?

1H MRS is useful as one can measure: NAA, theneuronal marker and CNS lactate, the end product ofnonoxidative metabolism (glycolysis). Lactate is oftenreferenced to internal water or NAA, a putative neuro-nal/axonal marker. Marked elevations of lactate can beobserved in patients with mitochondrial disease rangingfrom 3 to 11 mM [27, 28], especially in those withconfirmed defects in electron transport chain. Lactateelevation is a consistent feature observed in Leigh syn-drome, and the majority of patients with MELAS(Fig. 6). It is important to note that cerebral lacticacidosis occurs in the absence of plasma lactic acidemiain many patients, thus CNS lactate determination isoften a more specific marker of cerebral metabolismthan plasma lactate levels [28]. Elevated lactate shownby MRS often correlates well with other markers ofmitochondrial disease, but 1H MRS does not depictelevated lactate in all cases. In an instructive study byLin et al. [29], this group found abnormal CNS concen-trations of lactate may be undetected by 1H MRS,according to differences in the type of mitochondrialdisorder, timing, severity, or location of the affectedtissues and the site of interrogation. Often a voxel inthe ventricles proved more diagnostic than a voxelplaced in the parenchyma. Appearance of a lactate peakon MRS may be the first biomarker seen in a metabolicstroke and occurs even prior to changes that occur onDWIs [30–32].

Decreased NAA can indicate neuronal damage with thelevel of reduction taken as a marker of the extent of neuronalor axonal loss. In epilepsy, decreased NAA may be indica-tive of neuronal metabolic dysfunction. Clinicians workingwith pediatric populations should note that NAA levels arelower in the developing brain and typically peak at 10 to15 years of age [33].

Choline is an important molecule to consider in mi-tochondrial disorders. Choline containing compounds ortotal Choline is seen as a peak at 3.22 ppm. It iscomprised of several compounds, mainly phosphoryl-choline, phosphorylated cholines, and glycerophosphor-ylcholine. These are membrane and myelin markers, andcan increase with pathological changes to membraneturnover [34]. Fluctuations in total Choline compoundsoccur with diffuse axonal injury [25] among other find-ings. The major resonance of creatine is present at3.00 ppm, and in a healthy brain it is comprised ofequal proportions of free creatine and phosphocreatine,with creatine and phosphocreatin in constant enzymaticexchange. Phosphocreatin is a bioenergetic marker be-cause of its large role in the synthesis of ATP.

What are the Imaging Patterns in Mitochondrial Disorders?

The main imaging patterns seen in mitochondrial disordersinclude the more nonspecific findings of cortical and cere-bellar atrophy to the more specific T2 hyperintensities ofdeep gray matter nuclei, leukodystrophic changes, andstrokes in nonvascular territories [8, 35, 36]. Nonetheless,in children, imaging can be used in combination with theclinical, neurological, and biochemical data, as well as ped-igree to consider pursuing molecular diagnosis.

Patterns of Neural Injury in Mitochondrial DisordersThat Can Be followed by MRI

Focal Lesions in Deep Gray Matter Structures

Among child neurologists, bilateral lesions in the putamenand basal ganglia nuclei are recognized as being the sin quanon and most prevalent feature of many mitochondrial syn-dromes [8, 9, 27] (Fig. 4). Focal, bilateral, symmetric brainlesions involving basal ganglia and periaqueductal graymatterare typical of Leigh syndrome (subacute necrotizing encepha-lomyelopathy) [27–37], which can be caused by mutations inmtDNA and nuclear DNA encoded mitochondrial proteins.This disorder classically presents in infancy with brainstemdysfunction leading to abnormal cranial nerve findings, respi-ration, and oral motor dysfunction interfering with feeding,long tract signs leading to spasticity, multiple organ involve-ment, and global developmental delay.

Structural pathology of the brain lesions shows vascularproliferation, demyelination, gliosis, and necrosis [38–42].The typical anatomic distribution of the lesions is seen inbrainstem, diencephalon, basal ganglia, and cerebellum.Clinical severity is typically dependent on mutation typeand state of heteroplasmy in mtDNA subtypes.

Genotype predictions have been observed clinically, de-spite similar imaging features. Patients with Leigh disease dueto the nuclear mutations, surfeit protein-1 (SURF-1) muta-tions, and other cytochrome oxidase (COX) assembly geneshave an imaging pattern on MRI showing prominent T2-signal abnormalities in the deep brainstem nuclei, rather thanthe caudate and putamen lesions that are most characteristic inpatients with Leigh disease due to mtDNAmutations [33, 34].

Strokes in Nonvascular Territories

Patients with MELAS have a characteristic imaging patternthat should prompt further evaluation for a mitochondrialdisorder (Fig. 5). MELAS is a progressive neurodegenera-tive disorder associated with migraine-like headaches, re-fractory partial or generalized seizures, short stature, muscleweakness, exercise intolerance, sensorineural deafness,

Neuroimaging in Mitochondrial Disorders 279

diabetes, cardiac conduction defects, and slowly progressivedementia [43]. Patients may present with a seizure thatheralds a stroke. Approximately 80 % of patients withMELAS have a common mutation (mitochondrial (mt)3243 A>G within the (tRNA)leu transfer RNA (ribonucleicacid), leucine), however, other genotypes have beenreported [44]. Only approximately 50 % of patients whoharbor a mutation are symptomatic, with presentationdepending on tissue heteroplasmy. Urine sediment or bloodis the best tissue for diagnosis.

The most common finding in MELAS is an infarct in theparieto-occipital region, which may be unilateral or bilater-al. Vasogenic edema may be evident by hyperintensity ofthe lesion on T2-weighted images, diffusion-weighted im-aging, and apparent diffusion coefficient in the acute andsubacute or chronic stage, most frequently in the parieto-occipital region, surpassing vascular territories [27, 45–48].Figure 5b depicts a classical stroke in a patient withMELAS. Patients with MELAS may have recurrent strokesthat add to the morbidity of the disease. The strokes maycome back to back, or several years may separate events. It

has been suggested that adequate seizure control in patientswith MELAS syndrome may prevent the recurrence ofstroke-like episodes and may result in the disappearance ofstroke-like lesions on MRI [40].

There is a clear role for DTI and MRS in MELAS.Diffusion imaging is helpful to differentiate acute lesionswith restricted diffusion from older/chronic lesions that of-ten present with increased diffusion and to identify strokesbefore they are evident on T2-weighted images to initiatetherapy. In addition, repeated MRI using DTI may showprogressive spread of the cortical lesion to the surroundingcortex for a few weeks, even after the onset of symptoms. Inpatients with mitochondrial disease, significant widespreadreductions in FA values have been shown in white mattertracts, in some cases in normal-appearing white matter.Mean diffusivity values are increased [49]. It has beenreported on a limited basis in acute and long-term follow-up in which serial diffusion tensor imaging revealed astroke-like pattern with an initial strong reduction of theADC followed by elevated values after clinical recovery[50, 51]. ADCs in nonaffected areas of the brain on T2

Fig. 5 (a) A computedtomographic scan and (b) MRIof a patient with mitochondrialencephalopathy, lactic acidosis,and stroke syndrome (MELAS).(a) The computed tomographicscan shows typical location andappearance. (b) The MRI isshowing subacute left parieto-occipito temporal infarction.The MRI also shows right tem-poroparietal and bifrontal lobeencephalomalacia owing toprevious infarcts in this 15-year-old girl

Fig. 4 Examples of bilateralsymmetric T2 rightabnormalities in the basalganglia. (a) Symmetric in-creased T2 signal is evident inthe posterior and superior as-pect of the head of the caudate(blue arrow), the anterior puta-men (yellow arrows), and (b)the globus pallidus nuclei bilat-erally (red arrows)

280 Gropman

MRI in patients with MELAS may be higher than those ofnormal subjects. Pathological changes take place in the non-affected areas of the patients with MELAS and are bestdetected by these newer imaging methods [43].

Only single case reports have described the use ofserial MRI in patients with MELAS. Brain MRIs,obtained at symptom onset, at 3 weeks, and at 1 year,in a patient with stroke due to MELAS revealed migrat-ing T2-weighted hyperintensities in the temporal/parietaland occipital cortices and later revealed atrophy [52].MRS at 1 year poststroke revealed increased lactate inboth the occipital and temporal lobes [53]. At autopsy,there was gliosis found that mainly consisted of deepwhite matter. Typically there is sparing of U-fibers.

In another study, MRI was performed 2 days after thesudden onset of cortical blindness in a 25-year-old patientwith MELAS. FLAIR images demonstrated multifocal cor-tical and subcortical hyperintensities located bilaterally in thefrontobasal and temporo-occipital distributions. Diffusion-weighted images showed normal-to-increased apparent dif-fusion coefficient values in the acute left temporo-occipitallesion and increased values in the older stroke-like lesions.Based on these findings, 1 might concur that metabolic ratherthan the ischemic mechanisms account for the stroke-likeepisodes occurring in MELAS. The authors suggest thatnormal or increased apparent diffusion coefficient valueswithin 48 h of a neurological deficit of abrupt onset shouldraise the possibility of MELAS, particularly when theinfarcts are in the characteristic areas of cortex [52].

Newer imaging modalities will allow classification andelucidation of the pathomechanism. Currently, the 2 mainhypotheses that have been presented to explain these cerebrallesions invoke a vascular versus metabolic etiology. In thevascular hypothesis, metabolic damage of the endotheliumleads to small vessel occlusion and secondary neuronal death.Using routine imaging, it is supported to some extent bydisease course and the computed tomographic and MRI ap-pearance of the lesions and some pathological reports ofendothelium alterations in the brains of patients with MELAS[54]. On the other hand, the metabolism hypothesis states thatmitochondrial dysfunction results in anaerobic metabolismand neuronal death owing to lactic acidosis. To confirm orrefute this would require demonstration of hyperperfusion anddissociation between glucose and oxygen consumption,which would rely on positron emission tomographic andsingle photon emission computed tomographic studies.

Cortical and Cerebellar Atrophy

Cerebral atrophy is a common, but nonspecific finding inmitochondrial disease that may occur in multiple mitochon-drial subgroups [55]. Although tissue atrophy may be slow-ly progressing with time, rapid progression may be seen inassociation with the clinical decline [56] and may promptevaluation for Alpers hepatorenal presentation. Cerebellaratrophy might be the primary neuroimaging feature in somepatients with mitochondrial disorders [57–60], particularlychildren, and should figure into the differential diagnosis(Fig. 7). Mutations in polymerase gamma-1 (POLG1) (themitochondrial DNA replicase), can present with volume losspredominantly in cerebellar and occipital cortices, reflectingspecific loss of Purkinje cells and neurons in the dentate

Fig. 7 Cerebellar atrophy as a sign of a mitochondrial disorder. Theyellow arrow demonstrates cerebellar atrophy

Fig. 6 This proton 1H magnetic resonance spectroscopy shows a verylarge lactate peak. The characteristics of the lactate peak are doubletwith a 7 Hz split between the peaks (yellow arrow). The height of theleft side of the doublet is usually slightly higher than the right, givingan asymmetric appearance. These characteristics are required to differ-entiate this signal from noise in the background of the spectrum. Theconcentration must be at least 1 mmol/kg tissue for a signal to be seen.Although this spectrum was acquired at a long TE, a shorter TE willalso demonstrate a lactate peak. as well as other metabolites of interest

Neuroimaging in Mitochondrial Disorders 281

nucleus [61]. Cerebellar atrophy should be differentiatedfrom cerebellar hypoplasia, as a major metabolic disorderthat can mimic mitochondrial disorders with intellectualdelay, seizures, hypotonia, vision defects, and other organfindings are the congenital disorders of glycosylation.

White Matter/Leukodystrophy

White matter changes are often considered core features ofdistinct leukodystrophies. White matter changes are

frequently observed in association with many syndromesrelated to idiopathic developmental delay [62, 63]. Clinical-ly, leukoencephalopathies are clinically characterized by thepredominance of motor disturbances, with slow neurologicaldeterioration and a low (but not absent) incidence of seizures.Leukoencephalopathies are generally classified according toeither features on histopathology or location of the myelininvolvement, or alternatively by the causative enzyme defect.Cerebral white matter involvement is commonly seen inchildhood onset mitochondrial disorders. However, when

Fig. 8 Abnormal, bilateral T2-hyperintense signal is againidentified throughout the supra-tentorial and infratentorialwhite matter in these images ofchildren with mitochondrialdisorders

Fig. 9 Two examples ofdelayed myelination

282 Gropman

MRI denotes a leukodystrophy pattern, respiratory chaindefects are not often considered in the differential diagnosis.Genetic defects of oxidative phosphorylation should alwaysbe considered when the MRI is suggestive of leukoencephal-opathy in childhood. Although the literature on neuroimagingin mitochondrial disorders is not vast, there are many individ-ual case reports and case series of patients with primaryelectron transport chain (ETC). deficits who have white matterchanges [64–66] onMRI with clinical features that predict andconfirm this finding (Fig. 8).

Diffusion-weighted imaging and DTI are especially use-ful measures of axonal integrity in vivo by its ability tomeasure water movement and would be predicted to beuseful in the evaluation of patients with mitochondrial dis-orders. As previously reviewed, diffusion of water within amagnetic field gradient reduces the MR signal. Therefore,regions in which diffusion is restricted demonstrate lowersignal loss and brighter appearance on DWI, relative toregions of high diffusivity. Cytotoxic edema can be demon-strated on diffusion-weighted imaging, is an early changethat precedes T2 signal changes, and maybe a helpful dis-criminator of mitochondrial disease. The complexity ofwhite matter microarchitecture in mitochondrial disordersunderscores the importance of DTI as a tool; however, thefindings need to be interpreted in the context of clinicalparameters and other imaging findings. Depending on thenature of the pathology, either ADC or FA values may beincreased, decreased, or show a mixed pattern in mitochon-drial disorders.

Several of the mitochondrial disorders present withwidespread leukodystrophy in a pattern that should suggestthe underlying etiology. Patients with mitochondrial neuro-gastrointestinal encephalopathy due to mutations in thethymidine phosphorylase gene present with severe gastro-intestinal dysmotility, ophthalmoplegia, ptosis, cachexia,

and leukodystrophy, as well as peripheral neuropathy dueto involvement of peripheral nervous system myelin [64,67]. A mutation in the RRM2B gene that leads to neuro-gastrointestinal encephalopathy due to severe depletion ofmtDNA copy number also produces a leukodystrophy pat-tern [68].

Another common radiological finding is that of “delayedmyelination” (Fig. 9). This pertains to an MRI appearancethat is less mature than that expected by the chronologicalage of the child, which is corrected for any prematurity [21,69]. Although common teaching is that myelin delay due toa metabolic condition does not “catch up,” as it would whenthe delay is due to nutritional factors, as evidenced fromclinical cases suggests that myelination delay may “catchup” with a period of time [28, 70]. White matter changesmay be present at disease presentation or as a consequenceof the disease progression [65, 71].

Conclusions

Mitochondrial respiratory chain disorders can present at anyage with vast heterogeneity, and any organ or tissue may beaffected, although the CNS and the peripheral nervous sys-tem are most commonly involved [72]. As a group, theyrepresent a challenge to clinicians, especially in children(ages, birth to 16 years), in whom clinical presentation anddisease course show great variability [73].

In an attempt to navigate the complexities of mitochon-drial disorders, various consensus diagnostic criteria havebeen proposed, which mainly addresses adult disease [74,75] (Table 4). A modification to accommodate diagnosis ofpediatric cases has also been proposed [76]. These classifi-cations, in addition to clinical, genetic, histopathological,and metabolic criteria, include a scoring criterion for the

Table 4 Diagnostic Criteria inadults and Children*

COX 0 cytochrome oxidase;mtDNA 0 mitochondrion DNA;RC 0 respiratory chain;RRF 0 ragged red fibers;SD 0 succinate dehydrogenase;SSMA 0 subsarcolemmal mito-chondrial aggregates

*[80]

Major criteria Minor Criteria

Clinical presentation (↑lactate) Clinical presentation (+/−)

Histology Histology

>2 % RRF <2 % RRF (age, 30–50 years)

2-5 % COX-negative fibers >2 % SSMA (<16y)

Enzymology Abnormal mitochondria (EM)

<20 % RC in a tissue or Enzymology

<30 % RC ≥2 tissues 20-30 % RC in a tissue or 30-40 % RC ≥2 tissues

<30 % RC in a cell line 30-40 % RC in a cell line

Functional Functional

Fibroblast ATP synthesis rates >3 SD below normal Fibroblast ATP synthesis rates 2–3 SD belownormal

Molecular Molecular

Nuclear or mtDNA mutation of undisputedpathogenicity

Nuclear or mtDNA mutation of probablepathogenicity

Neuroimaging in Mitochondrial Disorders 283

results of biochemical studies for which no gold standard isgenerally accepted. In children, neuroimaging patterns arean important aspect of scoring systems and the recognitionof patterns seen in the developing brain, which may changein time, are necessary for familiarity with neurologists[77–79].

Common patterns of brain MR imaging can be identifiedin patients with mitochondrial disorders. However, using thenew modalities of DTI, volumetrics, MRS, and fMRI allowsan expansion of the field of knowledge in terms of combin-ing clinical and biochemical data with brain imaging toguide genetic studies.

Required Author Forms Disclosure forms provided by the authorsare available with the online version of this article.

References

1. Wallace DC, Fan W, Procaccio V. Mitochondrial energetics andtherapeutics. Annu Rev Pathol Mech Dis 2010;5:297–348.

2. DiMauro S. Mitochondrial medicine. Biochim Biophys Acta2004;1658:107–114.

3. Thorburn DR. Mitochondrial disorders: prevalence, myths, andadvances. J Inherit Metab Dis 2004;27:349–362.

4. Anderson S, Bankier AT, Barrell BG, de Bruijn MH, et al.Sequence and organization of the human mitochondrial genome.Nature 1981;290:457–465.

5. Dimauro S. A history of mitochondrial diseases. J Inherit MetabDis 2011;34:261–276.

6. Gropman AL. The neurological presentations of childhood andadult mitochondrial disease: established syndromes and phenotyp-ic variations. Mitochondrion 2004;4:503–520.

7. Brunetti-Pierri N, Selby K, O'Sullivan M, et al. Rapidly progres-sive neurological deterioration in a child with Alpers syndromeexhibiting a previously unremarkable brain MRI. Neuropediatrics2008;39:179–183.

8. Haas R, Dietrich R. Neuroimaging of mitochondrial disorders.Mitochondrion 2004;4:471–490.

9. Friedman SD, Shaw DW, Ishak G, Gropman AL, Saneto RP. Theuse of neuroimaging in the diagnosis of mitochondrial disease.Dev Disabil Res Rev 2010;16:129–135.

10. Masucci JP, Schon EA, King MP. Point mutations in the mito-chondrial tRNA(Lys) gene: implications for pathogenesis andmechanism. Mol Cell Biochem 1997;174:215–219.

11. Wallace DC. Mitochondrial diseases in man and mouse. Science1999;283:1482–1488.

12. Gropman AL. Expanding the diagnostic and research toolbox forinborn errors of metabolism: the role of magnetic resonance spec-troscopy. Mol Genet Metab 2005;86:2–9.

13. Gelfand JM, Duncan JL, Racine CA, et al. Heterogeneous patternsof tissue injury in NARP syndrome. J Neurol 2011;258:440–448.

14. Basser PJ, Mattiello J, LeBihan D. MR diffusion tensor spectros-copy and imaging. Biophys J 1994;66:259–267.

15. Basser PJ, Pierpaoli C. Microstructural and physiological featuresof tissues elucidated by quantitative-diffusion-tensor MRI. J MagnReson B 1996;111:209–219.

16. Ross BD, Bluml S. New aspects of brain physiology. NMRBiomed 1996;9:279–296.

17. Guye M, Bartolomei F, Ranjeva JP. Imaging structural and func-tional connectivity: towards a unified definition of human brainorganization? Curr Opin Neurol 2008;21:393–403.

18. Iwata O, Iwata S, Tamura M, et al. Periventricular low intensitieson fluid attenuated inversion recovery imaging in the newborninfant: Relationships to chronic white matter lesions. Pediatr Int2004;46:141–149.

19. Maillard P, Carmichael O, Harvey D, et al. FLAIR and DiffusionMRI signals are independent predictors of white matter hyper-intensities. AJNR 2012. doi:10.3174/ajnr.A3146.

20. Hoisington L, Miller RA, Vreibel B. Fast FLAIR techniques inMR imaging of the brain. Radiol Technol 1998;69:351–357.

21. Barkovich AJ, Kjos BO, Jackson DE Jr, Norman D. Normalmaturation of the neonatal and infant brain: MR imaging at 1.5T. Radiology 1988;16:173–180.

22. Roy CS, Sherrington CS. On the regulation of the blood supply ofthe brain. J Physiol 1890;11:85–108.

23. Logothetis NK. What we can do and what we cannot do withfMRI. Nature 2008;453:869–878.

24. Ashburner J, Friston K. Voxel-based morphometry— the methods.NeuroImage 2000;11:805–821.

25. Xu V, Chan H, Lin AP, et al. MR spectroscopy in diagnosis andneurological decision-making. Semin Neurol 2008;28:407–422.

26. Panigrahy A, Nelson MD Jr, Blüml S. Magnetic resonance spec-troscopy in pediatric neuroradiology: clinical and research appli-cations. Pediatr Radiol 2010;40:3–30.

27. Wilichowski E, Pouwels PJ, Frahm J, Hanefeld. Quantitative pro-ton magnetic resonance spectroscopy of cerebral metabolic distur-bances in patients with MELAS. Neuropediatrics 1999;30:256–263.

28. Dinopoulos A, Cecil KM, Schapiro MB, et al. Brain MRI andproton MRS findings in infants and children with respiratory chaindefects. Neuropediatrics 2005;36:290–301.

29. Lin DD, Crawford TO, Barker PB. Proton MR spectroscopy in thediagnostic evaluation of suspected mitochondrial disease. AJNR2003;24:33–41.

30. Abe K, Yoshimura H, Tanaka H, et al. Comparison of conventionaland diffusion weighted MRI and proton MR spectroscopy inpatients with mitochondrial encephalomyopathy, lactic acidosis,and stroke-like events. Neuroradiology 2004;46:113–117.

31. Bianchi MC, Sgandurra G, Tosetti M, et al. Brain magnetic reso-nance in the diagnostic evaluation of mitochondrial encephalopa-thies. Biosci Rep 2007;27:69–85.

32. Bianchi MC, Tosetti M, Battini R, et al. Proton MR spectroscopyof mitochondrial diseases: analysis of brain metabolic abnormali-ties and their possible diagnostic relevance. Am J Neuroradiol2003;24:1958–1966.

33. Pouwels PJ, Brockmann K, Kruse B, et al. Regional age depen-dence of human brain metabolites from infancy to adulthood asdetected by quantitative localized proton MRS. Pediatr Res 1999;46:474–485.

34. Bluml S, Seymour KJ, Ross BD. Developmental changes incholine- and ethanolamine-containing compounds measured withproton-decoupled 31PMRS in in-vivo human brain. Magn ResonMed 1999;42:643–654.

35. Barkovich AJ, Good WV, Koch TK, Berg BO. Mitochondrialdisorders: analysis of their clinical and imaging characteristics.AJNR 1993;14:1119–1137.

36. Barragan-Campos HM, Vallee JN, Lo D, et al. Brain magneticresonance imaging findings in patients with mitochondrial cytopa-thies. Arch Neurol 2005;62:737–742.

37. Medina L, Chi TL, DeVivo DC, Hilal SK. MR findings in patientswith subacute necrotizing encephalomyelopathy (Leigh syn-drome): correlation with biochemical defect. Am J Roentgenol1990;154:1269–1274.

38. Lee HF, Tsai CR, Chi CS, Lee HJ, Chen CC. Leigh syndrome:clinical and neuroimaging follow-up. Pediatr Neurol 2009;40:88–93.

39. Leigh D. Subacute necrotizing encephalomyelopathy in an infant.J Neurol Neurosurg Psychiatry 1951;14:216–221.

284 Gropman

40. Cavanagh JB, Harding BN. Pathogenic factors underlying the lesionsin Leigh’s disease. Tissue responses to cellular energy deprivation andtheir clinico-pathological consequences. Brain 1994;117:1357–1376.

41. Leshinsky-Silver E, Lebre AS, Minai L, et al. NDUFS4 mutationscause Leigh syndrome with predominant brainstem involvement.Mol Genet Metab 2009;97:185–189.

42. Farina L, Chiapparini L, Uziel G, Bugiani M, Zeviani M,Savoiardo M. MR findings in Leigh syndrome with COX defi-ciency and SURF-1 mutations. AJNR 2002;23:1095–1100.

43. Hirano M, Ricci E, Koenigsberger MR, et al. MELAS: an originalcase and clinical criteria for diagnosis. Neuromuscular Disord1992;2:125–135.

44. Wong LJ. Molecular genetics of mitochondrial disorders. DevDisabil Res Rev 2010;16:154–162.

45. Matthews PM, Tampieri D, Berkovic SF, et al. Magnetic resonanceimaging shows specific abnormalities in the MELAS syndrome.Neurology 1991;41:1043–1046.

46. Castillo M, Kwock L, Green C. MELAS syndrome: imaging andproton MR spectroscopic findings. AJNR 1995;16:233–239.

47. Renard D, Bonnaure H, Labauge P. Teaching neuroImages: diffuseposterior leukoencephalopathy in MELAS without stroke-like epi-sodes. Neurology 2010;75:e9.

48. Finsterer J, Barton P. Regression of stroke-like lesions in MELAS-syndrome after seizure control. Epileptic Disord 2010;12:330–334.

49. Ishak GE, Poliakov AV, Poliachik SL, et al. Tract-based spatialstatistical analysis of diffusion tensor imaging in pediatric patientswith mitochondrial disease: widespread reduction in fractional an-isotropy of white matter tracts. AJNR 2012. doi:10.3174/ajnr.A3045.

50. Brockmann K, Finsterbusch J, Schara U, Wilichowski E, Frahm J,Hanefeld F. Stroke-like pattern in DTI and MRS of childhood mito-chondrial leukoencephalopathy. Neuroradiology 2004;46:267–271.

51. Liu Z, Liu X, Hui L, et al. The appearance of ADCs in the non-affectedareas of the patients with MELAS. Neuroradiology 2011;53:227–232.

52. Clark JM, Marks MP, Adalsteinsson E, et al. MELAS: clinical andpathologic correlations with MRI, xenon/CT, and MR spectrosco-py. Neurology 1996;46:223–227.

53. Feng F, You H, Gao J, et al. Evaluation of mitochondrial encepha-lomyopathy with lactic acidosis and stroke-like episodes withmagnetic resonance imaging and proton magnetic resonance spec-troscopy. Chin Med Sci J 2006;21:234–238.

54. Ohama E, Ohara S, Ikuta F, Tanaka K, Nishizawa M, Miyatake T.Mitochondrial angiopathy in cerebral blood vessels of mitochon-drial encephalomyopathy. Acta Neuropathol 1987;74:226–233.

55. Lim BC, Park JD, Hwang H, et al. Mutations in ND subunits ofcomplex I are an important genetic cause of childhood mitochon-drial encephalopathies. J Child Neurol 2009;24:828–832.

56. Castro-Gago M, Gonzalez-Conde V, Fernandez-Seara MJ, et al.[Early mitochondrial encephalomyopathy due to complex IV defi-ciency consistent with Alpers–Huttenlocher syndrome: report oftwo cases]. Rev Neurol 1999;29:912–917.

57. Scaglia F, Towbin JA, Craigen WJ, et al. Clinical spectrum, mor-bidity, and mortality in 113 pediatric patients with mitochondrialdisease. Pediatrics 2004;114:925–931.

58. Scaglia F, Wong LJ, Vladutiu GD, Hunter JV. Predominant cere-bellar volume loss as a neuroradiologic feature of pediatric respi-ratory chain defects. AJNR 2005;26:1675–1680.

59. Valanne L, Ketonen L, Majander A Suomalainen A, Pihko H.Neuroradiologic findings in children with mitochondrial disorders.AJNR 1998;9:369–377.

60. Lincke CR, van den Bogert C, Nijtmans LG, Wanders RJ,Tamminga P, Barth PG. Cerebellar hypoplasia in respiratory chaindysfunction. Neuropediatrics 1996;27:216–218.

61. Harding BN. Progressive neuronal degeneration of childhood withliver disease (Alpers–Huttenlocher syndrome): a personal review. JChild Neurol 1990;5:273–287.

62. Hatten HP Jr. Dysmyelinating leukodystrophies: LACK ProperMyelin. Pediatr Radiol 1991;21:477–482.

63. Finsterer J. Central nervous system imaging in mitochondrial dis-orders. Can J Neurol Sci 2009;36:143–153.

64. Saneto RP, Friedman SD, Shaw DW. Neuroimaging of mitochon-drial disease. Mitochondrion 2008;8:396–413.

65. Weinstock A, Giglio P, Cohen ME, Bakshi R, Januario J, Balos L.Diffuse magnetic resonance imaging white matter changes in a 15-year-old boy with mitochondrial encephalomyopathy. J ChildNeurol 2002;17:47–49.

66. Burgeois M, Goutieres F, Chretien D, Rustin P, Munnich A,Aicardi J. Deficiency in complex II of the respiratory chain,presenting as a leukodystrophy in two sisters with Leigh syn-drome. Brain Dev 1992;14:404–408.

67. Nishino I, Spinazzola A, Hirano M. Thymidine phosphorylasegene mutations in MNGIE, a human mitochondrial disorder.Science 1999;283:689–692.

68. Shaibani A, Shchelochkov OA, Zhang S, et al. Mitochondrialneurogastrointestinal encephalopathy due to mutations inRRM2B. Arch Neurol 2009;66:1028–1032.

69. Lebel C, Gee M, Camicioli R, Wieler M, Martin W, Beaulieu C.Diffusion tensor imaging of white matter tract evolution over thelifespan. Neuroimage 2012;60:340–352.

70. Valanne L, Ketonen L, Majander A, Suomalainen A, Pihko H.Neuroradiologic findings in children with mitochondrial disorders.Am J Neuroradiol 1998;19:369–377.

71. Wolf NI, Seitz A, Harting I, et al. New pattern f brain MRI lesions inisolated complex deficiency. Neuropediatrics 2003; 34:156–159.

72. Munnich A, Rustin P, Rotig A, et al. Clinical spectrum and diag-nosis of mitochondrial disorders. Am J Med Genet (Semin MedGenet) 2001;106:4–17.

73. Haas RH, Parikh S, Falk MJ, et al. The indepth evaluation ofsuspected mitochondrial disease. Mol Genet Metab 208;94:16–37.

74. Walker UA, Collins S, Byrne E. Respiratory chain encephalomyo-pathies: a diagnostic classification. Eur Neurol 1996;36:260–267.

75. Wolf NI, Smeitink JA. Mitochondrial disorders: a proposal forconsensus diagnostic criteria in infants and children. Neurology2002;59:1402–1405.

76. Morava E, van den Heuvel L, Hol F, et al. Mitochondrial diseasecriteria: diagnostic applications in children. Neurology 2006;67:1823–1826.

77. van der Knaap MS, Valk J, de Neeling N, Nauta JJ. Pattern recogni-tion in magnetic resonance imaging of white matter disorders inchildren and young adults. Neuroradiology 1991;33:478–493.

78. Wray SH, Provenzale JM, Johns DR, Thulborn KR. MR of the brainin mitochondrial myopathy. Am J Neuroradiol 1995;16:1167–1173.

79. Cakmakci H, Pekcevik Y, Yis U, Unalp A, Kurul S. Diagnosticvalue of proton MR spectroscopy and diffusion-weighted MRimaging in childhood inherited neurometabolic brain diseasesand review of the literature. Eur J Radiol 2010;74:e161-e171.

80. Bernier FP, Boneh A, Dennett X, Chow CW, Clearly MA,Thorburn DR. Diagnostic criteria for respiratory chain disordersin adults and children. Neurology 2002;59:1406–1411.

Neuroimaging in Mitochondrial Disorders 285