Cerebellar ataxia is a prominent clinical symptom in patients with mitochondrial DNA (mtDNA) disease. This is often progressive with onset in young adulthood. We performed a detailed neuropathologic investigation of the olivary-cerebellum in 14 genetically and clinically well-defined patients with mtDNA disease. Quantitative neuropathologic investigation showed varying levels of loss of Purkinje cells and neurons of the dentate nucleus and inferior olivary nuclei. Typically, focal Purkinje cell loss was present in patients with the m.3243A>G mutation caused by the presence of microinfarcts, with relative preservation of neuronal cell populations in the olivary and dentate nuclei. In contrast, patients with the m.8344A>G mutation orrecessive POLG mutations showed extensive and global neuronal cell loss in all 3 olivary-cerebellum areas examined. Molecular analysisof mutated mtDNA heteroplasmy levels revealed that neuronal cell loss occurred independently of the level of mutated mtDNA present within surviving neurons. High levels of neuronal respiratory chain deficiency, particularly of complex I, were detected in surviving cells; levels of deficiency were greater in regions with extensive cell loss. Wefound a relationship between respiratory deficiency and neuronal cell density, indicating that neuronal cell death correlates with respiratory deficiency. These findings highlight the vulnerability of the olivary-cerebellum to mtDNA defects.
Neurodegeneration is an important cause of neurological disability in patients with mitochondrial DNA (mtDNA) disease (1). These diseases arise from either primary defects of mtDNA, such as point mutations or large-scale mtDNA deletions, or secondary defects, such as mtDNA depletion or deletions caused by defects in nuclear genes involved in mtDNA maintenance (2). The CNS has a high requirement for energy and is, therefore, dependent on mitochondria for adenosine triphosphate generation (3–5). Adenosine triphosphate generation occurs via oxidative phosphorylation through a series of 5 respiratory chain complexes embedded in the inner mitochondrial membrane. These complexes are partly encoded by the mitochondrial genome. It is not surprising, therefore, that in patients with mtDNA disease, neurons are especially vulnerable to degeneration.
Although neurological deficits in mtDNA disease may be manifold, cerebellar ataxia is a frequently reported symptom in patients with mtDNA defects (6–12). A recent United Kingdom–based MRC-funded cohort study estimates that the frequency of ataxia is high, with 225 of 345 patients with mitochondrial disease showing symptoms consistent with ataxia (Nesbitt, McFarland et al., unpublished observation). Cerebellar ataxia is often progressive in these patients and is a major cause of disability. Because cerebellar ataxia is commonly described in patients with mtDNA disease, there have been a number of attempts to characterize neurodegeneration using neuroimaging techniques. These are often useful as a noninvasive assessment of CNS degeneration that assists in the diagnosis of mtDNA disorders and enables documentation of progression. Typically, reduced cerebellar volume is reported and may be attributed to atrophy (13, 14). Indeed, a large retrospective study involving 113 patients showed that cerebellar atrophy might even be the primary neuroradiological finding in patients with mtDNA defects (15). However, imaging studies give only a limited insight into an ongoing neurodegenerative process for which understanding of the precise mechanisms remains undetermined.
Neuropathologic studies are limited in mtDNA disease because of the challenges associated with obtaining postmortem material and published studies tend to focus on single cases. Despite this, there is evidence of cerebellar involvement in the pathogenesis of mtDNA disease. Typically, studies have revealed cerebellar atrophy, Purkinje cell loss, abnormal dendritic arborization, and the presence of ischemic-like lesions (16–18). However, there have been no attempts to correlate the severity of clinical manifestations with neuropathologic changes associated with specific molecular defects.
Despite the known involvement of the cerebellar dysfunction in the genesis of ataxia in mtDNA disease, very little is understood about the molecular mechanisms underpinning the neurodegeneration. This study aims to increase our understanding of the relationship between clinical symptoms of ataxia, neuropathologic changes, and the molecular attributes of neurons within the olivo-cerebellum pathway in a large group of clinically and genetically well-defined cases with mtDNA defects. This work will provide an important platform from which we can begin to explore the mechanisms of neurodegeneration caused by mitochondrial dysfunction and could prove important for the development of novel therapeutic strategies for the treatment of ataxia. It could also have further implications for the understanding the role of mitochondrial dysfunction in other neurodegenerative disorders such as Alzheimer disease and Parkinson disease.
Materials and Methods
Clinical Assessment of Patients
Patients with clinically and genetically defined mtDNA disease were selected from the Newcastle Brain Tissue Resource (Table 1). Adult patients with a known molecular genetic defect were selected. The participants had received repeated clinical assessments annually until death and had agreed to donate their brain tissue for research purposes. Patients were rated using the Newcastle Mitochondrial Disease Adult Scale (NMDAS) to monitor disease progression (19). The rating scales given for cerebellar ataxia at final clinical assessment are included in Table 1, thereby confirming that all the included patients had ataxia.
All the brain tissue used in the current study was obtained from the Newcastle Brain Tissue Resource. The brain tissue collection for both mtDNA disease (n = 14) and control subjects (n = 15) was ethically approved by the Newcastle and North Tyneside Local Research Ethics Committee. For each region of interest investigated, suitable control tissue was acquired (Table, Supplemental Digital Content 1, ). Patients were between 20 and 68 years old, with a mean age of 42.4 ± 12.8 years (standard deviation). Controls used for the assessment of inferior olivary nucleus (control, 1–2), Purkinje cells (control, 3–7), and dentate nucleus (control, 8–15) had a mean age of 53.5 ± 0.7, 70 ± 7.6, and 52.13 ± 12.5 years, respectively. Whereas patient and control tissue were suitably age matched for inferior olivary nucleus and dentate nucleus, the controls used to investigate Purkinje cells were somewhat older. No statistical differences were determined for postmortem interval or length of fixation between patient and control tissue.
Histological Stains and Immunohistochemistry
Routine neuropathologic histological stains, including cresyl fast violet, hematoxylin and eosin, Loyez and Bielschowsky impregnation, were applied to formalin-fixed, paraffin-embedded tissues. Immunohistochemical (IHC) staining was performed on 5-μm-thick formalin-fixed, paraffin-embedded sections. Sections were deparaffinized and rehydrated through Histoclear (National Diagnostics, Charlotte, NC) and ethanol series. Antigen retrieval was performed using the optimal method for each primary antibody (Table, Supplemental Digital Content 2, ). Endogenous peroxide was blocked by incubating sections in a 3% H2O2 solution (Sigma-Aldrich, UK). A range of primary antibodies was used to identify various markers, including mitochondrial proteins, presynaptic terminal protein, and cytoskeletal components in neuronal and astroglial cells. Primary antibodies were diluted at the optimal concentration and applied to the tissue for 1 hour at room temperature. Sections were washed in 3 changes of Tris-buffered saline for 5 minutes and incubated with a Universal probe (Menarini Diagnostics, Wokingham, UK) to recognize the mouse or rabbit epitope of the primary antibody for 30 minutes. Sections were washed in 3 changes of Tris-buffered saline, and a horseradish peroxidase polymer (Menarini Diagnostics) was applied to the tissue for 30 minutes at room temperature. The sections were then visualized with 3,3′-diaminobenzadine (Sigma-Aldrich, UK), counterstained in Mayer hematoxylin, and the nuclei were blued in Scott tap water.
Two-Dimensional Neuron Counting and Calculation of Neuron Density in Olivary-Cerebellum
Because of limited tissue availability, a 2-dimensional neuronal cell counting protocol was used using a stereological workstation with a modified light microscope (Olympus BX51), motorized stage for automatic sampling, CCD color video, and stereology software (Stereo Investigator, MBF Bioscience, Williston, VT). The medulla oblongata was sectioned in the horizontal plane, and the cerebellum with the entire dentate nucleus was cut in the sagittal plane at 20 μm. This was sampled at approximately the same level in each patient and control to allow for direct comparisons. One section was sampled from each patient and control and stained with cresyl fast violet to allow identification of neuronal populations.
The total counts of Purkinje cells were performed on individual lobules of the cerebellum to calculate Purkinje cell density. The boundaries of the Purkinje cell layer in individual cerebellar lobules were identified at 10× magnification using the closed contour function of Stereo Investigator. Purkinje cells were defined as neurons if they had a visible nucleolus and clear cytoplasmic profile. Purkinje cells fulfilling these criteria were counted in each lobular area at 40× magnification, and the Purkinje cell densities (number per square millimeter) were then calculated for each lobule. Because there were no significant changes in cell density in specific lobules from the patients versus controls, the data were pooled to evaluate overall Purkinje cell density.
The gray matter ribbons of the inferior olivary and dentate nuclei were identified at 2.5× magnification, and the gray matter boundaries were outlined. Dentate nuclei were subdivided into quadrants to gauge the pattern of neuronal loss and provide some evidence of projections from the nuclei, which may be particularly affected by perturbed mitochondrial metabolism (20). This information was stored, and dentate nuclear areas were calculated. At 100× magnification, the entire neuronal cell population within the ribbon was counted using the criteria of a clear and visible nucleolus and cytoplasmic profile, and densities of olivary and dentate neurons were first calculated (number per square millimeter) in each quadrant. Because there were no significant differences between quadrants in any of the patients, the data were pooled to identify overall changes in neuron cell density.
Sequential Cytochrome c Oxidase/Succinate Dehydrogenase Histochemistry
A sequential histochemical assay to determine cytochrome c oxidase (COX) and succinate dehydrogenase (SDH) activities was used to identify COX-deficient/SDH-positive neurons in frozen tissue sections as previously described (21). Briefly, 10-μm-thick frozen sections mounted on glass slides were incubated with COX medium at 37°C for 50 minutes, followed by 3 washes with PBS, incubation with SDH medium at 37°C for 45 minutes, washed in PBS, and dehydrated in a graded ethanol series before being cleared in Histoclear and mounted with DPX (Raymond A. Lamb, UK). For those sections used for molecular analysis, the sections were dehydrated in ethanol and allowed to air-dry for 1 hour at room temperature before being stored at -20°C.
Quantification of Neuronal Respiratory Chain Deficiency
Respiratory-deficient neurons were identified either by IHC methods on formalin-fixed, paraffin-embedded sections or by histochemical methods on frozen sections, which reveal protein expression levels or enzyme activity, respectively. A neuron was judged respiratory deficient by an absence of 3,3′-diaminobenzadine staining in the neuronal cell cytoplasm after IHC or by blue staining of the neuronal cytoplasm after histochemical staining for COX/SDH. Numbers of respiratory-deficient cells and total number of cells were then counted and the proportion of respiratory-deficient neurons given as a percentage. Because this study has quantified the percentage of neurons harboring a respiratory chain deficiency for a number of oxidative phosphorylation markers, particularly complex I and complex IV, a severity index was devised. This depended on the following scoring system: 0% neurons deficient = 0, 1% to 9% neurons deficient = 1, 10% to 24% neurons deficient = 2, 25% to 49% neurons deficient = 3, 50% to 74% neurons deficient = 4, and 75% to 100% neurons deficient = 5. The score was applied to the IHC-stained nuclear-encoded subunits 19, 20, and 30 comprising complex I. The scoring system was also applied to the IHC of the mitochondrially encoded COX-I (complex IV subunit I) and to COX/SDH histochemistry because these provide a close correlation (22). A minimum score of 0, implying an absence of deficiency, and a maximum score of 15, implying severe deficiency, could be obtained. Because not all markers could be evaluated in some patients, percentage levels could then be derived from the complex I and complex IV scores (Table 2).
Laser Microdissection and DNA Extraction
For laser microdissection, 20-μm-thick cryostat sections were mounted onto polyethylene naphthalate-membrane slides (Leica Microsystems, UK). Because the levels of neuronal COX deficiency were low in these patients, COX/SDH histochemistry was performed. Individual neurons with a COX-deficient/SDH-positive cell body were microdissected from the inferior olivary nucleus, Purkinje cells, and dentate nucleus using a Leica laser microdissection microscope (Leica Microsystems LMD 6000, UK). DNA extraction was performed overnight using a standard lysis buffer (23).
Determination of Percentage Level Mutated mtDNA to Wild-Type mtDNA
Pyrosequencing for Point Mutations
The quantitation of mtDNA mutation load was performed using Pyromark Assay Design Software v.2.0 (Qiagen, West Sussex, UK) to design locus-specific polymerase chain reaction (PCR) and pyrosequencing primers that amplified a PCR product spanning the mutation region using a forward primer and a biotinylated reverse primer (Integrated DNA Technologies, Glasgow, UK; Table, Supplemental Digital Content 3, ). Pyrosequencing was achieved on the Pyromark Q24 platform according to the manufacturer’s protocol, using a mutation-specific pyrosequencing primer. Pyromark Q24 software was used to quantify mutated mtDNA heteroplasmy levels by directly comparing the relevant peak heights of both wild-type and mutant mtDNA at this site.
Real-Time PCR for mtDNA Deletions
A multiplex real-time PCR MT-ND1/MT-ND4 assay was used to quantify the levels of mtDNA deletions in individual neurons (24–26). The assay provides a deletion (heteroplasmy) level in the neurons. A total of 5 μL of sample was used and run in triplicate in a 96-well plate. A 24-μL master mix consisted of deionized water, TaqMan Universal Mastermix (Applied Biosystems, Paisley, UK), 100 nmol/L MT-ND1 VIC probe, 100 nmol/L MT-ND4 FAM probe, 300 nmol/L MT-ND1 FOR, 300 nmol/L MT-ND1 REV, 300 nmol/L MT-ND4 FOR, and 300 nmol/L MT-ND4 REV primers (Eurofins, Wolverhampton, UK; Table, Supplemental Digital Content 3, ). The PCR pro gram consisted of a 2-minute incubation at 50°C, 10 minutes at 95°C and 40 cycles of amplification, 15-second denaturation at 95°C, and 1 minute at 60°C for hybridization of probes and primers and DNA synthesis. Known deletion-level standards, a blood-positive control, and a blood-negative control, all run in triplicate, were added to the assays, as previously described (26).
Clinical evaluation using the NMDAS rating scale showed that all 14 patients displayed symptoms consistent with cerebellar ataxia. In these patients, the severity of ataxia varied without any relation to the mtDNA defect type (Table 1), with the highest levels of impairment observed in patients with either m.3243A>G (patient 6), m.8344A>G (patient 8), single large-scale mtDNA deletion (patient 11), or recessive POLG mutations (patients 12–14).
General Neuropathologic Findings
Most of the patients had variable degrees of cerebellar abnormalities; there was variable and selective loss of one or more neuronal types among Purkinje cells, olivary and dentate neurons, and/or granule cells. Evidence of myelin loss, dendritic abnormalities, and axonal degeneration were apparent in patients with a high degree of neuronal cell loss (Fig. 1). In accordance with previous findings in Kearns-Sayre syndrome (KSS), we found myelin loss and spongiform degeneration of the white matter tracts of patient 11 (Fig. 1Aiii) (18). We recently discussed the mechanisms underpinning the loss of myelin in this patient and attribute these changes to oligodendrocyte dysfunction (27). Other patients with different mtDNA defects also showed varying degrees of myelin loss, but this was likely secondary to axonal/neuronal loss because, in general, it appeared to occur in regions affected by severe neuron loss. Patients with the m.3243A>G mutation typically exhibited intact myelin, with only a mild myelin pallor in 1 case (patient 5; Fig. 1Aii) and selective myelin loss within regions of microinfarcts. Several patients had a profound myelin loss of the dentate nucleus outflow tract, which correlated with a marked neuron loss from the dentate nucleus (Fig. 1Aiv). These included patients with m.8344A>G, m.14709T>C, and recessive POLG mutations.
Synaptic disorganization assessed using synaptophysin IHC was prominent in many cases. The main changes were in the dentate nucleus. There was often an absence of presynaptic terminals on neurons in patients with extensive Purkinje cell loss compared with controls (Fig. 1Bi-iii), indicating a loss of input to the dentate nucleus. In contrast, patient 9 was the only case in which there were large intensely labeled synaptic-like structures lacking morphologically distinguishable postsynaptic neurons, possibly surrounding “ghosts” of degenerating neurons (Fig. 1Biv, v). This suggests relatively increased input to the dentate nucleus because the extent of the dentate neuron loss exceeded the Purkinje cell loss in this patient.
Axonal and dendritic abnormalities were observed in almost all patients and were particularly prominent in Purkinje cells. Most notably, axonal “torpedoes,” or swellings, were present in the proximal portion of Purkinje cell axons. These swellings largely consisted of neurofilaments and mitochondria and were localized in the Purkinje cell, granular cell, and molecular cell layers (Fig. 1Ci). Axonal spheroids in the deep white matter were often seen in conjunction (Fig. 1Ciii, iv). Dendritic abnormalities consisted of an increase in dendritic arborization, thickening of the dendrites, and evidence of “trapped” mitochondria (Fig. 1Cii). Often, the dendritic trees became more expansive in those Purkinje cells’ neighboring regions where cell loss had occurred, suggesting a compensatory mechanism.
There was evidence of cerebellar atrophy with increased interfolial spaces, mainly caused by widespread atrophy of the molecular layer and loss of granule cells. This was typically observed in patients with the m.3243A>G mutations (patient 5; Fig. 2Biv). The presence of small, shrunken, and eosinophilic “dark” neurons was the most widespread morphological alteration (patient 8; Fig. 2Aiii).
Variations in Neuronal Cell Loss in Different Areas of the Olivary-Cerebellum
Quantification of neuronal cell density revealed significant variations in cell loss across the olivary-cerebellum in relation to mtDNA defects (Fig. 2). Percentage levels of cell loss are shown in Table 1. Patient 5 (m.3243A>G) was the only patient who demonstrated an increase in cell density for the inferior olivary and dentate nucleus neurons versus controls. This was likely caused by a severe cerebellar atrophy; the percentage values for cell loss are listed as 0% in Table 1. Cell loss was typically most profound in all regions in association with the m.8344A>G and POLG mutations. For example, patient 12 (p.Ala47Thr and p.Trp748Ser POLG mutations) exhibited 81%, 91%, and 96% loss of neurons from Purkinje cells (Fig. 2Bv), inferior olivary, and dentate nucleus, respectively. Although patients with the m.3243A>G demonstrated interpatient variability in cell density, the main finding was of severe Purkinje cell loss (likely because of the presence of ischemic-like lesions), with an approximate average of 50% of cells lost (Table 1; Fig. 2Biii). Cell density was generally reduced in all other patients with different mtDNA defects. It appeared that for patient 9 (m.14709T>C), neuronal cell loss was most severe in the dentate nucleus, with 74% loss, whereas the inferior olivary nucleus and Purkinje cells were only moderately affected. Approximately 50% of neurons were lost from the olivo-cerebellum in patient 10 with an m.13094T>C mutation. Purkinje cells and inferior olivary neurons were predominantly affected in patient 11 because of the single large-scale mtDNA deletion where cell loss reached 52% and 60%, respectively. In that case, cell loss in the dentate nucleus was minimal (Fig. 2Cv).
Quantification of neuronal cell density revealed that patients with the common m.3243A>G mutation associated with the mitochondrial encephalopathy with lactic acidosis and strokelike episodes syndrome (MELAS) displayed microinfarcts or ischemic-like lesions in the cerebellar cortex (Fig. 2Biv). Typically, these lesions affect not only the Purkinje cells but also the molecular layer, granular cell layer, and adjacent white matter. In m.3243A>G patients, neuronal cell loss was predominately caused by the microinfarcts. In cases with the myoclonic epilepsy with ragged red fibers–associated mutation, m.8344A>G, and recessive p.Ala467Thr and p.Trp748Ser POLG mutations, microinfarcts were also found, but there was also more widespread Purkinje cell loss not associated with microinfarcts (Fig. 2Bv). Furthermore, assessment of the cerebellar lobules suggested that, in all cases evaluated, microinfarcts typically occurred in the posterior inferior lobules of the cerebellum.
Immunohistochemistry with glial fibrillary acidic protein, a marker of astrocytes, revealed proliferation of the astroglial cell populations in regions where cell loss was greatest. This was particularly obvious microinfarcts in which proliferation of the Bergmann glia was prominent. In addition, deficiency of components of the mitochondrial respiratory chain in the astroglia was evident in serial sections (Fig. 3). This suggests that astroglia may harbor respiratory-compromised mitochondria.
Correlation Between Neuronal Cell Loss and a High NMDAS Score for Cerebellar Ataxia
We observed a strong correlation between a high NMDAS score for ataxia and evidence of a low neuronal cell density in all 3 areas of the olivary-cerebellum, suggesting that the greater the impairment, the less chance that cellular density is maintained. Our data indicate that a high functional impairment is accompanied by a physical loss of neurons (because a low cell density indicates cell loss). This correlation was found to be significant in the inferior olivary nucleus (Pearson correlation = −0.902, p = 0.000), the Purkinje cell (Pearson correlation = −0.628, p = 0.029), and dentate nucleus (Pearson correlation = −0.577, p = 0.049) neuronal populations.
Molecular Genetic Investigation of Levels of Mutated mtDNA
Molecular genetic analysis was performed on individual isolated Purkinje cells and neurons of the inferior olivary and dentate nuclei to determine mutation load. Pyrosequencing of mtDNA point mutations revealed high percentage levels in remaining cells throughout the olivo-cerebellum (Table 3). High levels of point mutation did not correlate with the degree of cell loss. This was particularly evident in patient 9 where the m.14709T>C mutation levels were near homoplasmy, yet only dentate nucleus neurons reveal severe cell loss. Patients with the m.3243A>G mutation generally showed higher percentage levels of mutated mtDNA that are typically greater than 73%. A lack of correlation between a high mutation load and cell loss is clearly demonstrated by patient 4 (m.3243A>G) in which the level of m.3243A>G reached 88% in inferior olivary neurons despite having only 42% cell loss in this region (Tables 1, 3).
In contrast, patients with recessive POLG mutations and accumulated multiple mtDNA deletions showed relatively low levels of mtDNA deletion after real-time PCR. The highest level (50.7%) of mtDNA deletion was seen in the Purkinje cells from patient 13 and the lowest (19.9%) was in Purkinje cells from patient 12. This was despite the extreme cell loss seen in these patients. Neuron cell loss did not correlate with the level of mutated mtDNA in the remaining neurons. Therefore, the level of mutation alone does not appear to determine where neuron cell loss will occur. However, we did not evaluate mtDNA depletion in these patients.
Mitochondrial Protein Changes
Various mitochondrial markers were used to investigate the selective vulnerability of neurons. Porin IHC revealed punctate uniformly high staining within neuronal cytoplasm in both control and patients, indicating that the mitochondrial mass was abundant. However, neurons within the inferior olivary nucleus were the exception, showing unusual mitochondrial localization in both patients and controls. Generally, mitochondria within these neurons were distributed around the periphery of the cell body or were perinuclear, whereas the cytoplasm remained devoid of mitochondria.
Quantification of the respiratory chain proteins revealed a marked respiratory chain deficiency; there was complex I deficiency within remaining neuronal populations but only low levels of complex IV deficiency (Table 2). The percentage level of neuronal respiratory chain deficiency correlated with the degree of cell loss, particularly in the Purkinje cell and dentate nucleus neuronal populations. Typically, patients with extreme cell loss demonstrate high percentage levels of complex I–deficient neurons and moderate levels of complex IV–deficient neurons. The data obtained from patient 11 (m.14709T>C) support these findings, with a severe neuron loss from the dentate nucleus at 74% and with a complex I and complex IV severity index of 60% and 30%, respectively (Table 2). Conversely, patients with the m.3243A>G mutation where cell loss was moderate, a lower percentage of neurons was deficient for complex I, and complex IV was unaffected. For example, patient 3 (m.3243A>G) demonstrated a 26% reduction in neuron density in the dentate nucleus, with remaining cells revealing a complex I and complex IV severity index of 30% and 20%. This was also seen in patient 10 (m.13094T>C) and patient 13 (POLG mutations and multiple mtDNA deletions), where Purkinje cell and dentate nucleus cell loss reached 40% to 45% and the complex I and complex IV severity index revealed moderate and mild scores, respectively (Table 2; Fig. 4). Intriguingly, there was a stronger relationship between complex IV deficiency and cell loss, suggesting that the development of complex IV deficiency is particularly detrimental to neuronal viability (Purkinje cells, Pearson correlation coefficient = 0.736, p = 0.006; dentate nucleus, Pearson correlation coefficient = 0.606, p = 0.048).
The neurons within the inferior olivary nucleus showed widespread deficiency for complex I in the absence of neuronal cell loss (Fig. 4; patient 4). Therefore, all patients had a high complex I severity index, often reaching 100% (Table 2).
Despite the recognition of cerebellar ataxia as a frequently reported neurological deficit in patients with mtDNA disease, the neurodegenerative changes underpinning this disorder are largely unknown. The present study performs a detailed quantitative clinical, neuropathologic, and molecular investigation on a large number of patients with a variety of mtDNA abnormalities to determine the mechanisms of neurodegeneration in the olivo-cerebellar pathway.
Generalized cerebellar atrophy and eosinophilic dark neurons are commonly detected throughout the olivo-cerebellum in mtDNA disease. We also document degenerative changes in connectivity throughout these structures. These changes include the formation of axonal spheroids and “torpedoes,” axonal loss, synaptic changes, and demyelination. All cases investigated appear to show degenerative changes in axons and dendrites that correlate with the extreme cell loss in these regions. Evidence of Purkinje cell axonal degeneration has been reported in patients with single large-scale mtDNA deletions and the m.3243A>G mutation (21, 28). The current study extends this observation to other patients with different mtDNA mutations and suggests that axonal abnormalities are more widespread than previously thought in mtDNA disease. It is intriguing to note that similar swellings have been documented in patients with essential tremor, Alzheimer disease, and Parkinson disease (29–31). It is uncertain what these changes represent, but it is likely that they would result in axonal block in both an anterograde and retrograde direction and would likely lead to Purkinje cell death. The loss of myelin in these regions is likely a secondary consequence of axonal and neuronal demise. This would be true for all cases, except KSS, in which white matter abnormalities far outweighed gray matter abnormalities, consistent with previous reports (32).
Correlation of Cell Loss With Ataxia
Analyses of neuronal cell density and NMDAS scores revealed a strong association between severe ataxia and reduced neuronal density, although the olivo-cerebellum showed considerable variation in both the pattern and severity of neurodegeneration in the cases with mtDNA defects. The emergence of the infarct-like lesions is an important finding because this is not specific to MELAS patients with the m.3243A>G mutation. In fact, microinfarcts in the cerebellum and other brain areas have also been detected in patients with the m.8344A>G mutation and heterozygous POLG mutations (p.Ala467Thr and p.Trp748Ser; p.Gly848Ser and p.Ser1104Cys) herein and by others (33–35). Intriguingly, clinical assessment of patients with the m.8344A>G and POLG mutations showed strokelike episodes, which are associated with MELAS but also occur in these patients (36, 37). Whereas these infarct-like lesions resemble true microinfarcts, they do not conform to a particular vascular territory and have a cortical laminar distribution. Patients with the m.3243A>G suffer Purkinje cell loss, which tends to occur focally in regions of microinfarct, suggesting that the main mode of cell loss occurs through these lesions. The mechanisms behind the formation of these lesions are not known, but this type of lesion has been documented in cerebral ischemia and hypoxia that arises from localized energy deficiency (35).
This study used 2-dimensional neuron counting protocol because of limited availability of the patient postmortem CNS tissues and the need to make maximal use of available material. Because there is some debate in the scientific community regarding the usefulness of 3-dimensional stereology (38), and other studies have shown that 2- and 3-dimensional methodologies yield similar results (39), we used the former. Our results show a clear and obvious reduction of neuronal cell density in patient tissues versus control tissues. This is further strengthened in the assessment of Purkinje cells where the higher age of our control cases was documented, where Purkinje cell loss might be a feature (40–42).
Astrocytes in regions of cell loss are readily identified in mtDNA disease (18, 43). However, whether this alleviates or contributes to the pathophysiology of disease remains to be determined. Indeed, it has been postulated that dysfunctional astrocytes might be key players in gray matter pathology in KSS (43). Recent work in embryonic stem cells with genetic mutations in complex I–and complex IV–encoded mtDNA subunits differentiated into neurons also reveals the presence of astrocytes in the culture system. Severe mutant cells with complex I deficiency show an overall increase in the number of astrocytes; however, they exhibit a reduced mitochondrial membrane potential and lowered glutathione (44). It is possible that although mitochondrial dysfunction is evident in astrocytes, they are not as sensitive to mtDNA mutations because they can switch to glycolytic metabolism when adenosine triphosphate levels are lowered (45). It may be that the proliferation of astrocytes in mtDNA diseases might be a protective response to try to support neurons with dysfunctional mitochondria.
Heteroplasmy Level of Mutated mtDNA and Respiratory Chain Deficiency in Neurons
Previous investigations of heteroplasmic mutations, including the m.3243A>G mutation, have shown that the proportion of mutated to wild-type mtDNA in patients’ non-CNS tissue is crucial in the development of a biochemical defect and clinical abnormality (46). Therefore, evaluation of mutated mtDNA heteroplasmy levels in neurons from different regions of the CNS might account for the selective vulnerability in patients with mtDNA disease. However, combining our neuropathologic data, such as cell loss, with mutated mtDNA heteroplasmy reveals a lack of correlation. Typically, point mutations, including m.3243A>G and m.8344A>G, show a high heteroplasmy in the remaining neurons, whereas mtDNA deletions show relatively low heteroplasmy and the m.14709T>C mutation reaches near homoplasmy. This lack of correlation between cell loss and heteroplasmy has also been previously reported (21, 47) and suggests that there must be other mechanisms that determine which cells undergo cell death. These could include nuclear modifiers, neuronal mtDNA content, neuron subtype–specific neurotransmitter metabolism, and other factors.
We found high levels of complex I deficiency in surviving neurons in the olivo-cerebellum. A number of recent studies have shown elevated levels of neuronal complex I deficiency in CNS tissues from patients with POLG mutations (10). Here, deficiencies of complex I appeared to be most profound in nuclei with the greatest degree of cell loss. It is surprising that such high levels of complex I–deficient neurons exist without undergoing cell death, suggesting that complex I deficiency is not critical for determining the fate of neurons but is associated with cell loss. This also might indicate that complex I is affected earlier in disease pathogenesis. In contrast, complex IV deficiency was only found at very low levels within the neuronal populations and in regions where cell loss was most extensive. We believe that the mechanisms underpinning development of respiratory deficiency and neuronal cell loss are likely to be caused by the progression of the molecular defect. A high degree of heteroplasmy with associated low levels of wild-type mtDNA would initially lead to complex I deficiency and, as the molecular defect progresses, neurons develop complex IV deficiency. Neurons with a complex IV deficiency have a higher propensity to undergo cell death. This can be explained by the strong relationship between the presence of complex IV deficiency and lowered cell density.
This study highlights the importance of correlating clinical, molecular, and neuropathologic changes occurring in the CNS tissues from patients with mtDNA defects. Our results show that whereas there is a clear variability in the extent of neurodegeneration in the olivo-cerebellar pathway, the proportion of neurons with respiratory deficiency appears to correlate with the degree of cell loss. We also show a correlation between the clinical severity of disease and the severity of neuronal cell loss. This suggests that clinical assessments of patients involving the NMDAS is informative for determining the level of functional impairments but might also provide some insight into neurodegenerative changes occurring in the brain.