Dissecting the neuronal vulnerability underpinning Alpers’ syndrome: a clinical and neuropathological study

Abstract Alpers’ syndrome is an early‐onset neurodegenerative disorder often caused by biallelic pathogenic variants in the gene encoding the catalytic subunit of polymerase‐gamma (POLG) which is essential for mitochondrial DNA (mtDNA) replication. Alpers’ syndrome is characterized by intractable epilepsy, developmental regression and liver failure which typically affects children aged 6 months–3 years. Although later onset variants are now recognized, they differ in that they are primarily an epileptic encephalopathy with ataxia. The disorder is progressive, without cure and inevitably leads to death from drug‐resistant status epilepticus, often with concomitant liver failure. Since our understanding of the mechanisms contributing the neurological features in Alpers’ syndrome is rudimentary, we performed a detailed and quantitative neuropathological study on 13 patients with clinically and histologically‐defined Alpers’ syndrome with ages ranging from 2 months to 18 years. Quantitative immunofluorescence showed severe respiratory chain deficiencies involving mitochondrial respiratory chain subunits of complex I and, to a lesser extent, complex IV in inhibitory interneurons and pyramidal neurons in the occipital cortex and in Purkinje cells of the cerebellum. Diminished densities of these neuronal populations were also observed. This study represents the largest cohort of post‐mortem brains from patients with clinically defined Alpers’ syndrome where we provide quantitative evidence of extensive complex I defects affecting interneurons and Purkinje cells for the first time. We believe interneuron and Purkinje cell pathology underpins the clinical development of seizures and ataxia seen in Alpers’ syndrome. This study also further highlights the extensive involvement of GABAergic neurons in mitochondrial disease.


INTRODUCTION
Alpers' syndrome, more correctly described as Alpers-Huttenlocher syndrome, is a severe early-onset neurodegenerative disorder characterized by a clinical triad of intractable epilepsy, psychomotor regression and liver failure (2). Other symptoms include developmental delay, failure to thrive, hypotonia, spasticity and ataxia. The exact prevalence of Alpers' syndrome is unknown due to high mortality and early death but it is estimated to affect 1:100,000 births (38). A bimodal disease onset has been described with the first peak of onset affecting infants between the ages of 6 months and 3 years and a second peak, typified by an epileptic encephalopathy, during adolescence (16,48). The disorder is progressive, without cure and is invariably fatal caused by status epilepticus or hepatic failure. The typical clinical presentation of Alpers' syndrome is sudden explosive onset of medically refractory seizures in an apparently healthy and normally developing child or young adult. Initially, seizures may manifest as focal motor seizures and/or myoclonus and often progress to status epilepticus or focal motor status, electroenecephalogram (EEG) typically demonstrates a focus within the occipital lobe (49). Concomitant symptoms may also include cortical blindness, vomiting, headaches, and hemianopia. Cortical and thalamic involvement are seen on magnetic resonance imaging (MRI) and are characterized by areas of high intensity signal on T2 and T2-FLAIR weighted sequences corresponding to stroke-like lesions (44,46). Valproic acid was amongst a number of drugs used to treat the refractory epilepsy associated with Alpers' syndrome until it became evident that it was precipitating liver failure (7).
More than 90% of cases of Alpers' syndrome are thought to be due to autosomal recessive mutations in the nuclearencoded catalytic subunit of polymerase gamma (POLG) (7,8), the sole polymerase responsible for replication and repair of mitochondrial DNA (mtDNA) (14). Mutations in POLG cause a spectrum of clinical phenotypes, including myoclonic epilepsy myopathy and sensory ataxia (MEMSA), ataxia neuropathy spectrum (ANS) and both dominant and recessive chronic progressive external ophthalmoplegia (CPEO) (50). POLG defects result in either mtDNA depletion or accumulation of mtDNA deletions, and in Alpers' syndrome it is depletion of mtDNA which is responsible for the development of severe mitochondrial respiratory chain impairments, and consequent cellular dysfunction and cell death in brain and liver (4,32,46). Recent whole exome sequencing has identified other molecular genetic causes of Alpers' syndrome, including pathogenic mutations in FARS2 which encodes phenylalanyl-tRNA synthetase (10), NARS2 which encodes asparaginyl-tRNA synthetase and PARS2 which encodes prolyl-tRNA synthetase (40,41).
Neuropathological findings typically reveal symmetrical atrophy with laminar necrosis and neuronal degeneration affecting the cerebral cortex, hippocampi, olfactory bulbs and cerebellum (15,16,19). A recent neuropathological study of paediatric and adult POLG-related disease revealed chronic neurodegeneration with areas of acute focal neuronal necrosis involving frontal, temporal and occipital cortices, and the hippocampus, thalamus and cerebellar cortex (46). Subsequent evidence has revealed focal lesions provide no evidence of ischaemia, suggesting this loss is secondary to neuronal ATP depletion caused by a combination of underlying respiratory chain deficiency and epileptic activity increasing neuronal energy demands (45).
In the cerebellum, mild Purkinje cell loss has previously been documented in Alpers' syndrome (29,31). Diffuse Purkinje cell loss has been described in a cohort of adult patients with POLG mutations, and reduced expression of respiratory chain complex I has been noted in remaining Purkinje cells of patients (8,23). Numerous clinical and neuropathological studies reveal the occipital lobe as the main site of seizure genesis and a predominant area of neuronal degeneration in patients with Alpers' syndrome (3,11,19,29,35). Extensive respiratory chain deficiency and interneuron loss has also been documented in occipital lobe tissue of patients with mitochondrial disease including three adult patients harboring recessively inherited POLG variants (22).
In the current study, we aimed to further delineate the pathogenesis of Alpers' syndrome through a quantitative neuropathological investigation of a large series of postmortem brain tissues from thirteen patients with clinicallydefined Alpers' syndrome. Our approach involved quantifying the extent of neuronal-subtype specific cell loss in conjunction with assessing evidence of mitochondrial respiratory chain deficiencies, involving complexes I and IV, in occipital cortex and the cerebellum since these areas often show pronounced neurodegeneration.

Tissue material
Formalin-fixed paraffin-embedded (FFPE) brain tissue for the current study was acquired from the Newcastle Brain Tissue Resource, UK, BRAIN UK, Edinburgh Sudden Death Brain Bank, UK, University of Vienna, Austria, and University of Bergen, Norway. Where possible control cases were matched with our patients for age, sex, postmortem interval and length of tissue fixation (Supplementary Table 1). Not all brain regions were available for each patient and control and this is indicated by Supplementary Table 1. Ethical approval and full consent was granted for this work.
Immunofluorescent identification of mitochondrial respiratory subunits in interneurons, pyramidal neurons and Purkinje cells Immunofluorescent assays were developed to allow identification of subunits comprising complexes I and IV relative to a mitochondrial mass marker within specific neuronal populations. Since there are no reliable, commerciallyavailable antibodies raised against mtDNA-encoded subunits of complex I, commonly affected nuclear-encoded subunits of complex I (NADH dehydrogenase [ubiquinone] 1 beta subcomplex subunit 8 (NDUFB8) or NADH dehydrogenase [ubiquinone] 1 alpha subcomplex 13 (NDUFA13)) were selected in conjunction with a mtDNA-encoded subunit of complex IV (cytochrome c oxidase subunit 1 (COX1)) and a marker of mitochondrial mass (porin or cytochrome c oxidase subunit 4 isoform 2 (COX4I2)). While there are no reliable histochemical methods to asses complex I activity in tissue cryosections, previous work has shown that complex I NDUFB8 and NDUFA13 subunits are incorporated at late stages of complex I assembly and are an essential part of functional enzymatic activity (5,34). Downregulation of COX1 protein expression levels has been show to directly correlate to loss of COX activity upon cytochrome c oxidase (COX)/succinate dehydrogenase (SDH) histochemistry in tissue cryosections (28). Based on these studies, we infer that decreased protein levels can reliably be interpreted as reflecting enzymatic activity. We have optimized and validated these antibodies extensively within other studies of single muscle fibers (1,36) and various brain-specific cell types in tissue sections from patients with mitochondrial disease (8,13,22).
In occipital lobe tissue, these antibodies were used in conjunction with neuronal markers; anti-SMI-32P an antibody that recognizes non-phosphorylated neurofilaments in pyramidal neurons and anti-glutamic acid decarboxylase 65-67 (GAD65-67) an antibody that identifies GABAergic interneurons. The combinations of primary antibodies were selected on the basis of different IgG subtypes. Due to an overlap in IgG1 isotype for SMI-32P and NDUFB8 two triple-labeled assays were used: the first to identify complex I, mitochondria and pyramidal neurons and the second to identify complex IV, mitochondria and pyramidal neurons. These assays were applied to 8 patients and 4 controls were FFPE occipital lobe tissue was available. In the cerebellum, porin was used to define the Purkinje cell body due to their characteristic morphology therefore the assay did not utilize a neuronal marker; this was applied to 9 patients and 8 controls. All primary and secondary antibodies used are listed in Supplementary  Table 2. A no primary antibody (NPA) section was included alongside patients and controls for each brain region.
Immunofluorescence was performed on 5 μm FFPE brain sections as previously described (12,22). This briefly involved deparaffinization and rehydration of the sections followed by antigen retrieval and blocking in 10% goat serum for an hour. Primary antibodies were incubated overnight at 4°C, the following day and where applicable IgG subtype-specific biotinylated goat anti-mouse antibodies were applied to amplify either NDUFB8 or NDUFA13 antibody signal which has previously been shown to be weak. Sections were incubated for 2 h at 4°C, sections were washed with TBST and the appropriate IgG subtypespecific secondary anti-mouse or anti-rabbit antibodies conjugated with Alexa Fluor 405, 488, 546 and 647 antibodies applied for 2 h at 4°C (see Supplementary Table  2). When possible the sections were incubated with Hoescht (Sigma Aldrich; diluted 1:1200 with TBST) for 20 minutes at room temperature to label nuclei. All sections were incubated in 0.3% Sudan Black for 10 minutes to eliminate any autofluorescence. This was followed by washing in distilled water and mounting in Prolong Gold (Life Technologies, UK).

Confocal microscopy
Neurons were imaged using a confocal microscope (Nikon A1R, UK). Neurons were detected using an immersion oil ×60 objective with numerical aperture 1.4 by their SMI-32P or GAD65-67-positive signal. At least 40 neurons were randomly sampled per case using an electronic zoom of ×3.09. Purkinje cells in the cerebellum were detected using a ×20 objective on the basis of their porin signal and imaged at the same magnification. Microscope and laser settings were maintained throughout image capture.

Image processing and analysis
Neurons were identified and manually outlined according to the presence of either GAD65-67 405 nm (for interneurons), SMI-32P 647 nm (for pyramidal neurons) or porin 546 nm (for Purkinje cells) fluorescent signal using Volocity software (PerkinElmer); this defined the region of interest (ROI). For each neuron, the mean optical densities (OD) of each mitochondrial respiratory complex subunit (NDUFB8, NDUFA13 or COX1) and mitochondrial mass marker (porin or COX4I2) were measured within the defined ROI and subject to background correction from NPA section.
The control group was created by sampling equal numbers of neurons from each control subject and pooling the background corrected mean OD values. Box-Cox transformation identified the log transformation as optimal to normalise the data. Linear regression of transformed NDUFB8 (NDUFB8 T ) data against transformed porin (porin T ) data, transformed NDUFA13 (NDUFA13 T ) data against transformed COX4I2 (COX4I2 T ) data and transformed COX1 (COX1 T ) data against porin T data were performed to ensure the residuals of the regression were normally distributed. NDUFB8 T , NDUFA13 T and COX1 T values were then corrected for mitochondrial mass by dividing by the appropriate mitochondrial mass marker (porin T or COX4I2 T ). In order to derive z-scores for each pyramidal neuron, interneuron and Purkinje cell the mass-corrected NDUFA13 (NDUFA13 Z ), NDUFB8 (NDUFB8 Z ) and COX1 (COX1 Z ) values were calculated based on the parameters (mean and standard deviation) of the control group. Neurons were classified based on standard deviation (SD) limits (for NDUFA13 Z , NDUFB8 Z or COX1 Z : overexpression if z > 2SD, normal if z > -2SD, low if z < -2SD, deficiency if z < -3SD and severe deficiency if z < -4SD) and are presented as percentages to give the proportion of cells within each SD limit.

Statistical analysis
All statistical analyses were carried out using GraphPad Prism 7.0 (GraphPad Software, Inc., La Jolla, California). To determine statistical significant changes in complex I or IV protein expression in each patient relative to the control group, median z scores for each patient and control were compared using an unpaired t-test. To determine changes in porin or COX4I2, the mean OD per neuron in patient and control tissues was compared using a Mann-Whitney U test. To ascertain a correlation between neuronal cell loss, complex I or IV deficiency and age at death, Spearman rank correlation coefficient was performed. A significant P-value cut off was taken at *P ≤ 0.05, **P ≤ 0.01 and ***P ≤ 0.001.

Two-dimensional neuronal cell quantification
A two-dimensional neuronal cell counting protocol was employed using a modified light microscope (Olympus BX51), motorized stage for automatic sampling, CCD colour video and stereology software (StereoInvestigator, MBF Bioscience, Williston, VT) as previously described (22,23). This briefly involved outlining an area of at least 10μm 2 along the cortical ribbon for the occipital cortex encompassing cellular layers I-VI at ×2 magnification. Within this area GAD65-67-positive interneurons and SMI-32Ppositive pyramidal neurons in occipital lobe tissues were counted at ×20 magnification. Purkinje cell density was calculated by identifying the Purkinje cell layer using the closed contour function of StereoInvestigator, accounting for their linear arrangement. Purkinje cells with a visible nucleolus were counted within this, and density calculated. Neuronal cell densities were calculated as number per mm 2 .
Since only one measurement was provided per patient and control case, and due to the low numbers in each group, it was not possible to perform statistical analysis of neuronal cell density data.

Clinical features
In total thirteen patients with clinically-determined Alpers' syndrome were investigated and a diagnosis of Alpers' syndrome was made on the basis of the presence of refractory seizures/status epilepticus, psychomotor regression and/or liver failure (  (46), the clinical course and muscle biochemistry reported for patient 8 by Morris and colleagues (30) and the clinical and neuropathological findings reported for patient 10 by Jellinger and colleagues (17).

Gross neuropathological findings
The gross neuropathological findings from post-mortem are summarized by Table 2. The most common findings were significant atrophy affecting the cerebral hemispheres and focal areas of cortical depressions corresponding to areas of cortical laminar dehiscence caused by neuronal necrosis. Typically the occipital, parietal and cerebellar cortices were affected by severe, focal pan-necrotic lesions featuring profound neuronal cell loss and a strong astrocytic response. These findings are compatible with previous descriptions in the literature (15,45,46).

Respiratory chain deficiency in interneurons and pyramidal neurons
Quantitative quadruple-labeled immunofluorescence showed co-localization of NDUFB8, COX1 and porin within control interneurons and pyramidal neurons. In both the occipital lobe and cerebellum, patient neurons showed a significantly higher mean OD of porin per cell than control neurons (Mann-Whitney U test, P < 0.0001) suggesting that mitochondrial mass was much higher in patient neurons.
Interneurons in the occipital cortex of patients exhibited significant downregulation of NDUFB8 (unpaired t-test, P < 0.0001) and COX1 (unpaired t-test, P < 0.0005) protein expression relative to mitochondrial mass marker porin ( Figure 1A). In all patients a high proportion of interneurons demonstrated severe complex I deficiency (Figure 1Bi) while complex IV deficiency was more variable, with only patients 4 (p.Ala467Thr/p.Gly848Ser), 6 (p.Gly303Arg/p. Ala467Thr), 7 (pAla467Thr/p.Gly848Ser) and 10 (p.Thr748Ser/p.Thr748Ser) demonstrating severe complex IV deficiency in comparison to other patients (Figure 1Bii). All patient interneurons expressed complex I subunit NDUFB8 below the normal limits indicating that all interneurons exhibited severely deficient, deficient and low levels of NDUFB8 expression. While for complex IV subunit COX1 only two patients; patient 4 (p.Ala467Thr/p.Gly848Ser) and patient 6 (p.Gly303Arg/p.Ala467Thr), revealed downregulation of COX1 within all interneurons. indicates that there was only limited neuropathological information for these patients since they are historical cases.
Patient pyramidal neurons in the occipital cortex also demonstrated marked deficiency of NDUFA13 (unpaired t-test, P = 0.0358; Figure 2A,B). However, this is likely an underestimation of complex I deficiency since comparison of COX4I2 OD (which is used as a mitochondrial mass marker in this assay) in patient and control neurons revealed a lower density of COX4I2 expression in neurons implying a loss of this protein in patient neurons. Significant downregulation of COX1 expression levels in patient pyramidal neurons was also observed (unpaired t-test, P < 0.0063; Figure 3A,B). All patients showed high percentages of both complex I-deficient and severely deficient neurons ( Figure 2B). Patient 11 showed severe complex I deficiency in all neurons examined. Patient neurons showed milder complex IV deficiencies, with the most extensive deficiency being seen in patient 11 ( Figure 3B).

Respiratory chain deficient Purkinje cells of the cerebellum
The Purkinje cells in cerebellar cortex of patients exhibited decreased respiratory chain protein expression for both complexes I and IV (Figure 4). All patient Purkinje cells demonstrated a significant downregulation of complex I relative to control subjects (unpaired t-test,

Neuronal loss in Alpers' syndrome
The consequence of respiratory chain deficiency was investigated by quantification of GABAergic and pyramidal neuronal cell densities in the occipital cortex and Purkinje cell density in the cerebellar cortex. The grey matter cortical ribbon demonstrated atrophy, with laminar dehiscence observed in patients 4 (p.Ala467Thr/p.Gly848Ser), 6 (p.Gly303Arg/p.Ala467Thr), 8 (p.Ala467Thr/p.Thr914Pro),  Figure 5A) and pyramidal neurons ( Figure 5B) throughout patient cortical layers relative to controls suggestive of interneuron and pyramidal neuron loss. Neuronal loss did not correlate with either complex I or complex IV deficiency in interneurons or pyramidal neurons, nor did it correlate with age (Spearman rank correlation coefficient).
The cerebellar cortex was atrophic in patients relative to controls, with a fewer Purkinje cells observed in the majority of cases (Figure 6Aii). An area of focal neuronal necrosis, consistent with a stroke-like lesion, was seen in patient 13 (Figure 6Aiv). In other patients, Purkinje cells were selectively affected, with no apparent loss of granular cells or cells in the molecular layer on visual assessment. Quantification of Purkinje cell density revealed lower  Figure 6B) relative to controls. Purkinje cell loss did not correlate with age or with complex I or IV deficiency in surviving Purkinje cells (Spearman rank correlation coefficient).

DISCUSSION
Epilepsy, developmental regression and liver involvement are the classical features of Alpers' syndrome. Due to the fulminant epilepsy, this syndrome is associated with poor prognosis and a rapidly progressive disease course. In the current study, epilepsy was a salient feature in all patients, and the presenting symptom in six. Although epilepsy is common in Alpers' syndrome, the mechanisms underlying seizure generation and maintenance remain unclear. Using a novel quantitative approach, we investigated the development of neuronal respiratory chain deficiency and cell loss in a large cohort of patients with clinically defined Alpers' syndrome. In particular, we focused on potential mechanisms contributing to seizures. We showed that occipital cortex GABAergic interneurons and pyramidal neurons, as well as Purkinje cells in the cerebellum, exhibited extensive respiratory chain defects as evidenced by significant downregulation of complex I subunit proteins and milder downregulation of complex IV subunit proteins. This was despite evidence of maintained or increased mitochondrial mass which could represent a compensatory response. Although loss of interneurons and pyramidal cells was similar, the magnitude of respiratory chain deficiency was more pronounced in interneurons raising the possibility that the loss of inhibitory neural networks plays a role in the seizure development in these patients.
A major limitation of this study is the absence of a confirmed genetic diagnosis for eight of our patients due a lack of availability of suitable tissues for molecular genetic testing. Unfortunately many of these cases are historical and precede the development of sequencing technologies to identify the causative genes for Alpers' syndrome. Our criteria for inclusion of patient tissues was dependent on a clinical disease course and histopathological changes consistent with Alpers' syndrome.

Occipital lobe vulnerability in Alpers' syndrome
The occipital lobe appears to be the main site for seizure genesis in Alpers' syndrome and visual disturbances are also frequently described in both paediatric and adult patients harboring POLG mutations. Such visual phenomena, include flickering light, hallucinations, or loss of vision (11) and similar symptoms were described in seven of our patients. In this study, we confirm pathology affecting the occipital cortex with evidence of cortical atrophy, microvacuolation, and areas of focal neuronal necrosis in seven out of nine patients. We for the first time demonstrated quantitative evidence of extensive involvement of cortical interneurons in patients with Alpers' syndrome. This study supports a vulnerability of GABAergic interneurons to mtDNA defects in patients with mitochondrial disease and is in agreement with a recent study revealing similar findings in a cohort of adult patients with mitochondrial disease (22). Unfortunately due to limited tissue availability, it was not possible to ascertain whether the severity and type of pathological changes were restricted to the occipital lobes and it will be important to extend these investigations to other cortical regions in Alpers' syndrome in future studies.

Cerebellar vulnerability in Alpers' syndrome
In addition to the clear occipital involvement, we show a reduction in Purkinje cell density and respiratory chain defects, affecting subunits of complexes I and IV, in Alpers' syndrome. This is a novel albeit unsurprising finding given ataxia is common in both paediatric and adult patients with this syndrome. Reduced Purkinje cell density combined with compromised respiratory chain protein expression has previously been documented in adult patients with mitochondrial disease, including those with POLG mutations (8,23). Cerebellar atrophy and Purkinje cell loss are common histological changes in patients with epilepsy (9, 37) and many develop ataxia (51). The specific mechanisms underpinning this pathology are unknown, however it could be attributed to seizure activity, particular status epilepticus (42) or due to toxicity of anti-convulsants, such as phenytoin (9,25,26,33). It is challenging to determine the precise mechanism by which the cerebellum is affected since both mechanisms may act synergistically. Unfortunately, we are unable to comment on phenytoin use in our patient cohort due to the scarcity of clinical information, however given the short-nature of the disease, and extensive respiratory chain deficiency in remaining Purkinje cells we think it unlikely that the pathology is primarily caused by anticonvulsant toxicity.

Neuron-specific pathology
Our data show GABAergic interneurons and pyramidal neurons of the occipital cortex, and Purkinje cells in the cerebellum exhibit extensive respiratory chain defects with significant downregulation of complex I subunit proteins coupled with milder downregulation of complex IV subunit proteins despite evidence of maintained or increased mitochondrial mass. These findings provide further support for the hypothesis that focal neuronal loss seen in POLGrelated stroke-like lesions is a consequence of intrinsic neuronal energy failure rather than ischaemia (45). The consequence of the complex I deficiency within neurons we report in Alpers' syndrome is not well understood, however the findings here support suggestions that complex I deficiency could be a contributory factor in the development of epilepsy (20).
Quantitative analysis of neuronal density showed significantly reduced densities of interneurons and pyramidal neurons in the occipital lobe, and reduced Purkinje cell density in the cerebellum in patients relative to controls. Previous studies have highlighted areas of focal cortical necrosis in the pathogenesis of Alpers' syndrome; a finding observed in seven patients in this study (46). In the remaining six patients, neuronal loss appeared widespread throughout the occipital and cerebellar cortices. This could suggest two independent mechanisms of neurodegeneration are contributing to the pathology observed in Alpers' syndrome.
Our findings of severe respiratory chain defects and loss of interneurons in the occipital lobes of our patients are novel and potentially offer the first clue to what may be driving the epilepsy in this disease. Epilepsy often heralds the rapid deterioration in patients with Alpers' syndrome, and it was the only common symptom in all 13 patients in our cohort. It is striking, therefore, that patients 2, 4, 5 and 7 exhibited deficiency of complex I in all surviving occipital interneurons examined. Further, while neuropathological studies are limited by the fact that tissue demonstrates end-stage disease, it is remarkable that in all patients in whom occipital lobe tissue was studied, there were no normal interneurons remaining.