Delineating selective vulnerability of inhibitory interneurons in Alpers' syndrome

Abstract Aims Alpers' syndrome is a severe neurodegenerative disease typically caused by bi‐allelic variants in the mitochondrial DNA (mtDNA) polymerase gene, POLG, leading to mtDNA depletion. Intractable epilepsy, often with an occipital focus, and extensive neurodegeneration are prominent features of Alpers' syndrome. Mitochondrial oxidative phosphorylation (OXPHOS) is severely impaired with mtDNA depletion and is likely to be a major contributor to the epilepsy and neurodegeneration in Alpers' syndrome. We hypothesised that parvalbumin‐positive(+) interneurons, a neuronal class critical for inhibitory regulation of physiological cortical rhythms, would be particularly vulnerable in Alpers' syndrome due to the excessive energy demands necessary to sustain their fast‐spiking activity. Methods We performed a quantitative neuropathological investigation of inhibitory interneuron subtypes (parvalbumin+, calretinin+, calbindin+, somatostatin interneurons+) in postmortem neocortex from 14 Alpers' syndrome patients, five sudden unexpected death in epilepsy (SUDEP) patients (to control for effects of epilepsy) and nine controls. Results We identified a severe loss of parvalbumin+ interneurons and clear evidence of OXPHOS impairment in those that remained. Comparison of regional abundance of interneuron subtypes in control tissues demonstrated enrichment of parvalbumin+ interneurons in the occipital cortex, while other subtypes did not exhibit such topographic specificity. Conclusions These findings suggest that the vulnerability of parvalbumin+ interneurons to OXPHOS deficits coupled with the high abundance of parvalbumin+ interneurons in the occipital cortex is a key factor in the aetiology of the occipital‐predominant epilepsy that characterises Alpers' syndrome. These findings provide novel insights into Alpers' syndrome neuropathology, with important implications for the development of preclinical models and disease‐modifying therapeutics.


INTRODUCTION
Alpers' syndrome is a devastating mitochondrial disease characterised by intractable epilepsy, psychomotor regression and hepatic failure [1][2][3]. Focal seizures within the primary visual cortex are a typical early presenting feature and are often refractory to treatment [4].
Other seizure types including myoclonic, generalised tonic-clonic and focal motor seizures are also observed and may lead to status epilepticus or more frequently, epilepsia partialis continua [5]. Additional neurological impairments are common and include developmental delay, cerebellar ataxia, cortical blindness and hypotonia [6]. Symptoms often manifest during early infancy; however, a second peak of onset occurs in adolescence or early adult life [7,8]. Neurological deterioration is rapidly progressive, with fatal consequences occurring within a few months to years of initial presentation.
MtDNA depletion results in decreased activities of mitochondrial oxidative phosphorylation (OXPHOS) complexes and impaired generation of adenosine triphosphate (ATP), which leads to severe neural dysfunction and neurodegeneration [12]. These neurodegenerative changes are most profound within the primary visual cortex with focal necrotic cortical lesions characterised by severe neuronal dropout and gliosis [6,7]. Focal cortical lesions correlate with the clinical onset of stroke-like episodes (SLE) and are frequently observed on cranial magnetic resonance imaging as T2-weighted hyperintensities that are not confined to a single vascular territory [13]. Although the precise pathophysiology of these events remains unclear, they are hypothesised to result from prolonged focal epileptic activity and acute neuronal energy failure resulting in near-total neuronal loss [12].
Recent neuropathological investigations have shown a profound loss of gamma aminobutyric acid (GABA)-ergic inhibitory interneurons from the nonlesional primary visual cortex in Alpers' syndrome, accompanied by severe deficiencies of OXPHOS complexes I and IV within residual interneurons [14]. The severe dysfunction and degeneration of inhibitory interneurons likely underlies a loss of inhibitory neurotransmission that creates a permissive environment for the generation and perpetuation of seizure activity. Similar levels of interneuron loss and multiple OXPHOS deficiencies have been reported in other mitochondrial epilepsies including adult-onset POLG-related pathologies [15]. Different subtypes of inhibitory interneurons have distinct topographies and electrophysiological properties which mediate the differential roles of interneuron subtypes in regulating cortical networks and their involvement in epilepsy [16,17]. Parvalbumin-positive (+) interneurons, which are typically fast-spiking basket cells and underlie gamma frequency oscillations, are enriched with mitochondria and harbour particularly high levels of cytochrome c compared with other interneuron subtypes [18][19][20]. The increased metabolic requirements of parvalbumin+ interneurons to sustain their fast-spiking activity render them particularly vulnerable to OXPHOS defects and ATP depletion [21]. Incubation of rodent hippocampal slices with rotenone and potassium cyanide, inhibitors of complex I and complex IV respectively, abolishes gamma frequency oscillations in vitro [22]. This highlights the critical dependence of OXPHOS for optimal function of parvalbumin+ interneurons and implicates an increased vulnerability of these interneurons in Alpers' syndrome where OXPHOS function is severely impaired. The loss of inhibitory neurotransmission associated with impaired function of parvalbumin+ interneurons would alter physiological cortical rhythms and permit an unrestrained escalation of excitatory activity leading to epileptic seizures.
This study aimed to determine whether specific interneuron subtypes are differentially affected in Alpers' syndrome, with a particular focus on parvalbumin+ interneurons. This will provide a greater insight regarding the role of impaired inhibitory neurotransmission in the generation of epileptic activity and may explain the early and predominant involvement of the occipital cortex in Alpers' syndrome.
Better understanding of these pathomechanisms and cell-specific sensitivity to OXPHOS deficiency will aid the development of accurate, relevant, preclinical models of Alpers' syndrome and allow testing of novel antiepileptic therapies.

Key points
• Intractable epilepsy with occipital focus is a major presenting feature of Alpers' syndrome, but the aetiological factors underlying this focus in the primary visual cortex are not known.

Patient and control cohort
Formalin-fixed, paraffin-embedded human postmortem brain tissues were obtained from three brain regions: occipital cortex (Brodmann Immunohistochemistry for the identification of cortical neuronal subtypes Immunohistochemistry to identify parvalbumin+, calretinin+, calbindin+ and somatostatin+ interneurons and pyramidal neurons was performed using 5 μm thick formalin-fixed, paraffin-embedded sections as previously described (supporting information Table S2) [24]. These interneuron subtypes were investigated due to the known involvement in temporal lobe epilepsy (TLE) and the vulnerability of parvalbumin+ interneurons to mitochondrial dysfunction [22,[25][26][27][28][29].
Immunohistochemistry to identify reactive astrocytes and activated neurons (c-fos) was also performed to neuropathologically characterise cortical lesions.

Neuronal density (ND) quantification
Neuronal densities (neuron/mm 2 ) were quantified using a twodimensional neuronal cell counting protocol on an Olympus BX51 stereology brightfield microscope as previously described [14,15]. Quantification was performed in nonlesioned cortex; however, this was unavoidable for Patients 4-6 and Patients 10-11 as all cortical tissue was severely necrotic. Using control ND data, z-scores were calculated ([log-transformed NDmean control log-transformed ND]/SD control log-transformed ND) in order to make inferences about the severity of neuronal loss in patient tissues. ND classifications were based on the following standard deviation limits: increased density z > 2SD, normal density z < 2SD, mild loss z < -2SD, moderate loss z < -3SD and severe loss z < -4SD. Z-scores could not be calculated for patient tissues with complete loss of neuronal subtypes.

Quadruple immunofluorescence to visualise OXPHOS complexes within interneurons
Quadruple immunofluorescence was performed to identify the nuclear DNA-encoded complex I subunit (NADH:ubiquinone oxidoreductase subunit B8; NDUFB8) and mtDNA-encoded complex IV subunit I (cytochrome c oxidase I; COXI) within mitochondria, using porin (mitochondrial outer membrane protein, VDAC1) as a mitochondrial mass marker, in parvalbumin+ and calretinin+ interneurons. Loss of NDUFB8 and COXI protein levels has previously been reported in Alpers' syndrome GABAergic interneurons, indicative of decreased enzymatic activities of complexes I and IV [14].
Prolonged fixation of tissues in formalin is known to alter the antigenicity of target epitopes; therefore, only tissues which were fixed in formalin for less than 1 year were used for immunofluorescence experiments (unpublished observation; supporting information Table S1).
The immunofluorescence protocol has previously been described [14] but briefly involved heat-mediated antigen retrieval in EDTA and blocking sections in 10% normal goat serum for 1 h at room tempera-  Table S3). The following day, a biotinylated mouse IgG1 antibody was applied for 30 min at room temperature to amplify the signal of the NDUFB8 antibody. Appropriate Alexa Fluorconjugated secondary antibodies specific to the isotypes of the primary antibodies were applied to the sections for 2 h at 4 C (supporting information Table S3). Sections were finally incubated in 3.0% Sudan Black B for 10 min to minimise autofluorescence and were mounted in ProLong™ Gold antifade reagent to preserve fluorescent signals.

Confocal microscopy
Immunofluorescent sections were imaged using an inverted ZEISS LSM800 confocal microscope and ZEISS ZEN (blue edition) software.
A brightfield preview scan at x5 magnification was used to identify the same cortical region in which neuronal densities were quantified.
A second preview scan was captured at x20 magnification (405 nm laser) to identify interneurons for image capture. The tiled image was then used as a map to select individual interneurons to be imaged at x63 magnification (oil immersion lens) using a x2. 5  z < 2SD, low expression z < -2SD, deficiency z < -3SD and severe deficiency z < -4SD.

Statistical analysis
All statistical analysis was performed using GraphPad Prism 9.0 (GraphPad Software, Inc., La Jolla, California) and R (R Core Team, 2020). The assumption of normality was assessed by visual inspection of Q-Q plots and the Shapiro-Wilk test. Linear regression models were used to analyse ND data, total brain weights and porin data, with comparisons across groups by least square means, adjusted for multiple comparisons using the Tukey method. Immunofluorescence OXPHOS data were analysed using the Kruskal-Wallis test followed by Dunn's method for multiple comparisons or the Mann-Whitney U test. The level of significance (alpha value) was set at 0.05.

Clinical presentation
We investigated a postmortem cohort of eight patients with clinically and neuropathologically defined Alpers' syndrome and six patients with genetically confirmed Alpers' syndrome and POLG-related encephalopathy ( were the next most commonly reported symptom in 12 patients.
Seven patients showed signs of hepatic dysfunction and/or hepatic pathology, including hepatic failure in four patients. However, permission was not provided to assess the liver at postmortem for all patients. Seven patients had an ataxic phenotype suggestive of cerebellar dysfunction.

Neuropathological features
The neuropathological features observed within the occipital, frontal and temporal cortices of the patients with Alpers' syndrome are summarised by Table 2. The total weight of patient brains was significantly lower than control (p = 0.002) and SUDEP patient (P < 0.001) brain weights, indicating severe, generalised cerebral atrophy in Alpers' syndrome (supporting information Figure S1). Focal necrotic lesions, characterised by extensive neuronal loss, reactive astrogliosis, spongiosis and thinning of the cortical ribbon, show predilection for the occipital cortex [12]. Neuropathological changes including extensive neuronal loss were also prominent within the frontal cortex, while the temporal cortex was the least severely affected cortical region analysed.

Loss of cortical inhibitory interneurons and pyramidal neurons
Quantification of interneuron densities revealed topographical differences in the relative abundance of particular interneuron subtypes.
The occipital cortex had the highest density of parvalbumin+ interneurons relative to the frontal (P < 0.001) and temporal (P < 0.001) cortices (supporting information Figure S2). Within the occipital cortex, the mean density of parvalbumin+ interneurons was also higher than all other interneuron subtypes quantified.
Representative images of the morphology and density of neuronal subtypes quantified within the occipital cortex are presented in   Figure S3). The density of calbindin+ and somatostatin+ interneurons was also variably decreased, particularly in the occipital and frontal cortices ( Figure 2 and supporting information Figure S3). However, the percentage of parvalbumin+ interneurons remaining in Alpers' syndrome patient tissues compared with control tissues (15%) was consistently lower than the proportion of remaining calbindin+ and somatostatin+ interneurons (30%), indicating an increased degeneration of parvalbumin+ interneurons in Alpers' syndrome ( Figure 2).
Interestingly, calretinin+ interneurons were the least affected interneuron subtype, and were only significantly reduced in the occipital cortex of patients with Alpers' syndrome relative to controls at the group level (P < 0.05, Figure 2) but were preserved in many patient tissues (z > À2), even within cortex affected by focal lesions with a total loss of parvalbumin+ interneurons ( Figure 1B).
Patient tissues which were affected by widespread necrosis also frequently showed normal, or mildly decreased, calretinin+ interneuron density despite a total loss of parvalbumin+ and calbindin+ interneurons.
As comparison, pyramidal neuron densities were also quantified and demonstrated lower densities within the occipital cortex of patients with Alpers' syndrome relative to controls and SUDEP patients (P < 0.001, Figure 2), whereas pyramidal neuron densities were only severely reduced in the frontal (P < 0.01) and temporal (P > 0.05) cortices of patients who had an early-onset, rapidly progressive phenotype of Alpers' syndrome (supporting information Figure S3). Albeit the overall loss of pyramidal neurons as a percentage was consistently lower than the loss of parvalbumin+ interneurons.

Severe OXPHOS deficiencies are observed in parvalbumin+ interneurons
Since there was a consistent, severe loss of parvalbumin+ interneu- SUDEP patient z-scores were within the normal range ( Figure 3B,C) indicating intact NDUFB8 and COXI protein levels. Severe deficiencies were also observed in the temporal cortex and to a lesser extent, within the frontal cortex in Alpers' syndrome ( Figure 3B,C). Interestingly, parvalbumin+ interneurons from patients with Alpers' syndrome showed a trend towards greater loss of COXI protein expression relative to NDUFB8. This may suggest that there is a preferential impairment of complex IV within parvalbumin+ interneurons in Alpers' syndrome.
The mean optical intensity of porin was significantly increased in Alpers' syndrome parvalbumin+ interneurons (n = 135) compared with control interneurons (n = 240) (P < 0.01) (supporting information Figure S4), with most patient parvalbumin+ interneurons of the occipital cortex harbouring a porin z-score > 2 indicating increased mitochondrial mass (supporting information Figure S5), whereas the majority of parvalbumin+ interneurons within the frontal and temporal cortices in Alpers' syndrome showed normal or reduced levels of porin relative to controls (z < 2) (supporting information Figure S5).
In all cortical regions, there was a significant main effect of the mean optical intensity of parvalbumin+ calcium-binding protein (CBP) across all groups (Kruskal-Wallis, P < 0.05) with post hoc analyses revealing a significant loss of parvalbumin+ intensity within the majority of Alpers' syndrome patient parvalbumin+ interneurons relative to controls (P < 0.05) (supporting information Figure S6). Since parvalbumin expression is activity-regulated [36,37], this may suggest reduced activity of parvalbumin+ interneurons in Alpers' syndrome.

Milder OXPHOS deficiencies are observed in calretinin+ interneurons
Since calretinin+ interneurons demonstrated more intact neuronal densities in Alpers' syndrome, the immunofluorescence assay was adapted to directly compare OXPHOS protein expression in these cells. At the group level, calretinin+ interneurons from patients with Alpers' syndrome harboured a significantly higher mean optical intensity of porin relative to control interneurons (P < 0.001) (supporting information Figure S4); however, mean patient porin z-scores were mostly within the normal range (z < 2) indicating normal mitochondrial mass (supporting information Figure S5). Moreover, calretinin+ interneurons from patients with Alpers' syndrome did not show a consistent loss of calretinin+ CBP, with most patients showing either normal levels or increased levels of calretinin+ intensity relative to controls (supporting information Figure S6). Collectively, these findings indicate less severe affectation of calretinin+ interneurons in Alpers' syndrome.

Increased c-fos immunoreactivity indicates cortical hyperactivity
To investigate cortical hyperactivity within the primary visual cortex in Alpers' syndrome, immunohistochemistry to identify c-fos+ cells was performed. FOS is an immediate early gene which encodes the transcription factor c-fos and is transiently upregulated in response to neuronal activity [38]. Therefore, elevated c-fos abun-

Parvalbumin+ interneuron loss shows a predilection for the occipital cortex
The primary visual cortex is the predominant site of epileptogenesis in Alpers' syndrome; therefore, it is not surprising that neurodegenerative changes are most profound within the occipital lobes [6]. We within the temporal cortex of patients who had an early-onset of symptoms and died in childhood. The widespread loss of pyramidal neurons in these patients may reflect more severe seizure activity and rapidly progressive neurodegeneration. Although perhaps infrequently reported, frontal lobe involvement is common in patients with Alpers' syndrome, and neuronal loss within the dorsolateral prefrontal cortex may be associated with cognitive decline, cognitive disinhibition and anxiety in patients with Alpers' syndrome [39,40]. The overall preservation of neuronal densities in the SUDEP group suggests that seizure-associated neurodegeneration is not common within these cortical regions in SUDEP, albeit the cohort was small.

Parvalbumin+ interneurons harbour severe OXPHOS deficiencies
In addition to the consistent loss of parvalbumin+ interneurons and  [12,14,15,41], Parkinson's disease [42,43], Lewy body dementia [44] and TLE [45]. Since parvalbumin+ interneuron loss was extensive in Alpers' syndrome patient tissues, the severe loss of complex IV subunits may be more detrimental to the functioning and survival of parvalbumin+ interneurons. Complex IV-mediated dysfunction of parvalbumin+ interneurons may also underlie the vulnerability and impairment of these interneurons in other neurodevelopmental disorders associated with mitochondrial dysfunction, including autism spectrum disorder (ASD) and schizophrenia [46][47][48] and may also explain why epilepsy is a common comorbidity of these disorders [49,50].
A proportion of interneurons from SUDEP patients also harboured variably reduced levels of NDUFB8 and COXI protein levels, albeit less severe than the patients with Alpers' syndrome. Mitochondrial dysfunction has previously been reported in patients with TLE [45,51]. While mitochondrial dysfunction is the primary pathology underlying interneuron vulnerability and driving seizure activity in Alpers' syndrome, glutamatergic-induced neurotoxicity may underlie the secondary dysfunction of mitochondria in TLE and SUDEP [52,53]. Alpers' syndrome patients also showed a consistent increase in mitochondrial mass within parvalbumin+ interneurons of the occipital cortex which may represent a compensatory response to severe OXPHOS deficiencies. A loss of parvalbumin+ protein expression and altered mitochondrial mass has been reported in ASD, suggesting that overlapping mechanisms may mediate parvalbumin+ interneuron dysfunction in ASD and Alpers' syndrome [46,47].

Increased resilience of calretinin+ interneurons
It is tempting to speculate that parvalbumin+ protein expression may be reduced in an activity-dependent manner due to metabolic failure arising from extensive OXPHOS deficiencies and reduced excitatory input from pyramidal neurons, due to severe pyramidal neuron loss in Alpers' syndrome [36,37], whereas the majority of patient calretinin+ interneurons showed normal or increased levels of calretinin+ protein suggesting preserved activity of calretinin+ interneurons and intact calcium buffering capacity, which may partly underlie the increased resilience of these interneurons.

SLE are seizure-mediated events
We have provided further evidence of seizure-associated activity mediating SLE in patients with Alpers' syndrome. Neuronal energy failure due to severe OXPHOS deficiency is hypothesised to underlie the almost complete neuronal loss associated with stroke-like lesions [13,15]. We have demonstrated a near-total loss of inhibitory interneurons and an upregulation of c-fos within almost all neurons and glial cells within cortex affected by a recent SLE, suggesting hyperactivity of these cells. This supports the hypothesis of neuronal hyperexcitability and seizures precipitating SLE in patients with Alpers' syndrome and POLG-related encephalopathy [13,15].