Mitochondrial respiratory chain deficiency correlates with the severity of neuropathology in sporadic Creutzfeldt-Jakob disease

Mitochondrial dysfunction has been implicated in multiple neurodegenerative diseases but remains largely unexplored in Creutzfeldt-Jakob disease. Here, we characterize the mitochondrial respiratory chain at the individual neuron level in the MM1 and VV2 common molecular subtypes of sporadic Creutzfeldt-Jakob disease. Moreover, we investigate the associations between the mitochondrial respiratory chain and neuropathological markers of the disease. Brain tissue from individuals with sporadic Creutzfeldt-Jakob disease and age-matched controls were obtained from the brain collection of the Austrian Creutzfeldt-Jakob Surveillance. The mitochondrial respiratory chain was studied through a dichotomous approach of immunoreactivities in the temporal cortex and the hippocampal subregions of CA4 and CA3. We show that profound deficiency of all mitochondrial respiratory complexes (I-V) occurs in neurons of the severely affected temporal cortex of patients with Creutzfeldt-Jakob disease. This deficiency correlates strongly with the severity of neuropathological changes, including vacuolation of the neuropil, gliosis and disease associated prion protein load. Respiratory chain deficiency is less pronounced in hippocampal CA4 and CA3 regions compared to the temporal cortex. In both areas respiratory chain deficiency shows a predilection for the MM1 molecular subtype of Creutzfeldt-Jakob disease. Our findings indicate that aberrant mitochondrial respiration could be involved early in the pathogenesis of sporadic Creutzfeldt-Jakob disease and contributes to neuronal death, most likely via ATP depletion. Based on these results, we propose that the restricted MRI diffusion profile seen in the brain of patients with sporadic Creutzfeldt-Jakob disease might reflect cytotoxic changes due to neuronal respiratory chain failure and ATP loss.


Introduction
Sporadic Creutzfeldt-Jakob disease (sCJD) is a severe and lethal neurodegenerative disease that commonly manifests between the sixth and seventh decade of life with an overall incidence of 1-2 /million / year [1]. While it is most often recognized by rapidly progressive dementia and myoclonus, it has a broad spectrum of clinical presentations including ataxia, visual disturbances, tremor and other movement disorders. The neuropathology of sCJD is characterized by neuronal loss, vacuolation of the neuropil, reactive gliosis and disease-associated prion protein (PrP sc ) deposits [2]. PrP sc are pathological aggregations of prion protein (PrP c ), a naturally occurring protein of unclear physiological function encoded by the PRNP gene. sCJD differs from other neurodegenerative proteinopathies in that self-aggregating PrP sc is considered to be the primary disease-causing event, and that it is transmissible through nerve to nerve contact [1]. PrP sc aggregation is disease specific and is found both in subtle synaptic and perineuronal accumulations and more granular, perivacuolar and plaque-like accumulations depending on disease subtype [2]. PrP sc has been shown to exist in molecular subtypes that show differences in size and degree of glycosylation, degree of protease resistance, aggregation state and conformational stability; namely PrP sc type 1 and 2 [3,4].
Clinical and pathological features of sCJD are strongly influenced by the molecular subtype of PrP sc and the genotype of codon 129 of the PRNP gene, coding for either methionine or valine [2,5]. Individuals who are homozygous for the methionine codon (MM) and have type 1 PrP sc (MM1) typically develop rapidly progressive dementia and myoclonus. Pathological examination commonly shows involvement of neocortex, striatum and thalamus, while the hippocampus and brain stem are relatively spared [4]. Individuals who are homozygous for the valine codon (VV) and have type 2 PrP sc (VV2) often present with ataxia and develop dementia at later stages. Pathologically, individuals with VV2 show involvement of the deep neocortical layers, subcortical nuclei and the hippocampal formation [4].
Mitochondrial dysfunction, including mitochondrial respiratory chain (MRC) deficiency and mitochondrial DNA (mtDNA) damage, has been implicated in multiple neurodegenerative diseases [6], but has not been as extensively explored in sCJD. Mitochondrial abnormalities, including morphological alterations, MRC deficiency and impaired mitochondrial dynamics have been reported by studies in cell-and animal models of prion disease [7][8][9][10]. Furthermore, one study indicated decreased RNA and protein expression levels for several MRC subunits in human bulk brain tissue [11]. However, since disease-associated changes in cell-type composition have been shown to confound mitochondrial measurements in bulk brain tissue [12,13], it is unknown to what degree the reported findings reflected MRC depletion in individual cells, or altered cell composition in the CJD samples. While decreased immunofluorescent staining for subunits of complexes I, IV and V was also observed in sCJD [11], this has not been systematically studied. Thus, the question of whether aberrant mitochondrial function occurs within neurons of the sCJD brain remains largely unexplored.
In this work, we study the MRC complexes in single neurons of severely and mildly affected regions in the brain of individuals with either MM1 or VV2 molecular subtypes of sCJD. Our findings show that there is deficiency of all MRC complexes in the temporal cortex of sCJD. Interestingly, MRC deficiency has a predilection for the MM1 molecular subtype of sCJD and is strongly associated with neuropathological markers of the disease.

Patient cohorts and tissue samples
Brain tissue from individuals with sCJD (n = 20) and controls (n = 6) was obtained from the brain collection of the Austrian Creutzfeldt-Jakob Surveillance at the Institute of Neurology, Medical University of Vienna. We chose to study two regions showing different pathological involvement in sCJD: (i) the temporal cortex, which is severely affected, and (ii) the CA4 and CA3 subregions of the hippocampus, which are considered relatively spared of neuropathological changes [14]. Brains were collected at autopsy and sampling was performed as previously described [15]. All samples were collected using a standard technique and a mean fixation time of 30 days. Tissue with suspected prion disease was treated in 98% formic acid (FA) for one hour prior to paraffin embedding according to routine. Our material includes three neurologically healthy controls (Ctrl), and three disease-controls (Ctrl D ) in which CJD was initially suspected but excluded after neuropathological examination. Ctrl D showed signs of neurodegeneration including senile neurodystrophic changes and chronic vascular lesions, but an exact diagnosis was not available. Ctrl D were included to study the effect of FA treatment on the tissue, as neurologically healthy controls had not been FA treated. Individuals with sCJD were grouped according to codon 129 polymorphism of the PRNP gene and PrP sc type (n MM1 = 10, n VV2 = 10) based on neuropathological examination as previously described [16]. All included individuals were examined by a neuropathologist at autopsy and sections were stained with hematoxylin and eosin using a standard protocol and with antibodies against PrP sc (12F10sc, SPIbio #A03221, dilution 1: 2000). Sections were autoclaved at 121°C for 10 min following 5 min treatment in concentrated FA and proteinase K. All sections were stained using the Autostainer Link 48 by Agilent Dako. There was no difference in post-mortem delay, formalin fixation time or age between controls, and individuals with MM1-or VV2 molecular subtypes of sCJD. Furthermore, there was no difference in the disease duration between individuals in the MM1 and VV2 groups. Sex was not balanced between the controls (5 females and 1 male) and patients (9 females and 11 males). As previous studies have not shown gender-dependent differences in mitochondrial staining [17], we deemed this acceptable.

Immunohistochemistry analysis
Both FA treated and -untreated tissue was available from the Ctrl D -group, which allowed us to study the effect of FA pretreatment in single individuals. Only the FA treated sections were later included for comparisons between sCJD and the entire control group. To evaluate the effect of FA treatment, tissue from the Ctrl D -group that had either been FA treated or not was analysed using a semi-quantitative method to calculate the fraction of positive pixels per neuron as previously described [17]. All sections stained with antibodies against MRC complexes or VDAC1 were analyzed visually and the percentage of positive or deficient neurons was determined in the entire CA4, including the transition into CA3, and in a 2.5 mm 2 area in the temporal cortex in close relation to the transition from entorhinal cortex into the temporal cortex. Neurons were identified based on morphology, and only neurons where the nucleus and nucleolus were clearly visible were included. All sections were scanned using the NanoZoomer 2.0-HT While Slide Imager (Hamamatsu), and image analyses were conducted using the NDP.view2 (v2.7.25). Neurons from the CA4/CA3 regions were evaluated at 20X magnification, while neurons from the temporal cortex were evaluated at 40X magnification.

Neuropathological assessment
Hippocampal sections stained with hematoxylin and eosin (H&E) were scored with regard to gliosis and vacuolation based on a scale from 0 (no astrocytic gliosis/vacuolation) to 3 (severe astrocytic gliosis/vacuolation) as previously reported [5]. Reactive astrocytes were identified from their typically swollen eosinophilic cytoplasm. Only cells with a clearly visible nucleus were evaluated. The neuropathological scoring was performed by two investigators (IHF and GGK) and was done in a region-specific manner resulting in one score from the CA4 region and a separate score from the temporal cortex. In addition, region-specific morphometric analysis of immunoreactivity to PrP sc was calculated using Fiji (v.2.0.0) [18]. For the morphometric analyses, single images from the regions of interest were deconvoluted using the color deconvolution tool and an automatic threshold was set using the Isodata threshold mode. Next, all images were binarized and a fraction of measured pixels per image was used to give a score of percentage of PrP sc stain per image. The Isodata threshold mode was chosen as it was the best fit for strongly and intermediately stained images. Negative images were not analyzed (n = 3 in the CA4 of the MM1 group) due to an overestimation of immunoreactivity. Types of PrP sc deposits were not considered in the analyses. However, images from the temporal cortex were acquired in the deeper cortical layers (5 and 6). All images were acquired at 20X magnification. Examples of scores for the neuropathological assessment are shown in Fig. 1.

Statistical analysis
Statistical analyses were performed in SPSS (v.25.0.0) and Graphpad Prism (v7.0b). Clinical data were compared using t-tests, unless the data were not normally distributed in which case Mann-Whitney U-test was used. Normality was tested using the Shapiro-Wilk test. Equality of variances was tested using Levene's test. A one-way ANOVA was conducted between the groups (Ctrl, MM1 and VV2) to test differences of means of the MRC complex analyses. Due to the unequal sample sizes, the Welch's test for unequal variance was used. Eta squared (η 2 ) was used to determine effect size of any significant difference. Games-Howell post hoc test was chosen to decide at what level the groups were significantly different due to their unequal sample sizes and variance. The relationship between neuropathology and MRC deficiency was assessed by pairwise Spearman's correlations. Due to a small sample size, correlations were studied in the entire sCJD group. A linear regression was performed to evaluate covariation between the molecular subtype of sCJD and vacuolation as a marker of neuropathology. A P-value < 0.05 was considered statistically significant in all analyses. Data are presented as mean and 95% confidence interval (CI).

Effect of formic acid treatment on immunohistochemistry
Using semi-quantification of unstained cytoplasm, we found that FA treatment slightly skewed the measurements towards less MRC deficiency compared to non-FA treated tissue (Fig. 2a). To mitigate the confounder of FA treatment effects on the tissue analyses, we chose instead a visual binary analysis approach to categorize each neuron as either MRC positive or deficient. This approach produced substantially more consistent results between FA treated and untreated tissues (Fig. 2b). Specifically, FA treatment had no effect on the staining of VDAC1 or MRC complexes II, III and V. For the MRC complexes I and IV FA treatment had a minor effect on staining and was associated with a slight underestimation of deficient neurons in FA-treated tissue. Thus, FA treatment could potentially skew our results towards less deficient neurons for complexes I and IV in sCJD. As a result of this, the reported deficiency of MRC complexes I and IV in sCJD may be slightly underestimated in this work.

Neuronal MRC deficiency in sCJD
To assess the MRC in the sCJD-brain, we performed IHC for MRC complexes I-V in the CA4 region (Fig. 3) and temporal cortex (Fig. 4) of sCJD patients and controls. VDAC1 was used as a marker for mitochondrial mass. In the temporal cortex, sCJD patients had a significantly higher fraction of MRC deficient neurons compared to controls. Significant deficiency was found for all MRC complexes (complex I; P = 0.001 η 2 = 0.60, complex II; P = 0.001, η 2 = 0.55, complex III; P = 0.003, η 2 = 0.49, complex IV; P = 0.001, η 2 = 0.73, complex V; P = 0.003, η 2 = 0.40). A similar, but overall less pronounced deficiency was found in the CA4/CA3 region (complex I; P = 0.006 η 2 = 0.31, complex II; P = 0.019, η 2 = 0.20, complex III; P = 0.119, complex IV; P = 0.040, η 2 = 0.26, complex V; P = 0.003, η 2 = 0.33; Table 2). VDAC1 staining revealed no difference in mitochondrial mass between groups in either region (temporal cortex; P = 0.542, CA4/CA3; P = 0.621). The total number of evaluated neurons in each group, means with 95% CI and results of one way-ANOVA are summarized in Table 2. As reflected in their means and 95% CI's, there was no difference in staining intensity of neither mitochondrial marker nor MRC complexes between Ctrl D and Ctrl. Subject demographics of included individuals are summarised in Table 1.
MRC deficiency is more pronounced in the MM1 molecular subtype of sCJD Individual assessment of MM1-and VV2 sCJD molecular subtypes showed highly significant differences in terms of MRC deficiency. Compared to controls, tissue from MM1 molecular subtype of sCJD exhibited significant deficiency of all MRC complexes in the temporal cortex (complex I; P < 0.001, complex II; P = 0.001, complex III; P = 0.005, complex IV; P < 0.001, complex V; P = 0.007) and of complex I (P = 0.010) and complex V (P = 0.034) in the CA4/CA3 region. MRC staining was substantially less affected in VV2 molecular subtype of sCJD. With the exception of mild complex V (P = 0.019) deficiency in the CA4/CA3 region, neuronal MRC staining in the VV2 molecular subtype of sCJD was not significantly different from controls ( Table 2).

Complex I deficiency in the temporal cortex correlates with sCJD neuropathology
Next, we sought to investigate whether the MRC deficiency correlated with neuropathology markers of sCJD  (Fig. 5). To this end, we chose complex I deficiency as a representative marker of MRC deficiency. Individual scores of neuropathology markers and means with 95% CI are listed in Table 3. As expected, vacuolation and PrP sc -staining correlated significantly with gliosis in all examined areas (Fig. 5a). Intriguingly, in the temporal cortex, we found a significant positive correlation between the amount of complex I deficient neurons and vacuolation (P < 0.001, r s = 0.725), gliosis (P < 0.001, r s = 0.771) and PrP sc stain (P = 0.018, r s = 0.537; Fig. 5b). In contrast, the hippocampal regions CA4/CA3 showed no correlation with any of the neuropathology markers (vacuolation; P = 0.699, gliosis; P = 0.605, PrP sc stain; P = 0.109; Fig. 5c). Disease duration did not correlate with the percentage of complex I deficient neurons in either temporal cortex (P = 0.227) or the CA4/CA3 hippocampal regions (P = 0.951). Linear regression with vacuolation as covariate confirmed that molecular subtype of sCJD is a predictor of MRC deficiency (Beta = 0.40, P = 0.019), however, this difference could be partly confounded by severity of vacuolation (Beta = 0.57, P = 0.002).

Discussion
We show that neuronal MRC deficiency affecting all respiratory complexes occurs in the brain of patients with sCJD. This deficiency shows a strong association with the severity of pathological markers of disease and has a predilection for the MM1 molecular disease subtype. In spite of the pronounced MRC loss, the outer mitochondrial  [11].
Neuronal MRC deficiency in sCJD shows a strong correlation with the severity of the neurodegenerative changes. It has a clear regional predilection for the severely affected temporal cortex, whereas the relatively spared CA4/CA3 hippocampal regions show only a mild decrease in MRC staining. Moreover, the extent of MRC deficiency in the temporal cortex shows a positive correlation with the severity of neuropathological changes including gliosis, vacuolation and PrP sc load. We believe that the lack of association between MRC deficiency and severity of neuropathology in the CA4/CA3 region may reflect insufficient statistical power due to the substantially milder MRC involvement in that area. The severity of MRC deficiency is associated with the molecular subtype of sCJD. MRC deficiency is significantly more pronounced in neurons of individuals with MM1 type sCJD compared to VV2 type sCJD. In fact, with the exception of complex V, MRC deficiency in VV2 type sCJD was not statistically significant compared to controls, although a trend was seen for higher numbers of MRC-deficient neurons. Regression analysis shows that the difference between molecular subtype of sCJD is partly confounded by the more severe pathological changes in MM1 molecular phenotype. Spongiform changes are more pronounced in the neocortex of MM1 type sCJD compared to VV2, where the basal ganglia and thalamus are predominantly affected [19]. However, as disease subtype remains a significant predictor of MRC deficiency in our model, we cannot exclude the possibility of additional factors rendering MM1 neurons more susceptible to MRC impairment. Our findings suggest that MRC deficiency is an integral part of sCJD pathology and plays an important role in the pathogenesis of the disorder. While MRC deficiency correlates strongly with the severity of disease specific pathological markers, the fact that it is also present in morphologically normal neurons of the mildly affected CA4/CA3 region suggests it is not a terminal phenomenon in dying neuronal populations. The mechanisms underlying MRC loss in sCJD remain unknown. Since mitochondrial mass is unchanged, it is possible MRC is depleted via an active downregulation or increased degradation of the peptide subunits of the respiratory complexes. mtDNA encodes subunits of complexes I, III, IV and V and quantitative and/or qualitative mtDNA defects are a common cause of MRC deficiency [20]. However, mtDNA defects should not affect complex II, which is entirely encoded in the nucleus, and very rarely cause a uniform loss of the remaining respiratory complexes. Therefore, the MRC deficiency of sCJD does not fit the profile of pure mtDNA defects.
The observed pan-respiratory complex deficiency without a change in the mitochondrial outer membrane marker VDAC1 may reflect damage to the inner mitochondrial membrane. This is in line with previous  reports of abnormal mitochondrial cristae architecture in prion infected rodents [8,21,22]. Interestingly, while PrP c has been shown to localize mostly on the plasma membrane, it has been suggested that it may also be present in the inner mitochondrial membrane of healthy mice [23]. Furthermore, dysmorphic mitochondria with abnormal cristae morphology have been observed in PRNP knockout mice [24]. It is therefore possible that PrP sc aggregation injures the inner mitochondrial membrane, both directly and via loss of its physiological counterpart, PrP c . Irrespective of the cause, deficiency of the entire MRC is expected to compromise neuronal metabolism causing ATP depletion and a shift toward increased glycolysis [25]. Neurons are highly dependent on oxidative phosphorylation and use most of their ATP to maintain their membrane potential. This is achieved via the action of the sodium-potassium ATPase, which maintains the intraand extracellular concentrations of sodium and potassium ions against their electrochemical gradients and, by extension, regulates intracellular water balance [26]. Pump failure due to ATP deficiency causes a shift of water from the extracellular to the intracellular compartment (i.e. cytotoxic edema), leading to disruption of essential cellular processes and ultimately cell death. In fact, diffusion weighted imaging (DWI) of the brain typically shows evidence of restricted water diffusion in affected areas of the sCJD brain [27], suggesting that cytotoxic edema is indeed an important mechanism underlying neuronal death in sCJD. Based on our findings, we propose that cytotoxic changes in the sCJD brain are caused by ATP depletion due to neuronal MRC deficiency. Similar signs of restricted water diffusion on MRI are seen in other disorders involving neuronal energy failure such as mitochondrial disease [28][29][30], hypoxia/ischemia [31], hypoglycaemia [32,33] and carbon monoxide intoxication [34].
Another consequence of neuronal MRC impairment in sCJD would be a shift toward glycolytic metabolism, resulting in higher production of lactate. Elevated concentration of cerebrospinal fluid (CSF) lactate has indeed been reported in patients with CJD, corroborating this hypothesis [35]. In addition to decreased ATP production, a glycolytic shift could render neurons more susceptible to oxidative damage. Normally, neurons can consume glucose through the pentose phosphate pathway, which helps to regenerate reduced glutathione, an important component against oxidative stress. Upregulation of glycolysis may therefore cause increased levels of oxidized glutathione, in turn, increased susceptibility to the formation of reactive oxygen species in neurons [36]. Oxidative damage has indeed been implicated in cell models of prion disease [37][38][39].
While cause and effect cannot be confidently discerned by this type of study, our findings strongly suggest that MRC loss in sCJD is deleterious and actively involved in the neurodegenerative process. Interestingly, the pattern of MRC deficiency in sCJD is distinct from that of other neurodegenerative proteinopathies. In Parkinson's disease, MRC deficiency is selective for complex I (and to a much lesser degree complex IV) and does not correlate with the severity of neurodegeneration. In fact, unlike sCJD where the load of PrP sc correlates with MRC deficiency, in PD there is an inverse relationship between complex I loss and Lewy pathology leading to the hypothesis that it may be a partly protective event [17,40]. It is conceivable that downregulation of complex I levels without a decrease in the remaining respiratory complexes, as observed in PD, serves to limit reactive oxygen species (ROS) production and oxidative damage without a major compromise in energy transduction. In contrast, the global MRC loss seen in sCJD is highly likely to cause ATP depletion.
In conclusion, we show that mitochondrial dysfunction is an important mechanism underlying neuronal injury and neurodegeneration in sCJD. These findings provide an explanation for the clinical observations of restricted MRI diffusion and elevated CSF lactate in patients. Our findings suggest that mitochondria should be the focus of further study in sCJD and should be assessed as potential therapeutic targets for this incurable and devastating disorder. the work presented here was approved by the Regional Committee for Medical and Health Research Ethics, Western Norway (REK 2018/1666). The manuscript does not contain information or images that could lead to an identification of the individual or which could violate any personal rights.

Consent for publication
Not applicable.