Region-specific depletion of synaptic mitochondria in the brains of patients with Alzheimer’s disease

Tissue from eight clinically and pathologically diagnosed Alzheimer’s disease donors and nine control donors were used for this study (details in Online Resource 1). Use of human tissue for post-mortem studies has been reviewed and approved by the Edinburgh Brain Bank ethics committee and the ACCORD medical research ethics committee (approval HV-15-016; ACCORD is the Academic and Clinical Central Office for Research and Development, a joint office of the University of Edinburgh and NHS Lothian). The Edinburgh Brain Bank is a Medical Research Council funded facility with research ethics committee (REC) approval 16/ES/0084.

At post-mortem, the brain was removed and cut into coronal slices. Regions of interest were then dissected from each coronal slice. Samples from one hemisphere were dissected into smaller segments and processed for electron microscopy as described previously [29]. Fresh post-mortem samples from BA46 (dorsolateral prefrontal cortex) and BA41/42 (anterior/posterior transverse temporal cortex) were trimmed into small cortical blocks containing the six cortical layers and fixed in 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer (PB) for 48 h. Fixed tissue blocks were washed twice in 0.1 M PB and were exposed to osmium tetroxide (1% in 0.1 M PB) for 30 min (protected from light). Samples were washed twice for 15 min in 0.1 M PB and three times in previously boiled ddH₂O. Tissue blocks were then dehydrated for 15 min in 50% ethanol followed by exposure to 1% uranyl acetate in 70% ethanol for 40 min in the dark. Samples were further dehydrated in an ascending series of ethanol (95%, 100%, 100%) and propylene oxide before storage in Durcupan resin overnight at room temperature. Samples were embedded in Durcupan resin and allowed to polymerise for 48 h at 60 °C.

Resin-embedded tissue blocks were cut into 70-nm-thick sections using an ultracut microtome (Leica) equipped with a Jumbo Histo Diamond Knife (Diatome, Hatfield, PA, USA) and collected onto copper formvar-coated grids. Grids were stained with lead citrate in a CO₂-free environment for 2 min before imaging on a JEOL JEM-1400 Plus transmission electron microscope (TEM). For synapse analysis, an average of 50 images per case was taken at 6000× magnification in a systematic, random fashion from BA46 and BA41/42 (Fig. 1). Exclusion criteria for sampling images included the presence of a nucleus in the entire field of view, no synapses present in the field of view or the presence of an Aβ plaque or NFT in the entire field of view. TEM images were coded for blind analysis. Synapses were defined by a presynaptic terminal containing at least three synaptic vesicles adjacent to an electron-dense post-synaptic density. Mitochondria were defined by the presence of internal cristae and a defined outer membrane. 100 synapses per case for each brain region (where available, please see Online Resource 1) were analysed for a total of 3000 synapses analysed for the presence of mitochondria (individual or multiple) in the pre- or post-synapse and any unusual features in the synapse noted (abnormal mitochondrial morphology, presence of multivesicular bodies, or presence of fibrils). The length of the post-synaptic density opposing the presynaptic terminal was also measured (apposition length). Extra blocks from temporal, occipital, and frontal cortex were examined for plaques in three AD cases and all plaque-associated synapses pooled for apposition length measurements (80 synapses near plaques analysed). Single sections were imaged as opposed to reconstructing three-dimensional reconstructions of serial sections to allow imaging of large numbers of synapses. We have previously demonstrated that 2D EM quantification techniques performed on synapses, such as those used in our present study, generate near identical findings to parallel 3D quantification studies [19].

Images were analysed by an experimenter blind to diagnosis. For each variable, a single value was calculated per region per case. Statistics were calculated in SPSS and GraphPad Prism. Normality of the data was tested with a Shapiro–Wilk test. Normally, distributed data (apposition length, and percentages of pre- and post-synapses containing a single mitochondrion) were analysed by two-way ANOVA and Tukey’s post hoc tests and are shown as mean and standard errors. Non-normally distributed data (percentages of pre- and post-synapses containing more than one mitochondria and percentage of post-synapses containing multivesicular bodies) were analysed with non-parametric Kruskal–Wallis tests and are shown as median and interquartile ranges.

To test the hypothesis that mitochondrial localisation in synapses is affected in AD, synapses were analysed from AD and control brain samples using transmission electron microscopy (Fig. 1). Raw EM images are freely available from the University of Edinburgh Data Repository DataShare at https://​doi.​org/​10.​7488/​ds/​2417.

We systematically sampled cortex to analyse synaptic mitochondria in asymmetric synapses in two brain regions, superior temporal gyrus (BA41/42) and dorsolateral prefrontal cortex (BA46). These brain regions play an important role in memory encoding and recognition (BA46) and auditory working memory (BA41/42), which are disrupted in AD [2, 58], and by the end stages of disease, both of these regions have substantial plaque and tangle pathology [54]. We observed mitochondria in a subset of pre- and post-synapses in both regions of AD and control brains (Fig. 2). Two- to threefold more presynaptic terminals contained mitochondria (ranging from 14 to 37% in AD cases and 22–45% in controls) than post-synaptic terminals (ranging from 8 to 23% in AD cases and 5–17% in controls). In both AD and control cases, the higher percentage in pre-synapses than in post-synapses was significant (two-way ANOVA effect of pre vs. post in AD F(1, 26) = 68.84, p < 0.001; in control F(1, 26) = 105.3, p < 0.001). Interestingly, there was a trend towards a region effect in the difference between pre- and post-synaptic mitochondria only in AD (two-way ANOVA effect of region in AD F(1, 26) = 7.91, p = 0.093; control F(1, 26) = 0.57, p = 0.46). The presence of mitochondria in presynaptic terminals was lower in BA41/42 than in BA46 (Fig. 3a), whereas the percentage of post-synaptic profiles containing mitochondria did not differ by region or with disease (Fig. 3b). A small fraction (0–8%) of synaptic terminals contained more than one mitochondrial profile. In BA41/42, AD cases had over fourfold fewer presynaptic terminals with multiple mitochondrial profiles than control synapses from the same region (Figs. 3c, 4a). The percentage of pre-synapses containing multiple mitochondria was not different between AD and control in BA46, indicating that transport of mitochondria to presynaptic terminals may be impaired in a region-specific manner.

To ensure the significant decrease in presynaptic terminals containing multiple mitochondria was not an artefact of post-mortem degradation, we ran correlation analyses to confirm that there is not an association of PMI or brain pH with this outcome measure. None of the measures of mitochondria in synapses correlated with pH or PMI in either AD or control groups. When considering all cases (AD and controls) together, there is a strong correlation (r = − 0.76, p < 0.001) with disease status and % presynaptic terminals containing multiple mitochondria in BA41/42 as would be expected from our significant difference found with ANOVA.

Multiple mitochondrial profiles in a single presynaptic bouton in the single-section images that we analysed for the 3000 synapses in the study could correspond to multiple mitochondria or to different parts of the same mitochondrion that has curved out of the plane of view. Three-dimensional reconstructions of two pre-synapses with multiple mitochondria show separate mitochondria in presynaptic terminals (Online Resource 2, Online Resource 3).

Along with the localisation of mitochondria to synapses, we examined the ultrastructure of synapses and synaptic mitochondria. We noted occasional abnormal mitochondrial morphology in AD cases (Fig. 4b) which may indicate mitochondrial dysfunction. Further, we observed accumulation of multivesicular bodies in post-synaptic terminals in BA41/42 of AD cases which was rare but exhibited a trend towards significance (Fig. 4c, d). Apposition length was not changed by region or diagnosis (Fig. 4e). Previous data using array tomography indicated synapse shrinkage near plaques [31]. We did not find enough plaques in the BA41/42 and BA46 samples examined in the systematic random fashion for the main study to compare synapses near and far from plaques. However, we screened blocks for plaques from temporal, frontal, and occipital cortices from three of the AD cases to find more plaques to allow investigation of apposition length of synapses near plaques. A total of 80 synapses near plaques and 219 synapses far from plaques in the same blocks were observed across the three AD cases examined. The average apposition length for each case near and far from plaques was calculated and found to be 23% smaller near plaques than far from plaques (Fig. 4f, g).

It is also worth noting that of the 3000 synapses systematically sampled in BA41/42 and BA46 and in the 80 synapses near plaques from all regions, we did not find any pre- or post-synaptic terminals containing fibrils, despite the clear appearance of plaque and tangle fibrils in AD cases (Fig. 5). In our previous studies, we detect a substantial proportion of synapses near plaques that are positive for oligomeric Aβ using array tomography [31]. The current ultrastructural data indicate that these oligomers are not yet fibrillar.