Synaptic resilience is associated with maintained cognition during ageing

Abstract

Introduction: It remains unclear why age increases risk of Alzheimer's disease and why some people experience age-related cognitive decline in the absence of dementia. Here we test the hypothesis that resilience to molecular changes in synapses contribute to healthy cognitive ageing.

Methods: We examined post-mortem brain tissue from people in mid-life (n = 15), healthy ageing with either maintained cognition (n = 9) or lifetime cognitive decline (n = 8), and Alzheimer's disease (n = 13). Synapses were examined with high resolution imaging, proteomics, and RNA sequencing. Stem cell-derived neurons were challenged with Alzheimer's brain homogenate.

Results: Synaptic pathology increased, and expression of genes involved in synaptic signaling decreased between mid-life, healthy ageing and Alzheimer's. In contrast, brain tissue and neurons from people with maintained cognition during ageing exhibited decreases in synaptic signaling genes compared to people with cognitive decline.

Discussion: Efficient synaptic networks without pathological protein accumulation may contribute to maintained cognition during ageing.

Keywords: Alzheimer's; ageing; cognition; synapse.

FIGURE 1

High‐resolution array tomography (AT) reveals synaptic pathology in ageing and AD. (A) Synapses were quantified from AT image stacks and counted as a synaptic pair if the centroid of the presynaptic object (synaptophysin, magenta) was within 0.5 mm of the nearest postsynaptic object (PSD95, cyan). Arrows show examples of synaptic pairs. (B) The density of paired synapses decreases between mid life (ML), healthy ageing (HA), and Alzheimer's disease (AD) and a linear mixed effects model followed by ANOVA shows there was an effect of cohort F[2,36.79] = 7.02, p = 0.002. There were no differences in synapse density between males and females, or APOE4 carriers and non‐carriers. Pairwise post‐hoc comparisons showed synapses were significantly decreased in the AD cohort in comparison to ML and this was evident in both brain regions (BA20/21, t ratio = 3.02; p = 0.005, d = 68; BA17, t ratio = 3.04; p = 0.009, d = 59). Synapses were also decreased in HA in comparison to ML in BA17 (t ratio = 2.76; p = 0.02, d = 57). We examined the synaptic localization of Aβ (grey, C) and tau (yellow, D). In presynaptic terminals, there was a trend towards an increase in Aβ accumulation between midlife, healthy agers, and AD (E, F[2,34.95] = 3.30, p = 0.04) and a significant increase in presynaptic Aβ in regions containing plaques (F[2,164.15] = 15.39, p < 0.0001). Pairwise post‐hoc comparisons showed a significant increase in Aβ accumulation in AD BA17 (t ratio = 2.60; p = 0.03, d = 69). In post‐synaptic terminals, there was a significant increase in accumulation of Aβ between ML, HA, and AD (F, F[2,37.25] = 4.39, p = 0.01) and a significant increase in postsynaptic Aβ in regions containing plaques (F[2,157.72] = 12.23, p < 0.0001). Pairwise post‐hoc comparisons showed a significant increase in Aβ accumulation in AD BA17 (t ratio = 2.88; p = 0.01, d = 63). Presynaptic tau accumulation was significantly increased in AD (G, F[2,32.41] = 25.50, p < 0.0001) and significantly higher in BA20/21 than in BA17 (F[1,135.23] = 25.69, p < 0.001). There was also a significant interaction between cohort and brain region (F[2,127.08] = 5.75, p = 0.004). Pairwise post‐hoc comparisons showed a significant increase in presynaptic tau accumulation in AD in comparison to ML and HA cohorts and this was evident in both regions (BA20/21, ML v AD (t ratio = 5.67; p < 0.0001, d = 64); HA v AD (t ratio = 6.75; p < 0.0001, d = 52); BA17, ML v AD (t ratio = 3.95; p = 0.0006, d = 56); HA v AD (t ratio = 3.95; p = 0.0007, d = 50). Postsynaptic tau accumulation increases from ML to HA to AD groups (H, F[2,31.81] = 28.65, p < 0.0001) and is higher in BA20/21 than BA17 (F[1,135.56] = 31.78, p < 0.001). There is also an interaction between brain region and cohort (F[2,126.91] = 7.83, p = 0.0006). Pairwise post‐hoc comparisons showed a significant increase in postsynaptic tau accumulation in AD in comparison to ML and HA cohorts and this was evident in both regions (BA20/21, ML v AD (t ratio = 5.92; p < 0.0001, d = 63); HA v AD (t ratio = 7.40; p < 0.0001, d = 52); BA17, ML v AD (t ratio = 4.13; p = 0.0004, d = 55); HA v AD (t ratio = 4.15; p = 0.0004, d = 50). For box‐plots, each point represents case medians. Type III ANOVA with Satterthwaite correction were performed on the linear mixed effects models. Scale bar 1 μm for IMARIS reconstructions. * Represent p < 0.05 post‐hoc comparisons

FIGURE 2

Molecular changes in synapses in healthy ageing (HA) and AD indicate decreased synaptic function and increased inflammation compared to mid‐life (ML). (A) Comparing BA20/21 synaptic (synaptoneurosome) transcriptional changes between AD versus ML highlights many differentially expressed genes (9671 DEG's < FDR 0.05). (B) 5801 DEGs with FDR < 0.05 between HA and ML. (C) Fewer transcriptional changes are observed between HA and AD (293 DEG's < FDR 0.05). (D) Top 25 canonical pathways associated with total homogenate (TH) and synaptoneurosome (SN) brain preparations across both brain regions (BA17 and BA20/21) display similar profiles between HA or AD to ML cohorts. An inhibition or a decrease of canonical pathways (blue) associated with neurotransmission and memory is evident in both HA and AD in comparison to ML cohorts. Stress and immune response pathways appear to be activated (orange) or increased in both HA and AD in comparison to ML. (E) The top 25 canonical pathways associated with TH and SN brain preparations in BA20/21 between HA versus AD (BA17 showed no differences) show a decrease in only one biological pathway (blue) associated with stress response, whilst remaining canonical pathways associated with neurotransmission and memory were all increased (orange) in HA cohort. (F) 894 of the DEGs between AD and ML cohorts in BA20/21 synapses mapped to known synaptic proteins in the SynGO annotated database. 467 of these DEGs were associated with the post‐synapse and 401 with the pre‐synapse. (G) In HA versus ML BA20/21 synapses, 615 genes were mapped to SynGO of which 337 were associated with the post‐synapse and 285 with the pre‐synapse. (H) From the 293 transcripts identified in synapses of BA20/21 (HA vs. AD), an even number of both pre and post‐synaptic genes (n = 24) were identified using SynGo curated database, highlighting both synapse domains were adapting equaling in the healthy agers and/or not adapting in the AD brains. A‐C show Volcano plots of log2 fold change versus ‐log10 of the false discovery rate. Genes above solid grey line on volcano plots show FDR = 0.05 and dotted lines log2 fold change 1.2 (red) and ‐1.2 (green) respectively. Transcripts of interest are labelled in black

FIGURE 3

Maintained synaptic density and increased gliosis in people with lifetime cognitive decline. (A) Representative 3D reconstructions of AT stacks from Lifetime cognitive resilient (LCR) and Lifetime cognitive decline (LCD) cohorts. Serial sections of 70 nm sections from BA20/21 and BA17 were stained for synaptophysin (magenta), PSD95 (cyan), OC (grey) and Total tau (yellow). (B) Synapse density was lower in BA17 than BA2021 (F[1,133.40] = 4.05, p = 0.04); however there was no difference in excitatory synapse density between cognitive cohorts (F[1, 12.50] = 0.51, p = 0.48). There was similarly no difference between LCR and LCD in the percent pre or post synapses containing Aβ or tau, but co‐localization Aβ or tau with synapses showed regional increases in BA20/21 in comparison to BA17 (C F[1,110.53] = 4.84, p = 0.02; E F[1,89.94] = 6.38, p = 0.01; F F[1,88.17] = 4.67, p = 0.03). G Representative images of Aβ (BA4), microglia (CD68), astrocytes (GFAP), and P‐Tau (AT8) are shown from all five brain regions, BA20/21, BA17, BA24, BA46, and hippocampus across LCR and LCD groups. (H) Aβ burden measurements plotted across five brain regions show regional variation in Aβ burdens between both groups (F[4,60.00] = 3.80, p = 0.007) and an increase in APOE4 carriers (F[1,11.50] = 15.55, p = 0.002). (I) CD68 burden measurements show a significant increase in microglial burden in the LCD group (F[4,60.00] = 4.04, p = 0.005) which reaches post‐hoc pairwise significance in hippocampus (t ratio = 2.25; p = 0.02, d = 54). (J) GFAP burdens show no significant changes between groups or regions. (K) P‐Tau burdens plotted across five brain regions show regional variation in P‐Tau burdens between both groups (F[4,59.31] = 6.77, p = 0.0001) and an increase in APOE4 carriers (F[1,12.04] = 8.64, p = 0.01). For box‐plots, each point represents case medians. Type III ANOVA with Satterthwaite correction were performed on the linear mixed effects models. Scale bar 5 μm for IMARIS reconstructions, 150 μm for IHC images. * Represent p < 0.05 post‐hoc comparisons

FIGURE 4

People with lifetime cognitive resilience have dampened synaptic signaling pathways. (A) 363 DEG's (< FDR 0.05) were found when comparing transcription at the synapse in brain region BA20/21 between LCR and LCD cohorts. (B) Fewer DEG's (75, < FDR 0.05) were observed in the total homogenate fraction. (C) In BA17, 1116 DEG's < FDR 0.05 were identified between cognitive groups. (D) This was not reflected at a global level where only 112 DEG's were identified. (E) The top 25 canonical pathways changes indicate decreases in abundance (blue) of many pathways involved in synaptic function including neurotransmission and memory in people with lifetime cognitive resilience. Increased pathways (orange) were associated with stress and immune responses. (F) When looking at transcripts in the synapses of BA20/21, there are 363 DEGs of which 27 are known pre‐synaptic genes and 27 are known post‐synaptic genes in the SynGo curated database. The majority of these (23 of each) are downregulated. (G) SynGo analysis of the 75 DEGs in total homogenate of BA20/21 shows that people with better cognition had alterations in 10 synaptic genes, of which nine are pre‐synaptic and seven were post‐synaptic highlighting overlap between synaptic genes identified. (H) In BA17 synaptic fractions, there are 107 pre‐synaptic and 106 post‐synaptic genes changed. Of these, the vast majority (103 pre, 102 post) are downregulated. (I) In total four pre‐synaptic and three post‐synaptic specific genes were altered in total homogenate of BA17 between the cognitive groups. Volcano plot of log2 fold change versus ‐log10 of the false discovery rate. Genes above solid grey line on volcano plots show FDR = 0.05 and dotted lines log2 fold change 1.2 (red) and ‐1.2 (green) respectively. Transcripts of interest are labelled in black

FIGURE 5

iPSC‐derived neurons from LCR individuals have decreased expression of synaptic genes in response to Aβ challenge. (A) Overview of the process by which iPSCs from LCR and LCD individuals were differentiated to cortical neurons. Human AD brain homogenate enriched (Aβ+) or immunodepleted (Aβ‐) for soluble Aβ42 was added to cells at experiment time point. (B) Representative 3D reconstructions showing the effect of homogenate treatment on LCD and LCR neurons. (C) Automated quantification of homer1 puncta (post‐synaptic density) colocalized with MAP2 (dendrites) showed no significant difference between Aβ treatment or cell line group. (D‐E) RT‐qPCR revealed that SNAP‐25 and SYT1 expression significantly differed between Aβ treatments in the Lifetime Cognitive Resilient (LCR) group, but not in the Lifetime Cognitive Decline (LCD) group. (F) No difference between treatment or lifetime cognitive ageing groups was observed in TM4SF1 expression. For box‐plots, each point represents case medians. Type III ANOVA with Satterthwaite correction were performed on the linear mixed effects models. Scale bar 100 μm (A), 5 μm for IMARIS reconstructions (B). * Represent p < 0.05 post‐hoc comparisons