Abstract Body

Drivers of Disease Progression

Alzheimer’s disease is characterized by loss of neural system function in the setting of neuropathological lesions including tangles, plaques, neuroinflammation, and neuronal loss. As the disease process evolves over decades, neuronal dysfunction and eventual loss occurs in cytoarchitectural areas vulnerable for neurofibrillary tangles, although the amount of accumulate neuronal loss exceeds the number of tangles by ten fold. To examine the relationship between tangle formation and neuronal loss, we performed multiphoton microscopy longitudinal imaging of mice that develop tangles (Tg4510 and Thy22) using methods that allowed imaging of neuronal nuclei as well as tangles. Weekly imaging sessions revealed rare, but detectable loss of neuronal nuclei over time in both models. In the tg4510 mice, whose transgene can be downregulated by administration of doxycycline, we further showed that suppression of the tau transgene led to amelioration of continued cell death, despite continued presence of tangles. In fact, in both models, we surprisingly found that neurons that harbored tangles were several fold less likely to die than non-tangle bearing neurons. These data do not support the hypothesis that tau aggregation into a frank tangle is a critical mediator of neuronal death, and suggest instead that soluble, nonaggregated presumably misfolded tau is a driver of risk for neuronal death.

Aims

A significant proportion of individuals preserve cognitive function despite meeting neuropathological criteria for Alzheimer’s disease (AD) at autopsy, known as cognitive resilience. We aimed to define the molecular and cellular signatures of cognitive resilience against AD.

Methods

We integrated multi-modal data from the Religious Order Study and Memory and Aging Project (ROSMAP), including bulk (n=631) and multi-regional single nucleus (n=48) RNA sequencing. Subjects were categorized into AD, Resilient, and Control based on Aβ and tau pathologies, and cognitive status. We investigated genetic risk and transcriptomic changes in resilience and prioritized protected cell populations using genetic enrichment and cellular distribution estimation. We further characterized these populations using multiplex immunofluorescence.

Results

Resilient individuals exhibited an intermediate genetic risk profile between AD and Control. Remarkably, only GFAP (encoding glial fibrillary acidic protein) was differentially expressed between Resilient and Control in bulk tissue. It was upregulated in Resilient astrocytes in the dorsolateral prefrontal cortex compared to AD and Control. Inhibitory and excitatory neurons displayed distinct brain region-specific phenotypes in cognitive resilience. Somatostatin-positive interneurons were enriched for genes linked to protective rare genetic variants and showed vulnerability in AD but not in resilience. Specific excitatory neuronal populations in the entorhinal cortex exhibited resilience-like behavior and expressed genes previously associated with resilience against AD.

Conclusions

Our findings suggest that cognitive resilience does not exhibit marked bulk transcriptional differences from healthy aging, despite significant Aβ and tau neuropathology. Specialized excitatory neuronal populations may drive cognitive resilience, while a specific subset of interneurons may play a crucial role in other forms of AD protection. This study provides valuable insights into the molecular basis of cognitive resilience, offering the potential to convert natural protective systems into effective therapeutic interventions for AD.

Aims

Two main risk factors of Alzheimer’s disease (AD) are aging and APOE-ε4. However, some individuals are “cognitively resilient” because they are spared from developing cognitive impairment despite carrying these risk factors. This study aimed to identify molecular factors and pathways that confer cognitive resilience in APOE-ε4 carriers ≥ 80 years of age.

Methods

We applied weighted gene co-expression network analysis (WGCNA) to generate consensus co-expression networks using blood RNA-sequencing data from individuals from the Mayo Clinic Study of Aging (MCSA) (n=105), and blood microarray transcriptomic data from individuals from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (n=91). We associated the network expression with cognitive endophenotypes and hippocampal volume. We also identified blood consensus networks preserved in the brain transcriptome of AD patients and controls (n=883). Finally, we validated the findings in a mouse model of AD.

Results

We found six replicable consensus networks that were significantly correlated with a memory phenotype (logical memory delayed recall=LMDR) in both MCSA and ADNI. Among these networks, blood expression module M3 was most preserved with the brain transcriptome. M3 was enriched with NDUF hub genes that are involved in the mitochondrial respiratory chain. Expression levels of many blood NDUFs were also significantly associated with better memory scores. In the brain, NDUFs were upregulated in controls compared to AD, and their expression levels were associated with better global cognition and decreased AD neuropathology. Lastly, many NDUFs were significantly downregulated in the hippocampus of AD mice compared to controls.

Conclusions

Increased blood NDUFs expression may contribute to cognitive resilience despite the presence of AD risk factors. Upregulation of NDUFs in the brain may be protective against neurodegeneration. These results suggest that mitochondrial molecules NDUFs may represent therapeutic targets for AD.

Aims

Cardiometabolic diseases (CMDs) including type 2 diabetes, heart disease, and stroke have been linked to increased dementia risk. We examined whether high levels of cognitive (CR) can counteract the risk of dementia related to CMDs and explored the potential underlying mechanisms.

Methods

Within the UK Biobank, 216,185 dementia-free participants aged ≥60 years were followed for up to 15 years. CMDs and dementia were identified through linkage to medical records. Latent class analysis was used to generate an indicator of cognitive reserve (CR; low, moderate, and high) based on education, occupation, confiding in others, social connection, leisure activities, and television watching time. A subsample (n=13,664) underwent brain MRI scans 9 years after baseline. Volumes of total brain (TBV), white matter (WMV), gray matter (GMV), hippocampus (HV), and white matter hyperintensities (WMHV) were measured. Data were analyzed using Cox regression and linear regression.

Results

At baseline, 43,403 (20.1%) participants had at least one CMD. There were 6,600 incident dementia cases over the follow-up (median 11.7 years). In multi-adjusted Cox models, the hazard ratio (HR, 95% confidence interval) of dementia was 1.57 (1.48, 1.67) for participants with CMDs. In joint effect analysis, the HR of dementia was 2.13 (1.97, 2.30) for those with CMDs and low CR and 1.78 (1.66, 1.91) for those with CMDs and moderate to high CR (reference: CMD-free, moderate to high CR). Additionally, among people with CMDs, those with moderate to high CR had significantly higher TBV, GMV, WMV, and HV (p<0.001 for all) than those with low CR.

Conclusions

Moderate to high CR appears to mitigate the risk of dementia associated with CMDs by about one third. Lower levels of neurodegenerative and vascular pathologies might underlie this mitigating effect.

Aims

Memory encoding of temporal information represents a crucial cognitive function that deteriorates with age. Reservoir computing is a machine learning framework that uses complex networks to process time-series data, therefore offering unique avenues to evaluate temporal memory capacity. Here, we employ reservoir computing to assess the memory capacity of the human anatomical network.

Methods

We evaluated the network memory capacity in a lifespan cohort of healthy individuals (18-88 years old). We obtained individual anatomical brain connectomes using diffusion-weighted imaging and, modelling them as reservoirs, we examined their proficiency in replicating random functional signals. By analyzing memory capacity across different connection densities, we pinpointed the optimal density to predict an individual’s age using a deep neural network. We replicated our findings in an independent cohort of young and old individuals.

Results

Memory capacity declined with aging. It was an excellent predictor of individual age, reaching a 0.78 correlation between predicted and true age. In our replication cohort, it showed a high discriminative accuracy (AUC: 0.908 ± 0.054), being able to correctly classify 80.3% of the old and 92.7% of the young individuals. Memory capacity was associated with white-matter and locus coeruleus integrity, resting-state functional activation, as well as cognitive performance measured in clinical settings. Finally, we found that memory capacity is linked to genes involved in signal transduction processes in inhibitory neurons.

Conclusions

Network memory capacity emerges as a robust marker of aging, highlighting an important correlation between memory capacity and individual cognitive performance. By bridging the computational memory capacity of the brain with empirical cognitive assessments, our insights pave the way for using reservoir computing to investigate age-associated neurobiological changes and concomitant neurodegenerative disorders.

Aims

Cognitive reserve (CR) is considered a protective factor for cognitive function and may explain interindividual differences of cognitive performance given similar levels of neurodegeneration, e.g., in Alzheimer´s disease. Recent evidence suggests that CR is also relevant in Parkinson's disease (PD). Enhancing our grasp of how CR influences cognitive function can inform the creation of prevention strategies and targeted interventions to combat cognitive decline. We aimed to explore the role of life-stage specific CR for overall cognition and specific cognitive domains cross-sectionally and longitudinally in PD.

Methods

The cross-sectional analysis with data from the DEMPARK/LANDSCAPE study included 81 individuals without cognitive impairment (PD-N) and 102 cognitively impaired individuals (‘PD-MCI&PDD’). Longitudinal data covered 4 years with over 500 observations. CR was measured using the Lifetime of Experiences Questionnaire (LEQ), capturing complex lifestyle activities across distinct life-stages. Cognition was assessed using an elaborate neuropsychological test battery including the CERAD-Plus.

Results

Higher LEQ scores, particularly derived from mid- and late-life, were observed in PD-N compared to ‘PD-MCI&PDD’ [F(1,163)= 6.902, p = .009, η_p² = 0.041]. They were significantly associated with better performance in all cognitive domains (0.202 ≤ ß ≤ 0.266). Longitudinally, linear mixed effect models (0.755 ≤ conditional R² ≤ 0.886) revealed that LEQ scores were positively related to cognitive performance independent of time. However, the decline in overall cognition and memory over time was more pronounced with higher LEQ scores.

Conclusions

This study emphasizes the impact of complex lifestyle activities, especially in mid- and late-life, on cognition in PD. Data indicate that while CR may delay cognitive decline, individuals with high CR may experience a more pronounced drop in overall cognition and memory. Future studies will have to replicate these findings, particularly regarding domain-specific effects.

Aims

The prevalence of Alzheimer’s Disease (AD) increases with age, yet a small fraction of the population reaches ages beyond 100 years maintaining cognitive abilities, so-called cognitively healthy centenarians. We aimed to uncover the genetic factors associated with such protection against AD, and to highlight the effects of AD-associated variants during advanced aging.

Methods

Previous Genome-Wide Association Studies (GWAS) identified 86 single nucleotide polymorphisms (SNPs) associated with AD-risk. We investigated the frequency of each SNP in 346 cognitively healthy centenarians (mean age 101.05±2.51), compared to 2,281 AD cases (mean age 67.96±9.84), and 3,165 controls (mean age 62.57±8.66). We combined the SNPs into Polygenic Risk Scores (PRS) for each individual, and compared the PRS between groups. Finally, we characterized the functional properties of the SNPs enriched/depleted the most in the centenarians.

Results

At the SNP level, centenarians were depleted with risk-increasing SNPs and enriched with protective SNPs. The protective effect concentrated on the alleles located in/near TMEM106B, SORT1, GRN, EPDR1, PLCG2, RIN3, MS4A, TREM2, ABCA7 and APOE genes. This translated to a significantly lower PRS in centenarians (excluding APOE SNPs), not only compared to AD cases (OR=1.96, p=4.07x10^-26), but also to middle-aged controls (OR=1.29, p=2.61x10^-5). When including APOE SNPs these effects were even stronger (OR=5.07, p=6.15x10^-71, and OR=1.87, p=4.67x10^-17, for comparisons with AD and normal controls, respectively).

Conclusions

The ability to reach advanced ages without cognitive decline is influenced by common AD-associated genetic variants. An efficient maintenance of the brain’s endolysosomal and immune systems confers the strongest protection against AD and potentially other age-related diseases.

Abstract Body

Identification of regulators of microglia lipotoxicity

Authors: Martine Therrien, Nicolas Weider, Giulia Monti, Dan Meyer, Eugenio Mattei, Trevor Atkeson, Juan Lorenzo Pablo, Sulagna Ghosh, Steven McCarroll, Anna Greka, Beth Stevens

Emerging genetic studies of late-onset Alzheimer’s Disease implicate the brain’s resident macrophages in the pathogenesis of AD yet we know little about the underlying biology or how myeloid cells contribute to AD pathogenesis. Diverse microglial states are found in human and mouse brains, however, we lack models to track these states and understand the impact of environmental challenges or of genetic susceptibility. Using the iPSC-derived microglia platform our lab has recently developed (Dolan*, Therrien* et al Nature Immunology 2023), we characterized the impact of 48 different fatty acids on human microglia. We identified 5 distinct transcriptional signatures produced in response to various lipids, including a toxic lipotoxicity signature independent of TREM2 signaling. Together our data identifies genetic regulators and functional changes of microglia states. Our data identified key environmental elements altering microglia and how AD risk genes affect states and functions providing important insight on understanding how to protect synapses in Alzheimer’s and other neurodegenerative diseases that could lead to new therapies and biomarkers for early diagnoses.