Understanding the molecular basis of resilience to Alzheimer’s disease

Kathleen S. Montine, Eloïse Berson, Thanaphong Phongpreecha, Zhi Huang, Nima Aghaeepour, James Zou, Michael J. MacCoss, Thomas J. MontineFrontiers in Neuroscience 2023 · DOI

TL;DR

This paper explores why some individuals remain cognitively sharp despite having brains riddled with Alzheimer’s pathology—a phenomenon known as "resilience." By applying machine learning to clinical data and deep learning deconvolution to proteomic and epigenetic brain profiles, the researchers identified that resilience is a distinct biological process centered in the hippocampus, driven by synaptic integrity and cellular detoxification rather than just "avoiding" damage.

What Problem Does This Paper Address?

The traditional "direct model" of neurodegeneration is simple: stressors cause damage, damage leads to brain degeneration (like plaques and tangles), and degeneration results in clinical dysfunction (dementia). However, this model fails for 15–25% of the population. Some people have significant brain pathology but no symptoms, while others have minimal pathology but severe cognitive decline.

The authors argue that we need two "impedance terms" to explain this gap:

  1. Resistance: The ability to suppress the formation of damage/lesions despite stressors.
  2. Resilience: The ability to maintain cognitive function despite the presence of significant degeneration.

The challenge is that "resilience" has historically been treated as a binary (you have it or you don't) or conflated with educational attainment. This paper seeks to quantify resilience as a continuous variable and uncover its specific molecular and epigenetic signatures.

Methods

The researchers utilized a multi-pronged computational and molecular approach:

  • Quantifying Resilience: They developed a linear model where a Cognitive Resilience (CR) Score is derived by subtracting a machine-learning-estimated "Damage Estimate" (based on 17 different neuropathologic lesions) from actual cognitive test scores.
  • Predictive Modeling: They used Random Forest models to see if intra vitam (during life) features like education, cardiovascular health, and depression scores could predict these post-mortem CR scores.
  • Proteomics: They performed unbiased, data-independent peptide-based proteomics on four brain regions to identify differentially expressed proteins (DEPs) and protein co-expression modules.
  • Epigenetics & "Cellformer": To overcome the limitations of bulk ATAC-seq (which averages signals across all cell types), they used a novel deep learning algorithm called Cellformer. This allowed them to "deconvolve" bulk data into cell-type-specific open chromatin regions (OCRs) for neurons, microglia, and other glia.

Datasets

  • Primary Cohort: 43 highly selected, age- and sex-matched brain donations categorized into Healthy Control (HC), Alzheimer’s Disease Dementia (ADD), and Resilience to AD (RAD).
  • Validation Cohorts: Findings were validated using data from four independent external cohorts, totaling 689 human isocortical samples.
  • Clinical Data: Data included longitudinal neuropsychological tests and the Geriatric Depression Scale (GDS).

Key Insights and Ideas

  • Resilience is Independent of Damage: Surprisingly, the APOE ε4 allele—the strongest genetic risk factor for AD—correlates strongly with increased damage but does not correlate with lower resilience. This suggests that the pathways leading to plaque formation and the pathways allowing a brain to function despite those plaques are biologically distinct.
  • Health over Schooling: While "educational attainment" is often used as a proxy for resilience, this study found that cardiovascular health and mental health (low depression scores) were much stronger predictors of a high CR score.
  • The Hippocampus is the Hub: Epigenetic signatures of resilience were overwhelmingly (93%) localized to the hippocampus, specifically involving the regulation of excitatory and inhibitory neurons.
  • Molecular "Counterforces": Resilient brains showed lower levels of soluble amyloid-beta (Aβ) and enrichment in proteins related to actin filament organization (synaptic structure), cellular detoxification, and wound healing.

Key Results

  • Protein Signatures: The study identified 33 proteins specifically associated with resilience. A specific co-expression module (Module 5, containing 181 proteins) was able to distinguish resilient brains from demented brains across multiple regions.
  • Top Predictors: A decision tree classifier identified Aβ concentration in the hippocampus and PA1B3 concentration in the inferior parietal lobule as the top two features for distinguishing between the three clinical groups.
  • Chromatin Accessibility: Cellformer revealed that resilience is characterized by epigenetic upregulation that supports hippocampal neuronal processes and synaptic transmission, potentially preserving the number of neuronal projections even as pathology accumulates.

Limitations and Future Work

The authors acknowledge that their primary discovery cohort was relatively small (N=43), though they mitigated this through rigorous selection and external validation.

Future research needs to:

  1. Expand Diversity: Move beyond highly selected cohorts to include more diverse populations.
  2. Explore ADRDs: Apply this framework to Alzheimer’s Disease Related Dementias (ADRDs), such as Lewy body disease or vascular dementia, to see if resilience mechanisms are universal or disease-specific.
  3. Temporal Dynamics: Determine when these "reserve" and "compensation" mechanisms are built during a lifespan—is it early-life education, mid-life health, or late-life social engagement that matters most?

Summary updated

Abstract

The cellular and molecular distinction between brain aging and neurodegenerative disease begins to blur in the oldest old. Approximately 15-25% of observations in humans do not fit predicted clinical manifestations, likely the result of suppressed damage despite usually adequate stressors and of resilience, the suppression of neurological dysfunction despite usually adequate degeneration. Factors during life may predict the clinico-pathologic state of resilience: cardiovascular health and mental health, more so than educational attainment, are predictive of a continuous measure of resilience to Alzheimer's disease (AD) and AD-related dementias (ADRDs). In resilience to AD alone (RAD), core features include synaptic and axonal processes, especially in the hippocampus. Future focus on larger and more diverse cohorts and additional regions offer emerging opportunities to understand this counterforce to neurodegeneration. The focus of this review is the molecular basis of resilience to AD.
BibTeX
@inproceedings{montine2023understanding,
  author = {Kathleen S. Montine and Eloïse Berson and Thanaphong Phongpreecha and Zhi Huang and Nima Aghaeepour and James Zou and Michael J. MacCoss and Thomas J. Montine},
  title = {{Understanding the molecular basis of resilience to Alzheimer’s disease}},
  booktitle = {Frontiers in Neuroscience},
  year = {2023},
  doi = {10.3389/fnins.2023.1311157},
}
Figure