Multi-omics characterization of vascular, neurodegenerative, and mixed neuropathology in the aging human brain

This study utilizes multi-omics profiling of postmortem brain tissue to define conserved molecular programs that distinguish vascular, neurodegenerative, and mixed pathologies in late-life dementia, revealing that while individual gene and protein changes vary, pathway-level analyses—particularly involving immune-inflammatory upregulation and mitochondrial downregulation—provide a stable framework for understanding the molecular heterogeneity underlying cognitive impairment.

The Gist

Imagine the human brain as a bustling, ancient city. For decades, scientists have tried to understand why this city sometimes stops working properly, leading to memory loss and dementia.

For a long time, the focus was mostly on one specific type of disaster: Alzheimer's disease, which is like a slow, creeping fog of toxic waste (amyloid plaques and tau tangles) that clogs the streets and poisons the buildings.

However, this new study suggests that the reality is much more complex. In most aging brains, the city isn't just dealing with the toxic fog; it's also suffering from floods and road damage (vascular issues like blocked blood vessels and small strokes). Often, the city is hit by both at the same time.

Here is a simple breakdown of what the researchers found, using the "City of the Brain" analogy:

1. The Three Types of "City Disasters"

The researchers looked at the brains of nearly 1,000 people after they passed away. They sorted them into three groups based on what they found in the "city":

• The "Flood" Group (Vascular): The city has damaged roads and blocked water pipes, but the toxic fog is mostly absent.
• The "Fog" Group (Neurodegenerative): The city is covered in toxic waste, but the roads and pipes are relatively okay.
• The "Double Trouble" Group (Mixed): The city has both the toxic fog and the damaged roads. This turned out to be the most common scenario.

2. The Big Discovery: Two Different "City Management" Styles

The team didn't just look at the damage; they looked at the molecular "messages" the city's cells were sending. They compared the "Flood" group against the "Fog" and "Double Trouble" groups.

The "Fog" and "Double Trouble" groups spoke the same language:
When the brain has the toxic Alzheimer's fog (with or without the floods), the city's emergency response goes into overdrive.

• The Alarm Sirens (Immune System): The brain sounds the alarm constantly. It's like a city where the fire department and police are running around non-stop, trying to clean up the toxic waste. This creates a lot of inflammation.
• The Power Outage (Mitochondria): At the same time, the city's power plants (mitochondria) are shutting down. The cells are running out of energy. They can't keep the lights on or the traffic moving.

The "Flood" group was different:
Surprisingly, the brains with only vascular damage (the floods) didn't have the same power outage or the same level of immune panic. Their "power plants" were still running relatively well, and they weren't screaming for help as loudly.

3. The "Double Trouble" Group Has Extra Chaos

While the "Fog" and "Double Trouble" groups shared the same "Alarm + Power Outage" pattern, the "Double Trouble" group had some extra, unique problems:

• Construction Zone: Because of the vascular damage, the "Double Trouble" brains showed signs of heavy construction and repair work. The city was trying to rebuild its walls (extracellular matrix) and reroute traffic (vesicle transport) to cope with the damage. It was a messier, more chaotic version of the "Fog" group.

4. The "Recipe" vs. The "Ingredients"

One of the most fascinating parts of this study is how they found these patterns.

• Looking at single ingredients: If you look at just one specific protein or gene (like looking at a single brick in a wall), it's very hard to tell the difference between the groups. The bricks look almost the same.
• Looking at the recipe: But when you look at the whole recipe (the pathways and systems), the differences become crystal clear. The "Fog" and "Double Trouble" cities are following a recipe of "High Alarm, Low Power," while the "Flood" city is following a different recipe entirely.

The Takeaway:
This study tells us that we can't just treat dementia as one single disease.

• If a patient has vascular issues, their brain is struggling with blood flow but might still have energy.
• If a patient has neurodegenerative issues (even if they also have vascular issues), their brain is in a state of high inflammation and low energy.

Why does this matter?
It means that a "one-size-fits-all" drug might not work for everyone. A treatment that boosts energy might help the "Fog" group, but a treatment that calms the immune system might be needed for the "Double Trouble" group. By understanding these distinct "city management styles," doctors can eventually design better, more targeted therapies to keep our brain cities running smoothly for longer.

Technical Summary

1. Problem Statement

Late-life cognitive impairment is rarely caused by a single pathology; instead, it typically arises from a complex interplay of neurodegenerative processes (e.g., amyloid-β, tau, TDP-43) and cerebrovascular lesions (e.g., infarcts, arteriolosclerosis). While autopsy studies (such as the Religious Orders Study and Rush Memory and Aging Project, ROSMAP) have established that "mixed pathology" is the dominant biological state in the aging brain, the molecular programs distinguishing pure vascular, pure neurodegenerative, and mixed pathologies remain incompletely defined. Specifically, it is unknown whether these neuropathologically defined groups correspond to distinct, conserved molecular signatures across different brain regions and molecular layers (proteomics vs. transcriptomics).

2. Methodology

The study employed a neuropathology-stratified, cross-modal multi-omics approach using postmortem brain tissue from the ROSMAP cohort.

• Cohort & Stratification:
  • Participants were classified into three mutually exclusive groups based on comprehensive autopsy assessments: Neurodegenerative (ND) (high AD/TDP-43 burden, low vascular), Vascular (Vasc) (significant vascular lesions, no AD/TDP-43), and Mixed (significant burden of both).
  • Sample sizes: Proteomics (DLPFC, n=733); Transcriptomics (DLPFC n=938, Posterior Cingulate Cortex n=569, Anterior Caudate n=632).
• Data Generation:
  • Proteomics: Quantitative mass spectrometry (TMT-based) of the Dorsolateral Prefrontal Cortex (DLPFC), quantifying 10,030 proteins.
  • Transcriptomics: Bulk RNA-seq across DLPFC, Posterior Cingulate Cortex (PCC), and Anterior Caudate (AC).
  • Single-nucleus RNA-seq (snRNA-seq): Used to deconvolve cell-type-specific signals (specifically astrocytes) from bulk data.
• Analytical Pipeline:
  • Differential Expression: Linear regression models adjusted for age, sex, and technical covariates (batch, PMI, etc.) were used for pairwise contrasts (ND vs. Vasc, Mixed vs. Vasc, Mixed vs. ND).
  • Pathway Analysis: Gene Set Enrichment Analysis (GSEA) using Gene Ontology Biological Process (GO BP) annotations.
  • Network Analysis: Weighted Gene Co-expression Network Analysis (WGCNA) on proteomic data to identify modules associated with pathology and clinical traits.
  • Cross-omics Convergence: Comparison of enriched pathways between proteomic and transcriptomic layers to assess concordance at the gene vs. pathway level.

3. Key Contributions

• Definition of Molecular Signatures: The study defines a conserved "immune-metabolic" axis that distinguishes neurodegenerative and mixed pathologies from vascular pathology.
• Pathway-Level vs. Feature-Level Discordance: It demonstrates that while individual protein/gene differences between mixed and neurodegenerative groups are minimal, pathway-level organization reveals distinct remodeling programs specific to mixed pathology.
• Cross-Modal Robustness: It establishes that biological programs (immune activation, mitochondrial suppression) are robustly conserved across proteomic and transcriptomic layers, whereas individual gene-protein correlations are weak.
• Regional Conservation: It shows that these molecular programs are largely conserved across cortical (DLPFC, PCC) and subcortical (AC) regions, though with varying degrees of overlap.

4. Key Results

A. Proteomic and Transcriptomic Signatures

• Shared Immune-Metabolic Axis: Both Neurodegenerative and Mixed groups showed significant upregulation of immune/inflammatory pathways (leukocyte migration, chemotaxis, angiogenesis) and downregulation of mitochondrial/oxidative phosphorylation pathways compared to the Vascular group.
• Mixed Pathology Specificity: While Mixed and ND groups were molecularly similar at the single-protein level, pathway analysis revealed that Mixed pathology uniquely engages additional programs:
  • Upregulated: Extracellular matrix (ECM) organization, wound healing, epithelial morphogenesis, and vesicle-mediated transport.
  • Downregulated: Ribosome biogenesis and proteostasis pathways were more attenuated in Mixed than ND.
• Lack of Single-Feature Separation: No individual proteins were significantly differentially expressed between Mixed and ND groups (FDR < 0.05), highlighting the necessity of pathway-level analysis.

B. Co-Expression Network Analysis (WGCNA)

• Immune-Stress Modules (Salmon & Yellow): Upregulated in ND and Mixed vs. Vascular. Enriched for glial activation (GFAP, SDC4), AD-linked proteins (APP, CLU), and stress responses. Strongly correlated with amyloid burden, tau tangles, and cognitive decline.
• Mitochondrial Module (Turquoise): Downregulated in ND and Mixed vs. Vascular. Enriched for respiratory chain and bioenergetic proteins (TOMM40, SDHA). This module showed relative preservation in Vascular pathology, distinguishing it from the other groups.
• Mixed-Specific Modules: The "Green" module (vesicle transport, endosomal trafficking) was upregulated specifically in Mixed vs. Vascular, supporting the ECM/remodeling findings.

C. Cross-Omics and Regional Concordance

• Pathway Concordance: There was robust convergence between proteomics and transcriptomics at the pathway level (e.g., immune upregulation and mitochondrial downregulation were consistent across layers).
• Gene-Level Discordance: Cross-modal overlap at the level of individual genes/proteins was low (Jaccard index ≤ 0.12), indicating that systems-level organization is more stable than one-to-one transcript-protein correspondence.
• Regional Consistency: The immune-metabolic signature was conserved across DLPFC, PCC, and AC, with the strongest overlap between cortical regions (DLPFC and PCC).
• Cell-Type Resolution: snRNA-seq confirmed that the upregulation of GFAP (a marker of astrocyte reactivity) was distributed across multiple astrocyte subpopulations, rather than confined to a single reactive cluster.

5. Significance

• Redefining Mixed Pathology: The study provides molecular evidence that "Mixed Pathology" is not merely an additive combination of vascular and neurodegenerative states. Instead, it represents a distinct biological state characterized by a shared inflammatory-mitochondrial core plus unique tissue remodeling and vesicular transport programs.
• Biomarker Implications: The findings suggest that pathway-level signatures (systems biology approaches) are more robust biomarkers for stratifying late-life dementia than individual molecular markers, which may vary significantly between omics layers.
• Therapeutic Targets: The identification of preserved mitochondrial function in vascular pathology versus suppressed function in neurodegenerative/mixed states suggests that bioenergetic therapies might be specifically relevant for neurodegenerative and mixed cases, while vascular cases may require different interventions.
• Methodological Insight: The paper highlights the critical importance of analyzing molecular data at the pathway level to uncover biological truths that are obscured when focusing solely on individual differentially expressed genes or proteins.