Network subtypes of cortical similarity reveal molecular correlates of normative and compensatory ageing associated with longevity genes expression

By integrating structural MRI and cortical transcriptomics, this study identifies two distinct cortical ageing subtypes—a normative metabolic-immune decline and a compensatory stress-repair adaptation—linked to specific longevity gene expression profiles.

The Gist

Imagine your brain as a massive, bustling city. For decades, scientists have known that as people age, this city undergoes changes: some buildings (brain regions) get smaller, some roads (connections) get narrower, and the overall layout shifts. But until now, we didn't fully understand why some people's cities age gracefully while others crumble, or what the "blueprints" (genes) behind these changes look like.

This paper is like a new kind of urban planning report that combines satellite images of the city with the original architectural blueprints to find two distinct ways cities age.

Here is the breakdown of their discovery in simple terms:

1. The Tool: Measuring "City Similarity"

The researchers didn't just look at how big a building was; they looked at how similar different neighborhoods were to each other in terms of their shape, size, and texture. They called this "Morphometric Similarity."

Think of it like this: In a young, healthy city, the library, the school, and the park might look very different from each other. But as the city ages, these buildings might start to look more alike—perhaps they all get a bit smaller or develop similar cracks. The researchers mapped these "neighborhood similarities" to see how the city's layout was changing.

2. The Discovery: Two Different Aging Stories

By analyzing MRI scans of 952 people (from age 18 to 94), they found that the city doesn't just age in one straight line. Instead, the population splits into two distinct groups (or "subtypes") based on how their brain neighborhoods change:

Group A: The "Standard Decline" (Normative-Ageing)

• The Metaphor: Imagine a city where the infrastructure is slowly wearing out due to lack of maintenance. The roads get a bit potholed, and the buildings shrink naturally.
• What's happening in the brain: This group shows the "expected" signs of aging. The connections between brain regions become less distinct.
• The Genetic Blueprint: The "blueprints" driving this group are focused on metabolism and the immune system. It's like the city's power grid and sanitation workers are struggling to keep up with the daily wear and tear. The genes involved are related to how the body processes sugar (insulin) and fights off inflammation.
• The Outcome: This group's brain structure looks very similar to people who have started to show mild memory problems. It's the "typical" path of aging.

Group B: The "Resilient Repair" (Compensatory-Ageing)

• The Metaphor: Imagine a city that is under construction. Even though it's old, the city planners are actively reinforcing the foundations, patching holes, and building new support beams. The city looks different because it's fighting back against the decay.
• What's happening in the brain: This group actually keeps their brain neighborhoods looking more connected and organized than the "Standard" group, even at older ages. They are adapting.
• The Genetic Blueprint: The "blueprints" here are focused on emergency repairs and stress management. The genes involved are like a "fix-it crew" that repairs damaged DNA and cleans up cellular trash (proteins). They are the body's way of saying, "We are stressed, but we are fixing it."
• The Outcome: This group seems to be on a path of resilience. Their brains are reorganizing to stay functional, potentially protecting them from cognitive decline.

3. The Left-Handed Bias

One of the coolest findings is that this "repair crew" seems to be most active in the left side of the brain.

• Analogy: If the brain is a city, the left side seems to be the "administrative district" where the most intense genetic repair work is happening. The researchers found that the link between the genes and the brain structure was much stronger on the left side than the right. This might explain why the left brain is so crucial for things like language and memory as we get older.

4. Why This Matters

Before this study, we thought aging was just a slow, uniform slide into decline. This paper shows that aging is actually a fork in the road:

1. Path 1: You follow the standard metabolic decline (like a city slowly wearing out).
2. Path 2: You activate a stress-response repair system (like a city actively reinforcing itself).

The Takeaway:
The researchers suggest that the "Repair" group isn't just "aging slower"; they are aging differently. They are using a different set of genetic tools to keep their brain city running.

This is huge news because it means we might be able to identify which path a person is on before they lose their memory. If we can spot someone heading down the "Standard Decline" path, maybe we can give them treatments that switch them to the "Repair" path—teaching their brain to activate those stress-response genes and build better defenses against aging.

In short: Your brain isn't just getting old; it's choosing a strategy. Some brains are letting the paint peel, while others are actively repainting the walls. This study helps us see which strategy you're using and why.

Technical Summary

1. Problem Statement

Brain ageing is characterized by widespread structural changes, yet the biological mechanisms driving the heterogeneity of these trajectories remain poorly understood. Current neuroimaging studies often delineate normative patterns (e.g., cortical thinning) but fail to explain the system-level organization of these changes or their relationship to molecular ageing mechanisms. Specifically, there is a lack of frameworks to distinguish between normative decline (typical age-related deterioration) and compensatory adaptation (biological resilience or reorganization) at the network level. The authors hypothesize that integrating structural connectivity with transcriptomics can reveal distinct subtypes of ageing trajectories linked to specific genetic pathways.

2. Methodology

The study employed a multi-modal approach combining structural MRI, cortical transcriptomics, and normative modeling on a large cohort.

• Dataset:

  • Participants: 952 healthy adults (aged 18–94) from three datasets: NKI-Rockland, OASIS-1/2, and IEEE-OpenBHB.
  • Stratification: Divided into Young Adults (YA, 18–30, n=199) and Healthy Ageing (HA, 45–94, n=753).
  • Imaging: 3D T1-weighted MRI images preprocessed using FreeSurfer (v6.0).

• Morphometric Similarity Networks (MSNs):

  • The cortex was parcellated into 148 regions (Destrieux Atlas).
  • Nine morphometric features were extracted: surface area, gray volume, thickness (mean/SD), mean curvature, Gaussian curvature, folding index, and curvature index.
  • Three specific network configurations were constructed based on feature correlations:
    1. 4vf (4-volume-feature): Volume, surface area, thickness, and thickness SD (sensitive to age/neurodegeneration).
    2. 4cf (4-curvature-feature): Curvature and folding indices (developmentally constrained/stable).
    3. 5f (5-feature): A combination of volumetric and curvature metrics.
  • Connectivity Metric: Weighted node degree was calculated for three core cognitive networks: Default Mode Network (DMN), Central Executive Network (CEN), and Visual Network (VIS).

• Subtype Inference (SuStaIn):

  • The Subtype and Stage Inference (SuStaIn) framework was applied to the intra-network connectivity of the HA group.
  • Biomarkers were normalized to the YA group; decreasing biomarkers were inverted to model progression.
  • The model identified distinct subtypes and stages of ageing progression.

• Genetic Analysis:

  • Gene Selection: 251 longevity-associated genes from the GenAge database were matched with cortical transcriptomics from the Allen Human Brain Atlas (AHBA).
  • Association: Partial correlations were computed between intra-network connectivity and regional gene expression, controlling for cortical thickness and surface area.
  • Correction: False Discovery Rate (FDR) correction was applied ($p < 0.05$).

3. Key Contributions

1. Discovery of Two Distinct Ageing Subtypes: The study identified two robust subtypes of cortical ageing that emerge specifically from volumetric-feature-based MSNs (4vf), not from curvature-only or mixed-feature networks.
2. Molecular Characterization: The authors linked these structural subtypes to distinct molecular signatures involving longevity genes, providing a biological basis for "normative" vs. "compensatory" ageing.
3. Hemispheric Asymmetry: The analysis revealed significant lateralization in gene-network coupling, with the left hemisphere showing stronger associations (24% variance explained) compared to the right (10%).
4. Differentiation from Pathology: The study contextualized these subtypes against mild cognitive impairment (CDR 1/2), showing that the "normative" subtype shares structural profiles with early impairment, while the "compensatory" subtype represents a distinct, adaptive trajectory.

4. Results

A. Identification of Subtypes

• Subtype 1 (Normative-Ageing): Characterized by connectivity profiles consistent with typical age-related decline.
  • Molecular Signature: Enriched for genes involved in metabolism, insulin signaling, and immune regulation (e.g., IRS1, PPARGC1A, GSK3B).
  • Protective Factors: Negative associations with FOXO1 and PLCG2 suggest potential resilience mechanisms.
  • Spatial Distribution: Involved prefrontal and sensorimotor regions (CEN/DMN).
• Subtype 2 (Compensatory): Displayed more preserved intra-network connectivity and higher mean ages in the HA group.
  • Molecular Signature: Linked to stress-response, DNA repair, and proteostasis (e.g., DDIT3, UBB, STUB1, FEN1).
  • Spatial Distribution: Uniquely involved medial occipital and temporal regions (lingual/parahippocampal gyri), suggesting recruitment of visual and memory circuits.
• Overlap: Both subtypes shared genes related to oxidative stress and cell cycle regulation (E2F1, MXD1, GCLM, GSK3B), indicating a common core biology, but diverged in their specific stress vs. metabolic responses.

B. Network and Hemispheric Findings

• Network Specificity: Subtypes were most distinct in the DMN and CEN. In the Visual Network (VIS), differences were less pronounced, and a single subtype was often observed in the right hemisphere, consistent with the visual cortex's later vulnerability to ageing.
• Lateralization: Overlapping gene-associated nodes were exclusively located in the left hemisphere (e.g., superior frontal gyrus, cingulate), suggesting lateralized genetic contributions to network architecture, possibly related to verbal memory and executive control.
• Age Progression: Higher SuStaIn stages correlated with older mean ages. Subtype 2 consistently showed higher mean ages than Subtype 1, suggesting it represents a later or more resilient stage of the ageing process.

C. Comparison with Cognitive Impairment

• Normative Subtype (1): Showed morphometric similarity patterns resembling the CDR 1/2 (mild-moderate impairment) group, suggesting this trajectory may precede overt cognitive symptoms.
• Compensatory Subtype (2): Showed a distinct profile diverging from both the normative ageing and impaired groups, characterized by higher intra-network similarity, supporting the hypothesis of an adaptive reorganization.

5. Significance

• Novel Framework: This study establishes a biologically informed framework to distinguish between normative decline and compensatory adaptation, moving beyond simple "healthy vs. diseased" dichotomies.
• Biomarker Potential: The identification of specific gene-network couplings (metabolic/immune vs. stress/repair) offers potential biomarkers for early detection of pathological trajectories or for identifying individuals with high resilience.
• Mechanistic Insight: It demonstrates that volumetric morphometric features are more sensitive to age-related network reorganization than curvature-based features, refining the tools used in neuroimaging ageing research.
• Clinical Implication: By capturing the full spectrum of ageing trajectories, the findings support the development of personalized interventions targeting specific molecular pathways (e.g., metabolic support for Subtype 1 vs. stress-reduction strategies for Subtype 2) to delay or prevent cognitive decline.