Multimodal Ageing Biomarkers and Plasma Proteomic Signatures Associated with All-Cause Mortality

This study systematically benchmarks proteomic organ ages against established multimodal ageing biomarkers in the Lothian Birth Cohort 1936, revealing that while accelerated liver, immune, and heart ageing predict mortality, traditional markers like brain volume and cognitive function remain stronger predictors, and further identifies specific plasma proteins, notably GDF15, CST3, and COL18A1, as robust signatures of all-cause mortality.

The Gist

Imagine your body is like a massive, bustling city. For a long time, doctors have tried to guess how "old" this city is by looking at the calendar (your chronological age). But we all know that a 70-year-old city can look brand new and vibrant, while another 70-year-old city might be crumbling, full of potholes and failing infrastructure.

This paper is like a massive inspection report comparing different ways to measure how "old" and "worn out" that city really is, and which measurements are the best at predicting when the city might finally shut down (mortality).

Here is the breakdown of their findings, using some simple analogies:

1. The "Organ Clocks" vs. The "City Inspectors"

The researchers were excited about a new tool called Proteomic Organ Ages. Think of this as a high-tech scanner that looks at the blood and says, "Hey, your liver looks like it's 85, your heart looks like 78, and your immune system looks like 82."

• The Finding: These organ clocks do work. If your liver or heart looks older than your calendar age, you are at higher risk of dying sooner.
• The Catch: However, when they compared these organ clocks to other, more traditional "inspectors," the organ clocks weren't the strongest predictors. They were good, but not the best.

2. The "Super-Inspectors" (Who Won the Race?)

The study pitted the organ clocks against other well-known signs of aging. The winners were:

• The "Brain Map" (MRI Scans): Imagine looking at the city's central command center. If the brain has shrunk (less total volume) or has "rust spots" (white matter hyperintensities), that is a huge red flag. This was one of the strongest predictors.
• The "General Manager" (Cognitive Function): How well does the city run its daily operations? If the "General Manager" (your general intelligence or g) is slowing down, the whole city is at risk.
• The "Engine" (Physical Function): How fast can the city's delivery trucks move? If your walking speed is slow or your lungs (the engine) can't take in enough air, that's a major warning sign.
• The "DNA Report Card" (GrimAge2): This is a sophisticated blood test that looks at your genetic code. It was the single strongest predictor of all. It's like a report card that summarizes everything wrong with the city's infrastructure in one number.

The Big Takeaway: When you put all these inspectors in a room together, the organ clocks (liver, heart, etc.) stopped being the stars. The Brain, the General Manager (Cognition), the Engine (Walking/Lungs), and the DNA Report Card were the ones who could actually predict the future on their own. The organ clocks seemed to be just echoing what these other, stronger signs were already saying.

3. The "Smoking Guns" in the Blood (Proteins)

The researchers also looked at thousands of tiny messengers in the blood (proteins) to see which ones were shouting "Danger!"

They found 202 proteins that were linked to death. Two main themes emerged:

• The "Firefighters" (Immune System): Many of the dangerous proteins were related to the immune system. It's like the city's firefighters are either overworked, confused, or fighting a fire that won't go out. When the immune system is in a constant state of panic, the city suffers.
• The "Construction Crew" (Extracellular Matrix): This is the scaffolding that holds the city's buildings together. The study found that when this scaffolding gets messy, stiff, or broken (like old, rusted steel beams), the city becomes fragile.

The Top 3 "Villains":

1. GDF15: The loudest alarm bell. High levels mean the body is under massive stress.
2. CST3: A protein linked to inflammation and kidney issues.
3. COL18A1: A protein involved in that "scaffolding" mentioned above.

4. The "Protective Heroes"

Not all proteins were bad news. Some were like the city's maintenance crew working overtime to keep things running.

• HPGDS and NPS: These proteins were lower in people who died sooner. This means having more of them is actually protective, like having a strong security team or a well-oiled machine.

The Bottom Line for You

If you want to know how "biologically old" you are and how long you might live, don't just look at a single organ's age.

Think of your body as a complex city. The study suggests that to get the best prediction of the city's future, you need to check:

1. The Brain's structural integrity (Is the central command shrinking?).
2. The General Manager's sharpness (Is your mind still quick?).
3. The Engine's power (Can you walk fast and breathe deep?).
4. The DNA Report Card (What does your genetic code say about your stress levels?).

While knowing that your liver or heart is "aging fast" is useful, it's often just a symptom of the bigger problems happening in the brain, the mind, and the body's overall engine. The most powerful way to predict your future is to look at the whole city, not just one building.

Technical Summary

1. Problem Statement

The field of ageing research requires robust biomarkers that can predict mortality risk more accurately than chronological age alone. While various modalities exist (epigenetic, functional, neuroimaging), there is a lack of systematic benchmarking between emerging proteomic organ-age clocks (which estimate the biological age of specific organs like the liver, heart, and brain based on plasma proteins) and established multimodal biomarkers. Furthermore, the specific plasma proteins and biological pathways driving mortality risk remain to be fully characterized in a large-scale, unbiased manner. This study aims to:

1. Benchmark eleven proteomic organ-age gaps against a comprehensive set of existing ageing biomarkers (epigenetic, neuroimaging, cognitive, and physical) for predicting all-cause mortality.
2. Identify novel plasma proteomic signatures associated with mortality and elucidate their underlying biological processes and tissue origins.

2. Methodology

Study Population:

• Cohort: Lothian Birth Cohort 1936 (LBC1936), a longitudinal study of individuals born in 1936.
• Sample Size: 861 participants assessed at Wave 2 (mean age = 72.5 years).
• Follow-up: Up to 17.4 years (mean follow-up ~10 years for deceased participants).
• Outcome: All-cause mortality (444 deaths, 417 censored).

Data Modalities:

• Proteomic Organ Ages: Derived using SomaScan 11K plasma proteomics (8,699 unique proteins). Organ-age predictors (adipose, artery, brain, heart, immune, intestine, kidney, liver, lung, muscle, pancreas) were trained on the Knight-ADRC cohort and applied to LBC1936. Organ-age gaps were calculated as standardized residuals of predicted organ age regressed on chronological age.
• Epigenetic Age: GrimAge2 acceleration (DNA methylation-based).
• Telomere Length: Measured via qPCR.
• Neuroimaging: MRI-derived Total Brain Volume (TBV), Grey Matter Volume (GMV), White Matter Hyperintensity (WMH) volume, and MRI-derived brain age gap.
• Cognitive Function: General cognitive ability ($g$) derived from 13 tests via structural equation modeling.
• Physical Function: Grip strength, 6-meter walk time, and respiratory function (FEV, FVC, FER).
• Covariates: Chronological age, sex, smoking pack-years, alcohol units, and BMI.

Statistical Analysis:

• Survival Analysis: Cox proportional hazards models were used to assess associations between biomarkers and time-to-death.
• Benchmarking:
  • Univariate Models: Each biomarker adjusted for age and sex.
  • Multivariable Model: A "saturated" model including all 21 selected biomarkers (handling collinearity by excluding FER and GMV) adjusted for age, sex, and modifiable risk factors.
• Proteomic Discovery:
  • Stepwise adjustment: Age/sex $\rightarrow$ +eGFR $\rightarrow$ +modifiable risk factors.
  • Elastic Net-Penalized Cox Regression: Applied to all 8,699 proteins simultaneously to select a parsimonious set of predictors (5-fold cross-validation, $\alpha=0.5$, $\lambda=0.16$).
• Enrichment Analysis: Gene Ontology (Biological Process) and Tissue Enrichment (GTEx database) were performed on significant mortality-associated proteins.

3. Key Results

A. Benchmarking of Organ Ages vs. Other Biomarkers

• Proteomic Organ Ages: Accelerated ageing in the liver (HR = 1.43), immune system (HR = 1.42), and heart (HR = 1.38) showed the strongest associations with mortality among the organ clocks.
• Superior Predictors: Established biomarkers outperformed organ-age clocks. GrimAge2 acceleration had the highest hazard ratio (HR = 1.62). Other strong predictors included TBV (HR = 1.52), GMV (HR = 1.44), respiratory function (FEV) (HR = 1.51), and general cognitive function ($g$) (HR = 1.46).
• Multivariable Independence: In the fully adjusted model including all biomarkers, only four measures retained independent significance:
  1. Total Brain Volume (TBV)
  2. White Matter Hyperintensity (WMH) volume
  3. General cognitive function ($g$)
  4. Walk time
  • Implication: Proteomic organ ages did not provide incremental predictive value beyond these core measures, suggesting their effects may be mediated by brain health, cognition, and physical function.

B. Plasma Proteomic Signatures of Mortality

• Significant Proteins: After full adjustment, 202 proteins were significantly associated with mortality.
• Top Risk Factors (Positive Association):
  • GDF15 (HR = 1.53): Growth differentiation factor 15.
  • CST3 (HR = 1.48): Cystatin C.
  • COL18A1 (HR = 1.47): Collagen Type XVIII Alpha 1 Chain.
• Top Protective Factors (Negative Association):
  • HPGDS (HR = 0.72 reciprocal), NPS, and BAGE3.
• Elastic Net Model: A penalized regression model selected 37 proteins as the most predictive set. GDF15 remained the strongest positive predictor, while HPGDS was the strongest negative predictor. There was minimal overlap (only 23 proteins) between the mortality-associated proteins and those used to construct the organ-age clocks.

C. Biological Enrichment

• Pathways: The 202 mortality-associated proteins were enriched for immune-related processes (antimicrobial humoral response, chemotaxis, cell adhesion) and extracellular matrix (ECM) organization.
• Tissue Origin: The liver was the only significantly enriched tissue of origin, aligning with the strong liver-age signal in the organ-clock analysis.

4. Key Contributions

1. Systematic Benchmarking: This is the first study to directly compare proteomic organ-age clocks against a wide array of multimodal biomarkers (epigenetic, imaging, cognitive, functional) within the same population. It establishes that while organ ages are predictive, they are currently outperformed by GrimAge2, brain structure, and physical/cognitive function.
2. Identification of Core Predictors: The study identifies a "core" set of independent mortality predictors (TBV, WMH, $g$, walk time) that capture the majority of the variance explained by diverse ageing biomarkers.
3. Novel Proteomic Signatures: It provides a high-resolution map of plasma proteins associated with mortality, highlighting GDF15, CST3, and COL18A1 as top candidates.
4. Mechanistic Insight: The findings link mortality risk strongly to immune dysregulation and extracellular matrix deterioration, providing biological context for the proteomic signatures.

5. Significance and Limitations

Significance:

• Clinical Stratification: The results suggest that for general mortality risk stratification, focusing on brain health, cognition, and physical mobility (along with epigenetic clocks) may be more efficient than measuring specific organ-age clocks.
• Therapeutic Targets: The identification of immune and ECM-related proteins (e.g., GDF15, collagen pathways) offers new targets for interventions aimed at extending healthspan and longevity.
• Methodological Rigor: The use of elastic net regression on 11K proteins provides a robust, data-driven approach to biomarker discovery that accounts for high dimensionality and multicollinearity.

Limitations:

• Generalizability: The cohort consists of relatively healthy, ethnically homogeneous Scottish individuals, which may limit applicability to more diverse populations.
• Single Time Point: Analyses relied on baseline biomarkers; longitudinal trajectories of protein levels were not incorporated.
• Cause-Specific Mortality: The study focused on all-cause mortality; organ-age clocks might perform better for specific causes of death (e.g., liver disease).
• Platform Dependency: Results are specific to the SomaScan 11K platform.

Conclusion:
While proteomic organ-age clocks are valid markers of biological ageing, they do not currently offer superior predictive power over established multimodal biomarkers like brain volume and cognitive function for all-cause mortality. However, the study successfully uncovers a distinct set of plasma proteins (particularly GDF15 and ECM components) that serve as powerful, independent predictors of mortality, driven by immune and structural tissue pathways.