Longitudinal dynamics of organ-specific proteomic aging clocks over a decade of midlife

This study analyzes a decade of longitudinal proteomic data from middle-aged adults to demonstrate that organ-specific aging clocks are stable, dynamic indicators that track subclinical risk factors, reveal the central role of immune and adipose systems, and reflect the profound impact of menopause and medication on multi-organ aging trajectories.

The Gist

Imagine your body isn't just one big machine, but a bustling city with different neighborhoods: the Heart District, the Liver Industrial Zone, the Immune System Police Force, and the Brain Library.

For years, scientists have tried to build a "biological clock" to tell us how old these neighborhoods really are, compared to your actual birth date. Usually, they just take a snapshot of the city at one moment in time. But this new study is different. The researchers took a 10-year time-lapse video of 1,250 middle-aged people, checking their "city neighborhoods" twice, a decade apart.

Here is what they discovered, explained simply:

1. The Snapshot vs. The Movie

Most aging studies are like looking at a photo of a city and guessing how fast the buildings are aging based on how old the buildings look right now. This study asked: Does the photo actually match the movie?
The Answer: Yes! The researchers found that the "snapshot" (cross-sectional data) is a pretty good guess at how the city actually changes over time. If a protein looks old in a photo, it's likely actually aging fast in real life. This means we can trust the "snapshot" clocks we've been using.

2. The "Aging Chain Reaction"

They found that aging doesn't happen in isolation. If the Heart District starts to look a bit worn out, it often drags the Liver and Fat Tissue down with it.

• The Analogy: Think of the body like a relay race. The Heart and Lungs are the runners who start first. If they stumble (age early), they pass the baton of "wear and tear" to the Kidneys, Muscles, and Intestines.
• The Hubs: The Immune System and Fat Tissue act like the central train stations. If they get chaotic, the whole city gets disrupted.

3. The Great Gender Divide (and the Menopause Effect)

Men and women age differently, like two different cities with different traffic patterns.

• The Menopause Surprise: For women, the transition into menopause was the single biggest event in the 10-year video. It was like a sudden storm that accelerated aging across the whole city, especially for the Immune System.
• The White Blood Cell Mystery: As women went through menopause, their white blood cell counts dropped, and their "immune age" sped up. It turns out the drop in estrogen (the hormone that keeps the immune system calm) was the main culprit.

4. The "Fake" Aging (Medication vs. Lifestyle)

This is the most fascinating part. The researchers wanted to know: Can we slow down the clock?

• Lifestyle: Surprisingly, changing how much you smoke or drink didn't show a clear change in the 10-year video. (The researchers suspect this is because most people in the study didn't change their habits drastically, not that lifestyle doesn't matter).
• Medication: This is where it got tricky. When people started taking common heart medications (like statins or blood pressure pills), their "organ clocks" suddenly looked like they were aging faster.
  • The Twist: It wasn't that the drugs were hurting them! It was that the drugs were doing their job so well that they changed specific proteins the clock was watching.
  • The Analogy: Imagine a speedometer on a car. If you install a new engine that makes the car go faster, the speedometer needle jumps up. If you didn't know about the new engine, you'd think the car was broken. Similarly, statins lower cholesterol (APOB), and the clock saw that drop and thought, "Oh no, the liver is aging!" But it was actually a sign of the drug working.

5. Why This Matters

This study proves that these "organ clocks" aren't just static labels. They are dynamic dashboards.

• They can track if your body is reacting to a new disease (like diabetes making the pancreas age faster).
• They can tell the difference between a drug working (changing a specific protein) and a body breaking down.
• They show that if you take care of your heart and lungs early, you might protect your other organs later.

In a nutshell: Your body is a complex city. This study gave us a 10-year time-lapse showing that while some neighborhoods age together, the Heart and Lungs set the pace. It also taught us that sometimes, when our "clocks" look like they are speeding up, it might just be our medicine doing its job, not our body failing.

Technical Summary

1. Problem Statement

Organ-specific proteomic clocks are emerging tools for quantifying biological aging and predicting organ-specific disease risk. However, their development and validation have relied almost exclusively on cross-sectional data (single time-point measurements). This creates several critical knowledge gaps:

• Validity of Cross-Sectional Assumptions: It is unclear if cross-sectional age associations accurately reflect true within-person longitudinal aging trajectories.
• Dynamic Behavior: It is unknown whether organ age gaps (the difference between predicted and chronological age) are stable over time or if they track meaningful changes in organ health.
• Inter-organ Propagation: The temporal hierarchy of aging across different organs (i.e., whether aging in one organ drives aging in another) remains unexplored.
• Modifiability: The extent to which lifestyle changes or medication initiation alter organ age gaps is poorly understood.

2. Methodology

The study utilized a robust longitudinal design within the Asklepios cohort, a population-based study of generally healthy middle-aged adults in Belgium.

• Cohort & Data:

  • Participants: 1,250 adults (mean baseline age ~46, follow-up ~56) with paired plasma samples collected approximately 10 years apart (mean interval 9.9 years).
  • Proteomics: High-throughput profiling of 6,402 circulating plasma proteins using the SomaScan 7K assay (7,289 aptamers targeting human proteins).
  • Sample Size: After quality control, 1,114 participants had successful paired measurements (1,128 baseline and 1,228 follow-up samples).

• Statistical Framework:

  • Protein Trajectories: A within–between mixed-effects model was fitted for each protein to disentangle longitudinal (within-person) changes from cross-sectional (between-person) differences.
  • Clock Training: Organ-specific proteomic clocks were trained for 11 organ systems (e.g., heart, liver, kidney, immune) plus a full-body clock.
    • Organ Definition: Proteins were defined as organ-enriched if expression was >4-fold higher in one organ system compared to others, based on GTEx v8 RNA-seq data.
    • Algorithm: Ensemble of 500 LASSO regression models using nested 5-fold cross-validation. Both sex-combined and sex-stratified models were trained.
  • Age Gap Calculation: Organ age gaps were calculated as residuals from a LOESS regression of predicted age on chronological age, standardized to allow comparison.
  • Causal Discovery: A temporally constrained structural equation modeling (SEM) approach (Fast Greedy Equivalence Search) was used to infer directed cross-organ aging networks (baseline $\to$ follow-up).
  • Phenotype Association: Linear mixed-effects models and linear regressions were used to associate changes in organ age gaps with changes in clinical markers, lifestyle factors, menopausal status, and medication use.

3. Key Contributions

• First Longitudinal Validation: This is the first comprehensive study to evaluate the stability and dynamics of organ-specific proteomic clocks over a decade.
• Validation of Cross-Sectional Training: Demonstrated that cross-sectional age associations are a valid proxy for within-person longitudinal changes, justifying the common use of cross-sectional data for clock training.
• Discovery of Aging Hierarchy: Mapped a causal network showing how aging propagates across organs, identifying central hubs (immune, adipose) and upstream drivers (heart, lung).
• Mechanistic Interpretability: Showed that apparent "accelerations" in organ age due to medication are often driven by specific drug-targeted proteins rather than generalized organ deterioration, highlighting the importance of clock interpretability.

4. Key Results

A. Protein Trajectories and Sex Differences

• Cross-Sectional vs. Longitudinal: Cross-sectional and longitudinal age effects were highly correlated ($\rho = 0.91$), confirming that cross-sectional training captures true aging trends.
• Sex Interactions: Nearly half (45.9%) of age-associated proteins showed significant sex interactions. Age-related changes were generally more pronounced in women, likely due to the menopausal transition (ages 45–55).
• Key Proteins:
  • Increasing with age: Enriched for heparin binding (coagulation).
  • Decreasing with age: Enriched for potassium ion transport.
  • Sex-specific: FSH and LH increased sharply in women; MSMP (prostate-related) decreased in men.

B. Clock Performance and Stability

• Accuracy: General (sex-combined) clocks achieved out-of-fold correlations ranging from $r=0.91$ (full-body) to $r=0.31$ (lung). Sex-stratified clocks offered only marginal improvements.
• Stability: Organ age gaps were moderately stable over 10 years (median $r=0.58$ between baseline and follow-up).
• Coordination: Acceleration in one organ typically coincided with acceleration in others, indicating a shared systemic aging component.

C. Cross-Organ Aging Network

• Central Hubs: The immune and adipose systems occupied central positions, forming an immunometabolic axis.
• Upstream Drivers: The heart and lung emerged as upstream nodes. Early cardiorespiratory aging predicted subsequent aging in metabolic organs (liver, intestine, kidney, adipose).
• Feedback Loops: Bidirectional links were observed between liver–brain, brain–immune, and adipose–intestine.

D. Clinical and Lifestyle Associations

• Clinical Markers: Longitudinal worsening of cardiometabolic markers (BMI, glucose, uric acid, creatinine) tracked with worsening organ age gaps. For example, incident diabetes was associated with accelerated aging of the kidney, pancreas, and intestine.
• Menopause: The transition from pre- to post-menopause was a dominant driver of aging. Women who transitioned during the study showed significant acceleration in immune, full-body, liver, and arterial age gaps, driven largely by declining neutrophil counts and rising FSH/LH.
• Lifestyle: Unlike cross-sectional data, longitudinal changes in smoking or alcohol consumption did not significantly alter organ age gaps, likely due to limited variation in a general population cohort.

E. Medication Effects (Interpretability)

• Statin Initiation: Associated with increased liver and immune age gaps. However, sensitivity analysis revealed this was driven almost entirely by Apolipoprotein B (APOB) (a drug target) and KIR3DL1, not generalized organ damage.
• RAS Inhibitor Initiation: Associated with increased kidney age gaps, driven primarily by compensatory Renin elevation.
• Implication: When drug-targeted proteins were removed from the clock models, the associations with medication initiation were strongly attenuated. This proves the clocks are sensitive to specific molecular mechanisms rather than just "noise."

5. Significance

• Dynamic Monitoring: The study validates organ-specific proteomic clocks as dynamic indicators that track subclinical physiological changes, not just static risk predictors.
• Biological Insight: It elucidates the temporal hierarchy of aging, suggesting that interventions targeting cardiorespiratory health or the menopausal transition could have downstream benefits for metabolic organ aging.
• Clinical Utility: The findings emphasize that biological age clocks must be interpretable. Apparent "accelerations" can be mechanistic responses to therapy (e.g., statins lowering APOB) rather than pathological decline. This distinction is crucial for using these clocks in clinical trials and personalized medicine.
• Future Directions: The authors suggest that future clocks should be trained on clinical endpoints or functional decline rather than just chronological age to maximize clinical translatability.

In summary, this paper establishes that organ-specific proteomic clocks are stable, interpretable, and sensitive to physiological shifts (menopause, medication, metabolic health), providing a powerful toolkit for monitoring biological aging in midlife.