Plasma proteomics of APOE genotype: age-specific analyses in UK population-based cohorts

This study analyzes UK Biobank data to demonstrate that APOE ε4 and ε2 genotypes exert broad, often age-dependent effects on the circulating plasma proteome across diverse ancestries, revealing early biomarkers of neurodegeneration decades before typical Alzheimer's disease diagnosis.

The Gist

Imagine your body is a massive, bustling city. Inside this city, there are millions of workers (proteins) constantly delivering messages, fixing roads, and managing traffic. One of the most important managers in this city is a protein called APOE. Think of APOE as the city's "Traffic Control Chief."

This study is like a massive inspection of the city's traffic logs to see how different versions of this Chief affect the city's workers, and how those effects change as the city gets older.

Here is the breakdown of the research in simple terms:

1. The Three Versions of the Chief

The "Traffic Chief" (APOE) comes in three main uniforms, determined by your genes:

• The Standard Uniform (ε3): This is the most common version. It's the "average" Chief.
• The Risky Uniform (ε4): This version is known to be a bit chaotic. People with this version are much more likely to develop Alzheimer's disease later in life.
• The Protective Uniform (ε2): This version is like a super-efficient Chief. It actually lowers the risk of Alzheimer's.

2. The Big Investigation

Scientists wanted to know: How does wearing the "Risky" or "Protective" uniform change the behavior of the other workers in the city?

They looked at blood samples from over 42,000 people in the UK (the UK Biobank). They didn't just look at one or two workers; they checked 2,922 different proteins at once. It's like checking the performance of every single delivery driver, construction worker, and sanitation crew member in the city simultaneously.

They also split the people into three age groups:

• Middle-aged (40–50): The city is still relatively young.
• Pre-retirement (50–60): The city is getting busier.
• Senior (60–70): The city is older, and wear-and-tear is starting to show.

3. What They Found: The "Early Warning System"

The most exciting discovery is that the "Risky" and "Protective" uniforms start changing the city's traffic patterns decades before anyone gets sick.

• The "Risky" Chief (ε4) creates a specific traffic jam: Even in people in their 40s and 50s who are perfectly healthy, the ε4 version causes certain proteins to spike or drop. It's like the Chief is accidentally blocking the main highway, causing a backup that starts long before the accident (Alzheimer's) actually happens.
• The "Protective" Chief (ε2) clears the roads: The ε2 version changes the traffic in a different way, often keeping things flowing smoothly or reducing inflammation.
• The "Age" Factor: Some of these changes get worse as the city gets older. For example, the ε4 Chief seems to make the "cleanup crew" (immune system proteins) work overtime as people age, which might eventually lead to damage.

4. The "Cheat Sheet" (Biomarkers)

The study found specific proteins that act like smoke alarms.

• GFAP and NEFL: These are proteins that usually signal when brain cells are getting hurt. The study found that people with the "Risky" (ε4) uniform had higher levels of these alarms in their blood, even when they felt fine.
• The Surprise: They found that the "Protective" (ε2) uniform also changed these alarms, but in a way that suggested it might be helping the brain repair itself earlier than expected.

5. Why This Matters (The "So What?")

Think of Alzheimer's disease like a house fire.

• Old way: We used to wait until the house was burning down (symptoms appeared) to call the fire department. By then, it's often too late to save the furniture.
• New way (This study): This research shows us the smoke detectors going off in the kitchen (blood changes) 20 years before the fire starts.

Because we can see these changes in the blood of healthy people in their 40s and 50s, doctors might one day be able to:

1. Identify risk early: Tell someone, "Your Traffic Chief is wearing the risky uniform, and the city is starting to get congested."
2. Intervene sooner: Start treatments or lifestyle changes before the brain damage happens, rather than trying to fix it after the damage is done.

6. The Limitations (The "Fine Print")

• Different Tools: The scientists used two different types of "cameras" (proteomic platforms) to take these pictures. Sometimes the cameras saw slightly different things, which is a bit confusing but normal in science.
• The "Healthy" Bias: The people in the study were generally healthier and wealthier than the average person. The findings might look slightly different in a more diverse population.
• Snapshot vs. Movie: This study took a "snapshot" of people at one time. To see the full story, we need to watch a "movie" of the same people over many years to see how their traffic patterns change day-to-day.

The Bottom Line

This paper is a giant step forward in understanding how our genes shape our brain health long before we feel sick. It proves that the "Risky" and "Protective" versions of the APOE gene leave a distinct fingerprint in our blood, starting in middle age. This gives us a powerful new window of opportunity to catch Alzheimer's early and potentially stop it in its tracks.

Technical Summary

1. Problem Statement

The APOE locus is the strongest genetic risk factor for late-onset Alzheimer's disease (AD), with the $\varepsilon4$ allele increasing risk and the $\varepsilon2$ allele conferring protection relative to the common $\varepsilon3$ allele. While the effects of APOE on lipid metabolism are well-characterized, its broader impact on the circulating proteome across the human lifespan remains poorly understood. Most existing research focuses on older individuals or clinical AD cases, missing the critical window where molecular changes (decades before symptom onset) may begin. This study aims to characterize the specific, age-dependent effects of APOE $\vare4$ and $\vare2$ carriage on the plasma proteome in middle-aged and older adults to identify early biomarkers and pathophysiological pathways.

2. Methodology

Study Design & Cohorts:

• Primary Cohort: UK Biobank (UKB), a prospective cohort of ~502,000 adults. The study utilized plasma proteomic data from the UK Biobank Pharma Proteomics Project (UKB-PPP).
  • Sample Size: $N = 42,642$ participants after quality control (QC).
  • Ancestry: Primary analysis focused on European ancestry (EUR; $n=40,092$). Secondary analyses included African (AFR; $n=1,335$) and South Asian (SAS; $n=971$) ancestry groups. East Asian groups were excluded due to low power.
  • Age Range: 39.1 to 70.9 years.
• Replication Cohorts: Two independent UK-based cohorts profiled using SomaLogic SomaScan platforms:
  • INTERVAL: Used for ages 40–60.
  • NSHD (1946 British birth cohort): Used for ages 60–70.

Proteomic Measurement:

• Platform: Olink Explore 3072 proximity extension assay (measuring 2,922 proteins).
• Data Processing: Protein levels were normalized (NPX values), rank-based inverse normal transformed within ancestry/age groups, and modeled as continuous outcomes.

Genetic Definition:

• APOE genotypes were derived from whole-exome sequencing (rs7412 and rs429358).
• Reference Group: $\vare3/\vare3$ homozygotes.
• Exclusions: $\vare2/\vare4$ heterozygotes were excluded to avoid confounding opposing effects.

Statistical Analysis:

• Model: Multivariable linear regression testing associations between APOE carriage ($\vare4$ or $\vare2$ vs. $\vare3/\vare3$) and protein levels.
• Stratification: Analyses were stratified by age (40–50, 50–60, 60–70 years) to detect age-dependent effects.
• Covariates: Adjusted for age (within strata), sex, proteomic batch, sub-cohort, and ancestry-specific genetic principal components.
• Multiple Testing:
  • A priori sets (established biomarkers, AD risk loci proteins): No correction (unadjusted $P < 0.05$).
  • Hypothesis-free (all 2,922 proteins): Benjamini-Hochberg FDR correction at 5%.
• Trend Analysis: Meta-regression used to test for linear age-related trends in effect sizes.
• Pathway Enrichment: Gene Ontology (GO) Biological Process analysis using clusterProfiler and MSigDB.

3. Key Contributions

1. Largest Age-Stratified Proteomic Analysis: This is the largest study to date ($N > 42,000$) examining APOE-proteome associations across the mid-to-late life course, moving beyond cross-sectional snapshots of elderly AD patients.
2. Multi-Ancestry Characterization: While powered primarily for Europeans, the study provides the first large-scale proteomic characterization of APOE effects in African and South Asian populations, revealing both shared and ancestry-specific signals.
3. Decoupling of Age and Genotype: By stratifying by age, the study distinguishes between proteomic changes that are static (present from midlife) and those that emerge or intensify with aging, offering insights into the timeline of AD pathogenesis.
4. Platform Replication: Successful replication of key findings across two distinct proteomic platforms (Olink vs. SomaScan) in independent cohorts, validating the robustness of specific biomarkers.

4. Key Results

A. Established Neurodegeneration Biomarkers:

• GFAP (Glial Fibrillary Acidic Protein) & NEFL (Neurofilament Light):
  • $\vare4$ Carriers: Showed significantly elevated levels of both GFAP and NEFL compared to $\vare3/\vare3$, with effects increasing in older age groups (60–70). GFAP showed a significant age-related trend.
  • $\vare2$ Carriers: Showed lower NEFL levels in the youngest group (40–50), but this effect did not persist in older groups. No significant association with GFAP was found.

B. Proteins Encoded by AD Genetic Risk Loci:

• Overlap: 130 proteins were associated with both $\vare2$ and $\vare4$ carriage (often in opposing directions).
• Key Findings:
  • $\vare4$ Effects: Associated with lower levels of TREM2, GRN (Progranulin), IDUA, and SORT1 in older age groups. Notably, SORT1 levels reversed direction from higher in midlife to lower in late life.
  • $\vare2$ Effects: Associated with higher levels of TREM2 and GRN in younger groups, suggesting a potential neuroprotective mechanism that may attenuate with age.
  • Age Trends: Significant age-related trends were observed for IDUA, EPHA1, and MME in $\vare4$ carriers.

C. Hypothesis-Free Proteome-Wide Analysis:

• Discovery: Identified 480 proteins associated with $\vare4$ and 351 proteins associated with $\vare2$ (FDR-adjusted $P < 0.05$) in EUR ancestry.
• Consistent Signals:
  • $\vare4$ Specific: Lower APOE, PLA2G7, and ANGPTL3; Higher GFAP, HMOX2, IL32.
  • $\vare2$ Specific: Higher APOE, APOC1, LDLR.
  • Shared/Concordant: MENT and AGRP were elevated in both $\vare2$ and $\vare4$ carriers relative to $\vare3$.
• Cross-Ancestry Validation: APOE, MENT, and SNAP25 showed robust, concordant associations with $\vare4$ across EUR, AFR, and SAS groups.
• Age Trends: 25 proteins showed significant age-related trends, all observed in $\vare4$ comparisons. No strong age trends were found for $\vare2$-associated proteins.

D. Pathway Enrichment:

• $\vare4$ Carriers: Younger groups showed enrichment in lipid/lipoprotein metabolism. Older groups (60–70) shifted toward immune response, cell movement, and development pathways.
• $\vare2$ Carriers: Enriched for lipid metabolism in younger/middle age, shifting to cell-signaling and RNA-related pathways in older age.

E. Replication:

• In independent cohorts (INTERVAL/NSHD), 25 $\vare4$-associated and 26 $\vare2$-associated proteins replicated in the oldest age group (60–70).
• Discordance: Some proteins (e.g., NEFL, SNAP25, GRN) showed opposite effect directions in SomaScan data compared to Olink, likely due to platform-specific aptamer binding differences.

5. Significance and Implications

• Early Detection Window: The study demonstrates that APOE-associated proteomic perturbations are detectable decades before typical clinical AD diagnosis (starting in the 40s), highlighting a potential window for early intervention.
• Mechanistic Insights: The findings suggest that APOE $\vare4$ drives a progressive shift from lipid dysregulation to immune activation and neurodegeneration markers (GFAP, NEFL) as individuals age. Conversely, $\vare2$ appears to maintain higher levels of neuroprotective factors (e.g., TREM2, GRN) in midlife, which may wane in later life.
• Ancestry Considerations: The consistency of signals for APOE, MENT, and SNAP25 across ancestries suggests these are robust biomarkers, though sample sizes for non-European groups remain a limitation for full characterization.
• Clinical Translation: Identifying specific proteins (like PLA2G7 and MENT) that track with APOE genotype provides candidates for blood-based biomarkers to stratify risk in clinical trials and monitor therapeutic efficacy, potentially reducing reliance on invasive CSF sampling.

Limitations:

• Cross-sectional Design: Cannot establish individual aging trajectories; observed trends may reflect cohort effects.
• Platform Differences: Discrepancies between Olink and SomaScan highlight the need for standardized proteomic assays.
• Selection Bias: UK Biobank participants are generally healthier and wealthier than the general population, potentially underestimating disease severity.
• Power: Limited statistical power in non-European ancestry groups and East Asian populations.