Shared Genetic Architecture Between Kidney Function and Alzheimer Disease Across Ancestries

This multi-ancestry study reveals that while kidney function and Alzheimer's disease lack a genome-wide genetic correlation, they share specific pleiotropic loci—most notably APOE across ancestries and others like PICALM and EFTUD1 in Europeans—where distinct causal variants and mixed horizontal and vertical pleiotropy explain the complex kidney-brain genetic axis.

The Gist

The Big Picture: Two Different Rooms, One Shared Blueprint

Imagine the human body as a massive, complex house. For a long time, doctors and scientists treated the Brain (where Alzheimer's disease happens) and the Kidneys (which filter your blood) as two completely separate rooms. They thought, "If the kitchen sink is clogged, it doesn't mean the living room TV is broken."

However, we know from real-life observation that people with bad kidneys often get Alzheimer's earlier. Scientists have been trying to figure out why. Is it just bad luck? Do they share the same "blueprint" (genetics)?

This study is like a team of master architects who decided to look at the blueprints of thousands of houses to see if the Brain and the Kidney share any specific design plans.

The Main Discovery: It's Not the Whole House, It's Specific Rooms

The researchers started with a big question: "Do the genes for kidney health and Alzheimer's overlap everywhere in the genome?"

• The Old Way (The "Whole House" Scan): They first looked at the entire blueprint at once. The result? Nothing. It looked like the two rooms had totally different designs. If you tried to predict Alzheimer's risk just by looking at a person's general "kidney score," it didn't work.
• The New Way (The "Room-by-Room" Scan): The researchers realized that looking at the whole house at once was blurring the details. So, they zoomed in, room by room (locus by locus).
  • The Surprise: They found that while the whole house doesn't match, there are 16 specific "rooms" (genetic spots) where the blueprints for the Brain and Kidney are identical.
  • The Catch: In some of these rooms, the design helps both; in others, a design that helps the kidney might actually hurt the brain, and vice versa. This "tug-of-war" is why the big, whole-house scan showed zero connection.

The Star Player: The "APOE" Room

Out of all the rooms they checked, there was one famous room called APOE that was the same in both European and African ancestry groups.

• Think of APOE as the "Master Switch" in the house.
• One version of this switch (the ε4 allele) is like a faulty circuit breaker. It makes the brain more likely to short-circuit (Alzheimer's) and makes the kidneys less efficient at filtering.
• This was the only spot where the connection was consistent across different populations.

The Three Types of Connections

The researchers didn't just find matching rooms; they figured out how they were connected. They found three different types of relationships:

1. The "Domino Effect" (Vertical Pleiotropy)

• The Analogy: Imagine a leak in the roof (Kidney) that drips water onto the floor, eventually rotting the wood (Brain). The problem starts in one place and causes the other.
• The Science: At two specific spots (PICALM and EFTUD1), the genes seem to control how well the kidneys clean toxins. If the kidneys are bad, toxins build up and damage the brain. Here, fixing the kidney could theoretically help the brain.

2. The "Shared Contractor" (Horizontal Pleiotropy)

• The Analogy: Imagine a construction company (a gene) that supplies bricks to both the Roof and the Foundation. If the company sends bad bricks, both the roof and the foundation crumble. But the roof didn't cause the foundation to break; they just share the same bad supplier.
• The Science: At spots like CD2AP and SPI1, the genes control general body processes (like inflammation or cleaning up cellular trash). These processes affect both organs independently. Fixing the kidney won't necessarily fix the brain here because the root cause is a shared biological "contractor."

3. The "Opposite Directions" (Conflicting Signals)

• The Analogy: Imagine a thermostat. In one room, turning it up makes it hotter. In another room, turning it up makes it colder. If you look at the whole house, the average temperature change is zero, so you think the thermostat does nothing.
• The Science: For some genes, a variant that hurts the kidney might help the brain, or vice versa. When you average them all out, the signal cancels itself out, hiding the truth.

Why Does This Matter? (The "So What?")

1. Better Predictions: If you try to predict Alzheimer's risk using a "Kidney Score" that averages everything, you will fail. But if you look at the specific genes (like PICALM or EFTUD1), you might find new ways to predict who is at risk.
2. New Treatments: If we know that a specific gene causes kidney problems which then cause brain problems (The Domino Effect), we might be able to treat Alzheimer's by protecting the kidneys.
3. One Size Does Not Fit All: The study found that the "blueprints" are very different for people of European ancestry versus African ancestry. The only shared room was APOE. This means we can't use the same genetic tests for everyone; we need to tailor our medicine to the specific ancestry of the patient.

The Bottom Line

This study is like upgrading from a blurry, wide-angle photo of a house to a high-definition, zoomed-in inspection.

They discovered that the link between kidney disease and Alzheimer's isn't a vague, general connection. Instead, it's a specific, localized relationship happening in a few key spots in our DNA. Sometimes the kidney causes the brain issues; sometimes they share a common cause; and sometimes they fight each other.

By understanding these specific "rooms" in our genetic blueprint, we can finally start to fix the connection between the brain and the kidneys, leading to better treatments and predictions for both diseases.

Technical Summary

1. Problem Statement and Motivation

Epidemiological evidence consistently links chronic kidney disease (CKD) and reduced estimated glomerular filtration rate (eGFR) with an increased risk of Alzheimer's disease (AD). However, the underlying genetic mechanisms remain unclear. Previous genome-wide association studies (GWAS) have yielded inconsistent results, often showing weak or null global genetic correlations ($r_g \approx 0$) despite strong local biological signals.

Key challenges addressed in this study include:

• Masked Signals: Genome-wide averaging may obscure locus-specific effects where genetic variants exert opposing directions of effect on kidney function and AD risk.
• Ancestry Bias: Most prior studies focused on European (EUR) ancestry, despite African ancestry (AFR) populations bearing a disproportionate burden of both CKD and AD.
• Causality vs. Pleiotropy: It is unclear whether kidney dysfunction causally drives AD (vertical pleiotropy) or if both conditions share upstream biological pathways (horizontal pleiotropy).
• Bias: AD GWAS are susceptible to "competing risk bias" (survival bias), where individuals with severe kidney disease may die before developing AD, distorting genetic associations.

2. Methodology

The authors employed a comprehensive, multi-scale analytical framework integrating large-scale GWAS summary statistics from diverse ancestries.

Data Sources:

• AD: EUR (N≈462k, combining Kunkle et al. and Wightman et al.) and AFR (N=9,168, Ray et al.).
• Kidney Function (eGFR): EUR (N≈1.5M, meta-analysis of MVP, CKDGen, UK Biobank) and AFR (N≈145k, meta-analysis of Hughes et al., MVP, COGENT-Kidney).
• Bias Correction: Applied multivariable Mendelian Randomization (MRBEE) to correct AD effect sizes for competing mortality risks (e.g., heart failure, cancer) to mitigate collider bias.

Analytical Pipeline:

1. Global Genetic Correlation: Used LD Score Regression (LDSC) to estimate genome-wide $r_g$.
2. Polygenic Risk Scores (PRS): Calculated using SBayesRC to test if eGFR PRS predicts AD status.
3. Local Genetic Correlation: Used LAVA (Local Analysis of [co]Variant Association) to partition the genome into LD blocks and identify regional correlations.
4. Pleiotropy Detection: Applied Conjunctional False Discovery Rate (conjFDR) analysis to identify variants jointly associated with both traits, regardless of effect direction.
5. Fine-Mapping & Colocalization: Used SuSiE and coloc to distinguish whether overlapping signals arise from a single shared causal variant (vertical) or distinct variants in linkage disequilibrium (horizontal).
6. Causal Inference: Performed bidirectional cis-Mendelian Randomization (cis-MR) using cis-MRBEE and MAPLE to determine the direction of causality (kidney $\to$ brain vs. brain $\to$ kidney) at specific loci.

3. Key Results

A. Global vs. Local Architecture

• Global Null: No significant genome-wide genetic correlation was found between eGFR and AD in EUR ($r_g \approx -0.06, p > 0.1$) or AFR. PRS analysis confirmed that eGFR PRS does not predict AD status globally.
• Local Heterogeneity: Despite the global null, local genetic analysis revealed striking regional heterogeneity.
  • EUR: Identified 16 pleiotropic loci (conjFDR < 0.05) and 15 loci with significant local genetic correlations.
  • AFR: Identified 31 loci with significant local correlations, though many had wide confidence intervals due to sample size.
  • Ancestry Specificity: Only the APOE region was shared as a significant signal between EUR and AFR. Other loci were largely ancestry-specific, highlighting distinct genetic architectures.

B. Mechanistic Classification of Loci

The study categorized pleiotropic loci into distinct mechanistic classes based on colocalization and MR results:

1. Vertical Pleiotropy (Causal Mediation):

   • EFTUD1 (Chr15): Genetically predicted lower eGFR causally increases AD risk ($\beta = -0.97$). Suggests impaired kidney clearance of neurotoxic metabolites accelerates AD.
   • PICALM (Chr11): Genetically predicted lower eGFR causally decreases AD risk ($\beta = 0.71$). Suggests a complex, potentially protective relationship or distinct regulatory mechanisms.
   • Note: These loci showed opposing directional effects, explaining why global MR often fails.

2. Horizontal Pleiotropy (Shared Upstream Pathways):

   • CD2AP, MAT1A, SYMPK: Variants influence both traits via independent biological pathways (e.g., podocyte integrity, methylation, mRNA processing) without one trait mediating the other.
   • SPI1: Showed bidirectional signals, likely reflecting shared immune regulation (microglia/macrophages) affecting both organs.

3. True Biological Pleiotropy (Shared Causal Variant):

   • APOE: The $\epsilon$4 allele (rs429358) was the only variant identified with a shared causal variant driving both reduced eGFR and increased AD risk. It also showed evidence of reverse causality (AD liability affecting eGFR).

C. Ancestry Differences

• APOE Dominance: APOE was the sole overlapping significant locus between EUR and AFR.
• AFR Specificity: AFR analysis revealed unique signals on chromosomes 2, 3, 6, 8, 15, and 19 not detected in EUR.
• Directional Divergence: In AFR, positive local correlations (alleles raising eGFR associated with increased AD risk) predominated (64.5%), whereas EUR showed more negative correlations (53.3%).

4. Key Contributions

• Methodological Framework: Demonstrated that integrating conjFDR, local correlation (LAVA), colocalization, and cis-MR is essential for dissecting complex disease relationships that traditional genome-wide correlation misses.
• Resolution of Paradox: Explained the discrepancy between strong epidemiological links and null global genetic correlations: the signal is locus-specific with mixed directional effects (some loci increase risk, others decrease it), which cancel out at the genome-wide level.
• Ancestry Awareness: Provided the first large-scale multi-ancestry dissection of the kidney-brain axis, revealing that genetic risk models derived from EUR may not translate to AFR populations.
• Mechanistic Insight: Distinguished between vertical pleiotropy (kidney function causally influencing AD at specific loci like EFTUD1) and horizontal pleiotropy (shared pathways like inflammation or lipid metabolism).

5. Significance and Implications

• Clinical Translation: Polygenic risk scores (PRS) that aggregate eGFR variants genome-wide may fail to predict AD risk or even obscure it. Future models should utilize partitioned PRS focusing on vertically pleiotropic loci.
• Biomarker Interpretation: Routine adjustment for kidney function in blood-based AD biomarker studies may attenuate biologically meaningful signals if kidney dysfunction and neurodegeneration share causal pathways.
• Therapeutic Targets: The identification of EFTUD1 and PICALM as loci with causal links suggests that targeting kidney-related biological pathways (e.g., protein clearance, ribosome biogenesis) could offer novel therapeutic avenues for AD.
• Equity in Genetics: The findings underscore the critical need for ancestry-specific genetic models, as the shared genetic architecture is highly population-dependent, with APOE being the only universal signal.

Conclusion

This study redefines the kidney-brain axis not as a diffuse, genome-wide polygenic overlap, but as a collection of localized, ancestry-specific, and mechanistically diverse genetic connections. While most shared signals reflect horizontal pleiotropy, specific loci (EFTUD1, PICALM) provide evidence for causal pathways where kidney function directly modulates AD risk, offering new targets for intervention and risk stratification.