Population-scale burden analysis of rare damaging coding variants identifies novel risk genes for Alzheimer's disease and Parkinson's disease

By analyzing rare damaging coding variants in large-scale population biobanks and disease cohorts, this study confirmed known genetic risks and identified novel risk genes for Alzheimer's and Parkinson's diseases, with the strongest novel signal found in ANKRD27.

The Gist

Imagine your body is a massive, bustling city. The roads are your nerves, the traffic lights are your brain signals, and the delivery trucks are the proteins that keep everything running smoothly. Sometimes, in this city, a few delivery trucks break down or take the wrong route. When this happens, it can lead to traffic jams that eventually cause the city to shut down. This is essentially what happens in Alzheimer's and Parkinson's diseases.

For a long time, scientists have been looking at the "blueprints" of this city (our DNA) to find out why these traffic jams happen. They found that some common, small errors in the blueprint (common genetic variants) make the city slightly more prone to traffic. But they knew there were also some rare, catastrophic errors—like a truck completely missing a wheel or a driver who never shows up—that cause much bigger problems. The problem was, these rare errors are so rare that finding them was like trying to find a specific needle in a haystack the size of a mountain.

The Big Hunt: Searching the Haystack

In this new study, the researchers decided to stop looking at just a few haystacks and instead looked at the entire mountain. They combined data from massive databases (like the UK Biobank and "All of Us") containing the genetic blueprints of nearly 700,000 people.

To make their search even smarter, they used a clever trick called "proxy phenotypes."

• The Analogy: Imagine you want to find people who have a specific rare disease, but you don't have their medical records. However, you do know that their parents or siblings had that disease. In genetics, having a family member with the disease is a strong "proxy" (a stand-in) that suggests you might carry the same rare, broken blueprint, even if you haven't developed the disease yet. By using this "family history" clue, they could cast a much wider net and find the rare needles they were looking for.

What They Found: The Broken Trucks

By scanning this massive amount of data, they successfully identified two types of findings:

1. Confirming the Suspects (The Known Bad Guys)
They found the rare, broken blueprints in genes they already suspected were troublemakers.

• For Alzheimer's: Genes like TREM2 and SORL1 were confirmed. Think of these as the "famous criminals" the police had been chasing for years.
• For Parkinson's: Genes like GBA1 and LRRK2 were confirmed. These are the known troublemakers in the Parkinson's city.

2. Catching New Culprits (The Novel Suspects)
This is the exciting part. They found new genes that were previously unknown to be involved in these diseases.

• For Alzheimer's: They found genes like IMPA2 and PMM2.
  • Analogy: If the city's traffic lights were the problem, IMPA2 is like a new discovery that the power grid supplying electricity to those lights is also faulty. It suggests that how the brain handles energy and chemicals (inositol signaling) is a key part of the puzzle.
• For Parkinson's: They found genes like ANKRD27 and USP19.
  • Analogy: ANKRD27 is like a traffic controller that helps trucks recycle their cargo. The study found that when this controller is broken, the trucks get stuck, and waste piles up in the city. This waste is similar to the "Lewy bodies" (clumps of protein) that kill brain cells in Parkinson's.

The "Smoking Gun": ANKRD27

The strongest new discovery was a gene called ANKRD27.

• The Story: The researchers noticed that the broken parts of this gene were all clustered in a specific area, like a group of saboteurs all attacking the same bridge.
• The Bridge: This bridge connects to a system called the "retromer," which is the city's recycling plant. When this bridge is broken, the recycling plant stops working, and toxic trash builds up, destroying the city. This gives scientists a very clear new target for how to fix the problem.

Why This Matters

Think of this study as a massive upgrade to the city's maintenance manual.

1. It proves the method works: It shows that by looking at huge groups of people and using family history as a clue, we can find the "needles" that were previously invisible.
2. It opens new doors: Instead of just looking at the obvious traffic jams, we now know to check the power grid (IMPA2), the recycling plants (ANKRD27), and the waste disposal systems (USP19).
3. Future Treatments: Now that we know which trucks are broken and where they are broken, drug companies can start designing tools to fix them. Maybe a drug can boost the recycling plant, or another can fix the power grid.

The Bottom Line

This paper is like a detective story where the detectives finally found the hidden culprits behind two of the most devastating diseases of our time. By using a giant magnifying glass (huge data sets) and a clever trick (family history), they didn't just confirm who the bad guys were; they found brand new suspects and figured out exactly how they are breaking the city. This gives hope that in the future, we can repair the blueprint and keep the city running smoothly.

Technical Summary

1. Problem Statement

While Genome-Wide Association Studies (GWAS) have successfully mapped common genetic variants contributing to Alzheimer's Disease and Related Dementias (ADRD) and Parkinson's Disease and Related Disorders (PDRD), the role of rare, damaging coding variants remains incompletely characterized at the population scale.

• Limitations of previous studies: Rare variant discovery has historically been constrained by small sample sizes, heterogeneity in sequencing methods (e.g., varying exome capture kits), and the statistical power required to detect genes with low minor allele counts (MAC).
• The Gap: There is a need for large-scale, harmonized Whole-Genome Sequencing (WGS) analyses that leverage proxy phenotypes (family history) to maximize statistical power for late-onset neurodegenerative diseases without the biases of heterogeneous exome data.

2. Methodology

The authors performed a massive, multi-cohort gene-based burden analysis using a unified analytical framework.

Data Sources & Cohorts:
The study integrated WGS data from four major sources, totaling over 870,000 participants:

1. UK Biobank (UKB): N = 484,909 (WGS data).
2. All of Us (AoU): N = 222,274 (WGS data).
3. Alzheimer's Disease Sequencing Project (ADSP): N = 41,991 (Disease-focused WGS).
4. Accelerating Medicine Partnership in Parkinson's Disease (AMP-PDRD): N = 10,204 (Disease-focused WGS).
5. Holstege et al. (2022): Summary statistics from a previous exome study (N = 21,345) were meta-analyzed.

Phenotype Definition:

• Proxy Phenotypes: To increase power in population biobanks (UKB, AoU), the authors utilized "proxy cases" defined by a personal diagnosis or a first-degree family history of the disease. This approach is known to increase statistical power for late-onset conditions.
• Clinical Cases: Disease-focused cohorts (ADSP, AMP-PDRD) used clinically adjudicated diagnoses (biomarker or autopsy confirmed where available).

Variant Filtering & Annotation:

• Variants: Restricted to rare (Minor Allele Frequency < 1%) protein-coding variants.
• Types: Loss-of-Function (LoF) and in silico predicted deleterious missense variants.
• Deleteriousness Sets: Variants were grouped into four nested sets to test sensitivity:
  1. LoF only.
  2. LoF + Missense (REVEL score ≥ 0.75).
  3. LoF + Missense (REVEL score ≥ 0.50).
  4. LoF + Missense (REVEL score ≥ 0.25).
• QC: Variants were filtered using gnomAD v4.1 PASS flags, Hardy-Weinberg equilibrium, and cohort-specific genotype quality metrics (e.g., depth, allelic fraction).

Statistical Analysis:

• Burden Testing: Performed using REGENIE (v4.1), a two-step whole-genome regression method. Step 1 trained ridge regression models on common variants; Step 2 performed gene-based burden tests on rare variants.
• Meta-Analysis: Results were combined across cohorts using inverse-variance-weighted fixed-effects models.
• Significance Thresholds: Study-wide significance was determined via Bonferroni correction adjusted for the effective number of independent tests ($M_{eff}$) across the four deleteriousness sets.
  • ADRD Threshold: $P < 1.331 \times 10^{-6}$.
  • PDRD Threshold: $P < 1.342 \times 10^{-6}$.

3. Key Contributions

• Scale: This is one of the largest WGS burden analyses for neurodegenerative diseases to date, leveraging >870k individuals.
• Methodological Integration: Successfully combined population biobanks (using proxy phenotypes) with deep-phenotype disease cohorts to validate rare variant signals.
• Resource Creation: Developed an interactive Shiny app (adpd_burden) allowing researchers to query gene-level and variant-level burden signals, carrier counts, and effect estimates.

4. Key Results

A. Alzheimer's Disease and Related Dementias (ADRD)

• Replication: Confirmed rare variant burdens in established genes: TREM2, SORL1, ABCA7, GRN, PSEN1, ATP8B4.
  • TREM2: The strongest signal (OR=1.44), driven largely by the R47H variant.
  • APOE: Interestingly, the burden test showed a protective effect driven by rare protective alleles (R269G, V254E).
• Novel Associations (Study-Wide Significant):
  1. ADAM10: (OR=3.22 in European ancestry). Previously implicated via functional evidence but not significant in prior burden studies.
  2. IMPA2: (OR=2.22). Implicates inositol signaling and phosphoinositide metabolism; a potential link to lithium therapy mechanisms.
  3. PMM2: (OR=1.12). Highlights protein glycosylation pathways.
  4. SYNE1: (OR=1.13). Connects to nuclear-cytoskeletal coupling.
  5. CHRNA4: Showed a protective association (OR=0.84), suggesting damaging missense variants in this nicotinic receptor subunit may reduce risk.
  6. FCGR1A: Near-threshold signal involving a stop-gain variant (R92*).

B. Parkinson's Disease and Related Disorders (PDRD)

• Replication: Confirmed associations in GBA1 and LRRK2.
• Novel Associations (Study-Wide Significant):
  1. ANKRD27 (VARP): The strongest new signal (OR=1.21).
     • Mechanism: Variants cluster in domains mediating interactions with Rab GTPases and the retromer complex. This aligns with the endolysosomal recycling pathway, a known mechanism in PD (e.g., RAB32).
     • Validation: The lead variant (R21C) showed independent support in the GP2 clinical-only meta-analysis.
  2. CCL7: (OR=5.74). Implicates chemokine signaling and immune cell recruitment.
  3. SKP1: (OR=9.39). Implicates the SCF ubiquitin-proteasome system.
  4. USP19: (OR=4.03). A deubiquitinase linked to $\alpha$-synuclein handling and protein quality control.
  5. KANSL3: Near-threshold signal; links to chromatin regulation and the NSL complex.

5. Significance and Implications

• Pathway Expansion: The study moves beyond canonical lysosomal genes to identify novel biological axes:
  • ADRD: Metabolic regulation (inositol/lithium), glycosylation, and nuclear-cytoskeletal integrity.
  • PDRD: Vesicular recycling (retromer), immune recruitment, and ubiquitin-mediated proteostasis.
• Therapeutic Targets: The identification of genes like IMPA2 (lithium target) and USP19 (alpha-synuclein handling) provides immediate candidates for drug repurposing or novel therapeutic development.
• Protective Variants: The discovery of protective rare variants in APOE and CHRNA4 highlights the potential for "loss-of-function" therapies that mimic natural protection.
• Limitations: The authors note that proxy phenotypes may introduce heterogeneity and dilute effect sizes. Additionally, the study focused on coding variants, leaving structural and non-coding rare variants for future investigation.

Conclusion:
This study demonstrates that combining population-scale WGS with proxy phenotypes is a powerful strategy for dissecting the rare variant architecture of neurodegenerative diseases. It validates known risk genes and robustly identifies novel candidates that implicate specific, actionable biological pathways in both Alzheimer's and Parkinson's disease.