Local ancestry modulates gene expression: shifting our understanding of genetic regulation and disease association within and across populations

This study introduces a new statistical framework demonstrating that local ancestry significantly modulates gene expression through ancestry-specific regulatory variants, thereby explaining cross-population inconsistencies in genomic studies and improving the understanding of disease risk and health disparities in Hispanic/Latino and African American populations.

The Gist

Imagine your DNA is a massive library of instruction manuals for building and running your body. For a long time, scientists thought these manuals were universal: if a specific sentence (a genetic variant) said "make more of this protein," it would do the same thing in everyone, regardless of who they were.

This paper argues that that's not true. Instead, the "instruction manual" changes depending on the neighborhood (ancestry) where that sentence is written.

Here is the breakdown of what the researchers found, using simple analogies:

1. The Problem: The "One-Size-Fits-All" Mistake

Most genetic studies have been done on people of European descent. Scientists built a "Master Map" (like the GTEx project) based on these books. They assumed that if a genetic switch turns a gene "on" in a European person, it does the same thing in an African American or Hispanic/Latino person.

The Analogy: Imagine you have a recipe for a cake. In the "European" kitchen, the recipe says "add 2 cups of sugar." The researchers assumed that if you took that exact same recipe card to a "Mexican" or "African" kitchen, it would still mean "add 2 cups of sugar."

But in reality, those kitchens have different ingredients, different ovens, and different traditions. The same instruction might mean something totally different, or even be ignored, depending on the kitchen it's in.

2. The Solution: The "Local Neighborhood" Detective

The researchers studied two groups of people with mixed backgrounds:

• Hispanic/Latino individuals: A mix of Indigenous American, European, and African ancestry.
• African American individuals: A mix of African and European ancestry.

Because these populations are "mixed," their chromosomes are like a patchwork quilt. One part of a chromosome might be from an Indigenous ancestor, while the next part is from a European ancestor.

The team created a new tool called ancQTL. Think of this tool as a super-magnifying glass that doesn't just look at the genetic sentence; it also checks the "neighborhood" (the local ancestry) where that sentence is sitting.

3. The Big Discovery: Context is King

When they used this new tool, they found that the neighborhood matters immensely.

• The "Switch" Effect: A genetic variant might act like a light switch that turns a gene "ON" when it's sitting in a European neighborhood. But if that exact same switch is sitting in an Indigenous or African neighborhood, it might turn the gene "OFF," or do nothing at all.
• The "Opposite Day" Effect: In some cases, the same genetic change had opposite effects. In one neighborhood, it increased the risk of a disease; in the other, it decreased it.

The Metaphor: Imagine a word like "Cool."

• In a "European" context, "Cool" might mean "Low Temperature."
• In an "African American" context, "Cool" might mean "Stylish."
• If you only study the word in the first context, you will be completely confused when you hear it used in the second. The word didn't change, but the context changed the meaning.

4. Why This Matters for Medicine

The paper shows that ignoring this "local neighborhood" effect causes major problems:

• Missed Cures: Scientists might miss important genetic causes of diseases (like Type 2 Diabetes or heart disease) in minority populations because they are looking for the wrong clues.
• Wrong Predictions: If we try to predict disease risk using the "European Master Map" for everyone, we might get it wrong for mixed-ancestry people.
• Better Precision Medicine: By understanding that a gene behaves differently in different ancestral neighborhoods, doctors can finally tailor treatments that actually work for everyone, not just people of European descent.

5. The "Open Window" Clue

The researchers also found that these "context-dependent" switches are often located in open chromatin regions.

• Analogy: Think of DNA as a closed book. Most of the time, the pages are glued shut. But in "open chromatin" regions, the book is open, and the wind (cellular machinery) can blow through the pages.
• They found that when the "wind" blows through these open windows, the ancestral background of the room determines how the pages flutter. This suggests that the physical structure of the DNA in these specific neighborhoods is what changes how genes are read.

The Bottom Line

This paper is a wake-up call. It tells us that genetics is not just about the letters (A, C, T, G); it's about the story those letters are telling in their specific cultural and ancestral context.

To truly understand human health and fix health disparities, we can no longer use a single map for everyone. We need to map the "neighborhoods" of our DNA to understand how our genes really work.

Technical Summary

1. Problem Statement

Current expression quantitative trait locus (eQTL) studies, which map genetic variants to gene expression levels, largely ignore the ancestral origin of the gene haplotype.

• Limitation of Current Models: Standard eQTL mapping typically uses additive allelic models that treat variants as independent of their haplotype background. This approach fails to account for local ancestry in admixed populations (e.g., Hispanic/Latino and African American), where chromosomes are mosaics of distinct ancestral segments (e.g., Native/Indigenous American [NIA], European [EUR], African [AFR]).
• Data Bias: Major reference datasets like GTEx are heavily skewed toward European ancestry (~85%), leading to poor portability of regulatory models to non-European populations.
• Consequence: Ignoring local ancestry can obscure differential regulatory effects, reduce statistical power to detect causal variants, and lead to inconsistent or incorrect interpretations of Genome-Wide Association Study (GWAS) findings across diverse populations.

2. Methodology

The authors introduced a novel statistical framework called ancQTL to map Local Ancestry-based eQTLs (LAB eQTLs).

• Statistical Model (ancQTL):
  • The model leverages Allele-Specific Expression (ASE) data.
  • It extends the standard allelic imbalance framework by incorporating three key terms:
    1. Genetic Variant (GV): The main effect of the variant.
    2. Genetic Local Ancestry (GLA): The main effect of the local ancestry background.
    3. Interaction Term (GV × GLA): Quantifies the difference in eQTL effect sizes between different ancestral backgrounds.
  • Equation Concept: The model estimates the eQTL effect in a reference ancestry and captures the effect in alternative ancestries via the sum of the GV coefficient and the interaction coefficient.
• Study Cohorts:
  • BHRC (Border Health Research Cohort): 962 Hispanic/Latino individuals (primarily Mexican American) from Brownsville, TX. Average ancestry: ~49% NIA, ~46% EUR, ~5% AFR.
  • WHI (Women's Health Initiative): 1,094 African American individuals. Average ancestry: ~76% AFR, ~23% EUR, ~1% NIA.
• Data Processing:
  • Bulk RNA-seq was performed on peripheral blood samples.
  • Filtering: Only haplotypes with unbroken local ancestry spanning the cis-regulatory region (within 1Mb of the TSS) were used to ensure accurate assignment of reads to specific ancestries.
  • Fine-Mapping: Three tools (SuSiE, DAP-G, CAVIAR) were applied to identify Putatively Functional Regulatory Variants (PFRVs) with a posterior inclusion probability (PIP) > 0.8.
• Validation:
  • Replication was performed in independent subsets of the BHRC and WHI cohorts.
  • Colocalization analyses were conducted with GWAS data for complex traits (Type 2 Diabetes, lipid traits, height, BMI) to assess biological relevance.

3. Key Contributions

1. Novel Framework (ancQTL): The first statistical method to explicitly model eQTL effects as a function of local ancestry using ASE data, allowing for the detection of ancestry-specific regulatory mechanisms.
2. First Ancestry-Specific Regulatory Landscapes: Generated the first comprehensive maps of LAB eQTLs for NIA, EUR, and AFR ancestries in admixed populations.
3. Demonstration of Heterogeneity: Provided robust evidence that regulatory effects are not uniform across ancestries; a significant proportion of eQTLs exhibit different effect sizes or even opposite directions depending on the local ancestry background.
4. Improved Disease Mapping: Demonstrated that LAB eQTLs significantly outperform traditional eQTLs (and GTEx-based maps) in colocalizing with GWAS signals for complex traits in admixed populations.

4. Key Results

A. Prevalence of Ancestry-Specific Effects

• BHRC (Hispanic/Latino):
  • Identified LAB eQTLs for 8,523 genes.
  • 24.3% of all LAB eQTLs showed significant local ancestry-specific effects (GV × GLA interaction FDR < 0.05).
  • 52% of the evaluated genes had at least one local ancestry-specific eQTL.
  • Found variants with opposite directions of effect (e.g., risk allele in EUR background, protective in NIA background) for 1,411 genes.
• WHI (African American):
  • Identified LAB eQTLs for 5,281 genes.
  • 45.7% of genes had at least one local ancestry-specific eQTL.
  • Significant heterogeneity was observed between AFR and EUR haplotypes.

B. Fine-Mapping and Functional Variants

• Fine-mapping identified PFRVs likely to be causal.
• BHRC: 41.9% of genes with PFRVs showed ancestry-specific differences in effect.
• WHI: 32.7% of genes with PFRVs showed ancestry-specific differences.
• Patterns: PFRVs fell into three categories:
  1. Functional in one ancestry only (e.g., EUR-only or NIA-only).
  2. Functional in both but with different effect sizes.
  3. Functional in both with opposite directions of effect.

C. Functional Annotation

• LAB eQTLs were enriched in promoters and enhancers.
• Crucially, LAB eQTLs with significant ancestry-specific effect differences were specifically enriched in open chromatin regions, suggesting that epigenetic landscape differences drive these regulatory variations.

D. Genetic Correlation and Replication

• Correlation with GTEx: EUR LAB eQTLs showed high genetic correlation with GTEx (up to 0.91 for highly heritable genes), confirming consistency within European backgrounds.
• Divergence: Correlation between EUR and NIA (or AFR) LAB eQTLs within the same population was significantly lower (e.g., ~0.64), highlighting distinct regulatory architectures.
• Replication: LAB eQTLs showed high replication rates (~48% in BHRC, ~41-48% in WHI). PFRVs showed exceptionally high enrichment in replication sets (Odds Ratios > 30).

E. Colocalization with GWAS

• LAB eQTLs significantly improved the colocalization of genetic signals with complex traits compared to GTEx-derived maps.
  • Type 2 Diabetes (Hispanic/Latino): 10 colocalized regions with LAB eQTLs vs. 1 with GTEx.
  • Lipid Traits: 58 colocalized regions with LAB eQTLs vs. 19 with GTEx.
  • Height (African American): 29 colocalized genes with LAB eQTLs vs. 4 with traditional eQTLs.

5. Significance and Implications

• Mechanistic Insight: The study provides a mechanistic explanation for inconsistencies in genomic studies across populations. It reveals that "missing heritability" or failed replication in admixed groups may stem from unmodeled local ancestry effects.
• Precision Medicine: Ignoring local ancestry can lead to incorrect risk predictions. Incorporating LAB eQTLs is essential for developing accurate Polygenic Risk Scores (PRS) and transcriptome-wide association studies (TWAS) for diverse populations.
• Health Disparities: By elucidating how regulatory variants function differently across ancestries, this work aids in identifying population-specific therapeutic targets and addressing health disparities.
• Future Directions: The authors advocate for the integration of local genomic context into all genetic studies and call for larger, diverse transcriptome datasets to further refine these models.

Conclusion: The paper fundamentally shifts the paradigm of gene regulation studies by proving that the ancestral context of a gene is a critical determinant of its expression. The ancQTL framework offers a powerful tool to decode the genetic architecture of complex traits in admixed populations, moving beyond the limitations of Eurocentric reference datasets.