A sex-specific genome-wide association study of blood lipid levels in All of Us

Using whole-genome sequencing data from 124,920 diverse participants in the All of Us Research Program, this sex-stratified genome-wide association study identified five genes with significant sex-heterogeneous effects on blood lipid levels, highlighting the importance of sex-specific genetic contributions to cardiovascular risk.

The Gist

Imagine your body is a bustling city. In this city, cholesterol and fats (lipids) are the delivery trucks moving energy and building materials around. Sometimes, these trucks get clogged in traffic (plaque in arteries), leading to heart attacks or strokes.

For a long time, scientists knew that the traffic patterns in this city looked different for men and women. Women often had a "safer" traffic flow before menopause, while men tended to have more congestion earlier in life. But researchers were missing a crucial piece of the puzzle: Why? Is it just because of hormones, or is there a different "instruction manual" (DNA) being used by men and women to run these trucks?

This paper is like a massive detective story where the authors went through the genetic blueprints of 125,000 people (from the "All of Us" research program) to find the specific switches in the DNA that control these lipid trucks differently for men and women.

Here is the breakdown of their findings using simple analogies:

1. The Big Picture: Two Different Cities

The researchers looked at four main types of lipid "trucks":

• LDL-C: The "Bad" trucks that dump trash (cholesterol) on the road, causing clogs.
• HDL-C: The "Good" trucks that sweep up the trash and take it away.
• Triglycerides (TG): The fuel tanks.
• Total Cholesterol: The total count of all trucks.

They found that while men and women share many of the same genetic "traffic rules," there are specific 5 unique switches that work differently depending on whether the driver is male or female.

2. The Five Special Switches (The New Discoveries)

The study found five genes that act like dimmer switches for these lipid trucks, but the switch only works (or works differently) in one gender.

• The "Male-Only" Switch (CELSR2):

  • What it does: This gene is like a specialized mechanic for the "Good" trucks (HDL).
  • The Finding: In men, a specific version of this gene makes the "Good" trucks work much better. In women, this same gene version barely does anything. It's as if the mechanic only shows up to fix the trucks in the men's city.
  • Why it matters: This suggests that men might have a unique genetic way to keep their arteries clean that women simply don't have.

• The "Female-Only" Switches (GPAM, PLTP, ZPR1, CMIP):

  • GPAM & PLTP: These are like "fuel injectors" for the "Good" trucks (HDL). In women, specific versions of these genes make the trucks run more efficiently. In men, these switches are essentially turned off.
  • ZPR1: This gene acts like a gatekeeper for the "Bad" trucks (LDL). In women, a specific version of this gene helps keep the "Bad" trucks from clogging the roads. Men don't seem to use this gatekeeper in the same way.
  • CMIP: This gene is like a manager for the "Fuel Tanks" (Triglycerides). In women, this gene helps regulate how much fuel is stored. In men, this regulation is different.

3. The "Age" Factor: The Menopause Shift

The study also looked at how these lipid levels change as people age.

• Before age 50: Women generally have better traffic flow (higher "Good" trucks, lower "Bad" trucks) compared to men. It's like women have a VIP lane that men don't.
• After age 50 (Menopause): That VIP lane disappears. Women's "Bad" truck levels rise, and they start looking more like the men's traffic patterns. The study confirms that this shift is real and happens across the board.

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

Imagine if a doctor prescribed a traffic solution based on a rulebook written only for men. It might work for men, but it could be useless or even confusing for women.

• Current Risk Tools: Most heart disease risk calculators just ask, "Are you male or female?" and then apply a generic formula.
• The New Insight: This paper says, "It's not just about being male or female; it's about which genetic switches you have turned on."
• The Future: In the future, doctors might look at your DNA to see if you have the "Male-Specific" or "Female-Specific" switches. This could lead to personalized medicine, where men and women get different treatments or different lifestyle advice based on their unique genetic traffic patterns.

Summary

Think of your body as a city with complex traffic. This study found that while men and women share the same roads, they use different genetic traffic controllers to manage the flow.

• Men have a unique controller (CELSR2) that boosts their "Good" trucks.
• Women have a unique set of controllers (GPAM, PLTP, ZPR1, CMIP) that manage their "Good" trucks, "Bad" trucks, and fuel tanks differently.

By understanding these differences, we can stop treating everyone the same and start building better, safer roads for everyone, regardless of their gender.

Technical Summary

1. Problem Statement

Despite well-documented sex differences in lipid metabolism and cardiovascular disease (CVD) risks—such as the earlier onset of ischemic heart disease in males and distinct lipid profiles in pre- vs. post-menopausal females—the genetic architecture underlying these sex-specific differences remains incompletely characterized.

• Gap in Literature: Previous Genome-Wide Association Studies (GWAS) on lipid profiles have largely been conducted on pooled cohorts or lacked sufficient sample sizes and diversity to rigorously detect sex-heterogeneous genetic effects.
• Limitations of Prior Work: Existing studies often suffer from unequal sex representation, insufficient power to detect interaction effects, or a failure to explore sex-specific genetic associations, leading to inconclusive findings regarding the genetic drivers of lipid disparities between males and females.

2. Methodology

The study utilized data from the All of Us Research Program, a large-scale, diverse US population cohort.

• Study Population:

  • Cohort: 124,920 participants (75,127 females, 49,793 males) with short-read whole-genome sequencing (srWGS) data and lipid measurements.
  • Ancestry: Predominantly European (67.1%) and African (17.3%), offering greater diversity than many previous biobanks.
  • Phenotypes: Four lipid profiles: Low-Density Lipoprotein Cholesterol (LDL-C), High-Density Lipoprotein Cholesterol (HDL-C), Triglycerides (TG), and Total Cholesterol (TC).
  • Data Cleaning: Measurements were filtered for unit consistency (mg/dL), exclusion of negative values, and exclusion of extreme outliers. To mitigate statin bias, the highest recorded LDL-C value for each participant was selected, assuming it preceded the initiation of lipid-lowering therapy.

• Genetic Analysis Pipeline:

  • Quality Control (QC): Performed using PLINK2 and REGENIE. Filters included Minor Allele Frequency (MAF) $\le$ 1%, Hardy-Weinberg Equilibrium (HWE) $p < 1 \times 10^{-15}$, and genotyping rate $< 99\%$. Linkage Disequilibrium (LD) pruning was applied ($r^2 < 0.9$).
  • GWAS Execution: Conducted using REGENIE v2.0.2 in three distinct strata:
    1. Pooled cohort (with sex as a covariate).
    2. Female-only cohort.
    3. Male-only cohort.
  • Statistical Modeling:
    • Lipid values underwent rank-based inverse normal transformation.
    • Covariates included age and the first 16 genetic principal components.
    • Significance threshold: $P < 5 \times 10^{-8}$ (genome-wide), with Bonferroni correction applied for testing four lipid traits.
  • Gene-Based Analysis: Used MAGMA via FUMA to map significant SNPs to genomic risk loci and annotate genes.
  • Sex-Heterogeneity Assessment: Compared effect sizes ($\beta$) between sexes. Significant sex-specific loci were identified where an association was present in one sex but absent or significantly weaker in the other.

3. Key Contributions

• Scale and Diversity: One of the largest sex-stratified GWAS for lipid profiles to date, leveraging the diverse ancestry of the All of Us cohort.
• Sex-Stratified Approach: Explicitly modeled genetic effects separately for males and females rather than relying on interaction terms in a pooled model, allowing for the detection of loci with effects exclusive to one sex.
• Novel Sex-Specific Loci: Identified five genes exhibiting significant sex-heterogeneous effects that were not previously characterized as sex-specific in lipid metabolism contexts.

4. Key Results

A. Phenotypic Observations

• Age-Sex Interaction: Consistent with literature, females had higher HDL-C and lower LDL-C in early adulthood. However, post-menopause (approx. age 50+), females exhibited a sharp rise in LDL-C and Total Cholesterol, surpassing males of the same age. Males consistently showed higher TG levels, particularly in younger age groups.

B. Genetic Findings (Pooled vs. Sex-Specific)

• Pooled GWAS: Validated known lipid-associated genes (e.g., LPL, ABCA1, CETP, PCSK9, APOE, LDLR) but did not identify novel loci compared to the GWAS Catalog.
• Sex-Specific GWAS:
  • Female-Specific Associations:
    • HDL-C: GPAM and PLTP showed significant associations in females but not males.
    • LDL-C: ZPR1 showed a significant association in females (specifically for the heterozygous genotype) but not males.
    • TG: CMIP showed a significant association in females.
  • Male-Specific Associations:
    • HDL-C: CELSR2 showed a strong, significant association in males that was absent in females.

C. Detailed Locus Analysis

1. CELSR2 (Male-specific, HDL-C):
   • Variant rs7528419 (chr1:109274570:A:G).
   • Effect: In males, the alternate allele was associated with lower HDL-C ($\beta \approx -0.045$). This effect was not observed in females.
   • Mechanism Hypothesis: Linked to Endoplasmic Reticulum (ER) stress regulation, which may exhibit sexual dimorphism due to testosterone levels.
2. CMIP (Female-specific, TG):
   • Variant rs2925979 (chr16:81501185:A:G).
   • Effect: In females, the alternate allele was associated with higher TG ($\beta \approx 0.037$).
   • Mechanism Hypothesis: Potential interaction with female hormones and adiponectin pathways, influencing fat distribution and insulin sensitivity.
3. GPAM & PLTP (Female-specific, HDL-C):
   • Both genes showed significant associations with HDL-C in females only, suggesting female-specific regulatory mechanisms in lipid synthesis and transport.
4. ZPR1 (Female-specific, LDL-C):
   • Significant association observed in females carrying one alternate allele, but no dosage effect in males.

5. Significance and Implications

• Precision Medicine: The findings underscore that genetic risk scores for cardiovascular disease should ideally be sex-specific. A variant may be a strong risk factor for dyslipidemia in one sex but neutral in the other.
• Biological Insight: The identification of CELSR2 as male-specific and CMIP/GPAM/PLTP/ZPR1 as female-specific provides new targets for investigating the molecular mechanisms of sex hormones (estrogen/testosterone) on lipid metabolism.
• Clinical Risk Assessment: Current risk calculators (e.g., PREVENT, PCE) account for sex as a demographic variable; this study suggests that genetic risk stratification also requires sex-specific modeling to accurately predict cardiovascular events.
• Limitations & Future Directions: The study acknowledges limitations regarding calculated LDL-C (Friedewald equation) and missing statin usage end-dates. Future research should focus on functional validation of these sex-specific loci and the integration of more granular medication history data.

In conclusion, this study demonstrates that a substantial portion of the genetic architecture governing lipid metabolism is sex-dependent, necessitating a shift from pooled analyses to sex-stratified genomic evaluations for accurate cardiovascular risk prediction.