Multi-Ancestry Genome-Wide Association Study Reveals New Loci for Prostate-Specific Antigen Levels and Improves Prediction Across Ancestral Groups

Research Background and Problem Statement

Prostate-specific antigen (PSA) is a protein secreted by the prostate gland, commonly used for prostate cancer screening. However, PSA levels are influenced not only by prostate cancer but also by benign prostatic hyperplasia (BPH), local inflammation or infection, prostate volume, age, and genetic factors. Therefore, while PSA screening was approved by the US Food and Drug Administration (FDA) in 1994 for prostate cancer detection, there is ongoing debate regarding whether its benefits in reducing prostate cancer-specific mortality outweigh the harms from overdiagnosis and overtreatment.

Studies indicate that approximately 20% to 60% of screen-detected prostate cancers are overdiagnosed, meaning these cancers would not clinically manifest or lead to prostate cancer-related death. Additionally, it is estimated that 229 individuals need to be invited for screening and 9 diagnosed with prostate cancer to prevent one death. Consequently, countries like the United States, Canada, and the United Kingdom do not recommend universal population-based screening.

To improve the specificity and sensitivity of PSA screening, thereby reducing overdiagnosis and preventing more deaths, adjusting PSA for individual predispositions in the absence of prostate cancer becomes crucial. Twin studies estimate PSA heritability at 40% to 45%, and genome-wide evaluations suggest a heritability of 25% to 30%. This indicates that incorporating genetic factors into PSA screening could enhance its effectiveness.

Paper Source

This paper was authored by Thomas J. Hoffmann, Rebecca E. Graff, and others from multiple institutions, including the University of California, San Francisco, Stanford University School of Medicine, Argonne National Laboratory, etc. The paper was published in Nature Genetics with DOI: 10.1038/s41588-024-02068-z.

Research Process and Methods

Study Population and Sample Size

The study included 296,754 men from nine cohorts, comprising 211,342 of European ancestry (EUR), 58,236 of African ancestry (AFR), 23,546 of Hispanic/Latino ancestry (HIS/LAT), and 3,630 of Asian ancestry (ASN). An additional 95,768 independent individuals were used for replication analysis.

Data Processing and Quality Control

Participants were genotyped using conventional genome-wide association study (GWAS) arrays and imputed using various reference panels. Standard genotype and individual-level quality control procedures were implemented across each ancestral group, including removing low-quality genotypes, low-frequency variants, and variants not in Hardy-Weinberg equilibrium.

Genome-Wide Association Analysis

Researchers conducted linear regression analyses within each ancestral group, using log-transformed PSA as the dependent variable and additive genotypes as the independent variable, adjusting for age, genetic principal components, and other covariates. Meta-analyses were performed using inverse-variance weighted fixed effects models to identify independently associated genome-wide significant (p ≤ 5 × 10^-8) variants. To ensure new discoveries, researchers required new variants to have linkage disequilibrium (LD) less than 0.01 with previously reported variants.

Joint Meta-Analysis

By combining the discovery cohort (n = 296,754) with the replication cohort (n = 95,768), researchers analyzed a total of 392,522 individuals. The joint meta-analysis revealed 447 independent genome-wide significant variants, including 111 newly discovered variants.

Polygenic Risk Scores (PRS)

Polygenic risk scores (PRS) were constructed to evaluate different strategies for explaining PSA variance. PRS performance was assessed in four independent cohorts: Kaiser Permanente’s GERA, SELECT trial, PCPT trial, and All of Us project.

Key Findings

Newly Discovered Variant Loci

The study identified 447 independent genome-wide significant variants, including 111 newly discovered variants. These variants were predominantly found in individuals of European ancestry but also included unique variants in other ancestries. For example, 8 genome-wide significant variants were identified in African ancestry individuals, with only 2 common enough to be evaluated in European ancestry.

Variance Explained by PSAR Variants

Polygenic risk scores explained varying amounts of PSA variance across ancestries. For European ancestry individuals, PRS explained 11.6% to 16.9% of PSA variance; for African ancestry individuals, PRS explained 5.5% to 9.5%; for Hispanic/Latino ancestry individuals, PRS explained 13.5% to 18.6%; and for Asian ancestry individuals, PRS explained 8.6% to 15.3%. PRS predictive performance decreased with increasing age.

Genetically Adjusted PSA Levels

Genetically adjusted PSA levels showed stronger associations with overall and aggressive prostate cancer in midlife compared to unadjusted PSA levels. For instance, in European ancestry individuals, genetically adjusted PSA levels had an odds ratio (OR) of 3.92 for aggressive prostate cancer, compared to an OR of 3.46 for unadjusted PSA levels. In African ancestry individuals, genetically adjusted PSA levels had an OR of 5.39 for aggressive prostate cancer, compared to an OR of 4.72 for unadjusted PSA levels.

Conclusions and Implications

Scientific and Practical Value

Through a large-scale multi-ancestry genome-wide association study, this research identified 447 genome-wide significant variants related to PSA levels, including 111 newly discovered variants. These findings enhance our understanding of the genetic basis of PSA and improve the accuracy of genetic adjustment of PSA, particularly in African ancestry individuals who bear the highest burden of prostate cancer incidence and mortality.

Research Highlights

1. Newly Discovered Variant Loci: The study identified 111 new loci that exhibit different effect sizes across ancestries, contributing to the genetic architecture of PSA.
2. Multi-Ancestry Polygenic Risk Scores: PRS performance varied across ancestries, notably excelling in Hispanic/Latino ancestry individuals, underscoring the importance of studying diverse populations.
3. Genetically Adjusted PSA Levels: Genetically adjusted PSA levels in midlife showed stronger associations with aggressive prostate cancer, providing a scientific basis for personalized PSA screening.

This study marks a significant step towards leveraging genetic information for personalized PSA screening and significantly improves our understanding of PSA across different ancestries. Future research should continue to explore genetic factors specific to constituents of PSA to further optimize screening strategies.