This research paper, recently published in medRxiv, explores the genetic architecture of Prostate-Specific Antigen (PSA) levels. By looking at men without prostate cancer, the researchers identified 20 novel genes that influence baseline PSA levels, providing a clearer map of the genetic "noise" that can complicate cancer screening.
TL;DR
Researchers conducted a massive transcriptome-wide association study (TWAS) on nearly 96,000 men without prostate cancer to identify genes that regulate baseline PSA levels. They discovered 20 novel genes—including five with strong evidence of a direct causal link—that help explain why some men have naturally higher or lower PSA levels regardless of cancer status.
What Problem Does This Paper Address?
Prostate-Specific Antigen (PSA) is the primary biomarker used to screen for prostate cancer (PCa). However, PSA is a "noisy" signal. Levels can be elevated due to age, infection, or benign prostate enlargement, leading to frequent false positives and unnecessary, invasive biopsies.
We know that 30–45% of the variation in PSA levels is inherited. While previous Genome-Wide Association Studies (GWAS) identified 128 individual genetic variants associated with PSA, these only explain about 9% of the total variation. This paper aims to bridge that gap by moving from looking at single "letters" of DNA (variants) to looking at "sentences" (gene expression). By understanding the genetic baseline of PSA in healthy men, clinicians might eventually be able to "correct" a patient's PSA score based on their genetics, making screening far more accurate.
Methods
The team employed a Transcriptome-Wide Association Study (TWAS) using the MetaXcan framework.
Unlike a standard GWAS, which looks for correlations between single-nucleotide polymorphisms (SNPs) and a trait, a TWAS uses reference data to predict how a person's genetics will affect the expression (the activity level) of specific genes. They then correlate that predicted expression with PSA levels.
Key technical steps included:
- Tissue-Specific Modeling: They trained models using data from the Genotype-Tissue Expression (GTEx) project, focusing on whole blood and prostate tissue.
- Cross-Tissue Analysis (S-MultiXcan): They integrated signals across 45 different tissues to increase statistical power.
- Conditional Analysis (COJO): To ensure their findings were truly "novel," they statistically "subtracted" the effects of known GWAS variants to see which gene signals remained independent.
- Colocalization (fastENLOC): They used this to determine if the same genetic signal was responsible for both the change in gene expression and the change in PSA levels, strengthening the case for a causal relationship.
Datasets
The study utilized a massive multi-ancestry GWAS meta-analysis of 95,768 PCa-free men drawn from five major cohorts:
- UK Biobank
- Kaiser Permanente (GERA)
- The PLCO Cancer Screening Trial
- Vanderbilt University’s BioVU
- The Malmö Diet and Cancer Study
While the sample was primarily European (89.6%), it also included men of African, East Asian, and Hispanic/Latino ancestry. Gene expression models were trained using GTEx version 8 data.
Key Insights and Ideas
- The 4q27 Locus Discovery: One of the most significant "aha" moments was identifying a cluster of signals at the 4q27 genomic region. Previous GWAS had missed this area, but the TWAS highlighted two genes here—EXOSC9 and CCNA2—that appear to play a major role in PSA regulation.
- Constitutive vs. Cancerous PSA: By restricting the study to men without cancer, the researchers isolated "constitutive" PSA—the baseline level produced by healthy physiology.
- Pleiotropy: Some identified genes (like HIST1H2BN) are already linked to prostate cancer growth. This suggests that the same biological pathways that drive cancer might also regulate normal PSA production, making the two difficult to disentangle.
- Non-Cancer Genes: Conversely, six of the 20 novel genes have no known link to prostate cancer risk. These are the "pure noise" genes that could be used to adjust PSA screening thresholds without missing actual cancer signals.
Key Results
- 173 Total Genes: The study identified 173 genes whose predicted expression levels were significantly associated with PSA.
- 20 Novel Candidates: After rigorous filtering, 20 genes were identified as novel (not previously explained by GWAS).
- The "Top 5": Five genes showed high colocalization probability (>0.5), making them the strongest candidates for further functional study: EXOSC9, CCNA2, HIST1H2BN, RP11-182L21.6, and RP11-327J17.2.
- Biological Pathways: Pathway analysis pointed toward the renin-angiotensin-aldosterone system (typically associated with blood pressure) and extracellular matrix disassembly, suggesting PSA levels are influenced by broader systemic physiological processes.
Limitations and Future Work
The authors acknowledge several hurdles:
- Ancestry Bias: Nearly 90% of the participants were of European descent. Genetic signals can vary significantly across ancestries, so these results may not be fully generalizable to other populations.
- Cis-only Effects: The study only looked at cis-eQTLs (genetic variants near the gene they regulate), potentially missing "long-distance" (trans) genetic influences.
- Latent Disease: Although the men were "PCa-free" at the time of the study, some may have had undiagnosed, slow-growing cancer that influenced their PSA levels.
Future Directions: The next step is "wet lab" functional validation—actually manipulating these genes in cell lines to see exactly how they alter PSA production. Ultimately, this data could be used to develop polygenic transcriptome risk scores to personalize PSA screening.
Summary updated