Abstract
Although our understanding of Alzheimer’s disease (AD) has greatly improved in recent years, the root cause remains unclear, making it difficult to find effective diagnosis and treatment options. Our understanding of the pathophysiology underlying AD has benefited from genomic analyses, including those that leverage polygenic risk score (PRS) models of disease. In many aspects of genomic research the use of functional annotation has been able to improve the power of genomic models. Here, we leveraged genomic functional annotations to build tissue-specific PRS models for 13 tissues and applied the scores to two longitudinal cohort studies of AD. The PRS model that was most predictive of AD diagnosis relative to cognitively unimpaired participants was the liver tissue score: n = 1,116; odds ratio (OR) (95% confidence interval [CI]) = 2.19 (1.70-2.82) per standard deviation (SD) increase in PRS; P = 1.46 × 10⁻⁹. After removing the APOE locus from the PRS models, the liver score was the only PRS to remain statistically significantly associated with AD diagnosis after multiple testing correction, although the effect was weaker: OR (95% CI) = 1.55 (1.19-2.02) per SD increase in PRS; P = 0.0012. In follow-up analysis, the liver PRS was statistically significantly associated with levels of amyloid (P = 3.53 × 10⁻⁶) and tau (P = 1.45 × 10⁻⁵) in the cerebrospinal fluid (CSF) (when the APOE locus was included) and nominally associated with CSF soluble TREM2 levels (P = 0.042) (when the APOE locus was excluded). These findings provide further evidence of the role of the liver-functional genome in AD and the benefits of incorporating functional annotation into genomic research.
Introduction
Alzheimer’s disease (AD) has traditionally been viewed through a "brain-first" lens, focusing on the accumulation of amyloid plaques and tau tangles within neurons. However, the root causes of AD remain elusive, and current treatments offer limited efficacy. Emerging evidence suggests that AD is a systemic disease involving complex interactions between the immune system, the cardiovascular system, and metabolic processes.
Standard Polygenic Risk Scores (PRS)—which aggregate the effects of thousands of single nucleotide polymorphisms (SNPs)—typically treat the entire genome as a monolith. This approach can obscure important signals from specific organs. This study sought to determine if filtering the genome for "functional" regions—areas known to be active in specific tissues—could provide a clearer picture of AD risk and reveal which organs, beyond the brain, are driving the disease.
Methods
The researchers employed a technique called functional annotation using the GenoSkyline-PLUS framework. This tool uses epigenetic data to predict which parts of the DNA are active in specific tissues.
- Tissue-Specific Models: PRS models were constructed for 13 different tissues (e.g., brain, liver, heart, lung) by including only SNPs located in the functional regions of those tissues.
- Scoring and Validation: These models were applied to genetic data from two longitudinal studies. To account for the dominant effect of the APOE gene, a second set of analyses was performed with the APOE region removed.
- Biomarker Correlation: The scores were cross-referenced with cerebrospinal fluid (CSF) biomarkers, including amyloid-beta and tau, to determine if genetic risk translated into biological changes.
Study Population
The study utilized data from two major longitudinal cohorts based at the University of Wisconsin-Madison: the Wisconsin Registry for Alzheimer’s Prevention (WRAP) and the Wisconsin Alzheimer’s Disease Research Center (WADRC). The final analysis included participants of European ancestry categorized by clinical diagnosis.
As shown in the table above, the cohort included 79 AD cases, 48 individuals with mild cognitive impairment (MCI), and 1,037 cognitively unimpaired controls. The WADRC cohort provided the majority of AD cases, while the WRAP cohort consisted primarily of cognitively unimpaired individuals.
Key Results and Insights
The Liver Outperforms the Brain
Surprisingly, the liver-specific PRS was more strongly associated with AD diagnosis than the brain-specific PRS or the standard genome-wide PRS. With APOE included, the liver PRS showed an Odds Ratio (OR) of 2.19 (p = 1.46 x 10⁻⁹), meaning the odds of an AD diagnosis more than doubled for every standard deviation increase in the liver risk score.
Independence from APOE
While most AD genetic signals vanish once the APOE locus is removed, the liver PRS remained statistically significant (OR = 1.55, p = 0.0012). It was the only tissue-specific score to survive multiple testing correction without APOE. This suggests that the liver's contribution to AD involves a broad set of metabolic pathways beyond just the APOE gene's activity.
Biomarker Links
The liver PRS was significantly associated with classic AD markers in the CSF: lower amyloid-beta (indicating increased plaque deposition in the brain) and higher tau (indicating neuronal damage). It was also nominally associated with soluble TREM2 levels, a marker of neuroinflammation.
Discussion and Future Directions
These findings highlight the liver as a metabolic hub that may influence AD risk through cholesterol regulation and other metabolic traits. The study suggests that a "weak" liver genetic profile may lead to systemic metabolic imbalances that eventually manifest as neurodegeneration.
Primary limitations include the relatively small number of confirmed AD cases (79) and the focus on individuals of European ancestry. Future research should expand this functional PRS approach to larger, more diverse datasets to investigate the causal direction of these associations and evaluate the potential of the liver PRS as a clinical screening tool.
Summary updated