Endocannabinoid System Genes in Dementia: A Systematic Mendelian Randomization Study of Functional Variants and Tissue-Specific Expression Across Clinical, Molecular, and Neuroimaging Phenotypes

Abstract

BackgroundDementia is a leading cause of disability and death worldwide, and despite substantial research investment, disease-modifying treatments remain limited. The endocannabinoid system (ECS) has emerged as a promising candidate pathway, regulating key neurobiological processes implicated in dementia, including neuroinflammation, synaptic transmission, and cerebrovascular function. However, evidence linking ECS genetic variation to dementia risk in humans remains limited. MethodsWe conducted a systematic Mendelian randomization (MR) investigation of seven ECS genes (CNR1, CNR2, FAAH, DAGLA, DAGLB, NAPEPLD, MGLL) in relation to Alzheimers disease (AD), vascular dementia (VasD), and all-cause dementia. Primary analyses evaluated putatively functional variants within these genes, followed by complementary analyses of tissue-specific cis-expression quantitative trait loci (cis-eQTLs) from GTEx across brain and adipose tissues. Secondary outcomes included plasma and cerebrospinal fluid biomarkers and neuroimaging-derived phenotypes from the UK Biobank to explore potential biological mechanisms underlying dementia associations. Colocalization analyses were performed to assess evidence for shared causal variants. ResultsThe CNR2 missense variant rs4649124 was associated with increased vascular dementia risk (BH-adjusted P = 0.049) after correction at the 5% false discovery rate (FDR) threshold, with additional associations at the 10% FDR threshold for all-cause dementia risk (BH-adjusted P = 0.051) and lower circulating APP (BH-adjusted P = 0.088) and BDNF (BH-adjusted P = 0.083) levels. The FAAH missense variant rs324420 was associated with lower circulating BDNF levels (BH-adjusted P = 0.037) after multiple testing correction. The CNR1 missense variant rs12720071 was associated with reduced bilateral hippocampal volume (BH-adjusted P = 0.095), whereas the DAGLB functional variant rs1055430 showed no evidence of association with any of the outcomes examined (BH-adjusted P > 0.2). Tissue-specific cis-eQTL analyses identified associations between higher CNR1 cerebellar expression and increased Alzheimers disease risk (cerebellum: BH-adjusted P = 0.001; cerebellar hemisphere: BH-adjusted P = 0.043) and higher APP levels (cerebellar hemisphere: BH-adjusted P = 0.062), and between higher genetically predicted DAGLB expression in adipose and cerebellar tissues and increased all-cause dementia risk (visceral adipose: BH-adjusted P = 0.039; subcutaneous adipose: BH-adjusted P = 0.079; cerebellar hemisphere: BH-adjusted P = 0.079), higher BDNF levels (subcutaneous adipose: BH-adjusted P = 0.022), lower NfL levels (subcutaneous adipose: BH-adjusted P = 0.002; visceral adipose: BH-adjusted P = 0.011), and lower left and right amygdala volumes (subcutaneous adipose: BH-adjusted P = 2.24 x 10- and 0.026, respectively) and lower ventral striatal volumes (subcutaneous adipose: BH-adjusted P = 0.042). Higher cortical MGLL expression was additionally associated with increased all-cause dementia risk at the 10% FDR threshold (BH-adjusted P = 0.091). Colocalization analyses did not support shared causal variants for any of the observed associations. ConclusionsCommon ECS genetic variation is unlikely to exert a major causal effect on dementia susceptibility. However, several biologically plausible associations were identified, particularly involving CNR2, FAAH, CNR1, and DAGLB, suggesting that selected ECS components may influence neuroinflammatory, neurotrophic, cerebrovascular, and structural brain pathways relevant to dementia. Although colocalization analyses did not support shared causal variants, these findings highlight candidate neurobiological pathways that warrant further investigation in larger, well-powered genetic and functional studies.