Genome-Wide Meta-Analysis Identifies 80 Variants Across 74 Loci for Generalized Anxiety Symptom Severity
核心洞察
A genome-wide meta-analysis of 693,869 individuals of European ancestry identified 80 independent genome-wide significant variants across 74 loci associated with generalized anxiety disorder (搜索) symptom severity.
The strongest intragenic signal mapped to PCLO (搜索), a gene regulating presynaptic structure and neurotransmitter release, with additional support for SORCS3 (搜索) and CRHR1 (搜索).
SNP-based heritability was estimated at 5.90%, and polygenic risk scores explained 1.2–2.9% of variance in anxiety symptoms across European, African, and South Asian ancestry samples.
A landmark genome-wide association meta-analysis encompassing 693,869 individuals of European ancestry across 14 cohorts from 8 countries has identified 80 independent genome-wide significant variants at 74 loci associated with quantitatively measured generalized anxiety disorder (搜索) (GAD) symptom severity. The study, published in Nature, represents the largest genetic investigation of anxiety symptom dimensions to date and substantially expands the catalog of known genetic risk loci for this common mental health condition.
The top signal emerged from an intergenic locus near the long non-coding RNA RP4-598G3.1 on chromosome 1 (lead SNP rs7546305; P = 3.8 × 10⁻¹⁵; β = −0.014; 95% CI, −0.017 to −0.010), followed by a locus within an intron of PCLO (搜索) on chromosome 7 (lead SNP rs1476548; P = 4.9 × 10⁻¹⁵; β = −0.014; 95% CI, −0.017 to −0.010). Both loci have previously been associated with internalizing traits, including anxiety.
Of the 74 identified loci, 16 had no prior associations with internalizing trait GWAS—including neuroticism, anxiety, or depression—at any variant in linkage disequilibrium with the lead SNP. Thirty-nine loci were novel for anxiety specifically, whether measured as a diagnosis or symptom severity phenotype. Among the 58 loci previously identified in the PGC-ANX anxiety disorder study, all showed the same direction of association in the present analysis, with 19 (33%) reaching genome-wide significance and a further 33 (57%) showing nominal significance.
Heritability and Genetic Architecture
The SNP-based heritability estimate from SBayesRC was 5.90% (posterior s.d. = 0.146%). LDSC analysis indicated that genomic inflation was largely not attributable to confounding (intercept = 1.03; 95% CI, 1.01–1.05). Among the 6,012 genome-wide significant SNPs, between-study heterogeneity accounted for a moderate proportion of variance, with a median I² across cohorts of 29.7%. In contrast, the median I² for the 80 LD-independent genome-wide significant SNPs was 0%, indicating highly consistent effect size estimates across cohorts.
Genetic correlations between sufficiently powered cohorts, estimated using LD score regression, ranged from 0.64 to 0.97. The genetic correlation between the two most frequently used measures—GAD-7 and GAD-2—was high (rg = 0.86; 95% CI, 0.79–0.92; P = 9.6 × 10⁻¹⁶⁶).
Cross-Trait Genetic Correlations
The strongest genetic correlations were observed with quantitative internalizing traits, including neuroticism, depressive symptoms, and a genetic anxiety factor (rg = 0.83–0.86), and with case–control phenotypes for anxiety and depression (rg = 0.71–0.80). Moderate-to-strong estimates were also found for post-traumatic stress disorder (rg = 0.64), insomnia (rg = 0.49), irritable bowel syndrome (rg = 0.57), and chronic pain (rg = 0.56). Negative correlations were observed with socioeconomic status indicators including household income (rg = −0.43).
Conditional Analyses
Multi-trait conditional and joint analysis (mtCOJO) conditioning on neuroticism resulted in 9 remaining genome-wide significant SNPs across 2 loci. Conditioning on case–control anxiety yielded 66 significant SNPs across 6 loci, and conditioning on case–control depression produced 76 SNPs across 6 loci. Approximately half of SNP effect sizes were attenuated, and standard errors had a median inflation of 7–9%, suggesting the loss of significance was not solely attributable to reduced power.
Polygenic Risk Score Performance
Polygenic risk scores (PRS) derived from the meta-analysis demonstrated within- and cross-ancestry generalizability. The PRS significantly explained 2.9% of the variance in generalized anxiety symptom scores in an independent European ancestry sample (β = 0.55; 95% CI, 0.44–0.65; P = 8.9 × 10⁻²⁴), 1.4% in an African ancestry sample (β = 0.48; 95% CI, 0.28–0.67; P = 2.9 × 10⁻⁶), and 1.2% in a South Asian ancestry sample (β = 0.45; 95% CI, 0.27–0.64; P = 2.0 × 10⁻⁶). For case–control anxiety, assuming 20% prevalence, the PRS explained 1.8% of variance on the liability scale in European ancestry, 1.7% in African ancestry, and 1.4% in South Asian ancestry samples.
Functional Annotation and Gene Mapping
Functionally informed fine-mapping using PolyFun and SuSiE identified 4 putative causal variants with posterior inclusion probabilities ≥0.95. SNP-level gene annotation using FUMA highlighted PCLO (搜索), CRHR1 (搜索)—a key regulator of the stress response—and TMEM106B (搜索), which has been repeatedly associated with anxiety and depression. SORCS3 (搜索) was implicated by both positional and chromatin interaction mapping and has been reported in previous anxiety GWAS. Multiple genes previously implicated in depression (ERBB4, GRM7, VRK2, DCC, LRFN5, PCLO, NEGR1) and schizophrenia (ERBB4, VRK2, NEGR1) were also identified.
Gene-based association analysis using MAGMA identified 197 genes across 80 independent loci surpassing the Bonferroni-corrected significance threshold (P < 2.5 × 10⁻⁶). The top associated gene was PCLO (搜索) (P = 2.5 × 10⁻²¹).
Pathway and Tissue Enrichment
Six gene sets passed Bonferroni correction (P < 2.9 × 10⁻⁶): postsynaptic membrane (271 genes, P = 2.6 × 10⁻⁸), synaptic membrane (383 genes, P = 7.1 × 10⁻⁸), axon (627 genes, P = 5.3 × 10⁻⁷), neurogenesis (1,627 genes, P = 8.6 × 10⁻⁷), postsynapse (648 genes, P = 9.2 × 10⁻⁷), and generation of neurons (1,412 genes, P = 1.0 × 10⁻⁶). Gene–tissue expression analysis revealed significant enrichment in brain (P = 3.0 × 10⁻⁹) and pituitary (P = 4.9 × 10⁻⁵) tissues, with the strongest signals in the frontal cortex, cortex, cerebellum, anterior cingulate cortex, and nucleus accumbens.
Drug Target Analysis
DrugTargetor analysis did not reveal significant enrichment of anxiety associations for the 1,551 individual drug targets tested. However, at the drug class level, significant associations were identified for Anatomical Therapeutic Chemical classifications N06 'psychoanaleptics' (including N06A antidepressants (搜索)), N02A 'opioids (搜索)', and N06B 'psychostimulants' (Bonferroni-adjusted P < 0.05). The enrichment for psychostimulants likely reflects shared dopaminergic and noradrenergic pathways involved in arousal and vigilance.
The researchers note that the GAD-7—used by the majority (70%) of the sample—does not comprehensively assess all DSM-5 GAD symptoms, omitting sleep, fatigue, and concentration problems, and is not designed to capture fear-based anxiety or panic disorder. This limits generalizability across anxiety disorders. Additionally, the meta-analysis was restricted to cohorts of European ancestry, though PRS findings support a degree of shared genetic architecture with African and South Asian populations.
"Given the high and rising rates of anxiety, especially in young adults, it is more important than ever to improve our ability to identify and understand sources of risk," the authors state. "We hope our findings encourage genome-wide investigations leveraging existing but potentially underutilised anxiety severity data in genotyped cohorts, accelerating our progress in understanding the genetic architecture of anxiety."