Genetic Pathways Point to the Biology Underlying the Association Between Cannabis Use Disorder and Psychosis
Isabelle Austin-Zimmerman, Qiang Li, E. Johnson, Giulia Trotta, Edoardo Spinazzola, Jonathan R. I. Coleman, Zhi-Kun Li, Benjamin W. Bond, Diego Quattrone, Daniel F. Levey, Joel Gelernter, S. Jauhar, Robin Murray, Pak C. Sham, Evangelos Vassos, Marta Di Forti
Biological Psychiatry Global Open Science May 13, 2026 DOI: 10.1016/j.bpsgos.2026.100711 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Cross-disorder genome-wide association meta-analysis with Mendelian randomization and polygenic score analyses Peer reviewed |
|---|---|
| Sample size | 547,508 |
| Population | Participants in genome-wide association studies of schizophrenia and bipolar I disorder, plus a first-episode psychosis case-control sample (EU-GEI) |
| Topics | Cannabis |
| Key findings | The authors report that genetic liability for psychosis and cannabis use disorder are bidirectionally causally linked, with CUD-to-psychosis effects (β = 0.31) larger than psychosis-to-CUD effects (β = 0.19). They identified 553 psychosis loci, 122 novel, and three distinct causal clusters from CUD to psychosis. Glutamate signaling emerged as a key shared mechanism, and a glutamate polygenic score predicted psychosis status in both cannabis users and nonusers. |
Abstract
Background: Cannabis use is associated with increased risk of psychotic disorders, but the underlying biological mechanisms remain unclear. We used genetic data to investigate the molecular pathways contributing to the association between cannabis use disorder (CUD) and psychosis.
Methods: We conducted a cross-disorder meta-analysis of genome-wide association studies (GWASs) on schizophrenia and bipolar I disorder (N = 547,508) to define a combined-psychosis genetic liability. Using summary statistics from the latest cross-ancestry GWAS of CUD, we performed multitrait conditional analysis (psychosis conditioned on CUD), Mendelian randomization (MR), MR-clustering (MR-CLUST), and exploratory pathway-specific polygenic score (pPGS) analyses in the EU-GEI (European Network of National Schizophrenia Networks Studying Gene-Environment Interactions) first-episode psychosis case-control sample.
Results: We identified 553 independent genome-wide significant loci for psychosis, including 122 novel associations. More pathways were nominally associated with psychosis and CUD than expected by chance. MR analyses supported bidirectional causal effects (psychosis on CUD βIVW = 0.19, 95% CI, 0.15–0.22; CUD on psychosis βIVW = 0.31, 95% CI, 0.22–0.41). MR-CLUST identified 3 distinct clusters from CUD to psychosis, enriched for overlapping but functionally distinct pathways, including intracellular signaling, synaptic function, and neuronal development. In the EU-GEI sample, the glutamate pPGS consistently predicted psychosis status across the full sample and cannabis users and nonusers, explaining the most additional variance in each group.
Conclusions: Our findings highlight shared and distinct molecular pathways linking CUD and psychosis. Glutamate signaling emerged as a key mechanism across analysis strategies, while MR-CLUST revealed heterogeneous causal routes from cannabis use to psychosis. These insights may inform risk prediction and targeted interventions for cannabis-related psychosis.