Targeted Bayesian Prioritization Identifies Glutamatergic and Synaptic Plasticity Pathways as Key Contributors to Borderline Personality Disorder Risk

Background: Borderline personality disorder (BPD) combines pronounced affective lability with impulsive behaviour. Magnetic-resonance spectroscopy, post-mortem work and pilot pharmacological studies all point to abnormalities in glutamatergic transmission and synaptic plasticity. A recent genome-wide association study (GWAS) of European ancestry that excluded cases with bipolar disorder or schizophrenia showed subtle but consistent enrichment of risk signal in glutamate-related genes. Because those gene-set analyses could not pinpoint individual variants, we re-examined the same GWAS with a targeted bayesian prioritization approach. Methods: Summary statistics from the comorbidity-restricted GWAS (effective N ≈ 29,061) were analysed with SuSiE Bayesian Prioritization (modified fine-mapping approach). Four pre-specified gene sets were interrogated: (i) 23 core glutamatergic genes; (ii) an expanded panel of 130 glutamate/synaptic-plasticity genes; and two negative-control sets—(iii) monoaminergic genes and (iv) broadly expressed housekeeping genes. For each locus (gene ± 500 kb) we allowed up to ten causal variants, generated 95% credible sets and required minimum |r| ≥ 0.50 linkage between included SNPs. In the absence of a perfectly matched reference panel, linkage disequilibrium (LD) was approximated with an exponential distance-decay function. Results: Every one of the 33 loci in the two glutamatergic sets contained at least one single-nucleotide polymorphism (SNP) with posterior inclusion probability (PIP) ≥ 0.50. Ten genes—GRIA2, GRIN2B, GRIK2, SLC1A1, SIGMAR1, GRIA4, AKT1, ARC, GRIN1, and SHANK2—yielded variants with PIP = 1.00. By contrast, only six housekeeping or monoamine genes reached the 0.50 threshold. High-confidence signals clustered in AMPA-, NMDA- and kainate-receptor subunits, the glutamate transporter SLC1A1, the sigma-1-receptor modulator SIGMAR1, and key plasticity mediators (AKT1, ARC). Conclusions: Bayesian prioritization converts diffuse, sub-genome-wide association into discrete, high-probability causal variants that lie almost exclusively in glutamatergic and plasticity genes. Within a sample purged of major psychotic and bipolar comorbidity, common variation affecting glutamate receptors, transport and downstream signalling appears to carry disproportionate liability for BPD. The data reinforce mechanistic models of glutamatergic dysregulation in the disorder and support clinical development of NMDA/AMPA-modulating interventions.