Multi-voxel MR-Spectroscopy Signatures and Associations with EEG Network Hyperexcitability and Clinical Symptomatology in Borderline Personality Disorder

Andrea Schlump¹Bernd Feige¹Swantje Matthies¹Katharina von Zedtwitz¹Isabelle Matteit¹Kathrin Nickel¹Katharina Domschke¹Marco Reisert¹Thomas Lange¹Markus Heinrichs²Dominique Endres¹Ludger Tebartz van Elst¹Simon Maier^1*

• ¹University of Freiburg Medical Center, Freiburg, Germany
• ²Department of Psychology, Laboratory for Biological Psychology, Clinical Psychology, and Psychotherapy, University of Freiburg, Freiburg, Germany

Introduction: Previous studies have reported altered gamma-aminobutyric acid (GABA) and glutamate levels in borderline personality disorder (BPD), suggesting disruptions in excitatory-inhibitory neurotransmission. Electroencephalographic (EEG) research has indicated potential network hyperexcitability in BPD, evidenced by increased intermittent rhythmic delta and theta activity (IRDA/IRTA), which may reflect compensatory stabilization mechanisms. This study used multi-voxel magnetic resonance spectroscopic imaging (MRSI) to explore neurochemical abnormalities and their relationships with IRDA/IRTA, psychometric and neuropsychological measures. Methods: Sixty-six female patients diagnosed with BPD (mean age: 30.2 ± 9.7 years) and 29 age-matched female healthy controls (mean age: 27.8 ± 8.0 years) received spirally encoded 3D MRSI scans. GABA, glutamate plus glutamine (Glx), total creatine (tCr) and total N-acetylaspartate (tNAA) were quantified and reported as ratios relative to tNAA and/or tCr. The resulting spectroscopic images were analyzed using LCModel and FreeSurfer software, and IRDA/IRTA detection in a clinical EEG session was performed using independent component analysis. Metabolite ratios were analyzed using hierarchical linear mixed-effects models (ROIs nested within participants), with fixed effects of group or, in separate models, continuous predictors (IRDA/IRTA and psychometric/neuropsychological measures), ROI, and their interaction, and a subject-level random intercept. ROI-specific effects were quantified using estimated marginal means and within-ROI contrasts (emmeans). Results: No metabolite ratio showed a significant BPD vs. control difference. In BPD, IR-DA/IRTA-related measures were positively associated with Glx/tCr and Glx/tNAA in the accumbens, and with Glx/tCr in the right caudal anterior cingulate cortex (cACC). BSL-supplement scores showed ROI-dependent associations with tCr/tNAA, with positive ROI-specific effects in the bilateral caudate, right pallidum, and right putamen. Alertness measures were linked to GABA/tCr, Glx/tNAA, and tCr/tNAA (including caudate, pallidum/putamen, cACC, and thalamus), while divided attention omissions and working-memory errors were associated with higher Glx/tCr in the cACC, isthmus cingulate, and hippocampus. ROI-dependent associations were also observed for IQ and verbal learning/recognition with GABA/tCr and tNAA/tCr. Discussion: While no robust group differences emerged, mixed-models using continuous predictors linked higher Glx ratios in nucleus accumbens/cACC to IRDA/IRTA and higher striatal tCr/tNAA to BPD symptom severity and neurocognitive performance. These exploratory multimodal signatures point to cortico-striato-limbic mechanisms in BPD and should be confirmed in larger samples.