Deep Brain Stimulation Induces Antidepressant Effects by Restoring High‐Fidelity Communication in the BNST‐NAc Circuit

ABSTRACT Deep brain stimulation (DBS) for treatment‐resistant depression (TRD) is challenged by significant individual variability in efficacy and unclear neural circuit mechanisms. To address this, a cross‐species, multi‐level electrophysiological study was conducted to elucidate the core underlying pathophysiology and reveal the precise therapeutic mechanisms of DBS. Based on the clinical trial (NCT04530942), this study focuses on the bed nucleus of the stria terminalis‐nucleus accumbens (BNST‐NAc) circuit, and it is hypothesized that the fundamental pathology of the depressive state lies in the persistent hyperactivity of BNST neurons, which disrupts the high‐fidelity signal communication capacity of this circuit. In a mouse model, a key communication pattern, inhibitory period isolated spikes (IPIS), was first identified within the excitation/inhibition (E/I) cycle. This pattern involves slow‐wave oscillations creating a high signal‐to‐noise ratio window for the firing of single or few action potentials, thereby enabling efficient inter‐regional communication. Subsequently, it was found that chronic stress‐induced pathological hyperactivity of BNST neurons in stress‐susceptible animals specifically disrupts the inhibitory periods of network activity, thereby dismantling IPIS‐mediated cross‐regional neural synchrony and leading to circuit dysfunction. The therapeutic mechanism of DBS was verified to involve precisely suppressing the pathological hyperactivity of the BNST, thereby restoring the network's inhibitory periods and re‐establishing the efficient signal transmission pathway mediated by IPIS. In a closed‐loop DBS paradigm, only continuous stimulation and stimulation precisely locked to the inhibitory periods produced antidepressant effects and most effectively restored cross‐regional communication. Furthermore, in a cohort of human TRD patients, local field potential (LFP) data were recorded during BNST‐NAc DBS treatment, and LFP biomarkers corresponding to the restoration of circuit function were identified. To more directly validate changes in E/I cycles in the human brain, an innovative cross‐species algorithm was developed to decode functional excitatory and inhibitory periods from macroscopic human LFP signals. It was confirmed that the therapeutic response to DBS is associated with an increased proportion of inhibitory periods and the functional recovery of the BNST‐NAc circuit, providing direct quantitative evidence for the theory that DBS restores E/I balance in the human brain. Finally, a double‐blind, crossover randomized controlled trial (RCT) involving 18 participants confirmed that active DBS clinically alleviated depressive symptoms (an average of 9.4 reduction). Open‐label data were used for model development, while RCT data served as an independent validation set. Model ablation study within a deep learning framework confirmed that E/I cycle features provide significantly higher informative value than spectral models, establishing these dynamics as the primary electrophysiological determinants of the clinical state. This study integrates mechanistic research with clinical validation, providing evidence for precision and personalized closed‐loop DBS therapy.