High-resolution electrophysiological mapping of effective connectivity of lateral prefrontal cortex

The lateral prefrontal cortex (LPFC) serves as a critical hub for higher-order cognitive and executive functions in the human brain, coordinating brain networks whose disruption has been implicated in many neurological and psychiatric disorders. While transcranial brain stimulation treatments often target the LPFC, our current understanding of connectivity profiles guiding these interventions based on electrophysiology remains limited. Here, we present a high-resolution probabilistic map of bidirectional effective connectivity between the LPFC and widespread cortical and subcortical regions. This map is derived from intracranial evoked potential analysis of 48 797 intracranial direct electrical stimulation runs across 759 implantations in 724 patients with refractory epilepsy (368 male, 354 female, two unspecified; mean age ± standard deviation: 24 ± 13.5 years). We mapped probabilistic connectivity between brain parcels with adaptive resolution-higher resolution in the LPFC in the hemisphere of interest and lower elsewhere-maintaining statistical power while achieving 95% average confidence interval of ∼0.03 for connectivity probability estimates. In addition, the significance threshold (P-value) for probabilistic connectivity was obtained from surrogate distributions. Overall, we observed remarkable symmetry between afferent and efferent connectivity patterns of the LPFC, with a slight preference for efferent connections (mean slope = 0.92 ± 0.09, mean R2 = 0.93 ± 0.025). For example, connections between the inferior frontal gyrus (IFG) and anterior cingulate showed notable directional asymmetry. The IFG strongly projected to most brain networks compared to other LPFC regions, with the strongest connectivity to the ventral attention network (0.26 ± 0.01 compared to values between 0.15 and 0.21 in other LPFC regions). Posterior dorso-lateral prefrontal cortex (DLPFC) demonstrated stronger connectivity to brain networks compared to anterior DLPFC regions (e.g. 0.21 ± 0.01 versus 0.15 ± 0.01 for connectivity to ventral attention network), with the exception of the limbic cortex. All LPFC subregions strongly projected to the fronto-parietal (greater than 0.17) and ventral attention (greater than 0.15) networks, with moderate connections to the default network (between 0.1 and 0.15, with the maximum corresponding to superior DLPFC). Finally, latency analysis suggested that the left LPFC's influence on ipsilateral emotion-related regions is primarily polysynaptic, with particularly strong pathways from IFG to amygdala (0.16 ± 0.02) and hippocampus (0.12 ± 0.01). Taken together, these comprehensive connectivity maps provide a new detailed electrophysiological foundation for understanding the functional anatomy of LPFC and guiding targeted brain stimulation protocols.