Histologically Informed Multiscale Modeling of the Neuronal Elements Activated by TMS.

Background: The primary neural site(s) at which action potentials are initiated by transcranial magnetic stimulation (TMS) remain poorly understood. Multiscale computational models provide biophysically based hypotheses, but model accuracy is constrained by limited histological knowledge of the microscopic organization of neural tissue. Recent high-resolution electron microscopy, in particular the petavoxel H01 dataset, provides novel, detailed axon morphologies and myelination patterns within the human cortex and superficial white matter. Objective: To compare systematically multiple candidates for the neural elements activated by TMS using computer simulations informed by an extensive body of histological measurements, including neuron models directly reconstructed from the H01 dataset. Methods: We developed a novel modeling pipeline to extract individual morphologically realistic multi-compartment models with exact myelination from serial section electron microscopic segmentations. To assess candidate excitation sites, we simulated the extracted neuron models together with parameterized models of a "ball-and-two-sticks", bifurcation, termination, and bend under uniform electric fields. In addition, smooth and sharply bending myelinated axons were embedded in a realistic human head model to evaluate activation thresholds under anatomically realistic electric field distributions. Results: Axon terminations were only excitable by TMS when they were fully myelinated, which the histology suggested is unlikely. Partial myelination, even when separated by only 10 μm from the terminal, increased activation thresholds by more than 100%. Reconstructed H01 neurons exhibited correspondingly high activation thresholds at axon terminals due to a lack of myelination. Further, most other candidate structures exhibited low thresholds only for histologically unrealistic parameter choices. In contrast, myelinated axonal bends of fibers transitioning from cortex to superficial white matter consistently showed low activation thresholds for both uniform electric fields and in realistic head model simulations. These thresholds fell within physiologically realistic ranges and, for larger diameter fibers, approached experimentally measured motor thresholds. Conclusion: These results identify myelinated axons bending from cortex into superficial white matter as possible neural targets for transcranial magnetic stimulation, and demonstrate the relevance of detailed histological and biophysical information to support robust modeling results.