Computational Neural Modeling Using Human Neocortical Neurosolver (HNN)

Our group has developed a unique biophysically principled computational neural model of the neocortical circuitry that links human EEG or MEG signals to underlying cellular and circuit-level mechanisms, based on their biophysical origin (Jones et al., 2007, 2009; Ziegler et al., 2010; Lee and Jones, 2013; Sherman et al., 2016; Neymotin et al., 2018). The model represents a laminated cortical column with synaptically coupled inhibitory interneurons and excitatory neurons across layers (i.e., basket cells and pyramidal neurons), and includes exogenous excitatory synaptic drive to distinct layers. It also simulates the primary electrical current activity (i.e., current dipoles) that generates EEG/MEG sensor data from post-synaptic intracellular currents in large and spatially aligned dendrites of cortical pyramidal neurons (Hämäläinen et al., 1993; Okada et al., 1997; Murakami and Okada, 2006; see Figure ​Figure9A).9A). The model has been applied previously to study the origin of somatosensory oscillations and tactile evoked responses (Jones et al., 2007, 2009; Ziegler et al., 2010; Lee and Jones, 2013; Sherman et al., 2016).

Computational neural modeling suggests enhanced synaptic gain can account for the post-tACS evoked response. (A) Schematic illustration of the primary somatosensory cortex (SI) model network underlying the Human Neocortical Neurosolver (HNN) software. The network is represented by a canonical model of a layered neocortical column (top left), with synaptically coupled inhibitory (yellow circles, basket cells) and excitatory neurons (blue, multi-compartment pyramidal neurons) across layers (top right). It also includes two distinct pathways of exogenous excitatory synaptic network inputs that effectively drive proximal (red) and distal (green) dendrites of the cortical pyramidal neurons (bottom panels). HNN simulates the primary electrical currents underlying EEG/MEG signals (red and green arrows) from net post-synaptic intracellular currents within the large and spatially aligned cortical pyramidal neuron dendrites (see “Materials and Methods” for further details). (B) Simulation of the tactile evoked response from 60 to 130 ms (see Supplementary Figure S1 for additional description of the tactile evoked response simulation). A default model simulation of the SI threshold-level tactile evoked response (solid blue line), provided in HNN, was tuned to MEG data from a prior study (dotted blue line; Jones et al., 2007). Increasing local synaptic gain (i.e., maximal conductance of excitatory and inhibitory synapses) by a factor of 2 in the model simulation (cyan line) reproduced the enhanced ∼70–80 ms peak observed post-tACS. Increasing synaptic gain separately for excitatory or inhibitory populations, respectively, demonstrated that the simulated ∼70–80 ms peak was driven mainly by enhanced inhibitory synaptic gain (red line; compare to E), as opposed to enhanced excitatory synaptic gain (green line). (C,D) Spiking activity of each cell in the network (C) before and (D) after modifying synaptic gain demonstrates that increasing total synaptic gain resulted in decreased firing of layer 5 pyramidal neurons (red circles, yellow box) around the same time as the enhanced ∼70–80 ms peak. (E) EEG tactile evoked responses from 60 to 130 ms during pre- and post-tACS time blocks. The pre-alpha tACS EEG evoked response (blue line) closely resembles the simulation of a threshold-level tactile evoked response (compare to solid blue line, B), while the post-alpha tACS EEG evoked response most closely resembles the model simulation with enhanced inhibitory synaptic gain (compare to red line, B).

Our group has recently developed this model into a user-friendly software tool named Human Neocortical Neurosolver (HNN). HNN allows for the development and testing of specific hypotheses regarding the neural origins of EEG/MEG signals using a graphical user interface (GUI; Neymotin et al., 2018). A detailed description, tutorials, and open-source freely available distribution of this tool can be found at http://hnn.brown.edu.

For the purposes of the analysis presented here, we utilized the model parameters associated with the ‘ERP tutorial’ on the HNN website¹ (see Supplementary Materials). This tutorial provides a parameter set that simulates a SI-localized MEG evoked response potential (ERP) elicited by threshold-level taps to the finger, analogous to the tactile stimulation paradigm used in the current study. The ERP is simulated using a layer-specific sequence of exogenous excitatory synaptic drive to the local SI circuitry (as described in Jones et al., 2007, 2009). This sequence of drive reproduces early tactile evoked response peaks up to 165 ms post-stimulus (see Supplementary Figure S1). We specifically modified the model parameters representing synaptic gain, which can be accessed through the HNN GUI. This allows for adjustment of excitatory and inhibitory synaptic conductance within the model network by multiplying the targeted synaptic conductance weights (inhibitory:GABAA/GABAB or excitatory: AMPA/NMDA, respectively) by a specified amount. To test the specific hypothesis that tACS can affect synaptic plasticity, we increased total synaptic gain (i.e., both inhibitory and excitatory synaptic conductance weights), as well as inhibitory and excitatory synaptic gain parameters alone, by a factor of 2. Simulated evoked responses under various gain changes were compared with recorded EEG data to interpret potential effects of tACS on the EEG evoked response.

All simulations in this study can be reproduced by downloading the HNN software, running the ERP tutorial, and changing the synaptic gain parameters, as described. Parameter files used in the current analysis are included in the Supplementary Materials).