Multiscale dynamics and information flow in a data-driven model of the primary motor cortex microcircuit

Abstract

We developed a biologically detailed multiscale model of mouse primary motor cortex (M1) microcircuits, incorporating data from several recent experimental studies. The model simulates at scale a cylindrical volume with a diameter of 300 m and cortical depth 1350 m of M1. It includes over 10,000 cells distributed across cortical layers based on measured cell densities, with close to 30 million synaptic connections. Neuron models were optimized to reproduce electrophysiological properties of major classes of M1 neurons. Layer 5 corticospinal and corticostriatal neuron morphologies with 700+ compartments reproduced cell 3D reconstructions, and their ionic channel distributions were optimized within experimental constraints to reproduce in vitro recordings. The network was driven by the main long-range inputs to M1: posterior nucleus (PO) and ventrolateral (VL) thalamus PO, primary and secondary somatosensory cortices (S1, S2), contralateral M1, secondary motor cortex (M2), and orbital cortex (OC). The network local and long-range connections depended on pre-and post-synaptic cell class and cortical depth. Data was based on optogenetic circuit mapping studies which determined that connection strengths vary within layer as a function of the neuron's cortical depth. The synaptic input distribution across cell dendritic trees -likely to subserve important neural coding functions -was also mapped using optogenetic methods and incorporated into the model. We employed the model to study the effect on M1 of increased activity from each of the long-range inputs, and of different levels of H-current in pyramidal tract-projecting (PT) corticospinal neurons. Microcircuit dynamics and information flow were quantified using firing rates, oscillations, and information transfer measures (Spectral Granger causality). We evaluated the response to short pulses and long time-varying activity arising from the different long-range inputs. We also studied the interaction between two simultaneous long-range inputs. Simulation results support the hypothesis that M1 operates along a continuum of modes characterized by the degree of activation of intratelencephalic (IT), corticothalamic (CT) and PT neurons. The particular subset of M1 neurons targeted by different long-range inputs and the level of H-current in PT neurons regulated the different modes. Downregulation of H-current facilitated synaptic integration of inputs and increased PT output activity. Therefore, VL, cM1 and M2 inputs with downregulated H-current promoted an PT-predominant mode; whereas PO, S11 . CC-BY-NC-ND 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint . http://dx.doi.org/10.1101/201707 doi: bioRxiv preprint first posted online Oct. 11, 2017; and S2 inputs with upregulated H-current favored an IT-predominant mode, but a mixed mode where upper layer IT cells in turn activated PT cells if H-c...

NetPyNE, a tool for data-driven multiscale modeling of brain circuits

“…Computational needs for running much larger and more complex neural simulations are constantly increasing as researchers attempt to reproduce fast-growing experimental datasets (Bezaire et al, 2016; Markram et al, 2015; MindScope et al, 2016; Dura-Bernal et al, 2018; Hereld et al, 2005; Lytton et al, 2016). Fortunately, parallelization methods and high-performance computing (HPC, supercomputing) resources are becoming increasingly available to the average user (Hines, 2011; Hines et al, 2008; Migliore et al, 2006; Towns et al, 2014; Amunts et al, 2016; Sivagnanam et al, 2013; Krause and Thörnig, 2018).…”

“…The Neuroscience Gateway, which provides neuroscientists with free and easy access to supercomputers, includes NetPyNE as one of the tools available via their web portal. Larger-scale models – including the M1 model with 10 thousand multicompartment neurons and 30 million synapses (Dura-Bernal et al, 2018) and the thalamocortical model with over 80 thousand point neurons and 300 million synapses (Potjans and Diesmann, 2014; Romaro et al, 2018) – have been simulated on both the XSEDE Comet supercomputer and Google Cloud supercomputers. Run time to simulate 1 second of the multicompartment-neuron network required 47 minutes on 48 cores, and 4 minutes on 128 cores for the point-neuron network.…”

“…Our recent model of primary motor cortex (M1) microcircuits (Dura-Bernal et al, 2018; Neymotin et al, 2016b; Neymotin et al, 2017) constitutes an illustrative example where NetPyNE enabled the integration of complex experimental data at multiple scales: it simulates over 10,000 biophysically detailed neurons and 30 million synaptic connections. Neuron densities, classes, morphology and biophysics, and connectivity at the long-range, local and dendritic scale were derived from published experimental data (Suter et al, 2013; Yamawaki et al, 2015; Yamawaki and Shepherd, 2015; Harris and Shepherd, 2015; Sheets et al, 2011; Weiler et al, 2008; Anderson et al, 2010; Yamawaki et al, 2015; Kiritani et al, 2012; Apicella et al, 2012; Hooks et al, 2013; Suter and Shepherd, 2015).…”

“…Neuron densities, classes, morphology and biophysics, and connectivity at the long-range, local and dendritic scale were derived from published experimental data (Suter et al, 2013; Yamawaki et al, 2015; Yamawaki and Shepherd, 2015; Harris and Shepherd, 2015; Sheets et al, 2011; Weiler et al, 2008; Anderson et al, 2010; Yamawaki et al, 2015; Kiritani et al, 2012; Apicella et al, 2012; Hooks et al, 2013; Suter and Shepherd, 2015). Results yielded insights into circuit information pathways, oscillatory coding mechanisms and the role of HCN in modulating corticospinal output (Dura-Bernal et al, 2018). A scaled down version (180 neurons) of the M1 model is illustrated in Figure 8.…”

“…NetPyNE is publicly available from http://netpyne.org, which includes installation instructions, documentation, tutorials, example models and Q&A forums. The tool has already been used by over 50 researchers in 24 labs to train students and to model a variety of brain regions and phenomena (see http://netpyne.org/models) (Dura-Bernal et al, 2018; Romaro et al, 2018; Lytton et al, 2017; Neymotin et al, 2016b). Additionally, it has been integrated with other tools in the neuroscience community: the Human Neocortical Neurosolver (https://hnn.brown.edu/) (Jones et al, 2009; Neymotin et al, 2018), Open Source Brain (http://opensourcebrain.org) (Gleeson et al, 2018; Cannon et al, 2014), and the Neuroscience Gateway portal (http://nsgportal.org) (Sivagnanam et al, 2013).…”