Consiglio Nazionale delle Ricerche (CNR), I.B.F. (Italy)
Local Project ID:
HPC Platform used:
JUQUEEN of JSC
The main aim of this project was the development of the first detailed large-scale 3D model of the CA1 area of the hippocampus, a brain region well known for being involved in cognitive processes and deeply affected by aging and major brain diseases such as Alzheimer’s Disease and Epilepsy. Because of the current limitations in the experimental techniques, the cellular mechanisms underlying these processes remain relatively unknown. With our model, we maintain the 3D layout of the real system, and the neurons’ activity can be observed in exactly the same format as the in vivo recordings, with the fundamental advantage of being able to track network, cellular, and synaptic activity at any point of the network, and directly compare the results with experiments at all levels, including fMRI and EEGs. We expect that the model will significantly advance the state of the art in the field, and will help to predict and explain several experimental and behavioral data.
In this project, the focus was on the development of the first detailed and realistic large scale 3D model of the CA1 region of the hippocampus. The hippocampus (a latin word derived from greek to indicate a seahorse) is a small brain region located deep in the brain, in the medial temporal lobe, underneath the cortical surface (see figure below). Its structure is divided into two halves which lie in the left and right sides of the brain. The organ is curved with a shape that resembles a seahorse, explaining its name. It is well known that the processes related to higher brain function, such as memory, learning, and spatial navigation involve this region (Squire et al., 2004; Andersen et al., 2006; Morris 2006) that which, for this reason, is one of the most studied both experimentally and theoretically.
A major problem in interpreting the experimental findings in vivo, to understand the processes underlying memory functions in neurons, is that the recordings are usually obtained in single cells or in small (more or less randomly) selected sets of cells. However, a clear understanding of fundamental processes, such as the spatio-temporal organization of the hippocampal activity, requires simultaneous recording from a relevant subset of cells activated by specific sensory and cortical inputs. The functional effects of network-wide processes in relation to input patterns, therefore, remain relatively unknown and extremely difficult to explore experimentally. Addressing this critical step requires a large-scale realistic model maintaining the natural 3D layout of the real system, in which the neurons’ activity can be observed in exactly the same format as the in vivo recordings, with the fundamental advantage of being able to track network, cellular, and synaptic activity at any point of the network, and directly compare the results with practically all type of experimental data, including local field potentials and EEGs.
This was a big computational challenge, which required supercomputing power. To give an idea of the size of the problem one should consider that the region we were interested to model, called CA1, contains (in the rat) about 400000 neurons and approximately 400 million synaptic contacts. They were implemented with a system of approximately 109 nonlinear ordinary differential equations. The general method and algorithm used for the numerical integration of this large system was that used in the NEURON simulation environment, which is a de facto standard to model realistic neuronal networks. It is an open source package written in C++, freely available, does not require licenses or special libraries, it runs on parallel machines using pure MPI, and can be fully integrated into python code. In this case, a special version of the NEURON (coreneuron) was used, with the specific design objective of optimizing runtime and memory footprint for extreme scaling on the largest supercomputers in the world.
All simulations were carried out on the JUQUEEN HPC system of JSC. Typical snapshots from a movie created from one of the simulations are shown in the figures on the left. They represent the neuronal activity generated by spontaneous synaptic release in the entire system and in one “slice” of the system. Membrane potential in the soma and dendrites of each neuron is color coded (blue is close to resting potential, yellow represents the peak voltage reached during an action potential). From these figures, one can appreciate the unprecedented level of details with which our model can reproduce experimental recordings in vitroor in vivo. Worldwide, so far there are only three models (including this one) implemented in this way.
It is from models and simulations like this that both experimentalists and modelers can make direct comparison of their results and get a lot of insight and a better understanding of how the system works. For example, one of the current highly debated issues on how the hippocampus is involved in cognitive functions is to figure out the role of “traveling waves” of activity, experimentally observed to span the entire system at a frequency in the range of the theta rhythm, one of the characteristic brain rythms recorded at the EEG level during cognitive processes. In our model, traveling waves in the theta range emerge sponstaneously as soon as we put the system together, suggesting that this is an intrinsic network property rather than an intrinsic features of indivudual neurons. We are now exploring this issue in more details.
At this level of implementation, once we understand how cognitive functions can emerge from neuronal dynamics, we will also be able to model brain diseases and suggest to experimentalists and pharma industries new ideas on possible treatments. The results produced by this project have been presented in several international conferences, and a paper on the full system is in preparation. Interested readers can explore this circuit and its neurons through the set of online use cases dedicated to the hippocampus and available on the Brain Simulation Platform of the Human Brain Project.
References directly related to this project
Migliore R, Lupascu CA, Bologna LL, Romani A, Courcol JD, Antonel S, Van Geit WAH, Thomson AM, Mercer A, Lange S, Falck J, Rössert CA, Shi Y, Hagens O, Pezzoli M, Freund TF, Kali S, Muller EB, Schürmann F, Markram H, Migliore M. (2018), The physiological variability of channel density in hippocampal CA1 pyramidal cells and interneurons explored using a unified data-driven modeling workflow, PLoS Comp. Biol., 14(9):e1006423. doi: 10.1371/journal.pcbi.1006423
M Migliore1, A Romani2 , N Antille2 , LL Bologna1 , J Budd3 , JD Courcol2 , A Devresse2 , A Ecker2 , J Falck4,6 , C Favreau2 , E Giacalone1, M Gevaert2 , AI Gulyas3 , O Hagens7 , J Hernando2 , S Jimenez2 , L Kanari2, JG King2, S Lange4,5 , CA Lupascu1, R Migliore1, M Pezzoli2, A Povolotsky2, S Ramaswamy2, M Reimann2, CA Rössert2, S Sáray3,8, Y Shi2, WAH Van Geit2, L Vanherpe2, P Vitale1, TF Freund3,8, S Kali3, H Markram2, A Mercer4, AM Thomson4, EB Muller2
1 Institute of Biophysics, National Research Council, Italy;
2 Blue Brain Project (BBP), Brain Mind Institute, EPFL, Lausanne, Switzerland;
3 Institute of Experimental Medicine, Hungarian Academy of Sciences, Hungary;
4 University College London, United Kingdom;
5 University of Westminster, London, United Kingdom;
6 Deutsches Zentrum für Neurodegenerative Erkrankungen, Germany;
7 Laboratory of Neural Microcircuitry (LNMC), Brain Mind Institute, EPFL, Lausanne, Switzerland;
8 Faculty of Information Technology and Bionics, Pázmány Péter Catholic University, Budapest, Hungary
Consiglio Nazionale delle Ricerche (C.N.R.)
Istituto di Biofisica (I.B.F.)
Via U. La Malfa, 153
I-90146 Palermo (Italy)
e-mail: michele.migliore [@] cnr.it
NOTE: This project was made possible by PRACE (Partnership for Advanced Computing in Europe) allocating a computing time grant on GCS HPC system JUQUEEN of the Jülich Supercomputing Centre and by the Horizon 2020 Framework Programme for Research and Innovation under the Specific Grant Agreement No. 785907 (Human Brain Project SGA2).
JSC project ID: PRA098