Neuronal Spatial Arrangement Shapes Effective Connectivity Traits of in vitro Cortical Networks

Abstract

We studied effective connectivity in rat cortical cultures with various degrees of spatial aggregation, ranging from homogeneous networks to highly aggregated ones. We considered small cultures 3 mm in diameter and that contained about 2000 neurons. Spatial inhomogeneity favored an increase of metric correlations and connectivity among neighboring neurons. Effective connectivity was determined from spontaneous activity recordings using calcium fluorescence imaging. We used generalized transfer entropy as tool to infer the effective connectivity. We carried out numerical simulations to build networks that mimicked the experimental ones and to test the reliability of the connectivity-inference algorithm. Effective connectivity traits were investigated during the development of the cultures over two weeks, and along the gradual blockade of excitatory connections through CNQX. We observed that the average effective connectivity rapidly increased during culture development. At DIV 15 the average excitatory in-degree was measured ask in E 50 for homogeneous and semi aggregated networks, andk in E 120 for aggregated ones, and with 20% inhibition. Aggregated cultures exhibited assortative traits and a high resilience to chemical damage, while the other cultures were dissassortative or neutral, and less resilient. Our work illustrates the role of metric correlations in spatially embedded networks in shaping connectivity and activity traits in living neuronal networks.

Impact of Physical Obstacles on the Structural and Effective Connectivity of in silico Neuronal Circuits

“…We explored in silico neuronal networks with spatial constraints by considering different sets of obstacles arranged on a circular area of either 2 or 4 mm in diameter. This size was selected to mimic the characteristic size of small in vitro cultures (Orlandi et al, 2013 ; Tibau et al, 2020 ). In the simulations, neurons were laid out on the surface in a homogeneous manner and connected to one another following a geometric model as in Orlandi et al ( 2013 ), in which the axons grew as concatenated segments according to a biased random walk ( Figure 1A ) and that is known to mimic well the behavior of individual axons (Feinerman et al, 2008 ).…”

“…Our simulations reflect the importance of metric correlations in shaping connectivity and dynamics in neuronal circuits. Metric correlations appear naturally in spatially embedded networks (Orlandi et al, 2013 ; Tibau et al, 2020 ). As in our simulations, other studies pointed out the spatial distribution of neurons and the characteristic axonal length relative to system size as central ingredients in shaping local and global structural traits (Schmeltzer et al, 2014 ; Hernández-Navarro et al, 2017 ; Okujeni et al, 2017 ).…”

“…We used the same implementation as in (Orlandi et al, 2013 ; Tibau et al, 2020 ). Parameter values were similar to those used in (Orlandi et al, 2013 ) and were chosen so that the model reproduces typical behavior of cortical neurons.…”

“…Messé et al for instance used elaborate computational models and anatomical brain data to predict the activity patterns observed in resting-state functional magnetic resonance imaging, and concluded that the backbone of anatomical connectivity strongly shaped overall dynamical traits. Neuronal cultures, on the other hand, have helped elucidate the importance of spatial embedding and imposed metric correlations in shaping spontaneous activity (Orlandi et al, 2013 ; Hernández-Navarro et al, 2017 ; Okujeni et al, 2017 ; Tibau et al, 2020 ), the impact of modular organization (Shein-Idelson et al, 2011 ; Tang-Schomer et al, 2014 ; Yamamoto et al, 2018 ), the emergence of small-worldness (Downes et al, 2012 ; de Santos-Sierra et al, 2014 ), or the role of hubs (Schroeter et al, 2015 ).…”

Impact of targeted attack on the spontaneous activity in spatial and biologically-inspired neuronal networks

“…[14][15][16] Recent studies showed that spatial embedding in these systems naturally feature metric correlations that shape in great measure the architectural and dynamic traits of the neuronal network. 14,[16][17][18] To the best of our knowledge, there exist no computational studies for damage that take into account the spatial embedding of biologically-realistic neuronal microcircuits. Thus, in the present study, we aim at lling this gap by introducing a detailed numerical exploration of the di erences between targeted attack and failure in spatially-embedded networks.…”