Back to the abstracts page
Back to the NCCD 2015 home page
Neuronal noise correlations are caused by brief network transitions into silence.
Gabriela Mochol1,2 , Ainhoa Hermoso-Mendizabal1 , Shuzo Sakata3 , Kenneth Harris4 and Jaime de la Rocha1
1Institut D'Investigacions Biomediques August Pi I Sunyer, Barcelona, Spain. 2Nencki Institute, Warsaw, Poland. 3University of Strathclyde, Glasgow, UK. 4University College London, UK.

The spiking activity of cortical neurons is highly variable in a spontaneous and evoked conditions [1]. This variability is generally correlated among nearby neurons [2], an effect which potentially has impact on the computational capabilities of neuronal populations. Although correlations are commonly interpreted to reflect the co-activation of neurons due to anatomically shared inputs [3], recent findings indicate that correlations can be dynamically modulated [2, 4], suggesting that the underlying mechanisms are not well understood. We hypothesized that correlations reflect instead neuronal co-inactivation: the occurrence of brief silent periods during which all neurons in the local network stop firing. To test this, we recorded spiking activity from large population of neurons in the primary auditory cortex of anesthetized rats across different brain states ranging from synchronized to desynchronized. We found that, during spontaneous activity, the reduction of correlation accompanying brain state desynchronization was largely explained by a decrease in the density of the silent periods. The presentation of a brief stimulus caused an initial drop of instantaneous correlations followed by a rebound, a time-course that was mimicked by the instantaneous silence density across brain states.

To understand the mechanisms underlying correlations and their relation with silence density we assumed that neuronal variability has two sources: the first reflecting variations in the firing rate and the second reflecting the spiking stochasticity existent at constant rate [3, 5, 6]. The model with spiking statistics described by a Poisson process in which the trial to trial rate variations followed a bimodal distribution could explain spike count co-variations observed in the data across different brain states. The model with unimodal rate fluctuations failed to reproduce data variability especially during the epochs of synchronized brain state suggesting that silent periods reflected separate state in the circuit dynamics rather than the lower tail of a unimodal distribution.

To gain an insight into the dynamics of correlation and silence density we built a rate network model with fluctuation-driven transitions between a silent and an active attractor. Variations of the external input to the network altered the probability of transition to the silent attractor and reproduced the relation between correlation and silence density found in the data both in spontaneous and evoked conditions. Thus, correlations are mostly due to the occurrence of silent periods among otherwise uncorrelated activity and can be gradually and dynamically modulated by internal and external factors changing the probability that the circuit transitions into the silence mode.

[1] W.R. Softky and C. Koch J. Neurosci. 13:334-350 (1993).
[2] A. Kohn, A. Zandvakili and M. A. Smith Curr Opin Neurobiol. 19:434-438 (2009).
[3] M. N. Shadlen and W. T. Newsome J Neurosci. 18:3870-3896 (1998).
[4] M. A. Smith and A. Kohn J Neurosci. 28:12591-12603 (2008).
[5] M. M. Churchland et al. Nat Neurosci. 13:369-378 (2010).
[6] R. L. T. Goris, J. A. Movshon and E. P. Simoncelli Nat Neurosci. 17:858-865 (2014).