Flexible traffic control of the synfire-mode transmission by inhibitory modulation: Nonlinear noise reduction Takashi Shinozaki, 1, * Masato Okada, 2,1 Alex D. Reyes, 3 and Hideyuki Câteau 1,4 1 RIKEN Brain Science Institute, 2-1 Hirosawa, Wako, Saitama 351-0198, Japan 2 Graduate School of Frontier Sciences, University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, Chiba 277-8561, Japan 3 Center for Neural Science, New York University, 4 Washington Place, New York, New York 10003, USA 4 Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, 2-4 Hibikino, Kitakyushu, Fukuoka 808-0196, Japan Received 18 May 2009; revised manuscript received 2 December 2009; published 22 January 2010 Intermingled neural connections apparent in the brain make us wonder what controls the traffic of propa- gating activity in the brain to secure signal transmission without harmful crosstalk. Here, we reveal that inhibitory input but not excitatory input works as a particularly useful traffic controller because it controls the degree of synchrony of population firing of neurons as well as controlling the size of the population firing bidirectionally. Our dynamical system analysis reveals that the synchrony enhancement depends crucially on the nonlinear membrane potential dynamics and a hidden slow dynamical variable. Our electrophysiological study with rodent slice preparations show that the phenomenon happens in real neurons. Furthermore, our analysis with the Fokker-Planck equations demonstrates the phenomenon in a semianalytical manner. DOI: 10.1103/PhysRevE.81.011913 PACS numbers: 87.19.lm, 05.10.Gg, 87.18.Sn, 87.80.Jg I. INTRODUCTION Inhibitory input to network of neurons or more generally excitable media can modulate their activity in a way more complex than the simple activity suppression. For instance, inhibitory input delivered to a physiologically modeled neu- ron but not the leaky integrate-and-fire LIFneuron, can paradoxically increase the firing probability if it is delivered at a right timing 13. For a population of neurons, this increase in firing probability of a single neuron is interpreted as an increase in the number of firing neurons in a population due to inhibitory input 4. Here, we demonstrate a mecha- nism by which inhibitory input also enhance the synchrony of population firing. We carefully study this synchrony en- hancement, that was actually visible in our previous study 4but out of focus there. We indicate that a nonlinear mechanism is responsible for this synchrony enhancement. The synchrony enhancement by inhibitory input shows a clear contrast to the previously studied synchronization of neurons in mutually connected neural networks 5, which was observed even for a linear neuron model. For the best illustration of the synchrony enhancement in the feedforward setting, we take the so-called the synfire chain as an example 4,624which is stable propagation of population firing of neurons. Let us consider the population firing that is described with a bell-shaped pulse packet Fig. 1awhich indicates the number of firing neurons versus time 9. The pulse packet is specified with the number of total firing, a, the degree of synchrony measured with the width, , and the peak time of the population firing, t. The present study shows that properly timed inhibitory input can enhance synchrony of a pulse packet . This generalizes the previous observations 14that are interpreted as an increase in the number of firing neurons a in the present context. Our numerical simulations coupled with a dynamical sys- tem analysis reveal a mechanism through which the mem- brane potential histogram in a neural population gets sharp- ened by inhibitory input via a nonlinear effect. We refer to this effect as nonlinear noise reduction. Importantly, this nonlinear noise reduction occurs with a population of physi- ologically plausible neuron models but not the LIF model. Furthermore, our electrophysiological experiments with rodent brain slice preparations demonstrate that this mecha- nism works also in real neurons. Finally, we demonstrate semianalytically with Fokker-Planck FPequations how the mechanism works. Thus, the present study points out the * Present address: Center for Neural Science, New York Univer- sity, 4 Washington Place, New York, New York 10003, USA. inhibitory neuron excitatory neuron all-to-all connection time (ms) inhibitory excitatory 0 -20 (a) (c) (b) time FIG. 1. Definition of a pulse packet and simulation setup. a Firing time histogram is parametrized with its total area, a, width, and peak time, t a pulse packet9. bSchematic illustration of the time course of excitatory topand inhibitory bottomsynaptic currents. cFeedforward network of excitatory neurons which are modeled with Eqs. 1and 2. Each neuron in a layer sends input to all the neurons in the following layer as in 9. The excitatory neu- rons in the layer in the question centerreceive phasic inhibitory input with uniform strength at a specified timing. Open and filled circles, respectively, represent excitatory and inhibitory neurons. PHYSICAL REVIEW E 81, 011913 2010 1539-3755/2010/811/0119137©2010 The American Physical Society 011913-1