nature neuroscience • volume 3 no 10 • october 2000 1027 The cerebral cortex is constantly active. Even during slow-wave sleep, cortical and subcortical networks interact to generate rhyth- mic patterns of activity at a variety of frequencies 1–4 , including a ‘slow rhythm’ characterized by the recurrence of tonic activity in cortical neurons approximately once every three to five seconds 2–9 . This slow oscillation is generated within the neocortex and may coordinate other sleep rhythms (such as spindle waves and delta waves) into a coherent rhythmic sequence of recurring cortical and thalamocortical activities 2–5 . The slow oscillation is associ- ated with the arrival of a relatively steady barrage of excitatory and inhibitory postsynaptic potentials, and the discharge of both excitatory and inhibitory neurons (the ‘up’ state), interdigitated with periods of hyperpolarization and quiescence (the ‘down’ state) 2–6 . The propagation and synchronization of this slow oscil- lation depends at least in part on corticocortical connections, and is proposed to be generated by recurrent excitation among large networks of cortical neurons, interrupted by periods of ‘dis- facilitation’ in which the recurrent activity fails 5,6 . Although it was once thought that the generation of rhyth- mic activities in the electroencephalogram were due to the inter- action of very large networks of neurons and therefore could only be studied in vivo, we have shown that at least one network oscil- lation, spindle waves, can be maintained in small networks of neurons in vitro 10 . Here we demonstrate generation of the slow rhythm in neocortical slices and show that this activity is gener- ated through re-entrant excitation interspersed with failure of network activity. RESULTS Intracellular recordings from primary visual cortical neurons in halothane-anesthetized cats revealed the rhythmic re-occurrence of depolarized and hyperpolarized membrane potentials 2–7 at a periodicity of once every 3.44 ± 1.37 seconds ( n = 9; Fig. 1a–c). The depolarized phase appeared as a barrage of postsynaptic potentials that could reach action potential threshold and lasted an average of 1.08 ± 0.38 seconds ( n = 9; Fig. 1a and b). The mem- brane potential during the slow oscillation exhibited a bimodal distribution with peaks at –71 ± 3.6 mV and –59.4 ± 4.9 mV, rep- resenting the up and down states 6–8 . Intracellular and extracellular recordings from ferret visual and prefrontal cortical slices maintained in vitro in ‘traditional’ slice bathing medium (2 mM Ca 2+ , 2 mM Mg 2+ and 2.5 mM K + ) did not reveal spontaneous rhythmic activities. However, changing the bath solution to more closely mimic the ionic composition of brain interstitial fluid in situ 12,13 (1.0 or 1.2 mM Ca 2+ , 1 mM Mg 2+ and 3.5 mM K + ) caused the appearance of spontaneous rhythmic oscil- lations that were nearly identical to those occurring in vivo ( Fig. 1d–f; n = 87 visual cortical slices ; n = 5 prefrontal cortical slices). Once the slow oscillation developed, it was stable for the life of the slice (up to 12 h). It appeared as a depolarized state asso- ciated with action potential activity at 2–10 Hz (in regular spik- ing neurons), followed by a hyperpolarized state, recurring with a periodicity of once every 3.44 ± 1.76 seconds ( n = 24; Fig. 1d–f). The membrane potentials again exhibited a bimodal distribution, with peaks at –68.7 ± 3.8 mV and –61.2 ± 4.8mV ( n = 11). The depolarized state lasted for an average of 0.72 ± 0.43 seconds ( n = 25 cells) and occurred in regular spiking ( n = 13), intrinsi- cally bursting (n = 11) and chattering (n = 20) pyramidal neurons, as well as fast-spiking local interneurons ( n = 6) in phase with extracellularly recorded multi-unit activity (Figs. 1d and 6c; as also revealed with cross-correlograms; Methods). Propagation of the slow oscillation Extracellular multiple-unit recording with arrays of 8 microelec- trodes (with inter-electrode spacing of approximately 0.25 or 0.5 mm; Fig. 2a) revealed that the slow oscillation was most robust and occurred first in or near layer 5, followed after a short delay by activity in deeper layers (layer 6), and finally after an addition- articles Cellular and network mechanisms of rhythmic recurrent activity in neocortex Maria V. Sanchez-Vives 1,2 and David A. McCormick 1 1 Section of Neurobiology, Yale University School of Medicine, 333 Cedar Street, New Haven, Connecticut 06510, USA 2 Present address: Instituto de Neurociencias UMH-CSIC, Apartado 18, 03550 San Juan de Alicante, Spain Correspondence should be addressed to D.A.M. (david.mccormick@yale.edu) The neocortex generates periods of recurrent activity, such as the slow (0.1–0.5 Hz) oscillation during slow-wave sleep. Here we demonstrate that slices of ferret neocortex maintained in vitro gen- erate this slow (< 1 Hz) rhythm when placed in a bathing medium that mimics the extracellular ionic composition in situ. This slow oscillation seems to be initiated in layer 5 as an excitatory interaction between pyramidal neurons and propagates through the neocortex. Our results demonstrate that the cerebral cortex generates an ‘up’ or depolarized state through recurrent excitation that is regulated by inhibitory networks, thereby allowing local cortical circuits to enter into temporarily activated and self-maintained excitatory states. The spontaneous generation and failure of this self- excited state may account for the generation of a subset of cortical rhythms during sleep. © 2000 Nature America Inc. • http://neurosci.nature.com © 2000 Nature America Inc. • http://neurosci.nature.com