Neuronal Oscillations Enhance StimulusDiscrimination by Ensuring Action PotentialPrecisionAndreas T. Schaefer1, Kamilla Angelo1, Hartwig Spors2, Troy W. Margrie1*1 Department of Physiology, University College London, London, United Kingdom, 2 WIN Research Group of Olfactory Dynamics, Max-Planck-Institut fu¨r medizinischeForschung, Heidelberg, GermanyAlthough oscillations in membrane potential are a prominent feature of sensory, motor, and cognitive function, theirprecise role in signal processing remains elusive. Here we show, using a combination of in vivo, in vitro, and theoreticalapproaches, that both synaptically and intrinsically generated membrane potential oscillations dramatically improveaction potential (AP) precision by removing the membrane potential variance associated with jitter-accumulatingtrains of APs. This increased AP precision occurred irrespective of cell type and—at oscillation frequencies rangingfrom 3 to 65 Hz—permitted accurate discernment of up to 1,000 different stimuli. At low oscillation frequencies,stimulus discrimination showed a clear phase dependence whereby inputs arriving during the trough and the earlyrising phase of an oscillation cycle were most robustly discriminated. Thus, by ensuring AP precision, membranepotential oscillations dramatically enhance the discriminatory capabilities of individual neurons and networks of cellsand provide one attractive explanation for their abundance in neurophysiological systems.Citation: Schaefer AT, Angelo K, Spors H, Margrie TW (2006) Neuronal oscillations enhance stimulus discrimination by ensuring action potential precision. PLoS Biol 4(6): e163.DOI: 10.1371/journal.pbio.0040163IntroductionWithin neurons there exist various sources of membranenoise, including stochasticity of membrane conductances [1–3], stimulus nonspecific synaptic conductances [4–6], andvariable synaptic transmission [7]. While noise sources mayprove beneficial, for example, via stochastic resonance effects[8,9], they generally limit action potential (AP) precision andthus the fidelity of communication between cells [2,3]. This issupported by numerous experimental [10–13] and theoretical[14–16] studies that indicate that even for identical stimuli,the exact timing of AP discharge may differ substantiallybetween trials. Spike output, however, is all the information apostsynaptic neuron has available to potentially discernspecific patterns of activity occurring upstream in presynap-tic cells. Are there mechanisms in place that might increasethe robustness of spike discharge? Intrinsic events such asdendritic Naþ,Ca2þ,orN-methyl-D-aspartate (NMDA) spikes[17–22] might reliably signal the presence of a particular typeof event. Due to their all-or-none nature, they do not,however, readily permit the discrimination of more than asubset of stimuli. Furthermore, they appear insensitive tosubtle stimulus-specific differences in the temporal proper-ties of synaptic input patterns.In the scenario where individual synaptic events are largeenough to evoke spikes, it is well documented that specificpatterns of large fast input waveforms are one means ofproducing temporally precise AP discharge [10,13]. However,stimulus-evoked patterns of subthreshold activity oftenconsist of a series of rather small temporally dispersed eventsoccurring over a few to hundreds of milliseconds [5,23,24]. Inmost cells such inputs are typically processed in a highlynonlinear fashion in electrotonically dispersed dendriticlocations and result in patterns of AP discharge that reflectstimulus-specific properties of the input and subsequentintegrations performed within the cell [25]. Recently, it hasbeen shown that in small, electrically compact cells thatreceive very few synaptic contacts, even individual quanta arecapable of supporting reliable information transmission [23].Most cell types, however, receive an extremely high numberof synaptic inputs and a considerable fraction are needed toachieve threshold and encode a specific stimulus or motorcommand [5,26–28]. In such cases, many evoked currentsmust be integrated over time to produce AP discharge [28–30]—the patterns of which are thought to represent a specificsensorimotor signal.Membrane potential oscillations (MPOs) are a commonfeature of sensorimotor processing. They may be generatedby synaptic and/or intrinsic mechanisms and have beenattributed to synchronizing stimulus-relevant cell assemblies[31,32] and providing a phase ‘‘ tag’’ to individual spikes forefficient readout [33–39]. In the hippocampus and theolfactory system, MPOs, at the individual cell level, aresuggested to serve as an internal reference signal wherebyAcademic Editor: Mayank Mehta, Brown University, United States of AmericaReceived September 22, 2005; Accepted March 17, 2006; Published May 16, 2006DOI: 10.1371/journal.pbio.0040163Copyright: Ó 2006 Schaefer et al. This is an open-access article distributed underthe terms of the Creative Commons Attribution License, which permits unrestricteduse, distribution, and reproduction in any medium, provided the original authorand source are credited.Abbreviations: AHP, afterhyperpolarization; AMPA, a-amino-3-hydroxy-5-methyl-4-isoxazolepropionate; AP, action potential; EPSP, excitatory postsynaptic potential;InF, integrate-and-fire; MPO, membrane potential oscillation; NMDA, N-methyl-d-aspartate; PSTH, peristimulus time histogram; STDP, spike-timing–dependentplasticity; Vmpost, posthyperpolarization membrane potential; Vmpre, prehyperpo-larization membrane potential* To whom correspondence should be addressed. E-mail: [email protected] Biology | www.plosbiology.org June 2006 | Volume 4 | Issue 6 | e1631010PLoSBIOLOGYspike trains could encode information by their phase relativeto the background oscillation cycle [36,37,39].Here we have tested the hypothesis that MPOs facilitate thediscrimination of input patterns composed of small ampli-tude syna ptic events. We show that MPOs ensure APpreci sion in the olfactory bulb in vivo. Using in vitrorecordings, we subsequently describe the cellular mechanismsunderlying the optimization of AP precision by MPOs andhow this dramatically improves the discrimination of synapticinput patterns. Finally, using a variety of theoretical and invitro approaches, we show that for a broad range of MPOproperties, stimulus, and cellular parameters, ongoing oscil-latory activity ensures near-perfect discrimination of