Scientific Understanding of Consciousness
Dendritic Integration of Sensory and Motor Input
Nature, 492, 247–251 (13 December 2012)
Nonlinear dendritic integration of sensory and motor input during an active sensing task
Ning-long Xu, Mark T. Harnett, Stephen R. Williams, Daniel Huber, Daniel H. O’Connor, Karel Svoboda & Jeffrey C. Magee
Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, Virginia 20147, USA
Queensland Brain Institute, The University of Queensland, St Lucia, Queensland 4072, Australia
Active dendrites provide neurons with powerful processing capabilities. However, little is known about the role of neuronal dendrites in behaviourally related circuit computations. Here we report that a novel global dendritic nonlinearity is involved in the integration of sensory and motor information within layer 5 pyramidal neurons during an active sensing behaviour. Layer 5 pyramidal neurons possess elaborate dendritic arborizations that receive functionally distinct inputs, each targeted to spatially separate regions. At the cellular level, coincident input from these segregated pathways initiates regenerative dendritic electrical events that produce bursts of action potential output and circuits featuring this powerful dendritic nonlinearity can implement computations based on input correlation. To examine this in vivo we recorded dendritic activity in layer 5 pyramidal neurons in the barrel cortex using two-photon calcium imaging in mice performing an object-localization task. Large-amplitude, global calcium signals were observed throughout the apical tuft dendrites when active touch occurred at particular object locations or whisker angles. Such global calcium signals are produced by dendritic plateau potentials that require both vibrissal sensory input and primary motor cortex activity. These data provide direct evidence of nonlinear dendritic processing of correlated sensory and motor information in the mammalian neocortex during active sensation.
The unique integrative properties of cortical pyramidal neurons enable them to powerfully transform synaptic-input patterns delivered through structured network connectivity. Of particular interest is the ability of pyramidal-neuron dendrites to actively integrate input from spatially segregated and functionally distinct pathways (for example, sensory or motor pathways) when strong temporal correlations exist between these representations. Circuit computations based on active dendritic transformation of different streams of information by pyramidal neurons could underlie a variety of functions that include top-down cortical interactions, associative feature binding and predictive coding. To investigate the role of dendritic integration in behaviourally relevant network computations, we used two-photon microscopy to image dendritic Ca2+ signals from distal tuft branches of layer 5 pyramidal neurons of the barrel cortex (S1) labelled with a genetically encoded calcium indicator (GCaMP3) through a chronic imaging window while mice performed a task based on vibrissal active touch.
Recent recordings from populations of pyramidal-neuron dendrites suggest an active role for dendrites in information processing during the awake state. Here we report that novel long-duration plateau potentials and associated global dendritic Ca2+ signals are produced in the apical dendrites of individual layer 5 pyramidal neurons when mice perform an active sensing task. The Ca2+-mediated dendritic plateau potentials were probably evoked by correlated perisomatic input from ascending sensory drive and layer 1 synaptic inputs from cortico–cortical feedback connections. This active input processing in pyramidal-neuron dendrites seems to underlie a circuit computation for the tactile localization of salient objects. Thus, our results suggest that active nonlinear dendritic integration in layer 5 pyramidal neurons has a central role in the production of a behaviourally related computation in the barrel cortex. Because pyramidal neurons throughout the central nervous system (CNS) are capable of producing related forms of active dendritic integration, the above correlation-based computations may be a common feature of many neuronal circuits
[end of paraphrase]
Return to — Dendritic Trees
Return to — Neural Network