Synaptic scaling in vitro and in vivo

853 TO THE EDITOR—The recent study by Magee and Cook1 in CA1 pyramidal neurons in vitro (see also ref. 2) raises a fundamental issue. Is the dependence of somatic EPSPs on the location of the dendritic synapses, which is expected from dendritic filtering, a ‘bug’ that should be rectified (for example, by mechanisms that eliminate voltage attenuation in the dendritic tree), or is this dependence a ‘feature’ that enhances the computational capability of the neuron? Magee and Cook’s direct dendritic measurements show that the synaptic conductance change, gsyn, becomes larger as one moves along the apical dendrite, away from the soma. This progressive increase in gsyn counterbalances the voltage attenuation imposed by dendritic cable properties, and consequently, the amplitude of unitary somatic EPSPs is insensitive to its dendritic origin (‘location-independent’ somatic EPSPs). If the location dependence of soma EPSPs is indeed removed, then “...all synapses will have the same ability to initiate action potentials and to induce long-term synaptic plasticity regardless of their location in the dendritic arborization”1 and, functionally, the neuron could be treated as a ‘point neuron’. But is it valid to assume that if, in vitro, the size of individual somatic EPSPs is independent of the dendritic input location, this would also remain true when many synapses bombard the dendritic tree, as is the case in vivo? We show that in the latter case, the locationindependence found in the quiescent in vitro condition is lost, and distal synapses become weaker at the soma than do proximal synapses (Fig. 1; see web supplement, http://www.nature.com/neuro/ web_specials/, for detailed figure legend). This is the result of a several-fold increase in dendritic membrane conductance, Gm, due to the activity of many synapses in vivo3–6. In other words, precisely the same mechanism of synaptic conductance change that is used for scaling up distal synapses destroys the ‘location independence’ (it is ‘self defeating’) when the network is active. The general argument is that if, in some reference cases, the scaling of synaptic conductance gives rise to location-independent EPSP amplitude at the Synaptic scaling in vitro and in vivo